Direct and Indirect Speech

Trends in Linguistics
Studies and Monographs 31

Editor
Werner Winter

Mouton de Gruyter
Berlin · New York · Amsterdam
Direct and
Indirect Speech

edited by
Florian Coulmas

Mouton de Gruyter
Berlin · New York · Amsterdam
Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter & Co., Berlin.

Direct and indirect speech.

(Trends in linguistics. Studies and monographs ; 31)
1. Grammar, Comparative and general - Indirect
discourse - Addresses, essays, lectures. I. Coulmas,
Florian. II. Series.
P301.5.I53D57 1986 415 85-29795
ISBN 0-89925-176-5

CIP-Kurztitelaufnahme der Deutschen Bibliothek

Direct and indirect speech / ed. by Florian Coulmas.

- Berlin ; New York ; Amsterdam : Mouton, de Gruyter, 1986.
(Trends in linguistics : Studies and monographs ; 31)
ISBN 3-11-010599-3
NE: Coulmas, Florian [Hrsg.]; Trends in linguistics /
Studies and monographs

Printed on acid free paper.

© Copyright 1986 by Walter de Gruyter & Co., Berlin. All rights reserved, inclu-
ding those of translation into foreign languages. No part of this book may be re-
produced in any form - by photoprint, microfilm or any other means - nor trans-
mitted nor translated into a machine language without written permission from
Mouton de Gruyter, a Division of Walter de Gruyter & Co., Berlin.
Typesetting: Wagner GmbH, Nördlingen. - Printing: Ratzlow-Druck, Berlin. -
Binding: Lüderitz & Bauer GmbH, Berlin. - Printed in Germany.
Preface

The objective of this book is to clarify the notions of direct and indirect
speech. Its raison d'etre as a collection of articles is the fact that general
notions of this kind are best understood when investigated with respect
to several different languages. Only carefull comparative analysis can
show how general they really are.
The authors of this book all agree with me that reported speech is a
subject well worth the effort of a cross-linguistic co-operative approach.
It is thanks to them that this volume presents accounts of direct and
indirect speech in 14 different languages of almost as many language
families with occassional references to many others. I have learned a lot
from their contributions. Should this turn out to be true for other readers
too, then the book will have achived its purpose.
While working on this volume, I was a Heisenberg-Fellow of the
Deutsche Forschungsgemeinschaft, which I gratefully acknowledge.

F. Coulmas Tokyo, December 1984

Contents

Preface V
Contributors IX

Reported speech: Some general issues 1

Florian Coulmas

Direct and indirect speech: A functional study 29

Charles N. Li

Some remarks on direct and indirect speech in Slave (Northern

Athapaskan) 47
Keren D. Rice

Reported speech in Yoruba 77

Ay ο Bamgbose

Reported speech in Swahili 99

David P. B. Massamba

Speech reporting in the Caucasus 121

B. G. Hewitt and S. R. Crisp

Reported speech in some languages of Nepal 145

Karen Ebert

Direct and indirect speech in Japanese 161

Florian Coulmas

The particle -o and content-oriented indirect speech in Japanese

written discourse 179
Senko K. Maynard

Some semantic aspects of indirect speech in Hungarian 201

Ferenc Kiefer

Reported speech in Danish 219

Hartmut Haberland
VIII Contents

Reported speech in French and Hungarian 255

Ivan Fonagy

Introducing constructed dialogue in Greek and American conver-

sational and Literary Narrative 311
Deborah Tannen

Characteristics of direct and reported speech prosody: Evidence

from Spanish 333
Karen H. Kvavik

Index of subjects 361

Index of names 366

Contributors

Ayo Bamgbose Ibadan, Nigeria

Florian Coulmas Düsseldorf, FRG
S. R. Crisp Hull, England
Karen Ebert Marburg, FRG
Ivan Fonagy Antony, France
Hartmut Haberland Roskilde, Danmark
Brian G. Hewitt Hull, England
Ferenc Kiefer Budapest, Hungary
Karen H. Kvavik Minneapolis, U.S.A
Charles N. Li Santa Barbara, U.S. A
Senko K. Maynard New Brunswick, U.S. A
David P. B. Massamba Dar-es-Salaam, Tanzania
Keren D. Rice Toronto, Canada
Deborah Tannen Washington, D.C., U.S.A
Reported speech: Some general issues
Florian Coulmas

Summary

Reported speech has attracted the attention of scholars in several

different fields, such as, Linguistics, Poetics, Logic, and the Philosophy
of Language. In this paper I will review some of the literature most
relevant to a linguistic theory of speech reporting. In passing, I will also
refer to approaches quite outside Linguistics in a narrow sense, because
reported speech is one of those phenomena whose proper treatment
necessarily transcends departmental boundaries.
First, some of the fundamental differences between direct and indirect
speech are discussed with a special emphasis on a kind of sentence
meaning ambiguity only to be found in indirect speech. It arises because
in indirect speech the speaker is free to blend information about an
utterance with information about the world not conveyed by that utter-
ance. This ambiguity is thus known as the de dicto and de re ambiguity.
Next, a third kind of reported speech is introduced which has intri-
gued many authors as it shares essential features with both direct and
indirect speech.
Then I go on to discuss the question of 'what counts as the same,'
which is of central concern here because the purpose of speech reporting
is to convey what another speaker said, either in one's own words or by
reporting the same words that were used by that speaker. Prima facie
this distinction seems to be clear-cut. However, the notion of verbatim
rendition, that is, identity of form, seems to be culturally variable.
While differences between forms of speech reporting may, therefore,
be culturally determined to some extent, it is obvious that the grammati-
cal devices for integrating the speech of another into one's own differ
from one language to another. The remainder of the paper thus deals
with the question of how the function of speech reporting is grammati-
cally coded.
2 Florian Couhnus

1. Kinds of reported speech

Utterances can be made the subject of other utterances. They can be

criticized, questioned, commented on, or simply be reported. Language
can be used to refer to language. We can talk about talk. This is true for
all natural languages and is, indeed, a fundamental feature whose
absence disqualifies any sign system as a human language. However,
there are different ways of reporting the speech of another.
Traditionally, two kinds of reported speech have been distinguished:
oratio recta (direct quotation) and oratio obliqua (indirect quotation).
The former evokes the original speech situation and conveys, or claims
to convey, the exact words of the original speaker in direct discourse,
while the latter adapts the reported utterance to the speech situation of
the report in indirect discourse. The fundamental difference between the
two lies in the speaker perspective or point of view of the reporter:1 In
direct speech the reporter lends his voice to the original speaker and says
(or writes) what he said, thus adopting his point of view, as it were.
Direct speech, in a manner of speaking, is not the reporter's speech, but
remains the reported speaker's speech whose role is played by the
reporter.
(1) "Have you pray'd to-night, Desdemona?" asked Othello, and Des-
demona answered: "Aye, my lord."

The speaker of (1) phrases his report like a dialogue. Even though
Desdemona is not present in the report situation, Othello's utterance is
phrased as if she were, and, conversely, Desdemona's utterance presup-
poses the presence of Othello. The reporter thus steps back behind the
characters whose words he purports to report.
In indirect speech, on the other hand, the reporter comes to the fore.
He relates a speech event as he would relate any other event: from his
own point of view.
(2) Othello asked his wife whether she had said her nightly prayers,
which she affirmed.

(2) differs from (1) on several counts. First of all, there is no direct
address in (2) as direct address requires the presence of the addressee.
The second person address pronoun and the vocative, "Desdemona," in
Othello's utterance in (1) are replaced by a third person pronoun, "she,"
and a descriptive term, "his wife," both reflecting the reporter's point of
view. The tense of Othello's utterance is also changed: present perfect in
 Reported speech 3

(1) becomes ante-preterit in (2), a point to which we will return below.

Finally, Desdemona's utterance in (1) is reduced to a relative clause in
(2), where "which" refers to the content of the indirect question in the
preceding clause, "whether she had said her nightly prayers". Alternati-
vely, her utterance could be rendered, somewhat redundantly, by ma-
king the contents of her affirmation explicit: "She affirmed that she had
said her nightly prayer". Notice that Desdemona's utterance has no
propositional content of its own. Its function is purely communicative,
and this communicative function, the affirmation of the proposition of
Othello's question, cannot be expressed but only described in indirect
discourse. The expressive term, "aye," is thus replaced by a descriptive
term, "affirm," which gives a category label to her speech act.
Another apparent difference between (1) and (2) is the descriptive
term referring to Desdemona in (2). Othello, of course, knows that
Desdemona is his wife - this relationship is, after all, at the bottom of all
his troubles - but he doesn't say so. "His wife" is a term that is provided
by the reporter, who knows more about Othello and Desdemona than
can be inferred from (1). In indirect speech, the reporter is free to
introduce information about the reported speech event from his point of
view and on the basis of his knowledge about the world, as he does not
purport to give the actual words that were uttered by the original
speaker(s) or that his report is restricted to what was actually said.
Indirect speech is the speech of the reporter, its pivot is in the speech
situation of the report.

1.1. De dicto and de re

The possibility for the reporter to introduce additional information from

his own point of view in reporting what someone else said creates
potential ambiguities in indirect discourse, because it is not always
unequivocally clear whether a descriptive term testifies to the appraisal
of the original speaker or that of the reporter. This ambiguity is known
as the de dicto vs. de re interpretation of descriptive terms in opaque
contexts (cf. Partee 1973: 414f.). A classical example is that of Oedipus
and Jocasta.
(3) Oedipus said that his mother was beautiful.2
(3) has two readings, one in which the assertion is that Oedipus said
that an individual that he identified as his mother was beautiful, and
another where that individual is so identified only by the reporter but not
4 Florian Coulmas

by Oedipus. According to the former, the de dicto reading, the original

utterance by Oedipus must have been something like 3
(4) My mother is beautiful.
In the latter, the de re reading, Oedipus could have made any of the
following and several other utterances.
(5a) Jocasta is beautiful.
b) My wife is beautiful.
c) The mother of my children is beautiful.
If any of (5) rather than (4) was the original utterance (which must
have been the case, because Oedipus blinded himself in desperation
upon learning that the woman he found beautiful was his mother) then
only the de re reading of (3) would be true.
The general problem is that in opaque contexts codesignative terms
(Quine 1960: 151) cannot be exchanged salva veritate, that is, without
changing the truth or falsity of what is said. Thus the de dicto reading of
(3) would be false even if both (6) and (7) were true.

(6) Jocasta is Oedipus' mother.

(7) Oedipus said, "Iokaste is beautiful."

Similar problems arise when attitudinal or value loaded expressions

occur in the dependent clause of indirect speech.
(8) John says that this lunatic doesn't know what he is saying.
The italicized expression in (8) may be part of the report or an
addition by the reporter. In some cases one reading is preferred over the
other, as in (9),
(9) John asked me to dance with his hysterical wife.
where "hysterical" is likely to be the speaker's value judgement rather
than John's. Generally, however, it cannot be determined whether
identifications in indirect discourse belong to the original utterance or to
the report. This kind of ambiguity cannot occur in direct discourse,
because no interference of reporter's speech and quoted speech is
possible. Direct speech always has a de dicto interpretation.
The peculiarity of indirect discourse is that the reported utterance is
processed by the reporter. It presupposes his analysis. "Analysis is," as
Voloshinov (1973: 129) said, "the heart and soul of indirect discourse."
The reporter's analysis can be de dicto or de re. In the former case it
 Reported speech 5

consists in adjusting the original utterance to the deictic center of the

report situation without changing any other part of its linguistic form. De
re analysis, on the other hand, allows the reporter to alter the form of the
original utterance in accordance with what it means on the basis of this
knowledge of the world in such a way that it may include inferences of
which the original speaker is unaware. Grammatically, de dicto and de re
indirect discourse are not distinct in spite of the profound difference in
meaning. An utterance of the form

(10) X said that complement clause

in its de dicto reading implies the claim that the complement clause is
faithful to the form of X's original utterance except for the necessary
deictic adjustments. In its de re reading it implies a quite different claim
on the part of the speaker which might be paraphrased as follows.
(11) X said something which in conjunction with what I, the speaker,
know about the world, and about the subject of X's utterance in
particular, justifies the statement that
complement clause

What appears to be simply the alternative to direct discourse is thus a

complex assembly of ways of reporting another's speech or certain
aspects thereof, all labeled "indirect discourse." They differ with respect
to the faithfulness to the form of the original utterance and with respect
to the processing of its content by the reporter. The combination of these
two dimensions of variation make indirect speech a versatile mode of
speech reporting ranging from faithfully adapting the linguistic form of
the reported utterance to the deictic center of the report situation to a
summarizing paraphrase of an utterance irrespective of its linguistic
form.
Complement sentences in indirect speech can be qualified by com-
ments referring to the form of the reported utterance.
(12) Humpty Dumpty said that it was a most provoking thing when a
person didn't know a cravat from a belt, and those were his exact
words.

Humpty Dumpty's remark about the "most provoking thing" was

addressed to his interlocutor, Alice, who made the blunder of mistaking
his cravat for a belt. Thus, he couldn't very well have used the past tense
in making his accusation. Nevertheless, the comment "those were his
exact words" doesn't strike us as strange or illogical (cf. Banfield 1973:
9). What "those were his exact words" in an indirect speech context is
6 Florian Coulmas

apparently taken to mean is "his exact words, allowing for the changes
necessitated by grammatical rules and/or the obvious shift of deictic
center." In reporting another's speech the speaker can indicate in
various ways that he intends his report to be faithful to the form of the
reported utterance, in spite of its being phrased in indirect speech. On
the other hand, indirect speech can be introduced by phrases qualifying
it as a report of only the content of what another speaker said.
(13) The message of his book is that computerization is no salvation.
(14) The president made a statement to the effect that the present govern-
ment did not favor military intervention.

The italicized passages in (13) and (14) indicate that the speaker reports
content, not form.
While direct speech purports to give a verbatim rendition of the words
that were spoken, indirect speech is more variable in claiming to
represent a faithful report of the content or content and form of the
words that were spoken. It is important to note, however, that the
question of whether and how faithful a given speech report actually is, is
of a quite different order. Both direct and indirect speech are stylistic
devices for conveying messages. The former is used as if the words being
used were those of another, which are therefore pivoted to a deictic
center different from the speech situation of the report. Indirect speech,
in contrast, has its deictic center in the report situation and is variable
with respect to the extent that faithfulness to the linguistic form of what
was said is being claimed.

1.2. A third kind

The previous section has demonstrated that a simple dichotomy of direct

versus indirect cannot do justice to the complexities of reported speech.
In literary narratives a third kind of speech (and thought) reporting is
very common which makes its conceptual penetration even more com-
plex. Several nomenclatures have been proposed, behind which lie
conflicting theoretical notions of what its essence is.
As early as 1894, Tobler mentioned a third kind of speech reporting
which he defined as "a peculiar mingling of direct and indirect dis-
course".4 The ensuing discussion concentrated on the French language,
as French prose (La Fontaine, Zola, Merimee) seemed to offer the most
examples of this stylistic device, which exhibits features of both direct
 Reported speech 7

and indirect discourse. In Tobler's view, this style was a variant of direct
speech.
Kalepky (1899) proposed to treat the third kind as a completely
autonomous rather than a mixed style, on a par with direct and indirect
speech; for it, he introduced the term "veiled speech" (verschleierte
Rede). The stylistic veil covers the speaker, leaving it up to the reader to
determine whether the speaker of a given section of a narrative is the
hero or the author. Bally (1912) thought that this style was peculiar to
French and introduced the form "free indirect style" (style indirect libre),
thus classifying it as a kind of indirect speech. Lorck (1921) argued that
"indirect" was a misnomer for this kind of reported speech, which,
moreover, was to be found not only in French but in other languages as
well. He proposed the term "experienced speech" (erlebte Rede), which
was criticized by Jespersen, because "the writer does not experience or
"live" (erleben) these thoughts or speeches, but represents them to us"
(1924: 291f.). He therefore promoted the term "represented speech" for
this narrative style, which is typical for prose" where the relation of
happenings in the exterior world is interrupted - very often without any
transition like "he said" or "he thought" - by a report of what the person
mentioned was saying or thinking at the time, as if these sayings or
thoughts were the immediate continuation of the outward happenings"
(1924: 291). Jespersen, like Bally, conceived of represented speech as a
kind of indirect speech.
In a sense, the conflicting nomenclatures reflect a disparity of gram-
matical form and fictional content characteristic of this style as compared
with non-narrative speech. The omniscient author can freely invade
other minds and relate events, utterances, and thoughts as if he were a
witness to a scene to which no witness can have access. Grammatically,
much of what Jespersen called "represented speech" is phrased from the
point of view of the narrator, but contentwise it belongs to the hero's
speech, thought, or perception.5
Lerch (1919) emphasized the latter aspect and hence chose the term
"quasi-direct speech" (uneigentliche direkte Rede) which was also adop-
ted by Voloshinov (1929), who called it "the most neutral of all terms
proposed and the one entailing the least amount of theory" (1973: 141).
His concern was not with providing abstract grammatical descriptions
but rather to investigate reported speech from a historical point of view,
in order to demonstrate how languages at different historical stages of
their development perceive the words of another speaker. Forms of
reported speech are at the interface of grammar and style, and as such
8 Florian Coulmas

they can reveal, to the careful analyst, linguistic reflections of social

developments: "In the vicissitudes of the word are the vicissitudes of the
society of word-users" (Voloshinov 1973: 157). Unfortunately, Voloshi-
nov's approach did not inspire much systematic research into the interac-
tion of grammar and style or the societal preconditions of different
patterns of reported speech in different languages at different historical
stages. The chapters of this book cannot close this gap, but they facilitate
comparative analysis.
The above review of terms proposed for reported speech of the third
kind can be summarized as follows. In the following I will adopt Lech's
and Voloshinov's usage.
1894 Tobler direct speech indirect speech

mingling of direct and

indirect speech

1899 Kalepky direct speech indirect speech veiled speech

1912 Bally direct speech indirect speech

free indirect style

1919 Lerch direct speech indirect speech

quasi-direct speech

1921 Lorck repeated speech experienced speech reported speech

1924 Jespersen direct speech indirect speech

dependent represented
speech speech

1929 Voloshinov substituted discourse quasi-direct discourse indirect discourse

As an example of quasi-direct discourse consider now a passage from

Beckett.
(1) He begged her to believe him when he said he could not earn. (2) Had he
not already sunk a small fortune in attempts to do so? (3) He begged her to
believe that he was a chronic emeritus. (4) But it was not altogether a question
of economy. (5) There were metaphysical considerations, in whose gloom it
appeared that the night had come in which no Murphy could work. (6) Was
Ixion under any contract to keep his wheel in nice running order? (7) Had any
provision been made for Tantalus to eat salt? (8) Not that Murphy had ever
heard of. (Samuel Beckett. 1963. Murphy. London: Calder & Boyars, p. 18).
 Reported speech 9

Even though only third persons occur in this passage, both the narrator's
and hero's points of view are represented to the reader. Sentence (1) is,
seemingly, purely descriptive; it says something about the hero from the
narrator's point of view. (2) however, is phrased like a question whose
answer is known to the addressee, or rather, is made known to the
addresse by the question itself, a rhetorical question, that is. This cannot
be the reader as he is not directly addressed by the hero of the narration.
Hence, it can only be the hero's interlocutor which implies that the
question is to be understood as being phrased from the hero's point of
view. But, when the author lets the hero speak, one should expect a first
person subject: "Have I not already sunk . . . " Instead, we have a third
person subject which would accord with the narrator's perspective.
The change of perspective is brought about without any overt indica-
tion or explicit statement. In the sequel, too, both points of view are
intertwined, and it is left up to the reader to disentangle them. Gramma-
tically, it is not even clear that sentences (3) to (8) represent what
Murphy says rather than his thoughts. (3) surely allows for a descriptive
interpretation, that is, one from the narrator's point of view, and in the
other sentences the author continues to convey his point of view,
especially in (5) and (8) where he refers to the hero by name. On the
other hand, except for the proper name and the tense of the verb, (8) is
phrased like an expressive sentence suggesting direct discourse and, by
consequence, the hero's point of view. (6) and (7) are again rhetorical
questions. (8) cannot be embedded without substantial changes in the
wording into a matrix sentence as indirect discourse.

(15) *Murphy said that not that he had ever heard of.

The expressive "not that . . . " belongs to the hero, the proper name and
the ante-preterit to the narrator. Both points of view are fused in the
same sentence. The next sentence, following (8), then makes it clear that
the preceding sentences represent Murphy's utterances rather than his
thoughts, as they are a reaction by Murphy's interlocutor, Celia, in direct
speech:
(16) "But we cannot go on without any money", said Celia.

In quasi-direct discourse, the narrator lends his voice to the hero without
giving up his own identity. Rather than formally distinguishing direct
discourse from its introductory context and thus holding hero and
narrator apart, the role playing (direct discourse) is interwoven with, and
adjusted to, the more distanced viewpoint of the narrator. Tobler (1894)
10 Florian Coulmas

was not altogether wrong when he described quasi-direct discourse as "a

peculiar mingling of direct and indirect speech," but we can also under-
stand Voloshinov's emphatic rejection of this definition, because "quasi-
direct discourse is not a simple mechanical mixture or arithmetical sum
of two forms but a completely new, positive tendency in active reception
of another person's utterance" (Voloshinov 1973: 142).

2. What counts as the same?

As a conscious stylistic device, quasi-direct discourse is a fairly recent

phenomenon that developed in the 19th century, whereas the mixing, or
confounding, of points of view can be observed much earlier. As a
matter of fact, it seems to be typical of earlier forms of literature where
the narrator had not yet mastered the artful technique of choosing and
changing points of view. For Classical Greek, for instance, Rosier (1980:
303) observes the following: "Ob Autorenäußerung oder direkte Rede
einer handelnden Person, macht für Heraklit keinen Unterschied."6
Similarly, Old French texts are known for this kind of disarray of points
of view: "No clearly marked boundaries between direct discourse and
indirect discourse existed then. The Old French storyteller was as yet
unable to separate the figures of his fantasy from his own T" (Voloshi-
nov 1970: 150). Not every melange of points of view is, therefore, to be
regarded as quasi-direct discourse.
Voloshinov's approach to the analysis of speech reporting was rather
original, because (i) he was prepared to look for non-linguistic explana-
tions for the presence or absence of a given kind of speech reporting in a
language, and (ii) he admitted that "the borderline between grammar
and style is fluid" (1973: 126). However, what his approach has in
common with virtually all preceding and subsequent studies on reported
speech is that he concentrates on written language, literary prose in
particular.
Writing is an important cultural achievement which, in contradistinc-
tion to many other such achievements, is closely liked with language. A
cross-linguistic study of reported speech has to take into account the
possibility that writing itself influences the way how speech reporting is
carried out and understood. That writing is not simply speech written
down is by now a common place (see, e.g., Tannen 1982, Coulmas
 Reported speech 11

1981). Anthropological studies have shown that the introduction of

writing has profound effects on the development of languages and on the
ways they are perceived and conceptualized by their speakers (see
Goody and Watt 1968; Coulmas and Ehlich 1983).
One important point is the question of what counts as the same.
Anthropologists have argued that the same is not the same in oral and
literate cultures.
"Even the most standardised segments of oral sequences never become so
standardised, so formulaic, as the products of written man. Reproduction is
rarely if ever verbatim" (Goody 1977: 118).

Goody's observation seems to imply that reproduction in speech repor-

ting is verbatim in literate societies. And, indeed, the conventions for
using quotation marks, for instance in newspapers, are such that what is
enclosed in quotation marks is claimed to be a verbatim rendition of
what a speaker said. Direct discourse, in this understanding, means
fidelity not only to the content but also to the surface form of a reported
utterance. Notice, however, that the notion of fidelity to the surface
form of an utterance is not self-evident. Rather, it requires explanation.

2.1. Type-token identity

(17) It is too late.

(17) can be used to make a statement about the time of day, an

appointment, a schedule. But if the speaker continues, saying,
(18) This is what Othello said to Desdemona before he killed her.

then he, the speaker, makes no statement at all with (17) and does not
refer to any particular time, but only reports Othello's words. It is
Othello who is making the statement. The deictic word this in (18) refers
to (17), and (17), in the context of (18), refers to itself. Alternatively,
this could also be supplanted by (17) which would thus become the
subject of (18) as in (19).
(19) "It is too late," is what Othello said to Desdemona before he killed
her.

In (19) the expression enclosed in double quotes behaves much like a

demonstrative which in conjunction with a physical gesture of pointing is
used to identify an object. Likewise, (17) in (19) functions as a means of
naming a referent by demonstration. This is what quotation is all about:
12 Florian Coulmas

The speaker does not claim authorship for a part of his utterance which
he ascribes to another speaker or unspecified source. This part of his
utterance does not serve a regular referential function such that words
refer to things. Rather, they refer to words, not to any arbitrary words,
that is, but purportedly to those words that some other speaker uttered
at some other time.
This functional duplicity of words which quotation exhibits so clearly
and which is so important for natural languages poses a number of
challenges to linguists and philosophers. 7 Analysts in the philosophical
tradition, for instance, Linsky (1967), Plantinga (1969), and Quine
(1960, 1966) have focused on problems of reference. "Quotation,"
writes Quine (1966: 159), "is the referrentially opaque context par
excellence.'" Referring is, therefore, considered by some authors, no-
tably Searle (1969), as an act rather than a property of expressions.
Referring is something that people do when they use words, but not
when they mention them. In quotation, the normal referential function
of words is suspended, because the words that we utter when we quote
are not our own.
The above may look like a rather cumbersome way of saying that
quoting means to repeat the words of another, but it isn't all that easy.
Words are ephemeral entities and cannot be repeated in the same sense
as the showing of a slide on a screen can be repeated. We can, of course,
use a tape recorder and play the same recording over and over, but this is
not what we usually mean by to repeat one's own or another speaker's
words. What we mean is that we produce a word or words of the same
type as the ones uttered by the quoted speaker. The physical tokens are
singular events and as such not reproducible. Yet we can repeat the
words of others and reasonably make statements such as the following:

(20) Mary just said what Othello said to Desdemona before he killed her.

Clearly, (20) does not mean that Mary reproduced the physical events
that Othello once produced with his vocal tract before stifling Desde-
mona. Rather, what it means is that Mary produced an utterance token
of the type of which Othello's utterance was also a token. Quotation in
natural languages, and in other sign systems as well (see Goodman
1968), presupposes the structural possibility of establishing type-token
identity for utterances, a possibility that we would want to regard as a
universal feature of natural language.
To repeat the words of another in the sense just explained is some-
thing that in principle can be done in any language. But notice the
 Reported speech 13

qualification, "in principle." That something can be done in principle

does not imply that it is done, or that it is done in the same fashion in all
natural languages. In his book Semantics, Lyons offers some interesting
considerations in this connection:

(21) "John said X and so did Mary.

"John and Mary had produced two different tokens of the same type. So much
is clear. The difficulty lies in specifying precisely what X can cover and the
criteria for type-token identity between different instances of X . A s long as we
restrict our attention to some standardized written language or operate solely
with written representations of spoken forms (and especially so, if we make
use of non-cursive, printed representations in a alphabetic script), we may be
inclined to underestimate the difficulty of specifying the conditions under
which (21) would be true or false" (Lyons 1977: 17).

Type-token identity is a rather rigid notion whose meaning, Lyons

suggests, is easily understood in a literate culture where the relevancy of
this principle is omnipresent, since the technique of writing, and printing
in particular, is the technique of linguistic repetition. In other cultures, it
may be more difficult to specify exactly what counts as the same.
It is by virtue of the type-token relation that we can conceptually
distinguish between reporting form and reporting content. The question
of what counts as the same thus allows for different answers. It appears,
however, that the awareness of this distinction is, to some extent at least,
culturally determined. Not all speech communities make a rigorous
distinction between "same form" and "same meaning". In a paper about
oral poetry of the Limba of Sierra Leone R. Finnigan wrote:

When I asked a Limba assistant to elucidate the words I could not catch fully
while trying to transcribe taped stories, he could not be made to understand
that I wanted the exact words on the tape. A s far as he was concerned any
comparable phrase with roughly the same meaning would do. (Quoted after
Olson, Hildyard 1983: 293.)

It is, maybe, a matter of debate whether one such observation warrants

any general interpretation, but if it does, the implication seems to be that
the distinction of form and meaning with respect to linguistic utterances
is so unimportant in Limba culture that it is all but imperceptible to its
members. Such cultural differences may or may not coincide with
linguistic differences. So far we have no convincing theory about a causal
nexus between the two. But so much is clear: Languages differ widely as
regards overt distinctions between kinds of reported speech preserving
content or both content and form. 8
14 Florian Coulmas

3. How much is coded in the grammar?

"The changing sociolingual conditions of verbal communication are what

in fact determines the forms of reported speech" (Voloshinov 1973:
123). No matter whether or not one is inclined to subscribe to this
somewhat deterministic statement, it is an undeniable fact that "the
forms of reported speech" differ greatly across languages. We find
significant differences even in areas, which, at first sight, one would be
tempted to take for universal rather than language specific. There are
different ways of integrating reported speech into one's own; there are
differences as to how reporting devices get grammatically coded; and
there are differences concerning the kinds of reported speech that are
distinguished in a language.
Some languages have several different means of indicating whether a
given report is direct or indirect: tense, mood, complementizer, and
word order. German is an example where all of these grammatical
devices are employed for this purpose. The distinction between direct
and indirect speech is hence highly prominent. In other languages only
some of the above categories are made use of. English, for instance, has
no subjunctive mood, and changes in word order are fewer than in
German. Thus ambiguous sentences that could be read as both direct or
indirect speech are relatively more frequent than in German. They are
more frequent yet in a language such as Japanese where no changes in
word order or tense occur, and mood is not used to mark a piece of
discourse as the speech of another as French subjonctive or German
Konjunktiv. In Russian, too, indirect speech is not so clearly marked,
because it has neither a subjunctive mood nor a consecutio temporum.

3.1. The sequence of tenses

The so-called sequence of tenses is a grammatical feature of indirect

speech that many languages lack. In some, like for instance Russian, the
inflected verb in indirect speech remains in the same tense as would be
used in the corresponding independent utterance, no matter how its
original time reference relates to the report situation. The same is true of
Hungarian (cf. Kiefer in this volume). In other cases, however, there is
no back-shifting of tenses simply because there is no tense system.
Yoruba may be cited as an example of this type (cf. Bamgbose: in this
volume).
 Reported speech 15

An important conclusion to be drawn from the observation that there

are languages that have morphological tense systems but no sequence of
tenses rules is that back-shifting of tenses is a conventional phenomenon
rather than a necessary consequence of tense as a grammatical category.
Tense and sequence of tenses are independent features of individual
grammars. Special tense rules may be developed for particular styles,
such as, the Latin "epistolary tenses," in which the writer of a letter
adopts the viewpoint of his reader and therefore uses preterit or perfect
to refer to events that are present at the time of his writing. The back-
shifting of tenses in English must have evolved on the basis of similar
conventions. In his famous book The Philosophy of Grammar Jespersen
remarks that "the shifting of tenses in indirect speech is very natural and
in many cases even inevitable" (1924: 293). But if seen from the point of
view of languages with no tense system or no sequence of tenses rules it
is hardly so natural. Rather, for native speakers of such languages it may
be hard to understand that (22) does not mean that John referred to
Mary's prettiness as a matter which now, unfortunately, belongs to the
past, or why somebody's presence here is referred to in the past tense as
in (23).
(22) John said to Mary that she was pretty.
(23) He said you were here.

Shifting of tenses is "natural" within in the confines of a given grammati-

cal system but is by no means logically necessary or implied by the
presence of a tense system. Thus there is considerable cross-linguistic
variation as regards the rigidity of sequence of tenses rules.
Unlike English, German has both tense and mood shifts. Not all
morphological forms are, however, distinct. The 1. pers. sing, present
indicative form is by and large the same as the corresponding subjunctive
form (except for the verb sein (to be): Ich gehe, ich nehme, ich komme,
ich habe, etc.). In general, the past tense distinguishes the subjunctive
mood more clearly than the present and therefore is often used where
the present tense should be used if the sequence of tenses were the only
determining factor.
(24a) Er behauptet, ich habe das getan.
He claims I did it
b) Er behauptet, ich hätte das getan.
(24b) is preferred over (24a) although the latter harmonizes the tenses of
matrix and complement clauses in accordance with the consecutio tem-
16 Florian Coulmas

porum, i.e., present in the matrix clause and perfect in the complement
clause, whereas the complement clause in (24b) is in plusperfect. This
preference will be particularly strong when the speaker wants to express
doubt or, as in (24) is not in agreement with the proposition of the
reported clause. If, on the other hand, he is committed to the truth of the
proposition of the complement clause he might as well choose a form
that is indeterminate in its mood as in (25).
(25) Er glaubt nicht, daß ich das getan habe.
He does not believe that I did it.
The sequence of tenses in German is thus not a rigid principle that
requires tense forms in complement clauses to be adapted to those of
their matrix sentences following a mechanical procedure, but is rather an
instrument for differentiating subtle shades of meaning (cf. Wunderlich
1972). All possible combinations of matrix and complement clauses in
(26) are acceptable, but the complement clauses differ with respect to
the speaker's certainty about the truth of the proposition expressed.
They are ordered for declining certainty.
a. er kommt.
(26) Er sagt \ b. er werde kommen.
Er sagte 1 c. er komme.
Er hat gesagt J d. er würde kommen,
e. er käme.
Complement clause (a) (present indicative) in conjunction with any of
the matrix clauses expresses a high degree of certainty on the part of the
speaker; also wird er auch kommen ('that means he will come') would be
a likely continuation, (d) and (e) (preterit subjunctive), on the other
hand, indicate doubt and are more likely to be followed up by a negative
statement, such as, aber das ist recht unwahrscheinlich ('but it is rather
unlikely').
In English the sequence of tenses rules are more rigid as they do not
interfere with shifting of mood. Basically, there is a rather simple
grammatical rule whereby the verb in the complement clause must be
shifted back into the past wherever this is possible. A past tense form in
the complement clause is thus to be interpreted as present in the
purported original utterance, etc. The details of this rule have been
described by Jespersen (1924: 292ff.), Quirk et al. (1972: 786f.) and
recently Comrie (1984). From their accounts as well as from the above
observations that no back-shifting occurs in other tense languages it
becomes clear that sequence of tenses rules are independent not only
from tense as such but also from deictic changes.
 Reported speech 17

3.1.1. Temporal deixis

As tense is, among other things, a means of time reference, it would not
be alltogether unreasonable to assume that the deictic pivot determines
shifting of tenses in indirect speech, because, after all, the adjustment of
the subordinate clause to the deictic pivot of the reporter is one of the
most conspicuous features of indirect discourse. In a paper I read at the
XHIth International Congress of Linguists in Tokyo and which was
published later in the Journal of Pragmatics (Coulmas 1985) I wrote the
following.
Changes in tense and mood required by complementation are highly language
specific . . . Grammatical changes induced by embedding sentences in indirect
speech vary greatly across languages. As opposed to these switches, the deictic
switches have nothing to do with the grammatical form of sentences and are
not language specific. Rather they are required by the logic of indirect
quotation. . . . However much the grammar (and lexicalization) of deictic
expressions may differ from one language to another, deictic changes necessi-
tated by pivotal changes in speaker perspective are bound to follow the same
logical principles regardless of particular languages.

I now believe that things are more difficult, and that the distinction
between grammatical switches on the one hand and deictic switches on
the other is not as clear-cut as I first assumed. That grammatical switches
in indirect discourse vary greatly from one language to another is trivially
true, because the grammars themselves are so very different. It is also
true that deictic switches are conditioned by the speakers' location in
space and time, that is, by the ways their utterances are pivoted in a
given deictic space. However, my conclusion from these observations
was rash. The fact that deixis has to do with the physical world as it is and
the way a speaker is located in it and perceives it does not imply that
deictic terms work in a similar fashion independent of the languages to
which they belong. The consecutio temporum is where deixis and gram-
mar meet. Consider once again temporal deixis.
(27) John said that he would come tomorrow.

In English there is only one way to read this sentence: Tomorrow refers
to the day following the day on which (27) was uttered. The pivot of this
deictic term is, in other words, the speech situation of the report, not the
speech situation of John's original utterance. Indeed, it couldn't be
otherwise, because the pivoting in the actual speech situation is the very
essence of deictic terms, or is it not?
Tomorrow is one of the clearest cases of a deictic word. Yet, its
18 Florian Coulmas

equivalents in other languages do not need to be changed in indirect

reports of utterances in which they occur. As mentioned above, there is
no morphological distinction in Yoruba between present and past tense.
Thus, obviously no sequence of tenses rules apply. Less obvious are the
reasons why time adverbs do not require any changes either.
(28a) ό ηί, "ma ä lo lola."
he say I will go tomorrow
He said, "I will go tomorrow."
b) ό ni oun ä lo lola.
he say he will go tomorrow
(Bamgbose's example, this volume)
In (28b) pronoun reference has been changed in accordance with the
reporter's perspective. Lola ('tomorrow'), however, does not need to be
changed, as long as its time reference is contextually clear. One might
want to argue then that lola does not really mean 'tomorrow', but rather
means 'the following day.' But this assumption can hardly solve the
problem. Yoruba is not an exceptional case. Similar observations about
choice of temporal adverbs are reproted for other languages, too. Hewitt
and Crisp (this volume) have noted that the choice between the time
adverb appropriate at the time of the original utterance and that whose
deictic pivot is in the speech situation of the report is free in several
Caucasian languages as long as it is clear from the context what day is
under discussion.
In the paper quoted above I stated that "in indirect speech . . . deictic
expressions have to be adapted to the speech situation of the report in
such a way as to be consistent with the new speaker perspective." 9 But in
light of the examples just quoted this claim is no unconditionally tenable.
There is no compelling reason why tense in indirect speech should not be
used in accordance with the deictic center of the reported speaker. Thus,
while there may be a shift of deictic center for person reference in
indirect discourse relative to the purported original utterance, tense
may, nonetheless, be interpreted from the viewpoint of the reported
speaker. Consider the following Russian example from Comrie (1984).
"Let us imagine that today is the fifteenth of May. Kolya uttered the following
on the eighth of May: ja pridu cetyrnadcatogo maja Ί will arrive on the
fourteenth of May.' If I report this today in Russian, then I will say: Kolja
skazal, cto on pridet cetyrnadcatogo maja 'Kolya said that he would arrive
(literally: will arrive) on the fourteenth of May.' The tense in this sentence is
future pridet, although from the deictic centre of the report it has past time
reference, given its collocation with the adverbial cetyrnadcatogo maja, which
has past time reference."
 Reported speech 19

Notice that the reported sentence is clearly marked as indirect speech by

the complementizer cto. Moreover, the person reference is adjusted to
the deictic pivot of the reporter, the original speaker being referred to by
the third person pronoun on. Yet, the tense of the verb remains the same
as in Kolya's original utterance and is thus not treated as a deictic
element also to be adjusted. Obviously, then, Russian allows for senten-
ces with split deixis in indirect speech, the pronoun deixis being adjusted
to the report situation while temporal deixis by means of tense forms
keeps its pivot in the reported situation. Therefore, indirect speech
appears to be integrated into the reporter's speech to a lesser degree
than in other tense languages, such as, English or German, where tense
forms cannot, or not always, stay unchanged in indirect speech.

3.2 Grammatical integration

The general conclusion that follows from the above is that the extent to
which reporting and reported parts of indirect discourse are harmonized
and integrated with one another varies within and across languages. Li
(in this volume) uses the notion of 'fusion' to refer to this integration. If
we compare, for example, (29a) and (b), the Latin sentence exhibits a
higher degree of integration than its English counterpart.

(29a) Dico Ciceronem edere oleas.

b) I say that Cicero is eating olives.

In the English sentence what follows the complementizer that is a

potentially independent sentence whose subject is in the subject case. By
contrast, the reported sentence in (29a) has lost its independence com-
pletely by being put in an accusative with infinitive construction, typical
of indirect discourse in Latin, where the verb of the embedded clause is
non-finite and its subject appears in the accusative case.10
Greek exhibits a number of alternative constructions for indirect
speech ranging from low to high integration of matrix and subordinate
clause. (30a) resembles the English complement construction with that,
while (30b) ist much like the Latin ACI structure.
(30a) Aggelei höti Küros poreütai.
b) Aggelei Küron poreuesthai.
He reports that Cyrus is on the march.
20 Florian Coulmas

In (30a) hoti is a complementizer comparable with English that, and what

follows is like an independent sentence, which can be used to make the
statement 'Cyrus is on the march.' In (30b) poreuesthai is the infinitive,
its subject Küron being in the accusative case. The latter is thus more
highly integrated than the former. In Greek preference for one construc-
tion or the other is conditioned by the respective report verb, but some
verbs admit both.
An even higher degree of integration than in ACI sentences can be
found in constructions where the reported part is reduced to a noun
phrase as in (31).
(31a) He asked for permission.
b) She denied the existence of God.
c) They announced the victory.

It may seem far-fetched to subsume sentences such as these under the

notion of reported speech, but notice that they share some crucial
features with indirect speech: There is a report verb and a "propositio-
nal" part. The latter is syntactically reduced to a noun phrase in object
position, but then this is exactly the position that a complement sentence
occupies. It is surely no coincidence that complementizers such as
German daß, English that, or French que are identical in form with
demonstratives or relative pronouns that point to an object.

(32a) John told Danny that. Art is art.

b) Art is art. That is what John told Danny.
c) John told Danny that art is art.

Functionally the thats in (32a), (b) and (c) differ. In (a) and (b) that
behaves like a proform whose referent is another sentence as in That's
true. But in (c) that behaves like a syntactic connector within a sentence
by establishing a relation of subordination between its parts. To present
day native speaker intuition these two thats are rather distinct, but this
grammatical distinction has only gradually emerged from a common
source. The same is true of the German complementizer daß whose
spelling differentiation from the article and demonstrative das is of
recent origin. Both the English and German examples show what a
complement clause is: a sentential object. This is also very clearly
evidenced in Japanese where the direct object particle -o can be used for
embedding indirect speech (see Maynard in this volume). Thus nominal
objects of report verbs as in (31) are not all that different from comple-
 Reported speech 21

ment clauses. Yet a significant difference can be seen in the respective

degree of integration of reporting and reported parts of the sentence.
The reduction of a proposition to a noun phrase is the extreme form of
integration of reporting and reported parts in speech reporting. The
other extreme is direct quotation where reporting and reported parts are
distinct to the extent that they are phrased in separate sentences.
Inbetween lies the interesting transition zone of indirect speech forms of
varying degrees of integration. It is in this transition zone that languages
differ most. The important question to ask here with respect to any given
language is that of what and how much is coded in the grammar.

4. Conclusion

In the above I have touched upon several lexical and grammatical

devices for distinguishing the speech of another from one's own and for
integrating the former into the latter. In German, for instance, we find
that, in many cases, indirect speech is distinguished from direct speech
by means of the subjunctive mood, which is hence often cited as the
criterion of the "grammatical dependency" of indirect speech in Ger-
man. 11
The sequence of tenses is another obvious case in point. It may seem
indispensable for indirect speech in English or Latin, but when we look
at other tense languages we realize that it is a rather contigent feature.
A pronoun system enabling unequivocal reference seems to be essen-
tial for a proper interaction between reporting and reported parts of
speech. Particularly important in this respect is it to distinguish reference
to an individual whose speech or thought is reported from reference to
other individuals. Some languages have developed such forms that are
used to establish coreference with individuals whose point of view is
reported in indirect discourse. Hagege (1974) introduced the term
"logophoric pronouns." Their use permits a speaker to avoid ambiguity
of reference in many cases. Sentences such as (33), where it depends on
the context whether he refers to John, or Danny, Or a third person, may
seem intolerably ambiguous from the point of view of languages employ-
ing such logophoric pronouns. 12
(33) John said to Danny that he was a lady-killer.
22 Florian Coulmus

Reflexive pronouns are sometimes used in logophoric function, for

instance in English, as illustrated in the following example slightly
adapted from Kuno (1975).
(34) Mary said to John that lady-killers like himself were a pleasure to
meet.

But morphologically distinct logophoric pronoun systems are a rather

special property of a few languages.
Another feature worth mentioning are lexical or grammatical means
of distinguishing different kinds of reported speech with respect to
evidentiality, that is, the nature of the quoted source and the authenticity
of the report. Bloomfield (1927) reported about Navajo that "whatever
is hearsay and not the speaker's own experience has the predicate verb
or predicate in a special quotative form" (see also Collins 1983). Jakob-
son made similar observations for other languages including Kwakiutl,
Hopi, Tunica, and Bulgarian where the verb conjugation "distinguishes
two semantically opposite sets of forms: 'direct narration' vs. 'indirect
narration'" (1971: 135). Many languages use adverbs to differetiate
degrees of authenticity. Japanese has a sentence suffix, -sö, which in
conjunction with a form of the copula indicates hearsay quality. German
and Danish use modal verbs for the same purpose. When used in this
function, the verb sollen indicates lack of a specifiable source of informa-
tion and hence the speaker's unwillingness or inability to verify the
reported proposition. Its correlate wollen is used to show that the
speaker reports another speaker's statement but doesn't believe it.
(35) Er soll ein Schürzenjäger sein.
He is said to be (lit.: shall be) a lady-killer.

(36) Er will gestern Nacht bei ihr gewesen sein.

He claims (lit.: wants) to have been with her last night.

Some languages have grammatical devices to indicate that a given piece

of discourse represents another's speech or thought without explicitly
saying so. In French the report verb can be omitted, but tense and
pronouns can still be used in a manner appropriate to indirect speech.
Quasi-direct discourse is thus to some extent grammaticised in French
(cf. Fonagy in this volume).
The great variety of features that characterize reported speech across
languages leaves us with the question of whether any of them need to be
coded in the grammar of every language. To be sure, certain mechanisms
and features peculiar to reported speech are explicitly coded in the
 Reported speech 23

grammar of many languages. Especially the use of pronouns is subject to

seemingly rigid conditions (cf. Li in this volume). However, even in this
domain, which is so essential for communication, we find a great range
of variation. Indeed, the more we look into the details of individual
grammars, the more varied the picture becomes. Most of the features
that I have discussed above seem to be fairly arbitrary grammatical or
pragmatic conventions of individual languages rather than logical requi-
rements of integrating the speech of another into one's own. Mood,
tense, word order, complementation, do not necessarily get categorized
or categorized in the same manner in all languages. And the possibilities
of differentiating kinds of reported speech and the information about
speech acts encoded in report verbs seem to be even more culturally
specific.
However, speech reporting is essential to the nature of language
nevertheless. Without it, or rather, without the structural preconditions
referred to above in section 2.1., language would be fatally limited in its
potential. It would not allow us to construe sentences that refer to
themselves, such as the following.
(37) The sentence "The sentence may be quoted without permission" may
be quoted without permission.

By consequence, we would be deprived of the means - so essential to

scientific endeavors - to quote that another has quoted another as
quoting another as quoting . . .

Acknowledgement

Research for this paper was done while I held a Heisenberg-Fellowship of the Deutsche
Forschungsgemeinschaft, which I gratefully acknowledge.
For helpful comments on an earlier draft I am indebted to Danny Steinberg.
I am also grateful to the Language Institute of Gakushuin University, Tokyo, where I spent
the winter of 1984/85 when I wrote this paper.

Notes

1. Lanser (1981) offers an extensive discussion of point of view in fiction. Cf. also Leech
and Short (1981).
2. Example from Banfield (1973: 5).
3. I am disregarding here a third reading where his mother refers to someone else's
mother.
24 Florian Coulmas

I am also disregarding the obvious fact that Oedipus spoke Greek rather than
English, which is not at all a minor or trivial point because the status of an allegedly
verbatim quotation in a language which is not the same as that of the original utterance
is far from clear. In fiction it is a common practice to represent the direct speech of a
character who is identified as a foreigner using a foreign language in the same language
as the main body of the text. In order to remind the reader that a foreigner is speaking,
his speech is often marked with certain forms that are easily understood, such as terms
of address, greetings, etc. Joseph Conrad, at home in several languages, used this
technique extensively.
"The party he led were babbling excitedly in Italian and Spanish, inciting each other
to the pursuit. He put himself at their head, crying, 'AvantiF
'He has not stopped very long with us. There is no praise from strangers to be got
here,' Signora Teresa said, tragically. 'Avanti! Yes! that is all he cares for. To be first
somewhere - somehow - to be first with these English." (Joseph Conrad. 1963.
Nostromo. Harmondsworth: Penguin, p. 32.)
A peculiar situation obtains in some diglossic speech communities where two
markedly different varieties of a language coexist. To let characters speak the so-called
"high variety" in fictional narrative would be very unnatural. The "low variety",
however, is not usually written. Direct discourse is thus the only context where the
"low variety" is put in writing. Sinhalese is a case in point.
"The present situation is that all sentences within quotation marks in novels are
generally written in spoken idiom while the rest of the narrative is predominantly cast
in the classical format" (De Silva 1976: 98).
Speech reporting may thus eventually exercise an influence on the development of a
language by introducing the spoken vernacular into the written medium.
The interaction can also work the other way: Speakers may quote written forms in
the spoken language, as Keller has noted in his account of diglossia in German-
speaking Switzerland.
"Frequently one is not quite sure whether the speaker is not simply quoting
something in the 'High' variety. Quoting is in fact very often the first step to borrowing
and is a characteristic of the breakdown of the diglossia situation" (Keller 1982: 86).
See also Traugott (1981).
4. Cf. also Lorck (1921: 16):
"Auf die Tatsache, daß es außer der direkten und der indirekten Rede noch eine
dritte Darstellungsweise des gesprochenen Worts gebe, hat zuerst A. Tobler hingewie-
sen . . . Nach seinem Dafürhalten lag eine Variante der direkten Rede vor, die durch
eine eigentümliche Mischung dieser und der indirekten Rede entstanden sei."
5. Neubert (1957: 8f.) gives a useful chronological overview of the many different terms
proposed by writers on the subject of reported speech between 1894 und 1954. He
adopts Lorck's term erlebte Rede. See also McHale (1978).
6. Rosier (1980) investigates the development, or what he calls the discovery, of fiction in
Classical Greek which he explains as an outgrowth of literacy, that is, as part of the
transition from a oral to a literature culture.
7. This duplicity is generally referred to as the "use" vs. "mention" distinction. A term is
used whenever it refers to something other than itself. It is mentioned when it serves
the meta-linguistic function of naming itself. This distinction is vital as it enables
language to function as its own meta-language, a property which is also known as
"reflexivity."
 Reported speech 25

Hockett includes reflexivity in his list of universale of human language which he

deduced through comparison with animal sign systems: "In a language, one can
communicate about communication" (1963: 13). A more general way of capturing this
property of language, he suggests, is to say that in a language one can communicate
about anything. Such a position is, however, much harder to defend. There may be
things whose existence we cannot recognize because we cannot communicate about
them in natural languages. For our purposes, we can leave this philosophical problem
aside and register the fact that the metalinguistic function of natural languages is
logically necessary and hence universal.
Metalinguistic terms, for instance performative verbs, have been taken as the
vantage point for the analysis of speech acts by philosophers notably Austin (1965) and
Searle (1969). Investigations growing out of an ethnographic tradition which have
become known as ethnosemantics suggest that these philosophical treatments are
somewhat ethnocentric. Metalinguistic taxonomies reflect how speakers categorize
their verbal behavior and how speech events fit into a cultural matrix. Such taxonomies
differ widely across languages, see e.g., Sapir (1915), Frake (1962), Stross (1972).
There is no reason to believe that metalinguistic taxonomies of European languages,
English in particular, provide more general insights into universal principles of the
organization of verbal behavior than those of other languages which have received less
attention by speech act theorists.
For a summary of the extensive literature about use and mention see Zabeeh et al.
(1974: 20-31).
8. Two points should be carefully distinguished here: (1) the question of identity and (2)
the question of style. Whether type-token identity holds for two utterances is a
question not to be confused with the stylistic option that allows us to speak as if we
were repeating the "exact words" of another speaker. That exact repetion of form and
meaning is very rare in actual language use is well-known by psychologists and
rhetoricians. Experimental studies of the linguistic memory have shown that "memory
for form appears to be tied to memory for meaning" (Olson, Hildyard 1983: 291) and
that, therefore, long term memory for language "is rarely word for word, verbatim
memory" (ibid. 294). Moreover, there are different criteria for identity. While
semanticists may argue that, except for ironic utterances and other conventional
distortions of meaning, identity of meaning is implied by identity of form, rhetoricians
would not agree as is obvious from the following.
"Even the words of other people, when repeated by a speaker, have changed their
meaning, for in the process of repetition he always adopts toward them a position that
is in some way new, even if only in the degree of importance he attaches to them"
(Perelman, Olbrechts-Tyteca 1969: 317).
9. The necessity of adapting deictic terms in indirect speech to the reporter's viewpoint
was also stipulated by other authors. Wunderlich, for example, makes no qualification
as for the language for which his general statement is supposed to be true.
"Im indirekten Zitat müssen sich alle deiktischen Ausdrücke . . . auf den zitierenden
Sprecher selbst beziehen, er muß also die ursprünglichen Ausdrücke entsprechend
seiner Position transformieren" (Wunderlich 1972: 164).
10. Notice the accusative whom as a subject in the context of a verbum dicendi in
Shakespeare's sentence "Arthur whom they say is killed to-night" (John IV, 2.165).
Jespersen (1924: 117, 348) has described this phenomenon for which he cites a great
many examples as "an accusative + a finite verb dependent on a verb inserted after the
26 Florian Coulmas

accusative." Whom could not be the subject of the relative clause if it wasn't for the
"paranthetical" they say which thus exercises an influence on the grammatical structure
by reducing the embedded clause's independence.
11. "Die indirekte Rede im Deutschen muß jetzt als etwas grammatisch Abhängiges
betrachtet werden, und das Kennzeichen der Abhängigkeit dabei ist der Konjunktiv"
(Paul 1909: 146).
12. The study of logophoric pronouns has concentrated on West African languages where
their occurrence is common. See, for example, Clements (1975) for Ewe, Hyman and
Comrie (1981) for Gokana, and Perrin (1974) for Mambila; also Bamgbose in this
volume for Yoruba.

References

Austin, J. L. 1965. How to Do Things with Words. New York: Oxford University Press.
Bally, Charles. 1912. Le style indirect libre en f r a ^ a i s moderne. Germanisch-Romanische
Monatsschrift IV: 549ff.
Banfield, Ann. 1973. Narrative Style and the Grammar of Direct and Indirect Speech.
Foundations of Language 10: 1-39.
Bloomfield, Leonard. 1927. Literate and Illiterate Speech. American Speech 2, 10:
432-439.
Clements, George N. 1975. The Logophoric Pronoun in Ewe: Its role in Discourse.
Journal of West African Languages 10: 141-77.
Collins, James. 1983. Reported Speech in Navajo Myth Narratives. To appear in J.
Verschueren, ed., Linguistic Action: Some Empirical-Conceptual Studies..
Comrie, Bernard. 1984. Tense. London: Cambridge University Press.
Coulmas, Florian. 1981. Über Schrift. Frankfurt: Suhrkamp.
Coulmas, Florian. 1982. Some Remarks on Japanese Deictics. In: J. Weissenborn and W.
Klein, eds., Here and There: Cross-linguistic Studies on Deixis and Demonstration.
Amsterdam, Philadelphia: John Benjamins Publ.
Coulmas, Florian. 1985. Direct and Indirect Speech. General Problems and Problems of
Japanese. Journal of Pragmatics 9: 41-63.
Coulmas, Florian and Konrad Ehlich, eds. 1983. Writing in Focus. Berlin, New York,
Amsterdam: Mouton.
DeSilva, Μ. W. Sugathapala. 1976. Diglossia and Literacy. Mansagangotri, Mysore:
Central Institute of Indian Languages.
Frake, Charles. 1962. TTie Ethnographic Study of Cognitive Systems. In: T. Gladwin and
W. C. Sturtevant, eds., Anthropology and Human Behavior. Washington, D . C . :
Anthropological Society.
Goodman, Nelson. 1968. Languages of Art. An Approach to a Theory of Symbols.
Indianapolis, New York: The Bobbs-Merrill Comp.
Goody, Jack. 1977. The Domestication of the Savage Mind. Cambridge: Cambridge
University Press.
Goody, Jack and Ian Watt. 1968. The Consequences of Literacy. In: J. Goody, ed.,
Literacy in Traditional Societies. London: Cambridge University Press.
Greenberg, Joseph H., ed., 1963. Universals of Language. Cambridge, Mass.: MIT Press.
 Reported speech 27

Hagege, Claude. 1974. Les pronoms logophoriques. Bulletin de la Societe de Linguistique

de Paris 69 (1): 287-310.
Hockett, Charles F. 1963. The Problem of Universals in Language. In: J. H. Greenberg,
ed., 1963 Universals of Language: 1-29.
Hyman, Larry and Bernard Comrie. 1981. Logophoric Reference in Gokana. Journal of
African Languages and Linguistics 3: 19-37.
Jakobson, Roman. 1971. Shifters, Verbal Categories, and the Russian Verb. Selected
Writings II. The Hague, Paris: Mouton, 130-147.
Jespersen, Otto. 1924. The Philosophy of Grammar. London: Allen and Unwin.
Kalepky, T. 1899. Zur französischen Syntax. Zeitschrift für Romanische Philologie XXIII:
491-513.
Keil, Frank C. 1981. Constraints on Knowledge and Cognitive Development. Psychologi-
cal Review 88: 197-227.
Keller, R. E. 1982. Diglossia in German-speaking Switzerland. In: W. Haas, ed., Standard
Languages, Spoken and Written. Manchester: Manchester University Press.
Kuno, Susumu. 1972. Pronominalization. Reflexivization, and Direct Discourse. Lingu-
istic Inquiry 3: 161-96.
Kuno, Susumu. 1975. Three Perspectives in the Functional Approach to Syntax. In Papers
from the Parasession on Functionalism. Chicago: Chicago Linguistic Society, 276-336.
Lanser, Susan S. 1981. The Narrative Act: Point of View in Fiction. Princeton: Princeton
University Press.
Linsky, Leonard. 1967. Referring. London: Routledge and Kegan Paul.
Leech, Geoffrey N. and Michael Short. 1981. Style in Fiction. A Linguistic Introduction To
English Fictional Prose. New York: Longman.
Lerch, Gertraud. 1919. Uneigentliche direkte Rede. Diss. Munich.
Lorck, Etienne, 1921. Die erlebte Rede. Eine sprachliche Untersuchung. Heidelberg: Carl
Winter.
Lyons, John. 1977. Semantics I. Cambridge: Cambridge University Press.
McHale, Brian. 1978. Free Indirect Discourse: A Survey of Recent Accounts. PTL 3:
249-288.
Neubert, Albrecht. 1957. Die Stilformen der "erlebten Rede" im neueren englischen
Roman. Halle: Niemeyer.
Okutsu, K. 1970. Inyö kozö to kansetsuka tenkei. (Quotation Structure and Indirectifica-
tion). Gengo kenkyü 56: 1-26.
Olson, David R. and Angela Hildyard. 1983. Literacy and the Comprehension and
Expression of Literal Meaning. In: F. Coulmas and K. Ehlich, eds., 1983. Writing in
Focus, Berlin, New York, Amsterdam: Mouton.
Partee, Barbara Hall. 1973. The Syntax and Semantics of Quotation. In: S. Anderson, P.
Kiparsky, eds., A Festschrift for Morris Halle. New York: Holt, 410-18.
Paul, Hermann. 19094. Prinzipien der Sprachgeschichte. Halle: Niemeyer.
Perelman, Ch. and L. Olbrechts-Tyteca. 1969. The New Rhetoric. A Treatise on Argumen-
tation. Notre Dame, London: University of Notre Dame Press.
Perrin, Mona. 1974. Direct and Indirect Speech in Mambila. Journal of Linguistics 10,
27-37.
Plantinga, A. 1969. De Re et De Dicto. Nous III, 3: 235-58.
Quine, W. V. 1960. Word and Object. Cambridge, Mass.: MIT Press.
Quine, W. V. 1966. Three Grades of Modal Involvement. In: The Ways of Paradox. New
York: Random House.
28 Florian Coulmas

Quirk, Randolph et al. 1972. A Grammar of Contemporary English. London: Longman.

Rosier, Wolfgang. 1980. Die Entdeckung der Fiktionalität in der Antike. Poetica 12, 3-4:
283-319.
Sapir, Edward. 1915. Abnormal Types of Speech in Nootka. Ottawa.
Searle, John R. 1969. Speech Acts. London, New York: Cambridge University Press.
Stross, B. 1972. Speaking of Speaking: Tenejapa Tzeltal Metalinguistics. In: Explorations
in the Ethnography of Speaking. E. Baumann, J. Sherzer, eds. New York: Cambridge
University Press, 213-39.
Tannen, Deborah, ed. 1982. Spoken and Written Language: Exploring Orality and
Literacy. Norwood, N. J.: Ablex.
Tobler, A. 1894. Vermischte Beiträge zur französischen Grammatik II. Leipzig.
Traugott, Elizabeth Closs. 1981. The Voice of Varied Linguistic and Cultural Groups in
Fiction: Some Criteria for the Use of Language Varieties in Writing. In: Writing: The
Nature, Development, and Teaching of Written Communication. Vol. I, Variation in
Writing. Marcia Farr Whiteman, ed. Hillsdale, N.J.: Lawrence Erlbaum.
Voloshinov, V. N. 1973. Marxism and the Philosophy of Language. New York: Seminar
Press. (First Russian edition 1929.)
Wierzbicka, Anna. 1974. The Semantics of Direct and Indirect Discourse. Papers in
Linguistics 7, 3-4: 267-307.
Wunderlich, Dieter. 1972. Redeerwähnung. In: Pragmatik und sprachliches Handeln. U.
Maas, D. Wunderlich, Frankfurt: Athenäum, 161-88.
Zabeeh, F., E. D. Riemke, A. Jacobson. 1974. Readings in Semantics. Urbana, Chicago:
University of Illinois Press.
Direct speech and indirect speech:
A functional study*
Charles N. Li

1. Introduction

Direct speech and indirect speech are similar, yet different. Consider
two examples, one illustrating direct speech, the other indirect speech:
(1) John said, "I'm tired."
(2) John said (that) he was tired.
Let us first observe the similarities between (1) and (2). Both have the
same verb, "said" with the same subject, "John". Each contains a clause
signalling the content of a speech act. Although the two clauses are
different in form, they both convey the same message. Thus, at first
sight, the similarities between (1) and (2) appear so striking that early
transformationalists were motivated to propose that (2) should be derived
from (1) via an optional transformation called the "Indirect Discourse
Formation". 1
The differences between (1) and (2), on the other hand, are manifold.
The most readily observable ones are syntactic: first, the pronouns in (1)
and (2) are different; second, the tenses in (1) and (2) are different;
third, (2) but not (1) may have the complimentizer "that". Finally,
Longacre (1976) noted that the immediate constituent grouping in terms
of phonological pause and intonation pattern is different between (1)
and (2). But the phonologic grouping of the subject and the verb of
saying holds true only in verb-medial languages. In verb-final languages,
for example, a direct quote is often placed between the subject and the
verb of saying.
On the semantic level, Partee (1971) noticed that the surface form of
the direct quote, i.e. the exact wording of the quotation, is part of the
meaning of the whole sentence. Thus, even if we accept the synonymity
of (3) and (4),

* Preparation of this article was supported by National Science Foundation Grant

BNS 83-08220.
30 Charles Ν. Li

(3) John said that this theorem was false.

(4) John said that this theorem was not true,
we cannot consider (5) and (6) synonymous.
(5) John said, "This theorem is false."
(6) John said, "This theorem is not true."
In other words, even though "This theorem is false" is equivalent to
"This theorem is not true", these represent two different utterances. (5)
and (6) each attribute a different utterance to John, and, therefore,
cannot be synonymous. This observation led Partee (1971) and others to
reject the transformational derivation of indirect speech from direct
speech.
On the functional level, Wierzbicka (1974) pointed out that direct
speech is characterized by its "theatrical" nature. 2 Wierzbicka's analysis
of direct and indirect speech is insightful. In this article, I will pursue the
implications of Wierzbicka's study in light of evidence from a variety of
languages. Unlike Wierzbicka, however, I will not be concerned with
such questions of generative linguistics as "What is the underlying
representation of a sentence" or "Does direct speech underly indirect
speech". The theoretical thrust of this study is a functional one, i.e., I
will examine the structure of direct and indirect speech in terms of their
communicative functions.

2. The syntax of direct and indirect speech

2.1. Pronominalization

In section I, it was stated that the pronouns in sentences (1) and (2) are
different. Among all languages which have both direct and indirect
speech, the existence of different pronominalization strategies for the
two constructions is universal. I will state the various strategies as
follows.
A: The first and second person pronoun in a direct quote are respec-
tively co-referential with the reported speaker and the reported
addressee in the clause immediately outside of the quotation.
The following illustrate strategy A:
 Direct and indirect speech 31

(7) John said to Mary, "I love you."

(8) John told Mary that Peter said to Ellen, "I love you."

Sentences (7) and (8) demonstrate that the references of the first and
second pronouns in a quotation are conditioned by the reported speaker
and the reported addressee in the clause immediately outside the quota-
tion.
B: The third person pronoun in a direct quote must not be co-referential
with the reported speaker or the reported addressee in the clause
immediately outside the quotation.

Consider the following examples:

(9) John said to Peter, "He was sick."
(10) John said to Mary, "Peter told Ellen that he loved her."
(11) About Peter, John said to Paul, "He was sick."

In sentence (9), the third person pronoun in the quote cannot be co-
referential with either "John" or "Peter", the reported speaker and the
reported addressee respectively, in the clause immediately outside the
quotation. Sentence (10) contains an indirect quote embedded in a direct
quote. Again, the third person pronouns in (10) cannot be co-referential
with either "John" or "Mary", the reported speaker and the reported
addressee respectively, in the clause immediately outside the quote. In
sentence (11), the third person pronoun in the quote is co-referential
with "Peter", an NP which is neither the reported speaker nor the
reported addressee.
Let us now examine pronominalization strategies in indirect speech.

C: The I ^ r S t , 1 person in an indirect quote refers to either

( second J

® r Pr eeSPeaker}- 00
{ d s re
P°rted { SeJL } *

the direct quote in which that indirect quote is embedded.

For example,
(12) John told Mary that I love you.
(13) John said to Mary, "Peter told Ellen that I love you."
(14) You told me that I took you to San Francisco last year.

In (12), the first and second person pronouns in the indirect quote refer
respectively to the reporter-speaker and the addressee. In (13), the first
32 Charles Ν. Li

and second person pronouns are in an indirect quote embedded in a

direct quote, and therefore, refer respectively to the reported speaker
and the reported addressee of the direct quote. (14) shows that if the
clause with the verb of saying contains first and second persons, which
are co-referential with the reporter-speaker and addressee respectively,
then these pronouns are co-referential with the first and second person
pronouns respectively in the indirect quote.

D: Third person pronominalization in an indirect quote follows the

general rule of pronominalization in a language.

In other words, there is no constraint peculiar to indirect speech gover-

ning the reference of a third person pronoun in an indirect quote. If a
language allows a pronoun to refer only to an antecedent noun phrase,
then a third person pronoun in an indirect quote may be co-referential
with any antecedent noun phrase in the discourse. For example, in the
Mandarin Chinese sentence (15), the third person pronoun in the
indirect quote may be co-referential with any preceding NP in the
sentence or in the preceding discourse.

(15) Zhangsan dui Lisi shuo Wanger juede fa hen qiguai

Λ * ~t
Zhangsan to Lisi say Wanger feel s/he very strange.
Zhangsan said to Lisi that Wanger felt s/he was very strange.

If a language, such as English, allows the "preceding" rule of pronoun

reference to be superseded under certain conditions (e.g. a pronoun may
be co-referential with a following NP if the pronoun is in certain
subordinate clauses while the NP is in the main clause), the same
condition will allow the reference of a third person pronoun in an
indirect quote to supersede the "preceding" rule. Sentence (16) illustra-
tes this:

(16) John said to Peter that before he left, Paul was O.K.

I have gone to some length describing the referential strategy of

pronouns in direct and indirect speech because it is often the only
diagnostic criterion available in many languages for distinguishing a
direct quote from an indirect quote out of discourse context. Consider,
for example, the following sentence from Mandarin Chinese:
 Direct and indirect speech 33

(17) Zhangsan shuo wo lei le

Zhangsan say I tire Perfective
Zhangsan says I'm tired.

If the first person pronoun wo "I" is co-referential with Zhangsan, we

know that (17) contains a direct quote on the basis of A. However, if wo
"I" is co-referential with the reporter-speaker of sentence (17), we can
conclude on the basis of the principle stated in C that (17) has an indirect
quote.
Some languages have unusual pronominal strategies in direct and
indirect speech. Mambila, a Bantu language described by Perrin (1974)
is a good example. Mambila has a special pronoun neh which occurs only
in reported speech and which refers only to the reported speaker. The
following Mambila sentences are taken from Perrin (1974):
(18) woo jia neh a neh nda
you say me to reported go
speaker
You said to me you would go.

(18) alone cannot establish whether a clause containing neh is a direct

quote or an indirect quote since neh merely refers to the reported
speaker without distinguishing first, second or third person.
Sentence (19) shows that neh occurs in an indirect quote (on the basis
of Β) since the third person pronoun bu is co-referential with the
reported addresse, torn "rabbit".
Ψ ^
(19) Alhaji jia torn a neh kwuli bu
Alhaji say rabbit to reported thank third person
speaker pronoun
Alhaji said to the rabbit that he thanked him.

Similarly, (20) shows that neh occurs in an indirect quote (on the basis
of C (i)) since the second person pronoun chi "your" is co-referential
with the second person pronoun in the matrix clause, which, in turn, is
co-referential with the reporter-speaker.

(20) woo jia meh a heh neh derua a chi

you say me to give reported book genitive your
speaker
You told me to give you your book.

The evidence cited by Perrin (1974) indicates that neh occurs only in
indirect quotes, contrary to her claim that neh is used only in direct
quotes. 3
34 Charles Ν. Li

2.2. Deictics

Besides pronouns, deictics such as 'this/that', 'here/there', 'now/then'

and 'come/go' constitute another domain in which systematic differences
exist between direct and indirect speech. Let me illustrate with an
example. First, consider the speech context in which, at school, John
said to Paul in the presence of Peter, "Come here and take care of this
mess!". Secondly consider a different speech context in which Peter was
telling some people at home several days later what transpired between
John and Paul. If Peter were to give a direct quote of what John said, he
would say:
(21) John told Paul, "Come here and take care of this mess!"

However, if Peter were to render John's words in an indirect quote, he

would say:
(22) John told Paul to go there and take care of that mess.

The difference between (21) and (22) lies in the use of opposite members
of such deictic pairs as 'come/go', 'here/there', and 'this/that'. The
reason behind the choice of opposite members of these deictic pairs is
the different points of reference. In an indirect quote, the speaker
normally uses himself/herself as a spatial point of reference and the time
of utterance as a temporal point of reference. In a direct quote, the
speaker must suspend the normal practice and use the points of refe-
rence of the quoted speaker.

2.3. Syntactic and semantic role

An indirect quote is traditionally assumed to be a subordinate clause

serving as the direct object of the verb of saying. Partee (1973) suggested
that the direct quote is not a syntactic or semantic part of the sentence
containing it, but did not question the subordinate clause role of the
indirect quote. Munro (1982) pointed out a variety of syntactic differen-
ces between sentences containing indirect quotes and sentences contain-
ing direct quotes in many languages, further strengthening our impres-
sion of the different syntactic and semantic roles of the direct and
indirect quote. Whereas the fact that direct quote and indirect quotes are
different syntactically and semantically is incontrovertible, the nature of
their differences may vary from language to language. In English, for
 Direct and indirect speech 35

example, aside from pronominalization and deictics discussed above, we

can cite at least three syntactic differences between the direct and
indirect quotes: (i) the complementizer, "that", may occur with an
indirect quote but not with a direct quote; (ii) the indirect quote but not
the direct quote must bear the same tense as the verb of saying; (iii)
there is an absence of an intonation break between a direct quote and the
verb of saying, but not between an indirect quote and the verb of saying.
On the other hand, the English verb, "say", in its unstressed form, is
often semantically bleached and displays the characteristics of a hearsay
evidential. This hearsay evidential characteristic of "say" is especially
prominent when the indirect quote is fronted as in (23):
(23) He didn't want to play tennis, John said.

If we contrast (23) with (24) containing a direct quote,

(24) "I don't want to play tennis", John said.

we can observe that the differences between the syntactic-semantic roles

of the direct quote and the indirect quote are minimal. Neither the direct
quote of (24) nor the indirect quote of (23) seem to play the role of the
direct object or patient of the verb of saying. Since the semantic function
of "say" tends to lean in the direction of a hearsay evidential or an
epistemic quantifier, the new information transmitted by sentences such
as (23) and (24) rests in the quoted speech, whether it is direct or
indirect.
Let us consider another language, Tikar. 4 In Tikar, it is the direct
quote, but not the indirect quote, which must be introduced by the
complementizer /ε. Sentences (25) and (26) are examples of direct
quotes in Tikar:
f
(25)
t
SWE she Ιέ nü/ε
* .
wu limmi kä 1ε
* .
mü bäeni mwum Ιέ kirn
ant say to him that you know not that I caught man with force-Q
The ant said to him, "Don't you know that I caught a man by force?"

(26) Α she Ιέ kpulu ίε Kpulu wu yibä mü ndem

he say to turtle that Turtle, you stole me field
He said to the turtle, "Turtle, you stole my field!"

The pronominalization strategy indicated by the co-referentiality in

(25) and (26) shows that the sentences contain direct quotes.
According to Jackson (personal communication), one of the forms of
indirect speech in Tikar occurs without a verb of saying and without a
complementizer. Sentences (27) and (28) illustrate:
36 Charles Ν. Li

(27) Nü, nü nywael ime hje

he he escapes starvation
He (said/thought) he's escaped starvation.

Τ T T V V

(28) Foro nü, nü ywime Bs

Hyena he he F U T arrive
The hyena, he (says) he will come.
The use of the third person pronoun in (27) and (28) referring to the
reported speaker demonstrates that both contain indirect quotes.
The English and Tikar examples suggest that the differences between
direct and indirect speech can vary dramatically from language to
language and that their differences cannot be accounted for by the claim
that the indirect quote is a complement of the verb of saying, whereas
the direct quote is not a syntactic or semantic part of the sentence
containing it.
Haiman and Thompson (1984) reject the contention that indirect
quote is "subordinate" in a variety of ways that the direct quote is not.
To the extent that Haiman and Thompson (1984) are correct in showing
that the traditional notion of "subordination" is merely a cover term for
a number of independent properties describing clausal relations, it is best
not to subsume the syntactic and semantic differences between direct
and indirect quote under "subordination". Instead, it might be profita-
ble to consider those differences along a dimension of "fusion" between
clauses. Givon (1980) proposed a similiar notion, "binding", which
refers to the influence of the main clause agent over the complement
clause agent. "Fusion" can be interpreted as referring to the degree of
syntactic-semantic influence or control of one clause over another in
interclausal relationships. The following examples show an increasing
degree of syntactic-semantic influence of the first clause over the second
and therefore, display a higher degree of fusion between the two clauses
from sentence to sentence:

(29) John went to school and Mary stayed home.

(30) John lived in a bad district, but was robbed only once.
(31) John claimed that Peter went to school.
(32) John ordered Peter to go to school.
(33) John remembered to go to school.

In (29), the two clauses are independent syntactically and semanti-

cally. The only clausal relationship conveyed by (29) is that according to
 Direct and indirect speech 37

the view of the speaker, there is a vague pragmatic bond between the
events denoted by the two clauses. (30) demonstrates a higher degree of
the fusion between its two clauses than (29) because the subject of the
second clause must be interpreted as co-referential with the subject of
the first clause. In (31), the second clause is incorporated into the first
clause as a constituent, indicating greater fusion than either (29) or (30).
In (32) and (33), the second clauses are not only incorporated into their
respective first clauses, but they have also lost their independent clause
structure. For instance, the verb of the second clauses of (32) and (33) is
non-finite, and its tense and aspect must be interpreted according to that
of the verb of the first clause. Finally, (33) shows a greater degree of
fusion than (32) because the truth value of the second clause in (33), but
not in (32), is determined by the truth value of the first clause. Thus
"John remembered to go to school" implies "John went to school",
whereas "John didn't remember to go to school" implies that "John
didn't go to school". 5
Both the direct quote and the indirect quote are on the weak end of
the fusion scale in terms of their respective relationships with the verb of
saying. One reason for this is that the verb of saying tends to take on the
features of a hearsay evidential. However, the direct quote is even more
weakly fused with the verb of saying than the indirect quote. Thus, only
the direct quote has the full latitude of an independent sentence. For
instance, only the direct quote, but not the indirect quote, may take the
form of a performative speech act such as a command or a question:
(34) John said, "What did Mary buy?"
(35) John said, "Be quiet!"

In most languages, the indirect quote has some feature which signals
that it is more fused with the clause containing the verb of saying than
the direct quote. Examples of such features were cited by Munro (1982)
as evidence that the indirect quote, but not the direct quote, behaves as
the object of the verb of saying.

3. Functional characteristics of direct and indirect speech

Let us recall sentences (1) and (2), the two simple examples of direct and
indirect speech in English:
(1) John said, "I'm tired."
(2) John said (that) he was tired.
38 Charles Ν. Li

Consider the two sentences in a communicative context. Both involve

four elements:
i) the reported speaker or the original speaker
ii) the reported speech
iii) the reporter-speaker or simply the speaker
iv) the hearer or the addressee

In direct speech, the reporter-speaker plays the role of the reported/

original speaker. The reporter-speaker intends for the hearer to believe
that the form, the content and the non-verbal messages such as gestures
and facial expressions of the reported speech originate from the reported
speaker. That is where the theatrical feature noted by Wierzbicka comes
in: the reporter-speaker acts as the reported speaker when s/he utters the
direct quote. The full communicative picture of sentence (1), then, is
composed of two parts. In one part, the reporter-speaker identifies the
reported speaker. In the other part, the reporter-speaker acts as the
reported speaker. The full communicative situation may be represented
by (36):
(36i) reporter-speaker identifies reported speaker
reporter- acts as reported [form] direct
ϋ) speaker => speaker - [content] reported
" non-verbal speech
.messages .

In indirect speech, the reporter-speaker does not play the role of the
reported speaker.6 The form and the non-verbal messages of the report-
ed speech belong to the reporter-speaker. The reporter-speaker intends
for the hearer to believe that only the content of the reported speech
originates from the reported speaker. Thus, the communicative situation
involving a sentence such as (2) may be depicted in (37).

(37i) reporter-speaker identifies reported speaker

[form]
ii) reporter-speaker non-verbal
.messages indirect
reported
reported speaker [content] speech

Given (37), we can understand how the reporter-speaker may commu-

nicate his own feelings through the form (e.g. intonation) and non-verbal
messages of the reported speech as a comment on the content of the
reported speech. Sentence (2), for example, may be uttered with a great
deal of disgust, anger, or sarcasm. Such emotional overtones would be
 Direct and indirect speech 39

communicated by the intonation, gestures and facial expression of the

reporter-speaker. The hearer will understand that those emotional over-
tones do not originate from the reported speaker, but constitute a
comment by the reporter-speaker on either the reported speaker or the
content of the reported speech or both.
In direct speech, the reporter-speaker does not have the option of
communicating a comment on the content of the reported speech as s/he
utters the direct quote, because as (36) shows, not only the form and the
content of the reported speech, but also the non-verbal messages accom-
panying it, originate from the reported speaker. Thus, if the reporter-
speaker utters the reported speech with anger, for instance, the hearer
will interpret the anger as part of the message originally communicated
by the reported speaker.
The analysis provided in (36) and (37) should make it clear why
sentences such as (5) and (6) cannot be synonymous.
(5) John said, "This theorem is false."
(6) John said, "This theorem is not true."

The analysis in (36) and (37) should also make it clear why a direct quote
should have the full latitude of an independent sentence: the form (not
only the substance) the direct quote also originates from the reported
speaker.

4. Corroborative evidence

In this section I will cite four pieces of evidence which corroborate the
preceding functional analysis of direct and indirect speech.

4.1. Languages without indirect speech

Direct speech is universal; indirect speech is not. Paez,7 for example, is a

language without indirect speech. Gerdel and Slocum (1976) noted that
"indirect quotation sentences" in Paez only employ verbs such as
"know", "see", . . . etc., as their matrix verb. There is no Paez indirect
quotation sentence whose matrix verb is a verb of saying. Other known
examples of languages without indirect speech are Navajo and Amharic.
40 Charles Ν. Li

The analysis in Section III shows that indirect speech is a more

complex communicative strategy than direct speech. Direct speech in-
volves reproducing or mimicking the speech of the reported speaker,
whereas indirect speech involves rephrasing or paraphrasing the speech
of the reported speaker. Clearly, mimicking is a simpler undertaking
than paraphrasing. Hence, it is not surprising that for reported speech
the mimicking strategy occurs in all languages, whereas the paraphrasing
strategy does not. After all, mimicking is employed from the onset of
first language acquisition. It is an innate ability in human beings, pongid
and many simian species.
There is also a cultural reason for the non-occurrence of indirect
speech in languages such as Paez. In Paez culture, a person is not
allowed to quote by taking the responsibility for the quote himself/
herself - which is the implication of an indirect quote. A quote must be
entirely attributable to the original speaker, in form and content - which
is the strategy of a direct quote.

4.2. The use of direct quote in narrative

Several of the linguists who participated in the project "Study of

Discourse from Folk Literature in Aboriginal Languages of Colombia,
Panama, and Ecuador" reported that direct quote, but not indirect
quote, typically occurs as the peak of narrative. Mansen and Mansen
(1976) noted that quotes and dialogue fill the content of the peak of
Guajiro 8 narrative. Waltz (1976) observed that direct speech is often
used at the peak of Guanano 9 narrative. Borman (1977) stated that
quotes, including onomatopoetic quotes, and dialogue lend vividness to
the peak episodes of Cofan 10 narrative. Witte (1977) pointed out that
quotes and dialogue occur most prominently at the peak of Andoke 11
narrative and that they contribute to the vividness at the narrative peak.
The observations cited above corroborate Wierzbicka's theatrical
feature of direct speech. Since direct speech requires the reporter-
speaker to act out the role of the reported speaker, it is a natural vehicle
for vivid and dramatic presentation. Thus, direct quote is the most
common mode of expression at the peak of oral narrative in many
languages.
 Direct and indirect speech 41

4.3. Speaker's involvement

Chafe (1982) pointed out that direct quotes express the reporter-spea-
ker's involvement in the events s/he is reporting. Schwartz (personal
communication) has noted that when his children become very involved
in the events they are recounting, they use direct quotes exclusively,
often with "go" as the verb of saying. A natural consequence of
involvement in the event one is reporting is to act out the event. That is
precisely what direct speech is: the reporter-speaker acting out the role
of the reported speaker.

5. Evidentials and direct/indirect speech

Evidentials specify the nature of the source of information in speech: Is it

hearsay or the speaker's own previous experience? Is it imagination or
conjecture? The function of evidentials is to enable the speaker and
hearer to gauge the authenticity of a statement. Thus, direct quote and
indirect quote are forms of evidentiality. A direct quote communicates a
more authentic piece of information than an indirect quote in the sense
that a direct quote implies a greater fidelity to the source of information
that an indirect quote. Languages which do not have indirect speech may
be viewed as languages which do not allow adulteration of the source of
information in reported speech. Some languages which have both direct
and indirect speech allow the use of indirect speech only if the reporter-
speaker is unsure about the original speech. Cuiva12 is such an example,
although in her study of Cuiva discourse, Kerr (1977) states that Cuiva
indirect speech also occurs when the reporter-speaker translates an
original speech into another language for the benefit of the hearer. Kerr
(1977) writes: "Because [the reporter-speaker] cannot use a direct quote
in such circumstances, he uses an indirect quote to relay the substance of
what was said in the other language" (p. 155).
Thurgood (1981) pointed out that direct speech and indirect speech
are manifestations of evidentiality in English grammar. Modern English,
however, has also developed other forms of reported speech. One
frequently cited form is "free indirect speech" which contains the
features of both direct and indirect speech. An example cited by Polanyi
(1982) is,
42 Charles Ν. Li

(38) And he was telling Dolly, I don't want Dolly.

The co-reference between the reported speaker and the first person
pronoun in (38) indicates direct speech. But the presence of "Dolly"
instead of a second person pronoun "you" points to indirect speech.
Free indirect speech, then, is a device which simultaneously presents the
third person perspective of the reporter-speaker and the first person
perspective of the reported speaker. Such a device sacrifices the distinc-
tion between direct and indirect speech in terms of evidentiality.13

6. Conclusion

I have presented a functional analysis of direct and indirect speech in this

study. The analysis corroborartes a number of phenomena observed in a
wide variety of languages. Problems such as "What is the underlying
representation?" or more specifically, "Does direct speech underly
indirect speech?" might have motivated earlier investigators to examine
the relation between meaning and structure. However, they do not lead
us to a better understanding of either the structure or the function of
language. In addition, looking at grammar in terms of problems of that
sort may lead to pvseudo-problems. For example, Wierzbicka's insightful
study (1974) incorporated into her notion of underlying representation
pragmatic information describing the communicative functions of direct
and indirect speech; but she also created some pseudo-problems, be-
cause her study was carried out within the theoretical framework of
generative linguistics. Thus, after postulating elaborate underlying re-
presentations which capture most of the features of direct and indirect
speech, she conceded that those underlying representations could not
account for such utterances as

(39) John told Peter something.

But (39) is simply a case of indirect speech in which the reporter-speaker

has suppressed the content of the reported speech. In terms of the
analysis provided in (37), the utterance (39) may be represented as
follows:

(40i) reporter-speaker identifies reported speaker

 Direct and indirect speech 43

[form] indirect
reporter-speaker —» non-verbal reported
.messages speech

reported-speaker —> [content]

The reason for preferring a functional approach to the study of

language is straigthforward: language serves homo sapiens as their
means of communication. If a language is a means of communication,
then, it is clear that its structure must be adapted to its communicative
function.

Acknowledgement

I wish to thank Sandra S. Thompson for bringing to my attention various items of literature
on the subject of direct and indirect speech. I also wish to thank Arthur Schwartz for
discussing with me the nature of direct and indirect speech. Finally, I am grateful to Talmy
Givön, Ellen Jackson, Petr Sgall and Sandra Thompson for their important comments,
suggestions and criticism of an earlier version of this paper.

Notes

1. See Kuno (1972) in which the direct speech sentence is postulated not only as the
underlying representation of indirect speech sentences with such matrix verbs as
"expect", "claim", "worry", "believe", etc. . . .
2. Wierzbicka (1974) considers the theatrical nature of direct speech as part of the
meaning of direct speech. Clearly, she uses the term 'meaning' in a broad sense to
include not only the lexical and grammatical meanings, but also the communicative
function of the utterance.
3. Perrin claims that neh is used in direct quote because (i) it only refers to the reported
speaker in the matrix clause, and (ii) only direct speech takes its pronoun orientation
from the clause in which it is immediately embedded. Since (ii) is clearly false (see Β
und D above), Perrin's claim is vacuous unless sentences such as the following [(a)] are
acceptable in Mambila:

(a) Alhaji jia torn a neh kwull woo

Alhaji say rabbit to reported thank you
speaker
The co-referentiality between torn "rabbit" und woo "you" would establish (on the
basis of a above) that the clause neh kwull woo is a direct quote.
4. Tikar is a Bantoid language of the Benue-Congo group spoken in Cameroon. The
Tikar data in this paper was provided by Ellen Jackson.
5. See Karttunen (1971) for more discussion on implicative verbs.
44 Charles Ν. Li

6. Jackson (personal communication) cited Tikar as an exception. According to Jackson,

the Tikar reporter-speaker plays a partial role of the reported speaker in indirect
speech, i.e. while the pronominalization is that of indirect speech, intonation and
tense-aspect indicate direct speech. In the following example, the pronouns indicate
indirect speech, but the question is posed with reported speaker intonation and
question particle:
ä sh£ Ιέ myiö 1ε Si §i yenä swum ss nü neemzi yi mbyi
he say to his wife that they Quote saw thing that he sent them before-Q
He said to his wife, have they (wife and children) seen what he sent them before?

This type of Tikar example posses a challenge to the traditional dichotomy of direct
speech vs. indirect speech seen as two distinct categories where pronominalization
serves as the diagnostic signal for category classification.
7. Paez is a member of the Macro-Chibchan linguistic family of South America.
8. Guajiro belongs to the Arawakan family. Its speakers live in the border region between
Colombia and Venezuela.
9. Guanano is a member of the Tucanoan family. Its speakers number fewer than 1000 in
Colombia and Brazil.
10. Cofan has about 600 speakers along the Ecuador-Colombia border in the eastern
foothills of the Andes Mountains. It has been claimed to be a language isolate or a
member of the Chibchan family. Borman (1977) remarked that Cofan exhibits Chib-
chan features as well as features of Western Tocanoan.
11. The Andoke language is an isolate according to Witte (1977). It has less than one
hundred speakers all living near the Caquetä River of Colombia.
12. Cuiva is a Guahiban language of Colombia and Venezuela.
13. Polanyi (1982) cites several other forms of reported speech. I will not delve into them
here since they are beyond the scope of this study.

References

Borman, Μ. B. 1977. Cofan Paragraph Structure and Function, in Discourse Grammar,

Part 3, ed. by R. E. Longacre. pp. 289-338.
Chafe, Wallace L. 1982. Integration and Involvement in Speaking, Writing and Oral
Literature, in Spoken and Written Language, ed. by D. Tannen, pp. 35-44.
Gerdel, Florence L. and M. C. Slocum. 1976. Paez Discourse, Paragraph and Sentence
Structure, in Discourse Grammar, Part 1, ed. by R. E. Longacre. pp. 259-443.
Givön, Talmy. 1980. The Binding Hierarchy and the Typology of Complements. Studies in
Language, 4.3, pp. 333-377.
Haiman, John and Sandra A. Thompson. 1984. 'Subordination' in Universal Grammar, in
Proceedings from the 10th Annual Meeting of the Berkeley Linguistics Society.
Hopper, Paul and Sandra A. Thompson, (ed.) 1982. Studies in Transitivity. New York:
Academic Press.
Karttunen, Lauri. 1971. Implicative Verbs. Language, 47, pp. 340-58.
Kerr, Isabel. 1977. The Centrality of Dialogue in Cuiva Discourse Structure, in Discourse
Grammar, Part 3, ed. by R. E. Longacre. pp. 133-174.
 Direct and indirect speech 45

Kuno, Susumu. 1972. Pronominalization, Reflexivization, and Direct Discourse. Lin-

guistic Inquiry, III.2, 161-196.
Longacre, Robert E. 1976a. An Anatomy of Speech Notions. Lisse, Belgium: Peter de
Ridder Press.
Longacre Robert E., ed., 1976b. Discourse Grammar: Studies in Indigenous Languages of
Colombia, Panama, and Ecuador, Part 1. Published by the Summer Institute of
Linguistics and the University of Texas at Arlington.
Longacre, Robert E., ed., 1977. Discourse Grammar: Studies in Indigenous Languages of
Colombia, Panama, and Ecuador, Part 3. Published by the Summer Institute of
Linguistics and the University of Texas at Arlington.
Mansen, Richard and Karie. 1976. The Structure of Sentence and Paragraph in Guajiro
Narrative Discourse, in Discourse Grammar Part 1, ed. by R. E. Longacre. pp. 147-258.
Munro, Pamela. 1982. On the Transitivity of 'say' Verbs, in Studies in Transitivity, ed. by
P. Hopper and S. A. Thompson, 1982.
Partee, Barbara H. 1973. The Syntax and Semantics of Quotation, in A Festschrift for
Morris Halle, ed. by P. Kiparsky and S. Anderson. New York: Holt, 410-18.
Perrin, Mona. 1974. Direct and Indirect Speech in Mambila. Journal of Linguistics 10:
27-37.
Polanyi, Livia. 1982. Literary Complexity in Everyday Storytelling, in Spoken and Written
Language, ed. by D. Tannen, pp. 155-170.
Thurgood, Graham. 1981. The Historical Development of the Akha Evidentials System, in
Proceedings from the Seventh Annual Meeting of the Berkeley Linguistics Society,
pp. 295-302.
Waltz, Nathan E. 1976. Discourse Functions of Guanano Sentence and Paragraph, in
Discourse Grammar, Part 1, ed. by R. E. Longacre. pp. 21-146.
Wierzbicka, Anna. 1974. The Semantics of Direct and Indirect Discourse. Papers in
Linguistics, 7:3/4, 267-307.
Witte, Paul. 1977. Functions of the Andoke Copulative in Discourse and Sentence
Structure, in Discourse Grammar, Part 3, ed. by R. E. Longacre. pp. 253-288.
Some remarks on direct and indirect discourse in
Slave (Northern Athapaskan)*
Keren D. Rice

Summary

In the Athapaskan languages, three discourse types, direct discourse,

indirect discourse, and quotation, have been distinguished (Schauber
1975, Rice 1983).1 Most complement-taking verbs require indirect dis-
course complements, a few occur with direct discourse complements,
and a very small number can take direct quotation. The different
discourse structures correspond to different points of view from which
the complement is interpreted.
Complements of direct discourse verbs are interpreted from the point
of view of the subject of the direct discourse verb: first person refers to
the subject of the direct discourse verb. Complements of indirect dis-
course verbs are interpreted from the point of view of the discourse
speaker: first person decribes the speaker's situation, second person the
hearer's situation, and third person the situation of others. Direct
quotation is simply that: an evocation of the original speech situation.
In this paper, I shall be concerned with the properties of indirect
discourse and direct discourse determining verbs in Slave,2 a northern
Athapaskan language. These discourse types are distinguished in a
number of ways. First, there are differences in interpretation of pro-
nouns in the complement sentence. Second, there are structural differen-
ces in complement type between direct and indirect discourse determi-
ning verbs. Finally, question movement can move question words out of
complements of direct discourse verbs but not out of those of indirect
discourse verbs. I will not attempt in this paper to look at the properties
of Slave in terms of universal proposals of differences between discourse
types; I will be concerned solely with describing the distinguishing
characteristics of the discourse types in Slave. Section 1 is an overview of

* This research was funded by the Northern Social Research Division, Department of
Indian and Northern Affairs, Ottawa, Canada. My deepest thanks to all of the Dene
people who worked with me.
48 Keren D. Rice

the differences between discourse types, section 2 a discussion of pro-

noun use, section 3 presents structures, and section 4 syntax.
The discussion in this paper is set within the extended standard theory
model proposed by Chomsky (1981). The main relevance of this model
to the paper is that I consider semantics (logical form) to be an
interpretive component that assigns readings to syntactically wellformed
strings. Any sentence that cannot be assigned an interpretation by the
rules of logical form is filtered out.

1. An overview

An example of a verb which takes an indirect discourse complement is

shown in (1).
(1) John ?erakie?ie wihsj gu kodihshc?3
parka l.sg. made comp 3.sg. know
John knows that I made a parka.

In this sentence, the embedded sentence is interpreted from the point of

view of the speaker of the entire sentence. This contrasts with sentence
(2), which has a direct discourse determining verb as the matrix verb.
(2) Rosie ?erakie?ie wishj sedeyjdi
parka l.sg. made 3.sg. told l.sg.
Rosie ; told me that she; made a parka.

Here the complement is interpreted from the point of view of the subject
of the direct discourse determining verb, 'tell.' This can be clearly shown
by comparing the pronouns in the complement sentences. Both comple-
ments contain the verb wihsi Ί made.' In (1), with the indirect discourse
determining verb, the subject of this verb is understood from the point of
view of the speaker: the maker and the speaker are the same person. In
(2), the subject of wihsi is understood from the point of view of the
subject of the dominating verb; it is thus Rosie who made the parka.
A further difference between indirect and direct discourse determi-
ning verbs is illustrated in (1) and (2). While indirect discourse determi-
ning verbs take complementizers, as illustrated by the presence of the
complementizer gu. in (1), direct discourse determining verbs never take
a complementizer.
Another difference between indirect and direct discourse has to do
 Direct and indirect discourse in Slave 49

with the location of prominent question words. With indirect discourse

determining verbs, prominent question words always occur within the
sentence with which they are logically associated, as in (3).
(3a) beyighöne ke judone shptj gü ?akjhw?
3.parent Q when 3.sleep comp 3.pi make 3.sg
When do hisj parents make hircij go to bed?
b) * judone beyighöne ke shptj gu ?akjhw$

With direct discourse determining verbs, on the other hand, prominent

question words can, and in fact usually do, occur outside the sentence
with which they are logically associated.
(4a) yenene judone ri Albert ghp beghärayuhdä sudeli
nurse when PQM about l.sg will see 3.sg 3.sg want l.sg
When does the nurse want to see me again about Albert?
b) judöne ri yenene Albert ghp beghärayuhdä sudeli

With this brief overview of the differences between direct and indirect
discourse, we can now go on to examine the characteristics of the
discourse types in more detail.

2. Pronouns

2.1. Indirect discourse

Most complement-taking verbs in Slave require indirect discourse com-

plements. A few of the verbs are given in (5) below. All verbs are shown
in the third person imperfective form. 4
(5) -egodjhshp 'he knows'
?uweh?p 'he hears, listens'
goghahureht? 'he teaches'
-ghäyeda 'he sees'
?ekahededi 'he says thus'
-eradi 'he remembers'
-eguh?a 'he finds'

These verbs fall into different semantic classes: There are verbs of
perception, of saying, and of knowing.
Several examples of the use of indirect discourse verbs are given
below.
50 Keren Ό. Rice

(6) judenj rähgwe kodjhshp yfle

where l.sg live 3.sg know NEG
He doesn't know where I live.
(7) judenj dewpyä nj nenp ?ekähedewpdi
2.sg will go Comp 2.mother 2.sg say this
You sg. tell your mother where you're going.
(8) sü go were jp yidä keranedi
Q before here l.sg sat 2.sg remember
Do you sg. remember that I was here before?
(9) raxets'? ?awodie kegodihshp
l.pl to 3.sg will go l.sg know
I know that he will come to us.
(10) John hayi k'ähba m{ ?ah?t'e gu. begh<? ?tiweh?o
NM Colville Lake 3.sg Ν comp it about l.sg hear
I heard that John is living in Colville Lake.
(11) ?ariyiine hahshä ?ahahja
all 2.pi be big 2.pi became
You pi. all got big.

In all these examples, complement first persons refer to the speaker,

second persons to the hearer, and third persons to others. Given that all
verbs in Slave must be inflected for subject, there is nothing at all
unexpected about such pronoun markings.
The pronouns in the complement are summarized in (12).
(12) marking of embedded role reference of embedded role
1 speaker
2 hearer
3 other
Pronouns embedded under direct discourse verbs are thus interpreted
from the point of view of the speaker: this is 'regular' discourse.

2.2. Direct discourse

Complements of direct discourse verbs are interpreted from the point of

view of the subject of the direct discourse verb. While this is the primary
factor involved in interpretation of complements of direct discourse
verbs, a second factor plays a role: interpretation also depends on
whether the complement reports what was said to someone or what was
said about someone.
There are four direct discourse determining verbs in the Hare dialect
of Slave, as listed in (13).
 Direct and indirect discourse in Slave 51

(13) hadi 'he says' (intransitive)

-edi ~ -ededi 'he tells, asks' (transitive)5
yenjw? 'he wants, thinks' (intransitive)
hudeli 'he wants, thinks' (transitive)

These are all verbs of thinking and saying. However, while these verbs
do form a semantic class, indirect discourse determining verbs can also
occur in this semantic class. Verbs that allow direct discourse comple-
ments must be marked as such in the lexicon.

2.2.1. Reporting on statements made to someone

With the verb -di 'tell,' the complement reports directly what was said to
someone. The pronouns in the complement are interpreted according to
who is speaking, or the subject of the dominating sentence, and who is
being spoken to, or the object of the dominating sentence. Several
examples of this type of direct discourse are given below.
(14) John ?aranjla sehdi
2.sg go home 3.sg told l.sg
John told me to go home.
(15) John ?aranjla nehdi
2.sg go home 3.sg told 2.sg
John told you sg. to go home.
(16) John ?aranjla yehdi
2.sg go home 3.sg told 4.sg
John told him to go home.
(17) sine ?ahaht'e kudjdi
quiet 2.pi are 2.sg tell 3.pi
You sg. tell them to be quiet.
(18) segha räwpdi sedjdi yjle
1.sg for 2.sg will buy 2.sg tell l.sg PAST
You sg. told me to buy it for you.
(19) w'ilada sets'? ?anet'i yfle hedesj
again l.sg to 2.sg come NEG l.sg tell 3.sg
I told him not to visit me again.
(20) ?eghalayjda nedesj
2.sg work l.sg tell 2.sg
I told you sg. to work.

The surface subjects of the complements in (14) through (20) are all
second person. This second person does not refer to the hearer, but
rather to the one that the subject of the dominating sentence (i.e. John in
52 Keren D. Rice

(14) through (16), the hearer in (17) and (18), and the speaker in (19)
and (20)) is speaking to, or the hearer from the dominating subject's
point of view. The complement is a report of what the subject of the
direct discourse verb said to the object of this verb.
In the examples above, the person or persons being reported about in
the complement are identical to the object of the direct discourse verb.
In (22) through (24), the person(s) being reported about in the comple-
ment is identical to the subject of the dominating verb rather than the
object. The use of first person in the complement leads to an interpreta-
tion of identity to the subject of the direct discourse verb.

(21) le g^hljnj gots'e räwit'a naxökedi yjle

Norman Wells there to l.d will go 3.pi told l.pl past
They told us that they were going to Norman Wells.
(22) ?jlu yerehw? hedeyidi
apple l.sg want l.sg told 3.sg
I told her that I wanted an apple.
(23) kani ?awohle raxedeyinedi
like l.sg will do 2.sg told l.pl
You sg. told us that you would do it.

In the examples above, we see that second person in the complement

is interpreted as coreferential to the object and first person as p r e f e r -
ential to the subject of the direct discourse verb. When third person is
used in the complement, it is interpreted as non-identical to either the
higher subject or the higher object, as in (24) and (25).

(24) John dula sehdi

3.sg will go 3.sg told l.sg
John told me that you sg. were coming or John told me that she was
coming.

(25) ts'ödani hek'ewodi hedeyidi

child 3.sg keep l.sg told 3.sg
I told him that you will keep the kids, or
I told him that she will keep the kids.

As the glosses above illustrate, such sentences are often ambiguous since
complement third persons can be interpreted in more than one way.
Indirect discourse verbs are frequently used instead of direct discourse
verbs in such situations, as in (7) above.
The complement pronouns used when the matrix verb is a direct
discourse determining verb reporting on statements made to someone
are summarized in (26).
 Direct and indirect discourse in Slave 53

(26) marking of embedded role reference of embedded role

1 subject of direct discourse (DD)
verb
2 object of D D verb (hearer of
discourse from subject of D D
verb's point of view)
3 other

2.2.2 Reporting on statements made about someone

The remainder of the direct discourse determining verbs take comple-

ments that report what was said about someone. The pronouns in these
complements are interpreted differently than those discussed in section
2.2.1 because of this difference in reporting.

2.2.2.1 say
The simplest case of reporting about something in direct discourse is
illustrated with the verb hadi 'say', an intransitive verb. First person in
the complement is interpreted as coreferential to the subject of the direct
discourse verb.
(27) John hjdowedzine k'e deshjta duhla hadi
tomorrow on bush 3.sg will go 3.sg say
Johnj says that hej is going to the bush tomorrow.
(28) seni ?awohle hahdi
l.sg l.sg will go l.sg say
I said I'll do it.

Second persons in the complement are interpreted as in indirect

discourse or from the point of view of the speaker rather than that of the
subject of the direct discourse verb. This is illustrated in (29) and (30).
(29) Simon räsereyineht'u hadi
2.sg hit l.sg 3.sg say
Simon said that you sg. hit him.
(30) William neghp ?anjw$ hadi
2.sg for 3.sg like 3.sg say
William said he loves you sg.

Third person roles in the complement are interpreted from a direct

discourse point of view: they refer to neither the subject of the direct
discourse verb nor the hearer of the discourse. They are interpreted as
having 'other' reference: reference to the speaker or to some other third
person.
54 Keren D. Rice

(31) behshine räyuhdi hadi

sled 3.sg will buy 3.sg say
He said that I will buy a sled or He said that he will buy a sled.
(32) neji hadi
3.sg is scared 3.sg say
He said that I am scared or He said that she is scared.

With the verb 'say,' the pronouns in the complement are interpreted
from the point of view of the subject of the direct discourse verb except
in the case of second persons, which are interpreted from the point of
view of the speaker. These pronoun markings are summarized in (33).
(33) marking of embedded role reference of embedded role
1 subject of direct discourse verb
2 hearer of discourse
3 other

2.2.2.2 want 1 - intransitive

The verb yen{w% 'want' (intransitive) can have interpretation of comple-
ment pronouns as outlined in (33). A second interpretation of corefer-
entiality which may be used is discussed in section 2.2.2.5.
As summarized in (33), complement first persons are interpreted as
coreferential to the subject of the direct discourse verb.
(34) deshjta duhshä yerehwe
bush l.sg will go l.sg want
I want to go to the bush.
(35) SÜ leshuyie k'eguhw'e yerinewe
Q spoon l.sg will lick 2.sg want
Do you sg. want to lick the spoon?
(36) hjdowedzine k'e rirawohja yeniwe
tomorrow on l.sg will return 3.sg want
He wants to come back tomorrow.
(37) segha ts'ödani hek'£wodi yerehw?
l.sg for child 2.sg will keep l.sg want
I want you sg. to keep the child for me.
(38) sets'? räwodi yerinew?
l.sg to 3.sg will help 2.sg want
Do you sg. want him to help you?

Second persons are interpreted as discourse hearers.

(39) deshita dewoyä yerehwe
bush 2.sg will go l.sg want
I want you sg. to go to the bush.
 Direct and indirect discourse in Slave 55

(40) denexare wpj? yenjwe

sister 2.sg will sing 3.sg want
Sister wants you sg. to sing.
(41) nenp kuhe ?awpdie yenjwe
2.sg mother 3.pi with 2.sg will go 3.sg want
Your mom wants you sg. to go with us/them.

Finally, third persons are interpreted as either the discourse speaker or

other third persons.
(42) ?ehdinj- wani deshjta duyä yerehw^
NEG alone bush 2.sg will go l.sg want
I don't want him to go to the bush alone.
(43) wo je yerinewe
3.sg will sing 2,sg want
Do you sg. want him/me to sing?
(44) gpsho ?eghälayudä yenjw^
hard 3.sg will work 3.sg want
He wants me/her to work hard.

When there is a third person pronoun in the complement, it is ambiguous

in reference between the speaker and other third persons.
Reflexive and reciprocal pronouns in the complement are interpreted
as such and must have an antecedent within the complement.
(45) ?elegh<? näwit'a kenjw?
RECIP to l.d will go 3.pi want
They want to visit each other.
(46) ?edets'? guhdie yenjwe
REFL to l.sg will talk 3.sg want
He wants to talk to himself.

With the intransitive verb 'want,' coreference interpretation is accor-

ding to the rule summarized in (33).

2.2.2.3 want 2 - transitive

There is a second verb 'want,' -udeli, which is transitive. 6 This verb is
often used if the complement contains a pronoun that cannot be inter-
preted as coreferential to the matrix subject. Interpretation of comple-
ment roles is done according to the schema outlined in (33). Several
examples of this transitive verb are shown below.
(47) ?abä yeghäyudä sudeli
dad 3.sg will see 4.sg 3.sg want l.sg
My dad wants me to see it.
56 Keren D. Rice

(48) sü bek'e duh4a sudjli

Q 3.sg after l.sg will go 2.sg want l.sg
Do you sg. want to go with me?
(49) beghäyewpdä nudehli
2.sg will see 3.sg l.sg want 2.sg
I want you sg. to see it.
(50) neghäyuhdä nudeli
l.sg will see 2.sg 3.sg want 2.sg
He wants to see you sg.
(51) ?abäsets'? räwodi yudeli
dad l.sg to 3.sg will help 3.sg want 4.sg
My dad; wants her to help him.
(52) sehlege segha gonihkie rärulu yudeli
l.sg friend l.sg for slippers 3.sg will sew 3.sg want 4.sg
She; wants her, friend to sew slippers for her;.
(53) su kuk'e duhl4a kudili
Q 3.pi after l.sg will go 2.sg want 3.pi
Do you sg. want to talk with them?
(54) ?abä deshjta duyä hudehli
dad bush 3.sg will go l.sg want 3 sg.
1 want my dad to go to the bush.

First person complement roles are interpreted as coreferential to the

subject of the direct discourse verb, second persons as coreferential to
the discourse hearer, and third persons as referring to the speaker or to
other participants.
As can be seen in the examples above, the transitive verb 'want'
occurs with an object pronoun. The role of the particular object pro-
nouns is discussed in section 2.2.2.4. The object pronouns in Hare have
the following forms.
(55) singular plural
1 se- raxe-
2 ne- raxe-
3 0- ku-
4 ye- go-
areal go-
reflexive ?ele-
reciprocal ?ede-

As mentioned in section 2.2.2.2 above, with the intransitive verb

'want,' ambiguities often arise when there are third person pronouns in
the complement. The use of the transitive verb greatly reduces these
ambiguities between speaker interpretation and other participant inter-
 Direct and indirect discourse in Slave 57

pretation. This can be seen by comparing sentences: (47), (48), (51),

(52) and (53) are unambiguous while comparable sentences (38), (41),
and (43) with the intransitive verb are ambiguous between speaker
reference and other reference for third person complement pronouns.
The use of the transitive verb reduces but does not eliminate ambigui-
ties. In sentences with two third persons in the complement and the first
person object pronoun on the direct discourse verb, ambiguities arise.
(56) yegha ?ela ke wo?ie sudjli
4.sg with boat in 3.sg will go by boat 2.sg want l.sg
Do you sg. want him to go with me/me to go with him by boat?

The first person object pronoun signals that one of the third persons in
the complement is to be interpreted as identical to the discourse speaker;
it does not, however, specify which of the third persons has this refe-
rence and the sentence is ambiguous.
All of the examples above show non-anaphoric pronouns used as
objects. The anaphoric pronouns reflexive ?ede- and reciprocal ?ele- also
occur as objects on the transitive verb 'want.' Some examples of recipro-
cals are shown in (57) and (58).
(57) ?elegh9 näwit'a ?elekenegw?
RECIP l.d will go 3.pl want RECIP
They want to visit each other.
(58) ra?ekudlo lekudedli
3.pi will laugh 3.pi want REFL.
They want each other to laugh.7
While generally the transitive verb is not used if there is only a first
person pronoun in the complement, it can be used with areflexive
pronoun as object as in (59).
(59) guhshä ?edenegw?
l.sg will be wise3.sg want RECIP
He wants to be wise.
cf. guhshä yenjwe
l.sg will be wise 3.sg want
He wants to be wise,
(intransitive)

So far, the facts about the interpretation of complement pronouns

have been presented. There is no difference in interpretation of these
pronouns that is linked to transitivity of the direct discourse verb. The
transitive verb functions to reduce the number of potential readings
assigned a sentence with third person pronouns in the complement. In
58 Keren D. Rice

the next section, the form of the object pronoun used with the transitive
verb is discussed.

2.2.2.4 The object pronouns

With the transitive verb 'want,' the verb is inflected for object and the
form of the object pronoun controls the readings assigned to the comple-
ment pronouns. The object pronouns work on a hierarchy, as summari-
zed in (60).
(60) 1
2
areal
3
This hierarchy is interpreted as follows. If the object pronoun on the
direct discourse verb is first person, then the sentence is interpretable if
the subject of the direct discourse verb is not first person and there is a
third person pronoun in the complement. This third person pronoun is
linked to, or is interpreted as coreferential to, the object pronoun and
thus to the speaker. This can be seen in an example like (61).
(61) sets'4 rawpdi sudeli
l.sg to 3.sg will help 3.sg want l.sg
He, wants me to help him;.

Here the first person oblique object pronoun se- in the complement is
interpreted as referring to the subject of the matrix sentence, as predic-
ted by (33). The third person subject of the complement is interpreted as
referring uniquely to the discourse speaker since it must be linked to the
first person object in the matrix, which has discourse speaker reference.
The presence of this first person matrix object demands that there be a
third person pronoun in the complement that it controls.
If the object pronoun is second person then the sentence is interpreta-
ble if the subject of the matrix is non-second person and if there is a
second person pronoun in the complement sentence. Third persons in
the complement must be interpreted as having reference to other per-
sons than the discourse speaker; the only way they can be interpreted as
coreferential to the discourse speaker with the transitive verb is if the
direct discourse verb is inflected for object by the first person pronoun.
Consider the following example. With the intransitive verb, the sentence
is ambiguous.
(62) bets'? räwq>di yenjw?
3.sg to 2.sg will help 3.sg want
He wants you to help me/her.
 Direct and indirect discourse in Slave 59

With the transitive verb, only one reading is possible. Which reading is
assigned depends on the pronoun used on the direct discourse verb. If
the first person pronoun is used, the complement object must refer to the
discourse speaker since the first matrix pronoun controls the interpreta-
tion of the third person in the complement.

(63) bets'? rawpdi sudeli

3.sg to 2.sg will help 3.sg want l.sg
He wants you sg. to help me.

If the second person pronoun is used, only the reading where the third
person complement pronoun refers to some other third person is obtai-
ned; the third person pronouns are not controlled and must have other
reference.

(64) bets'? rawpdi nudeli

3.sg to 2.sg will help 3.sg want 2.sg
He wants you to help her.

With an areal object pronoun the sentence is interpretable if there is an

areal pronoun in the complement, no second pronoun in the comple-
ment and third persons are interpreted as uncontrolled. 8

(65) gots'? du4a godeli

area to 3.sg will go 3.sg want area
He wants her to go there.

Finally, if the third person pronoun 0- or fourth person pronoun ye- is

present on the direct discours verb, the sentence is interpretable if the
complement contains pronouns that can refer to other than speaker,
hearer, or matrix subject. This is illustrated in the examples below. Here
the third or fourth person matrix object controls interpretation of third
person complement pronouns, requiring that they be interpreted as
having other reference.

(66) third person object

SÜ kuk'e duh4a kudjli
Ο 3.pi after l.sg. will go 2.sg want 3.pi
Do you sg. want to walk with them?

(67) third person object

?abä deshjta duyä hudehli
dat bush 3.sg will go l.sg want 3.sg
I want my dad to go to the bush.
60 Keren D. Rice

(68) fourth person object

ts'<idani tidirta yile yudeli
child 3.sg will go out NEG 3.sg want 4.sg
He doesn't want the child to go out.

The pronoun hierarchy can be summarized as follows. If the first

person object pronoun is used, the complement must contain a pronoun
that can be interpreted as coreferential to the discourse speaker; if the
second person object pronoun is used, there must be a pronoun in the
complement interpretable as having hearer reference and no comple-
ment pronoun can be interpreted as having discourse speaker reference;
if there is a third or fourth person object pronoun, the complement
cannot contain pronouns interpreted as coreferential to the speaker,
hearer, or matrix subject. The matrix object thus controls the interpreta-
tion of complement pronouns. First person matrix objects require that a
pronoun in the complement be interpreted as discourse speaker; if there
is some other pronoun in matrix object position, discourse speaker
reference in the complement is not allowed and so on.
Notice that this hierarchy is independent of the syntactic role of the
complement pronoun. No matter what grammatical function the pro-
noun has, this hierarchy is used to determine interpretation. Thus in (69)
below, the first person object is linked to the complement subject while
in (70) it is linked to the complement oblique object.
(69) ?ama tjch'a räguyie sudeli
mom outside 3.sg will play 3.sg want, l.sg
My mom wants me to play outside.
(70) su bek'e rawoh^a sudjli
Q 3.sg. after l.sg. will go 2.sg. want l.sg.
Do you sg. want to go with them?

The hierarchy serves to place conditions on interpretation since the

matrix object must control pronouns in the complement. The use of the
transitive 'want' thus limits the number of possible readings assigned a
sentence by placing further conditions beyond the rule of interpretation
given in (33) on reference interpretation.

2.2.2.5 want 3 - indirect discourse

Discussion so far has focussed on the direct discourse interpretation
of complements of 'want.' Complements of this verb can also be inter-
preted from the point of view of the speaker, or from an indirect dis-
course point of view.
 Direct and indirect discourse in Slave 61

In the indirect discourse style, interpretation of complement pronouns

is done following (12): first persons refer to the speaker, second persons
to the hearer, and third persons to others.
The examples in (71) through (75) below show the transitive verb
'want' with indirect discourse interpretation. The matrix object pro-
nouns are linked to complement roles as described below.
(71) nehe ?awoht'e sekudeli
2.sg. with l.sg. will stay 3.pi. want l.sg.
They want me to stay with you sg.
(72) bets'? räwohdi sekudeli
3.sg. to l.sg. will help 3.pi. want l.sg.
They want me to help him.
(73) raxegha ?awoht'e nudeli
1.pl. with l.sg. will stay 3.sg. want 2.sg.
He wants you sg. to stay with us9.
(74) nek'e räwoh a sudjli
2.sg. after l.sg. will go 2.sg. want l.sg.
Do you sg. want me to go with you?
(75) sek'e rawo4a sudili
l.sg. after 2.sg. will go 2.sg. want l.sg.
Do you sg. want to go with me?
In these examples, it is clear that indirect discourse interpretation is
used: complement first persons refer to the discourse speaker rather than
to the matrix subject and complement third persons refer only to others,
never to the speaker.
The matrix object pronouns are linked to complement pronouns in
indirect as well as in direct discourse; however, the linking is slightly
different. The matrix object controls the subject of the complement
unless the subjects of the two sentences are coreferential, in which case it
controls a first person (if there is one), a second person (if there is no
first person), and otherwise a third person. The case with coreferential
subjects is illustrated in (75); here the matrix object is first person, linked
to a complement oblique object. In all the other examples, the matrix
object is linked to the complement subject. This is particularly noticea-
ble in (73), where with direct discourse the matrix object would be first
person as in an example like (63).
With the transitive verb 'want,' the matrix objects make it clear
whether the discourse is to be interpreted as direct or indirect discourse.
The intransitive verb 'want' can also occur with a complement that
receives an indirect discourse interpretation, as in (76) below.
62 Keren D. Rice

(76) Larry hayi sharon raxeda ?awot'e yenjw?

NM l.pl. for 3.sg. will stay 3.sg. want
Larry wants Sharon to wait for us.

This sentence receives an indirect discourse interpretation, with the first

person pronoun in the complement referring to the discourse speaker
(or, in this case, the group that the discourse speaker is part of). If it
were interpreted as direct discourse, (76) would be glossed 'Larry wants
Sharon to wait for them,' where Larry is a part of the group that Sharon
will wait for.
With the intransitive verb, a sentence can always be assigned either a
direct or an indirect discourse interpretation. Perhaps for this reason,
indirect discourse interpretation is rather uncommon with the intransi-
tive verb. It is found predominantly in two situations. 10
First, indirect discourse interpretation is found in sentences such as
(76) where a singular form and a plural form must be interpreted as
coreferential under a direct discourse interpretation. Use of the indirect
discourse interpretation avoids this case where the singular must be
interpreted as belonging to a larger group.
Second, indirect discourse interpretation with the intransitive verb is
also often found when a complement contains a possessive pronoun. In
the following sentence, the direct discourse interpretation marks the first
person possessive pronoun se- in the complement as coreferential to the
matrix subject while the indirect discourse interpretation marks it as
coreferential to the discourse speaker. This is indicated by the two
different glosses given to (77).

(77) John beya räwozie yudeli

l.sg. son 3.sg. will hunt 3.sg. want 4.sg.
a) John s wants hiSj son to go hunting, (direct discourse)
b) John wants my son to go hunting, (indirect discourse)

If the indirect discourse interpretation is assigned to (77), then the

sentence glossed (77a) is rendered as in (78) or (79).

(78) John beya räwozie yudeli

3.sg. son 3.sg. will hunt 3.sg. want 4.sg.
Johnj wants his; son to go hunting, (or John's son wants him to go
hunting.)
(79) John deya räwozie hudeli
R E F L son 3.sg. will hunt 3.sg. want
Johnj wants his; son to go hunting.
 Direct and indirect discourse in Slave 63

2.2.2.6 Summary
There are several verbs whose complements generally receive a direct
discourse interpretation. An indirect discourse interpretation is also
possible, especially with the transitive verb 'want.' An interpretation
holds for an entire complement: no complement is interpreted partly
from the point of view of the speaker and partly from the point of view of
the matrix subject.

2.3 Discourse shifting

As discussed above, the complement of a direct discourse verb is

interpreted from the point of view of the subject of this verb while the
complement of an indirect discourse verb is interpreted from the point of
view of the discourse speaker. When more than one complement sen-
tence is involved, it can be clearly seen that the interpretation of a
complement depends on the point of view of interpretation required by
the immediately dominating verb.
Sentences with more than one direct discourse verb show that it is the
subject of the immediately dominating sentence that controls the inter-
pretation of the complement clause.

(80) John [Susan [ e g^hljni ?awodie] yeniwe ] hadi

Norman Wells l.sg. will go 3.sg. want 3.sg. say
John said that Susan wants to go to Norman Wells.
(81) Frederick [Bill [sehe ?awodie] sudeli] hadi
l.sg. with 3.sg. will go 3.sg. want l.sg. 3.sg. say
Frederick said Bill; wants me to go with himj.
(82) Joanne [Tom [yehe ?awodie] sudeli] hadi
4.sg. with 3.sg. will go 3.sg. want l.sg. 3. sg. say
Joannej said Tom wants me to go with herj/him
or Joanne; said Tom wants herj/him to go with me.

The first person pronouns in (80) and (81) refer to the subject of the
immediately dominating sentence, in these examples the subject of
'want.' Thus, in (80), the one who will go is Susan and in (81) the one
who the speaker will go with is Bill. In all three examples, the pronouns
of the complement of 'want' are interpreted from the point of view of the
subject of this verb; those of 'say' from the point of view of its subject.
Shifts of interpretation can also be seen when both direct and indirect
discourse determining verbs are present.
64 Keren D. Rice

(83) si[menj s2 [?edihk'e ] S2 sedhi yile]si kodish<? yile

2.sg. shoot 3.sg. told l.sg.PAST l.sg. known NEG
I don't know who told me to shoot.

The highest verb requires an indirect discourse interpretation of its

complement, so in the verb sehdi the third person subject refers to
another person and the first person object refers to the speaker. The
verb sehdi, on the other hand, takes a direct discourse interpretation of
its complement so the second person object of 'shoot' is interpreted as
coreferential to the object of sehdi, or the hearer of the discourse from
the point of view of the subject of this sentence. Thus, SI is interpreted
from the point of view of the speaker while S2 is interpreted from the
point of view of the subject of sehdi.
Some further examples of discourse shifting are given below.

(84) si[s2[judeni gots'<? yjlj] S2 kegodushshä] S1 yenjw^

where there to 3.sg. is l.sg. will know 3.sg. want
He wants to know where she is.

in (84), the subject of 'want' and the subject of 'know' are interpreted as
coreferential.
(85) S1 [ S2 [däri ?ets'edehke'e] S2 kegoduhshä] S1 yerehwe
l.sg. will know l.sg. want
I want to know how to shoot.
(86) ?eyi dene [[ se esasone ?oneduh?ä] kegoduhshä] yenjw^
that man l.sg. kicker l.sg. will sell l.sg. will know 3.sg. want
That man wants to know if I'll sell my kicker.

The highest verb in (85) and (86) is a direct discourse determining verb,
so the first person subject of 'know' is interpreted as coreferential to the
subject of 'want.' The next verb, 'know,' is an indirect discourse determi-
ning verb, requiring that its complement be interpreted from the point of
view of the discourse speaker. The first person subject of 'sell' in (86) is
thus interpreted as coreferential do the discourse speaker.
A discourse determining verb controls the interpretation of only the
sentence embedded directly beneath it. This is shown in the examples
above. One further example, involving a relative clause structure, is
given in (87).
(87) John S i[seno 52t t'iere ghp ?aniw? i ] S2
l.sg. mother 3.sg. like Rel
ghäyudäjsi yenjw?
3.sg. will see 3.sg. want
John; wants hisj mother to meet the girl he^ loves.
 Direct and indirect discourse in Slave 65

The interpretation of SI is direct discourse since the matrix verb is direct

discourse determining. The first person possessive pronoun is thus
coreferential to John. S2 is controlled by the indirect discourse determi-
ning verb 'see, meet;' thus the third person pronoun can refer to John.
Point of view of interpretation of pronouns switches from one clause
to another in a complex sentence as a verb controls the interpretation of
only the sentence it most immediately dominates.

2.4 Lexical features

Verbs allowing direct discourse interpretation of their complements

must be diacritically marked as such in the lexicon as there does not seem
to be any feature that this follows from. Since there are two different
kinds of direct discourse determining verbs, two features are needed,
one for direct statements (statements to someone) and one for indirect
statements (statements about someone).
This diacritic feature is apparently not a property of the verb theme, 11
as the following examples show. The verb theme -di 'say' requires a
direct discourse interpretation of the complement in some cases (e.g.
hadi 'say', -ededi 'say, tell'), while in other cases it requires an indirect
discourse interpretation (e.g. ?ekdhededi 'say thus'). The verb theme \νς
is similar: the base yen(wq 'want' takes a direct discourse complement
while the base ?ahqhwq 'cause' requires an indirect discourse comple-
ment. It appears that the diacritic indicating direct discourse is a feature
of prefixes rather than of themes since a given theme can take comple-
ments of either discourse type.

2.5 Summary

Complement-taking verbs in Slave are of two types: those requiring

direct discourse interpretation of the complement and those requiring
indirect discourse interpretation. These types of interpretation differ in
the point of view from which the pronouns of the complement are
interpreted. The possible coreference relationships between matrix and
complement roles are determined by rules of control and of direct
discourse interpretation, as given in (26) and (33).
66 Keren D. Rice

3. Subcategorization

In this section, the grammatical characteristics of direct and indirect

discourse determining verbs are explored. We will see that there is one
major structural difference between these types of verbs: the direct
discourse verbs occur with S-complements, meaning they do not take
complementizers, while the indirect discourse determining verbs occur
with S' complements, or can occur with complementizers. This is shown
by illustrations in this section; in section 4 some of the syntactic ramifica-
tions of this difference in subcategorization are discussed.
The complements of indirect discourse determining verbs can occur
with final complementizers in Slave. Two of the complementizers in the
Hare dialect are n\ and gü. Both of these (and all other complementi-
zers) occur exclusively with the indirect discourse determining verbs.
Some examples are given below.
(88) gah [dek'ale gu] ?arat'i
rabbit 3.sg. is white Comp 3.sg. become
Rabbits turn white.
(89) Mary denp [judenj duyä ni] ?ekähedeyjdi
REFL mother where 3.sg. will go Comp3.sg. told
Mary told her mom where she is going.
(90) [gie gä rayeda go] begowit'o
road along 3.sg. is walking back Comp l.pl. found 3.sg.
We found him walking along the road.
(91) dene detiie [hühshu ni] gots'e yih?a
man REFL daughter3.sg. will marry Comp 3.sog. forced
He forced his daughter to marry.
(92) [segoyä htile gü] ?asedalie
l.sg. mind it is gone Comp 3.sg. caused l.sg.
He made me crazy.
(93) Mary [ setue ?etl'u nj] gäyurehtg
l.sg. daughter 3.sg. knit Comp 3.sg. taught 4.sg.
Mary taught my daugther to knit.

These examples show a number of different verbs which take indirect

discourse complements. With all these verbs, a complement final com-
plementizer n\ or gu, is allowed.
With direct discourse determining verbs, a complementizer is not
allowed. The following sentences have direct discourse determining
verbs and are ungrammatical with a complementizer.
 Direct and indirect discourse in Slave 67

(94) *beta [yah?0ne ?awohdie nj/gu] hadi

3.sg. father over there l.sg. will go Comp 3.sg. say
His dadi said that he; is going there.
(95) "Marie hayi [Susan goda?awot'e nj/gu] yenjw?
NM 3.sg. will wait to 3.pi. Comp 3.sg. want
Marie wants Susan to wait for us.
(96) *John [ sehe ?awodie nj/gu] yudeli
l.sg. with 3.sg. will go Comp 3.sg. want 4. sg.
John; wants her to go with him;.

Based on these data, it appears that indirect discourse verbs comple-

ments dominated by S', while direct discourse verbs take complements
dominated by S. Given a phrase structure rule S'—» S COMP, we can
account for why only indirect discourse determining verbs take comple-
mentizers: it is only these verbs that have a structural position for the
complementizer to fill.
One might propose an alternative analysis, that in sentences such as
(94) through (96) there is actually a complementizer that is phonetically
null. In the next section, we will see evidence that different subcategori-
zation frames are required by direct and indirect discourse determining
verbs.

4. Questions

There are two strategies used in Slave for forming direct content
questions. The prominent question word may either appear in the
position where it is generated by the phrase structure rules or it may
occur sentence initially. Some examples of question words occurring
where generated are shown in (97) through (100).
(97) Lucy yeri räyehdi
what 3.sg. bought
What did Lucy buy?
(98) ?eyi dene judeni ka?0
those people where 3.pi are going
Where are the people paddling?
(99) John men{ ghp ?anjw?
who to
Who does John love?
68 Keren D. Rice

(100) netä judöne radujä

2.sg father when 3.sg will return
When will your sg. dad return?

These question words can also occur sentence initially. When the nomi-
nal question words (yeri 'what' and men{ 'who') are placed sentence
initially, one of the pronouns, be-, third person or ye-, fourth person,
must occur within the sentence. The examples in (101) through (104)
below should be compared with (97) through (100).
(101) yeri Lucy räyeyehdi
what 3.sg bought 4.sg
What did Lucy buy?
(102) judeniteyi dene ka?0
where there people 3.pi are going by boat
Where are the people paddling?
(103) meni John yeghp ?anjw$
who 4.sg for 3.sg like
Who does John love?
(104) judöne netä radujä
when 2.sg father 3.sg will return
When will your sg. dad return?
In these example the question words are sentence initial.
The examples above all show simplex sentences. When prominent
question words are in complex sentences, there is an interesting differ-
ence between direct and indirect discourse with respect to the placement
of the question words.
When a prominent question word is in a complex sentence with a
direct discourse verb, the question word can occur to the left of the
sentence containing the direct discourse verb as well as in the embedded
sentence. This is shown in (105) through (107) below. In the (a) versions,
the prominent question word occurs in the embedded sentence, the
sentence with which it is semantically associated. In the (b) versions, it
occurs at the beginning of the sentence containing the direct discourse
verb.
(105a) Mary hayi [John juden{ ri raraxuwo?i] hadeyidi
NM where PQM 3.sg will wait for l.pl 3.sg said.
Where did Mary say John will wait for us? 12
b) juden{ ri Mary hayi John raraxuwo?j hadeyjdi
(106) Denise yeri gha sjka gudie sudeli
why l.sg to 3.sg will talk 3.sg want l.sg
Why does Denise want me to call her?
 Direct and indirect discourse in Slave 69

b) yeri gha Denise sjka gudie sudeli

(107a) nurse [ judoni ri Mary ghp beghärayuhdä] sudeli
when PQM about l.sg will see 3.sg 3.sg want l.sg
When does the nurse want to see again about Mary?
b) judöne ri nurse Mary gho beghärayuhdä sudeli
(108a) Margaret [ yeri segha wpshi] nehdi
whatl.sg to 2.sg will make 3.sg told 2.sg
What did Margaret tell you to make for her?
b) yeri Margaret segha wpshi nehdi

A question word can occur to the left of a direct discourse verb even
when the sentence it is semantically part of is deeply embedded under
the direct discourse verb. In other words, long distance movement of
prominent question words is possible so long as the dominating verbs are
all direct discourse determining verbs.
(109a) John hayi [Lucy hayi [Alfred juden( ri shuwote] yenjw?] hadi
NM NM where 3.sg will sleep 3.sg want 3.sg
PQM
Where did John say Lucy wants Alfred to sleep?
b) juden{ ri John hayi Lucy hayi Alfred shuwote yenjw^ hadi
With prominent questions, the preferred placement of the question
word in direct discourse is to the left of the sentence containing the direct
discourse verb. When the question word is syntactically part of the
sentence with which it is semantically associated, the interpretation is
often as an indirect question. For instance, a second reading for (109a) is
given in (110) and for (106a) in (111).13
(110) Margaret said to me 'What will you make for me?'
(111) Denisej wonders why I/he should call her;.

While the embedded question words can yield two interpretations, as

either a direct question or an indirect question, when the question word
occurs to the left of the sentence containing the direct discourse verb,
only a direct question interpretation is possible.
When the matrix verb is a direct discourse determining verb, promi-
nent question words can occur to the left of it. With indirect discourse
determining verbs, non-nominal prominent question words occur in the
sentence with which they are semantically associated.14 This is shown in
(112) and (113), where the (b) versions illustrate that ungrammatical
sentences arise when the question words occur outside the sentence they
are semantically part of.
70 Keren D. Rice

(112a) Raymond [Jane juden{ ri yjlj] kodjhsQ

where PQM 3.sg is 3.sg know
Where does Raymond know Jane to be?
b) *juden( ri Raymond Jane yjlj kodjhsQ
(113a) beyighone ke [ judone shptj] ?akjhw^
3.sg parent PL when 3.sg sleep 3.pi make 3.sg
When do his parents make him go to bed?
b) *judonä beyighöneke shptj ?akjhw$

We thus see that adverbial question words can occur either in the
sentence they are semantically part of or to the left of the dominating
sentence when the dominating verb is a direct discourse verb; with an
indirect discourse verb only the first of these positions is avialable. Why
should this be the case?
In Rice (1983), I argue that adverbial question words are all base
generated in the sentence with which they are semantically associated.
When the dominating verb is a direct discourse verb, the question word
can (and usually does) move while in indirect discourse such movement
is blocked. Let us adopt this transformational account of movement and
examine now the reasons for the differences between direct and indirect
discourse. We will see both a formal syntactic and a functional explana-
tion for the movement.
A sentence such as (107) has the following deep structure if we accept
that direct discourse determining verbs take complements dominated by
S, a proposed in section 3.15
(114) / S

Q Comp

ADV
judöne ri beghärayuhdä

Mary gha

When the question word is moved into question complementizer posi-

tion, the structure in (115) results.
 Direct and indirect discourse in Slave 71

(115)

Ο Comp

ADV
judönö ri

V
beghärayuhdä

Mary gha

judone in question complementizer position is bound to the adverb in

ri
the lowest sentence.
Now consider a sentence such as (112) where the question word does
not move. If we asume that indirect discourse determining verbs take S'
complements, as proposed in section 3, then (112) has the structure
shown in (116).
(116) ^-S

Q Comp

Here question movement fails to apply.

If we compare the structures in (114) and (116), we see one major
difference between them: the question word moves in (114) where no S'
is involved, but in (116) it cannot move out of S'.
Thus, given a difference in subcategorization between direct and
indirect discourse determining verbs, the placement of question words
follows from a general principle: no lexical item can move leftwards out
of S'. 16
72 Keren D. Rice

There is rather interesting confirmation of this analysis. In section

2.2.2., it was shown that the verb 'want,' while usually taking a direct
discourse complement, can take an indirect discourse complement as
well. When the complement is indirect discourse, the placement of
question words is like that for other indirect discourse determining
verbs: the question word cannot be moved to the front of the matrix
sentence. This is illustrated in (117), where (117a) has a direct discourse
complement and (117b) an indirect discourse complement.
(117a) yenene judenj ?awot'e sudeli
woman where 3.sg will stay 3.sg want l.sg
b) yenene judenj ?awotht'e sudeli
woman where l.sg will stay 3.sg will want l.sg
Where does the woman want me to stay?

An alternative version for (117a), with a fronted question word, is

possible. No such alternative exists for (117b).
(118a) judenj yenene ?awot'e sudeli
b) *judenj yenene ?awoht'e sudeli
If direct discourse complements are dominated by S and indirect dis-
course complements by S', these facts follow.
The analysis makes a second prediction. Since there is no complemen-
tizer present in the complement of a direct discourse determining verb,
one would predict that a question word cannot move to the front of a
direct discourse complement. With an indirect discourse complement,
the presence of such a complementizer position should allow question
words to occur at the front of the sentence. This indeed is the case. In
(119), the complement is indirect discourse: the question word can occur
at the front of the complement (119b) but not of the matrix (119c).
(119a) John [ beya judenj räwozie] sudeli
3.sg son where 3.sg will hunt 3.sg want l.sg
Where does John want my son to go hunting?
b) John [judenj beya raowzie] sudeli
c) *judenj John beya räwozie sudeli

We see that there is a formal syntactic explanation for the placement

of prominent question words: they can occur in question complementizer
position but cannot be moved over an S' to achieve this position. 17
Different structures for direct and indirect discourse complements
provide a syntactic explanation of conditions on question movement.
Schauber (1975) examines a functional explanation for unbounded ques-
 Direct and indirect discourse in Slave 73

tion movement with direct discourse verbs in Navajo. The facts of

Navajo and Slave appear to be very similar and the following discussion
relies on Schauber's excellent analysis.
As we have seen, unbounded question word movement is allowed
only in sentences containing direct discourse verbs, or verbs which
govern the point of view from which the sentence is interpreted. What is
contained in the direct discourse part of the sentence is interpreted from
the point of view of the matrix subject but the question word itself is
interpreted from the speaker's point of view. There is thus a conflict in
points of view for interpretation when a prominent question word is
within the scope of a direct discourse verb. When the question word is
moved leftward, it moves out of the scope of the direct discourse part of
the sentence and is thus interpreted from the point of view of the
speaker. Movement is out of the scope of a verb interpreted from the
subject's point of view into the scope of one interpreted from the
speaker's point of view.
Question movement occurs only when a constituent falls in the scope
of a verb but cannot be interpreted from the point of view of that verb.
This conflict arises only in cases of direct discourse complements; thus no
movement occurs with indirect discourse determining verbs.
Syntactically, a question word cannot move over an S'. Since only
direct discourse determining verbs take S complements, it is only with
these verbs that question movement occurs. The function of this move-
ment rule is to move the question word out of the scope of a direct
discourse verb, thus resolving point of view of interpretation conflicts.

5. Summary

In Slave, there is a difference between direct discourse, indirect (or

regular) discourse, and direct quotation. While direct discourse reports
on events from the subject of the direct discourse verb's point of view,
indirect discourse is speaker oriented. There are two major differences
between direct and indirect discourse, pronouns and structure. First,
pronouns are interpreted differently under the discourse types due to
this difference in point of view of interpretation. Second, the comple-
ment types required by direct and indirect discourse requiring verbs
differ. From this difference in complement type, it follows that direct
74 Keren D. Rice

discourse determining verbs allow unbounded question movement while

indirect discourse determining verbs do not. This question movement
results in a resolution of conflicts of point of view of interpretation.

Notes

1. The terms direct discourse and indirect discourse are not used in the standard way in
the Athapaskan literature. 'Indirect discourse' refers to 'regular' discourse while 'direct
discourse' refers to reported speech or indirect quotation. The term 'direct quotation'
is used in the standard way.
2. Slave is an Athapaskan language spoken in northern Canada. All the data in this paper
is drawn from the Hare dialect of Slave. The facts about direct and indirect discourse
are similar in all the other dialects.
3. All examples are given in the practical orthography, this differs from standard
transcription symbols as follows:
j =[J]
ch' = [C]
sh = [$}
gh = [γ]
ie = [i?] (a diphthong)
Υ = nasalized vowel
V = high tone
V - low tone
The following abbreviations are used.
1 first person NM noun marker
2 second person Q question marker
3 third person PQM prominent question marker
REL relative clause complementizer
Comp complementizer
NEG negative
4 fourth person (third person object when subject is third person)
sg. singular
d.dual
pi. plural
rec. reciprocal
refl. reflexive
Verbs are glossed as follows: X.V.Y, where X is the subject and Υ the object. This
reflects English word order rather than Slave morpheme order.
4. Some of these forms are morphologically perfective forms, but they function as
imperfectives. The final syllable of the verb is the verb stem; prefixes represent various
kinds of derivational information not relevant to the discussion here.
5. The stem -di is used in all persons but the first person singular. In this form, the
suppletive stem -s( is used.
6. This verb -udeli is used only in the Hare dialect. In the other dialects a transitive form
that is morphologically related to the verb yeniwe discussed in section 2.2.2.2 is used.
The properties described for -udeli in Hare hold for this verb in the other dialects.
 Direct and indirect discourse in Slave 75

7. This particular sentence can be rendered only with the transitive verb. With the
intransitive verb, the sentence obtained is as in (i).
(i) rä?ekudlo gonjw?
They; want thenij to laugh.
A reciprocal reading is not possible.
8. The pronoun go- areal is used to refer to items that occupy space or time and are
perceived as permanent in some sense - rivers, mountains, houes, and other buildings
(but not tents).
9. The direct discourse interpretation of this sentence is 'he wants you sg. to stay with
them' where the subject of 'want' is part of the group that the hearer will stay with.
10. Direct discourse interpretation seems to be less marked with the intransitive verb in all
cases.
11. The verb theme is the lexical entry for a verb. A base is formed from a theme by adding
derivational prefixes to the theme. The forms given in the text as themes are actually
simplified; the bases given in the text are inflected for third person.
12. The particle ri marks that the question word is a prominent question word and thus that
the question is to be interpreted as a direct question rather than as an indirect question.
This can be seen very clearly in examples with matrix indirect discourse determining
verbs, such as (i).

(i) Mary Joe judenj ri yjlj kodjhsh?

where PQM 3. sg Ν 3. sg know
Where does Mary know Jane to be?

(i) has only the direct question interpretation. This contrasts with (ii), where ri is not
present, which is interpreted as an indirect question.

(ii) Mary Joe judenj yjlj kdjhshp

Mary knows where Joe is.
See Rice (1983) for further discussion.
13. (105a) and (107a) must be interpreted as direct questions because of the presence of
the particle ri. See note (12).
14. I restrict the following discussion to non-nominal question words. In Rice (1983) I
argue that non-nominal question words achieve sentence initial position by movement
while nominal question words are base generated in that position.
15. There is evidence for a sentence-initial question complementizer position: the yes/no
question particle sti occurs sentence initially, as shown in the sentences below.

(i) sü Mary teht'ie weht'ie

Q bread 3.sg baked
Did Mary bake bread?

(ii) su [ts'$dani ke tjch'a rägokeyie gu] kughäyjda

Q child pi outside 3.pi play Comp 2.sg see 3.pi
Do you sg. see the children playing outside?
There is no evidence for a verb phrase node in Slave; the trees shown in the text are
thus basically fan-like in structure. See Rice (1983) for further discussion.
16. Further confirmation of this analysis comes from topicalization. Topics occur in the
following structure:
76 Keren D. Rice

TOPIC S'

With direct discourse verbs in S', non-nominal question words can be moved from a
complement into topic position, with indirect discourse verbs in S', such movement is
blocked. See Rice (1983) for details.
17. This analysis further predicts that no topics should be allowed in the complements of
direct discourse determining verbs. In other words, structures such as the following
should not occur.

direct discourse verb

All my data is consistent with this prediction.

References

Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris.

Rice, K. 1983. A Grammar of Dene (Slave). Manuscript.
Schauber, E. 1975. Theoretical Responses to Navajo Questions. Ph. D. diss. Cambridge,
Mass.: MIT.
Reported speech in Yoruba
Ay ο Bamgbose

Reporting is an essential speech act, particularly in the context of

communication. Since not everyone can be present in every speech
situation, it follows that what is observed by one person may have to be
reported to others. Indeed, even the participants in a speech situation
may have cause to report a speech event to one another.
It is usual to refer to a report incorporating the exact words of the
speaker as direct speech and a report in which the speaker's utterance has
been adapted and rendered in the reporter's own words as indirect (or
reported) speech (Quirk et al. 1972: 785). For example, in the following
sentences.
(la) Οΐύ ηί, "mo ή bo"
PN say I PROG come1
'Olu said, "I am coming'"
b) Οΐύ ηί oun ή bo
'Olu said that he was coming'
(2) mo ή bö
Ί am coming'

(la) is taken to be the direct speech, and (lb), the indirect or reported
speech. Strictly speaking, (la) is not really the direct speech. It is a
report of the direct speech by the reporter. The speaker's exact words
are (2), the so-called "quotation content" (Perrin 1974: 27) of (la). In
this paper, the term direct speech will be limited to the exact words of the
speaker as in (2). A report incorporating the direct speech as in (la) will
be referred to as a direct-speech report, while a report in the reporter's
own words as in (lb) will be referred to as an indirect speech report or,
just simply, indirect speech. Both the direct-speech report and the
indirect-speech report are two kinds of reported speech (Leech: 1971:
99).
In the speech situation, the speaker may address himself to all the
persons present or to one or more persons among those present. In order
to give an accurate report, the reporter may wish to specify the speaker's
addressee. This is done by employing an appropriate verb which makes
the indication of a noun or pronoun object possible as in
78 Ayo Bamgbose

(3a) Οΐύ so fiin Ayö ρέ ki ό mäa bö

PN say give PN COMP, let he PROG, come
O l u told Ayo that he should be coming'
b) Οΐύ dä mi lohün pe öun ή b<)
'Olu answered me that he was coming'

If a reporter is an eyewitness of the speech event he is reporting, his

role in the speech situation may have been that of a mere onlooker (i.e. a
passive participant) or that of an active participant, either as speaker or
addressee. In the majority of eyewitness reports, the reporter is an
onlooker. In some, the report is vague about the role of the reporter. For
example, the reporter in (1) could have been an onlooker or an addres-
see. However, where a report specifies the role of the reporter through
the choice of an appropriate pronoun, it is possible to distinguish
formally an active participant reporter from a passive one. The report in
(3a), for instance, is that of a passive reporter, while that in (3b) is that of
an active reporter who happens to have been the addressee in the
original speech situation. It is even possible for a reporter to report his
own speech. For example, in
(4) mo ni mo fee lo
I say I want go
Ί said I wanted to go'

the reporter is the same person as the original speaker.

1. Direct and indirect speech reports

In some languages of the world, direct und indirect speech reports are
kept apart by a number of devices. For instance, in English, indirect
speech expressions are backshifted in tense, time adverbials are modified
to make them more distant, and questions are altered to look like
statements.2 In Yoruba, these devices are non-existent. For example, as
there is no morphological distinction between present and past tense, the
use of back-shift is automatically eliminated. Hence, the change from
direct speech in (2) to indirect speech in (lb) does not involve any
change in the form of the verb. Similarly, indirect questions need not
undergo a change of form, apart from pronoun deixis, nor do time
adverbs need to be changed. Compare, for example, (5a) with (5b), and
(6a) with (6b):
 Reported speech in Yoruba 79

(5a) ό ηί, "ki ni η mäa je?"

he say what FOC. I will eat
'He asked, 'What am I going to eat?"
b) ό ηί ki ni öun mäa je
'He said what was he going to eat?'
(= 'He asked what he was going to eat')
(6a) ό ηί, "mä ä lo löla"
he say I will go tomorrow
'He said, Ί will go tomorrow"
b) ό ηί oun ä lo 101a
'He said that he would go tomorrow'
(= 'He said that he would go the next day')
There are, however, four main ways of identifying an indirect-speech
report in contrast to a direct-speech one:
a) There is the pronoun reference by which the first person pronoun in
the direct speech automatically changes into a third person pronomi-
nal whenever it is co-referential with the subject of the reported
speech (other than the first or second person pronoun subject). 3
Examples of this change are to be seen in (lb) as compared with (la),
(5b) as compared with (5a), and (6b) as compared with (6a).
b) Certain verbs are associated only with indirect speech. For example,
a possible alternative indirect question to (5b) is:
(7) ό bere ρέ ki ni oun mäa je
he ask COMP, what FOC. he will eat
'He asked what was he going to eat?'
(= 'He asked what he was going to eat')
In this sentence, bere 'ask' has been made to replace the neutral
report verb ni 'say' and, when it is used together with the comple-
mentizer pe 'that', it becomes clear that what we have here is an
indirect question.4
c) Certain sentence types have their own distinctive indirect speech
forms. For example, indirect imperatives always have the particle ki
'let' following the report verb. See, for instance, (3a).
d) When two or more sentences are embedded in an indirect statement,
the complementizer ρέ 'that' with or without the conjunction ati 'and'
can occur between such sentences. Such occurrence is not possible in
a direct speech report.
(8a) ό ηί, "mä ä lo; mä ä si tfete padä"
he say I will go 1 will and. quickly return
'He said, Ί will go and I will return quickly"
b) ό so pe öun ä lo, (äti) pe öun ä tete padä
'He said that he would go and that he would return quickly'
80 Ayο Bamgbose

The direct speech incorporated in (8a) consists of two co-ordinate

sentences. Notice that these sentences, when they occur in the indirect
speech in (8b), are now treated as sentential complements which may be
conjoined in the same way as noun phrases.
In written Yoruba, direct speech is always differentiated from indirect
speech by the familiar device of quotation marks. This can be seen in all
the examples above. However, in spoken Yoruba, where there are no
such signposts, it is quite easy for ambiguities to arise between direct and
indirect-speech reports in those cases where the four distinguishing
characteristics of indirect speech discussed above are not in evidence.
The possibility of such ambiguity becomes quite strong when it is
realized that some verbs are common to both direct and indirect speech
reports. For example, using the verb ni 'say', it is only the context that
can determine whether (9) is a direct-speech report or an indirect one in
the spoken language:

(9) Οΐύ ni ko burii

PN say NEG bad
O l u says, 'It isn't bad"
'Olu says that it isn't bad'

2. Pronoun Deixis

The possible participants in a speech situation are the speaker, the

addressee, and the other. Since no participant can report to himself,
there are six possibilities of reporting. However, each of the three
participants may report to someone not present in the speech situation.
Let us call this person the outsider. This allows for three additional
possibilities of reporting. Taking (10) as the direct speech, the nine
possibilities of reporting are as shown in (11), (12), and (13):
(10) mo fee ri e
I want see you
Ί want to see you'
(11a) Speaker to Addressee: mo ni mo fee ri e
Ί said I wanted to see you'
b) Speaker to Other/ mo ni mo fee ri i
Outsider: Ί said I wanted to see him'
 Reported speech in Yoruba 81

(12a) Addressee to Speaker: ο η ί ο ίέέ ri mi

'You said you wanted to see me'

b) Addressee to Other/ ό ηί öun ίςέ ri mi

Outsider: 'He said he wanted to see me'
(13a) Other to Speaker: ο ηί ο f§e ri i
'You said you wanted to see him'

b) Other to Addressee: ό ηί öun ίές ri e

'He said he wanted to see you'

c) Other to Outsider: ό ηί öun fee ri i

'He said he wanted to see him'

The indirect-speech reports in (11)—(13) when compared with the

direct speech in (10) indicate the kind of shifts that take place in the
pronouns. The generalization that can be made is that pronoun reference
is oriented towards the active participants i.e., either the speaker or the
addressee. In all cases, the subject of the indirect-speech report is
oriented towards the speaker, but the orientation of the pronouns in the
embedded sentence depends on the original pronoun in the direct
speech. In the case of the pronouns in (11)—(13), the pronoun subject of
the embedded sentence, like the subject of the matrix sentence, is
oriented towards the speaker, while the pronoun object is oriented
towards the addressee.5 The pronoun choice is illustrated in the follow-
ing table:

Reporting Reported to Neither

Subject Choice
Speaker: 1st Person 1st Person 1st Person
(No shift) —» 2nd Person —*• 3rd Person

Object Choice
Addressee: 2nd Person 2nd Person 2nd Person
—> 1st Person (No shift) —* 3rd Person

The above table shows that when the speaker is reporting, the original
first person subject of the direct speech remains unchanged; but when he
is reported to, the first person subject shifts to the second person. In all
other cases, the first person subject shifts to the third person. Similarly,
when the addressee is reporting, the second person object shifts to the
first person, but when he is reported to, the second person object
remains unchanged. In all other cases, the shift of the second person
object is to the third person.
82 Ayo Bamgbose

When the subject of both the matrix and the embedded sentence in an
indirect-speech report is a third person pronoun, the question arises
whether such pronouns of the same number are co-referential. For
example, in English, only the context can determine whether the two
pronouns in each of (14a) and (14b) are co-referential or not.
(14a) He said that he would like to go
b) They said that they would like to go

The situation in Yoruba is different. A clear distinction is made between

subjects that are co-referential and those that are not, except in the case
of the third person plural of the embedded sentence which may or may
not be co-referential with the subject of the matrix sentence (Bamgbose
1966: 107). Consider the following sentences:

(15a) ό ηί ό fee Ιο
'He said he would like to go'
(he = someone else)
b) ό ηί öun fee Ιο
'He said he would like to go'
(he = he himself)
(16a) w<5n ηί wön fee lp
'They said they would like to go'
(they = other people or they themselves)
b) wön ηί äwon fee Ιο
'They said they would like to go'
(they = they themselves)

When co-referentiality is intended, the strategy is to choose a third

person pronominal and match it with a pronoun of the same person and
number.
As simple as it may look, determining co-referentiality may, in fact,
be quite complicated in the case of multiple embedding. For example, in

(17) ό ηί ό so pe ki öun wä bä öun Ιο

he say he say COMP let he come accompany he go
'He said that he asked that he should come along with him'

there are two third person singular pronominals. Which pronoun subject
is each one of them co-referential with? To determine this, one has to go
into the derivation of the indirect speech as follows:

(18a) ό ηί, "wä bä mi Ιο"

'He said, "Come along with me"
 Reported speech in Yoruba 83

b) ό so pe ki η wä bä oun Ιο
'He said that I should come along with him'

In the direct-speech report in (18a), the addressee is reporting the

speaker to some other person using his exact words. Hence, he has to
choose a third person pronoun subject in referring to him while retaining
the first person pronoun mi 'me' in the direct speech. In the indirect
speech in (18b), the first person has to shift to a third person pronominal
that is co-referential with the subject. When (18b) is further reported by
a passive reporter to another person as in (17), the initial pronoun
subject no longer refers to the original speaker, but rather to the earlier
addressee-reporter. Hence, the first person pronoun in (18b) has to shift
to a pronominal that is co-referential with the initial pronoun subject in
(17). Thus the co-referentiality in (17) is a follows:
(19) ό] ηί ό 2 so pe ki öuni wä bä öun2 Ιο
'Hej said that he 2 asked that he! should come along with him2'

One other type of pronoun deixis that has its motivation in Yoruba
culture is what may be called taboo reference. It arises from a deliberate
avoidance of attributing to oneself unpleasant things such as death,
illness, and other types of misfortune. Suppose someone says to me
(20) ara e ö da
body your NEG well
'You are sick'

In reporting this statement to another person, the expected indirect

speech should have been (21a). But the chances are that I will rather
produce (21b) to avoid attributing the undertones of a mental illness to
myself:
(21a) ό ηί ara mi ö da
'He says that I am sick'
b) ό ηί ara öun ö da
'He says that he is sick'

Obviously (21b) now becomes ambiguous, and the person being repor-
ted to has to work out from the context whether I am reporting truly that
the speaker says that he is sick or whether I am accusing him of claiming
that I am sick.
84 Ayo Bamgbose

3. Report verbs and their complements

There are two types of report verbs: the neutral report verb which is a
purely locutive verb merely stating the fact of the speaker saying
something, and the marked report verb which conveys the reporter's
description of the manner or content of what is said.
The neutral report verbs are: ni 'say', wi 'say', and so 'say'. Of these
verbs, only ni 'say' can take a sentential complement without an overt
complementizer. The others require the complementizer ρέ 'that'. The
distribution of the verbs according to the type of report in which they
may occur is as follows:

Direct-Speech Indirect-Speech Report

Report

All Types Indicative Imperative Interrogative

Verb
ni 'say' + + + +
wi 'say' Ί- + + +
so 'say' Ο + + ο

There is some controversy as to whether ρέ is a report verb meaning

'say' or a complementizer meaning 'that'. In support of its verbal status,
sentences may be found in which it occurs without a preceding verb 6 , but
this appears to be the exception rather than the rule. Consider the
following sentences:
(22a) <?mo nää ό tile pe oun ri mi
child the ΝEG even say-that he see me
'The child didn't even say that he saw me'
( = 'The child pretended not so see me')

b) o m o nää ö tile so pe öun ri mi

'The child didn't even say that he saw me'

c) *omo nää ö tile pe, "mo ri e "

'The child didn't even say that "I see you'"

d) *οιης> nää ö tile so pe, "mo ri e "

'The child did not even say that, "I see you'"

e) Qmo nää ö tile ni pe oun ri mi

'The child didn't even say that he saw me'
 Reported speech in Yoruba 85

f) omo nää tile so pe öun ri mi äti pe öun mo mi

child the even say say-that he see me and say-that he he know me
'The child even said that he saw me and that he knew me'
g) ό dära pe won jo lo
it be-good say-that they assemble go
'It is good that they went together'
h) pe wön jo lo dära
'That they went together is good'
In (22a), pe, which normally occurs with a preceding verb, now occurs
without such a verb. Hence, in this context, it is believed to be a verb
meaning 'say'.
There is little doubt that historically pe must indeed have been a
report verb; but in a synchronic description of Yoruba, there is over-
whelming evidence that it has lost its verbal status and is now no more
than a complementizer. 7 Five reasons may be advanced in support of this
claim:
a) In the overwhelming majority of cases, pe occurs in conjunction with
a preceding verb.
b) In the few cases, such as (22a), where it so occurs, it is an alternative
to soρέ 'say that' as in (22b), and the same restrictions apply to them.
For example, neither of them can be followed by direct speech. This
suggests that ρέ in (22a) is not really a verb like ni 'say', but rather a
residual form of so ρέ 'say that' after the deletion of so 'say'.
c) Although the verb ni 'say' need not have an overt complementizer,
pe may be added to it in indirect speech only. See, for example,
(22e). Again, this restriction of ηί ρέ 'say that' to indirect speech
confirms the complementizer status of ρέ.
d) Since complementation is akin to nominalization, it is not surprising
to find that sentential complements introduced by ρέ can be con-
joined by a typical noun-phrase conjunction. See (22f). An analysis
that makes ρέ a verb will not be able to account for this otherwise
inexplicable nominal behaviour.
e) Sentential complements introduced by ρέ occur with verbs other than
report verbs; and, in the case of those verbs that take the impersonal
pronoun subject, the entire complement may be extraposed to re-
place this subject (Bamgbose 1975: 204). For example, see (22g) and
(22h). Again this nominal behaviour supports the status of pe as a
complementizer.

A look at the distribution of the neutral report verbs shows that two of
them can occur in indicative, imperative, and interrogative sentences in
both direct and indirect-speech reports, and the third in more than one
86 Ayo Bamgbose

sentence mood. This means that there must be some device for indica-
ting which sentence mood occurs in any given indirect-speech report.
The indirect interrogative is marked by the presence of a question
item such as a question marker or a question nominal. The only
difference between a direct interrogative and an indirect one is the
pronoun deixis. Examples of the indirect interrogative are:
(23a) ό ηί se öun le lo
he say QM he can go
'He said could he go?'
( = 'He asked if he could go')
b) ό wi ρέ s£ öun le lo
'He asked if he could go'
c) ό ηί ki ni ä ή wä
he say what FOC we PROG seek
'He said what are we seeking?'
( = 'He asked what we were looking for')
d) ό ηί ta ni ä ή wä
'He asked who we were looking for'
If (23a) and (23b) were direct interrogatives, the pronoun in the direct
speech would have referred to the speaker as mo T . Similarly, for (23c)
and (23d), the pronoun in the direct speech would have referred to the
addressees as e 'you (plural)'.
The indirect imperative is marked by the absence of a question item
and the occurrence of the particle ki 'let' immediately after the comple-
mentizer ρέ 'that', or alternatively in the case of ni 'say', after the report
verb itself. Should a question item occur with ki 'let', the complement
would be an indirect interrogative. Consider the following sentences:
(24a) ό so pe ki η wä
he say COMP let I come
'He says that I should come'
b) ό wi pe ki η wä
'He says that I should come'
c) ό ηί pe ki η wä
'He says that I should come'
d) ό ni ki η wä
'He says that I should come'
e) ό ηί s6 ki öun lo
'He asked if he should go'
In (24a-d), ki 'let' follows directly after the complementizer pe 'that' or
the report verb. Hence each of them is an indirect imperative. In (24e),
 Reported speech in Yoruba 87

however, a question item precedes ki 'let', which makes the complement

an indirect interrogative.
The indirect indicative is marked by the absence of the markers of the
two other moods. For the complement in an indirect-speech report to be
indicative, it must not contain a question item nor the particle ki 'let'.
For example,
(25a) ό ηί (pe) wön ti lo
'He said that they had gone'
b) ό wi pe wön ti lo
'He said that they had gone'
c) ό SQ ρέ wön ti lo
'He said that they had gone'
Since, in the direct-speech report, the verb wi 'say' must be followed
by the complementizer pe 'that', it follows that unless some other
distinguishing festure of indirect-speech report (such as pronoun refe-
rence or coordination) is employed, there may be no way of telling
whether a sentence in which this verb occurs is a direct or indirect-speech
report in the spoken language. (25b) above is an example of such
ambiguity.
A summary of the neutral report verbs together with the distinguish-
ing features of their complements may be shown as follows:

Direct-Speech Report Indirect-Speech Report

Indicative Imperative Interrogative

Markers -Q - o
-ki +ki +Q

0) ni 'say' ηί, ηί ρέ ni ki, ηί ρέ ki ni + Q, ni ρέ + Q

(ϋ) wi pe 'say that' wi ρέ wi ρέ ki wi ρέ + Q
(Hi) - so ρέ 'say that' so ρέ ki -

The marked report verbs are always followed by the complementizer

pe 'that', and they almost invariably occur in indirect speech. 8 Examples
of these verbs and the sentence moods with which they are associated are
as follows:
Indicative
debi fun 'blame' fow<5 söyä 'assure, pledge'
fi csun kän 'accuse' büra 'swear'
puro mö 'lie against' jewo 'confess'
88 Ayo Bamgbose

säläye 'explain' jiyan argue

seleri, leri 'promise' tahün 'complain'
tara 'worry' se 'deny'
jäkä 'boot' ko 'refuse'

Imperative
däbää 'suggest' wääsü 'plead with, condole'
kilo fiin 'warn' bä kedün 'sympathize with'
gbädurä 'pray'
däjo 'judge' pe 'summon'
ki 'greet' darf ji 'forgive'

Indicative!Imperative
sepe 'curse' gbä 'agree'
gegün-ύη 'curse' fönte 16 'decree'
taku 'insist, refuse' be, b£be 'beg'
ro 'beseech' päse, pa ä läse 'order'
tenu mo 'emphasize' kede 'announce'

Interrogative
bere, bi leere 'ask' wädili 'enquire, investigate'

AU Moods
dähün, da lohün'answer'
fesi, ίύη lesi 'answer'
kigbe 'cry out' shout'
logun 'shout'
ke mo 'snap at'

The above classification of marked report verbs shows that some of

them occur in more than one mood. This happens because the verbs are
sensitive to the semantics of the complement sentence. For example, one
would expect a report verb like ρ äse 'order' to take only an imperative
sentence complement. However, since there are indicative sentences
that can be used to express prohibitions, it follows that such sentences
can also occur as complements to this verb as in:
(25a) ό päse pe ki η mä jäde
he order COMP let INEG go-out
'He ordered that I should not go out'
 Reported speech in Yoruba 89

b) ό pä§e pe n o gbödö jade

'He ordered that I must not go out'

The complement of (25a) is an indirect imperative sentence while that of

(25b) is an indicative sentence.
In the case of marked report verbs that occur with all sentence moods,
it can be observed that they are closest semantically to the neutral report
verbs in that they concentrate mainly on the speech act, describing its
manner or indicating its conversational status. Given this minimal se-
mantic colouring, it is not surprising that they can occur with all sentence
moods. Consider, for example,
(26a) ό dähün pe öun le lo
'He answered that he could go'
b) ό dähün pe ki w$n fi öun sfl?
'He answered that they should leave him alone'
c) ό dähün pe se öun Ιέ bä w<?n lo
'He answered that could he go with them?'
(= 'He answered asking whether he could go with them')

In (26a) the complement is indicative. In (26b) it is imperative, and in

(26c) it is interrogative.
Even when marked report verbs occur in the same category, the
semantics of each verb will determine the kind of complement it can
take. For example, debifün 'blame' can only refer to an event that has
taken place in contrast to säleri 'promise' which necessarily refers to a
future event as in (27a) and (27b) respectively.
(27a) ό d£bi fun un pe ό lo
'He blamed him that he went'
(= 'He blamed him for going')
b) ό §eleri ρέ öun ä 1ς>
'He promised that he would go'

In written Yoruba, there is a device for making practically any report

verb introduce a direct speech. It consists of either the insertion of the
adverbial bäyii 'thus' between the verb and complementizer or the
addition of another sentence containing an anaphoric pronoun and the
neutral verb ni 'say'. For example, as in
(28a) ό kigbe bäyii pe, "fi mi βίΐφ"
'He shouted thus, "Leave me alone"
b) ό seien bäyii pe, "η ό bä ο lo"
'He promised thus, "I will go with you"
90 Ayo Bamgbo$e

c) ό ki i bäyii pe, "ό digbä"

'He greeted him thus, "Goodbye".
(29a) ό ke m6 mi, ό ηί, "padä sfle"
'He snapped at me and (he) said, "Go back home"
b) Ade takii, ό ηί, "emi ö Ιο"
"Ade refused and (he) said, "I won't go"
Apart from report verbs, there are several other verbs in Yoruba that
take the complementizer pe. The question then arises, how does one
know when a verb that takes this complementizer is not a report verb? It
seems that there are two criteria for determining this:

a) A report verb must be capable of being substituted for any of the

neutral report verbs. Consider the following examples:
(30a) ό ηί, "mä binu"
'He said, 'Don't be angry"
b) ό b£ e pe ki ό mä binü
'He begged him that he should not be angry'
c) ό pe ki ό mä bimi
'He wished that he should not be angry'
(31a) ό ηί, "mo le se e"
'He said, "I can do it".
b) ό jäkä pe öun le se e
'He boasted that he could do it'
c) ό rö pe öun le se έ
'He thought that he could do it'
Whereas (30b) is a plausible indirect-speech report of (30a) and (31b) of
(31a), the same cannot be said for (30c) and (31c). This is because the
verb fe 'want, wish' and ro 'think' are not true report verbs. 9
b) A report verb must convey the report of a speech addressed to
someone, whether the addressee is specified or not. In fact, practi-
cally all report verbs, with the exception of ηί 'say', are capable of
having the addressee specified. In contrast to this, whatever informa-
tion is conveyed in the complement of a non-report verb is not
addressed to anyone. Rather, it represents the private thought,
observation or comment of the speaker. Compare (30b), where the
addressee is indicated by the third person pronoun object, with (30c)
where no addressee can be meaningfully introduced.
 Reported speech in Yoruba 91

4. Content of reported speech

In reporting someone else's speech, a reporter can take certain liberties

particularly when the report is in indirect speech. Among such liberties
are: choice of report verb, paraphrase (Zwicky 1971: 73-77, Leech 1980:
34-35), and injection of his own knowledge or view of the world (Leech
1980: 51-52).
In an indirect speech, a neutral report verb is often replaced by a
marked report verb which semantically represents the reporter's descrip-
tion of the content of the speech. Thus, instead of merely reporting that
the speaker said this or that, we are told that he shouted, complained,
begged, asked, warned, announced or agreed. Even in direct-speech
reports, the same liberty may be taken in the written language by
employing the devices illustrated in (28) and (29).
It is well-known that a reporter has the freedom of paraphrasing what
a speaker says. The only constraint is that such a paraphrase must
preserve the spirit of the original speech. Consider the following senten-
ces:
(32a) ό ηί, "e fi mi sfle"
'He said, 'Leave me alone"
b) ό ηί ki η fi öun sfl§
'He said that I should leave him alone'
c) ό ηί ki η mä yo öun lenu
'He said that I shouldn't bother him'
d) ό ηί egbön öun ti 1<?
'He said that his brother had gone'

The expected indirect-speech form for (32a) is (32b), but (32c) is a close
paraphrase. If a reporter were to give (32d) as the paraphrase of (32b),
this would clearly be inaccurate and therefore unacceptable. It is because
of this presumption of accuracy when reporting that even something
completely illogical can be accepted as long as it is presented as an
accurate report of someone else's utterance. Take, for example, the
following sentence:
(33) ό ηί öun päde pkunrin οΐόη-meji
'He said he met a man with two heads'

Irrespective of whatever reservation one may have about a two-headed

male human, we do not doubt the fact that the speaker being reported
claimed that he met one.
92 Ayo Bamgbose

Fidelity in reporting what is said applies to the content rather than the
form of what is said. If a speaker uses an ill-formed sentence, a reporter
may only repeat such a sentence in a direct-speech report, but not in the
indirect speech (Zwicky 1980: 50). Consider the following sentences:
(34a) ό ηί "gbe ägüngün wä"
'He said 'Bring the medicine"
b) *6 ni ki η gbe ägüngün wä
'He said I should bring the medicine'
c) ό ni ki η mu öögün wä
'He said I should bring the medicine'

The direct speech in (34a) contains the exact words of someone who is
not proficient in Yoruba. In asking for a charm, he uses the expression
gbe ... wä 'bring' normally used for fetching heavy objects, instead of
mu.... wä 'bring' which is used for light objects which can be held in one
hand. He also uses the wrong lexical item for 'medicine'. In spite of this,
since (34a) is a direct-speech report, the sentence is perfectly acceptable.
However, when this wrong expression occurs in an indirect speech as in
(34b), it becomes unacceptable. The correct report has to eliminate the
errors as in (34c).
In making a paraphrase of what a speaker says, the reporter is at
liberty to introduce his own knowledge or view of the world. For
example, suppose a speaker were to say that he was going to buy aspirin.
The reporter, from his experience of the use to which this analgesic is
commonly put, could report that the speaker was going to buy a
medicine for headache. Similarly, suppose a speaker were to produce the
direct speech in (35a), the reporter could report it as (35b).

(35a) Olü ni, "iyä Düpe ö f^ran omode"

PN say mother PN NEG like children
'Olu says, 'Dupe's mother does not like children"
b) Οΐύ ni aj6 ni iyä Düp£
'Olu says that Dupe's mother is a witch'

Notice that the speaker did not call Dupe's mother a "witch". This was
the conclusion of the reporter based on the Yoruba world view that a
woman who does not like children must be a witch.
The intrusion of the speaker's world view as in (35b) raises the
question of referential opacity (Leech: 1980: 51-52). Suppose the report
in (35b) represented what the speaker actually said? This sentence would
then be ambiguous as to whether it was the speaker that called someone
 Reported speech in Yoruba 93

a witch or the reporter. If it was the reporter, this would amount to the
transparent reading, but if it was the speaker, it would mean the opaque
reading (Heringer 1969: 91).

5. Reporters and types of reporting

Reference has earlier been made to two types of reporters: an active

(participant) reporter and a passive (participant) reporter. As the sen-
tences in (11)—(13) show, this is important in the pronoun deixis of
indirect speech.
A participant reporter, whether active or passive, is necessarily an
eyewitness to the speech act being reported. Two kinds of eyewitness
reports are possible: eyewitness-to-eyewitness and eyewitness-to-outsider.
The direction of the reports may be illustrated as follows:

a) Eyewitness-to-Eyewitness
Speaker ^ > Addressee

b) Eyewitness-to-Outsider

In the eyewitness-to-eyewitness report, both reporters and those

reported to are participants in the original speech act. Any of them can
report to any of the others. Hence, there are six possibilities of such
reporting as in ( l l a - b ) , (12a-b), and (13a-b). This type of report is the
marked one, for the reason that it is not usual for an eyewitness of a
speech act to report it to another eyewitness. When this happens, it is
94 Ayo Bamgbose

generally done for some special purpose such as reiteration, emphasis,

request for confirmation, reminder, recapitulation etc.
In the eyewitness-to-outsider report, only the reporters are partici-
pants in the original speech act. The outsider to whom they are reporting
is absent from the original speech situation. Since each of the three
participants can report to an outsider, there are three possibilities of
reporting as in (lib), (12b) and (13c). The eyewitness-to-outsider report
is the more usual and more common one, as can be observed in the
majority of reports of this type in this paper. This is so because it is only
natural to expect an eyewitness to a speech act to report it to someone
not present.
Apart from eyewitness reports, there are hearsay reports. What
distinguishes the latter from the former is that the reporter is repeating
something which has been passed down from one or more previous
reporters. There are two types of hearsay reports: general hearsay and
specific hearsay.
In the general hearsay report, the original speaker is a nonedescript
character unknown to the reporter. In Yoruba, this character is repres-
ented by the third person plural pronoun won 'they' as in:

(36a) wön ni m^tö ögä wä ti bäje

they say vehicle boss our PERF. spoil
'They say that our boss's car has broken down'
b) WQn ni enik^ni 6 gb<jdö jäde 101a
they say anyone NEG must go-out in-tomorrow
'They say that no one should go out tomorrow'
c) wön ni göminä ö Ιο
'They say that the governor did not go'

The fact that this pronoun does not refer to any persons in particular is
illustrated by (i) the possibility of the reporter answering, "I don't
know" to the question, "Who says?" and (ii) the impossibility of this
pronoun subject being co-referential with the subject of the embedded
sentence as in (16b). This contrasts with an eyewitness report in which
this pronoun stands for an identifiable subject. Since the report verb ni
'say' generally used in general hearsay reports also occurs in eyewitness
reports, it follows that any report of the type in (36) is potentially
ambiguous as between an eyewitness report in which the subject is
identifiable, and a hearsay report in which it is not. It must be noted,
however, that the time reference of the matrix sentence in a general
hearsay report is always understood to be the present (i.e. 'they say' and
 Reported speech in Yoruba 95

never 'they said') even when the time reference of the embedded
sentence is past. See, for example, (36c).
Another characteristic of a general hearsay report is that the reporter
does not accept any responsibility for the accuracy of the report. Should
the report turn out to be false, the reporter can take refuge in its
indefinite source.
The specific hearsay report differs from the general one in the sense
that the subject of the report is identifiable. However, the report is one
that incorporates another report as in
(17) ό ηί ό so pe ki öun wä ba öun Ιο
he say he say COMP let he come accompany he go
'He said that he asked that he should come along with him'

The reporter of the specific hearsay report in (17) was not an eyewitness
of the original speech act. But the eyewitness of this speech act who was
also the addressee reported it to him. He now takes it upon himself to
report what has been reported to him to someone else. The grammatical
device used in presenting specific hearsay reports is multiple embedding.
A type of reporting which is found in literary narrative texts is the so-
called free indirect speech. It consists of the omission of the reporter's
introduction such as ό ηί 'he says/said' and some other characteristics of
indirect speech such as the complementizer pe 'that', but the pronoun
deixis typical of indirect speech is retained. Consider the following
examples:
(37a) ki ni ki öun se bäyii?
'What should he do now?'
b) ό ηί "ki ni ki η se bäyii"?
"He said, 'What should I do now?"
c) 6 bere pe ki ni ki öun se bäyii
'He asked what should he now do?'
( = 'He inquired what he should then do')

(37a) is the free indirect speech version of the direct and indirect-speech
reports in (37b) and (37c) respectively. As can be seen, the reporter's
introduction and the complementizers are absent in (37a), but the form
of the indirect speech is otherwise preserved. The use of the free indirect
speech is not common in Yoruba literature. It is generally confined to
the description of a character's inner thoughts (i.e. the so-called "inte-
rior monologue").
The above account of reported speech in Yoruba indicates that a
language must have some way of distinguishing direct-speech reports
96 Ayo Bamgbo$e

from indirect ones. In the absence of back-shift in tense, the factors that
seem to count most in the language are pronoun deixis, choice of a
marked report verb (which almost invariably indicates an indirect-speech
report), the separation of sentence moods, and to some extent, the
occurrence of a complementizer.

Notes

1. Abbreviations used in this paper are COMP, (complementizer), FOC. (focus marker),
NEG. (negation marker), PN (personal name), PROG, (progressive aspect marker), Q
(question item), and QM (question marker).
2. The complementizer is not mentioned here because although both English and Yoruba
have a complementizer 'that', its distribution differs considerably in both languages. For
example, the Yoruba complementizer 'that' occurs in sentential complements of all
moods and sometimes in direct-speech reports. Apparently the latter feature also occurs
in other languages (Jespersen 1924: 299).
3. A pronominal resembles a pronoun in having a system of person and number but
otherwise, it behaves entirely like a noun. Some people refer to it as an independent
pronoun.
4. An alternative form for an indirect yes/no question is bire bi... 'ask if . . . ' . For the
WH-type question, a noun phrase may be substituted for the question item as in 6 bird
rtnkan ti dun mäa je 'He asked (for) the thing he was going to eat'.
5. The embedded sentence is the sentential complement following the report verb, while
the matrix sentence is the rest of the reported speech.
6. This appears to be the strongest of the arguments in Oyelaran (1982: 111-119) in
support of the status of ρέ as a verb.
7. Evidence in support of the diachronic status of pi as a verb may be found in the
analogical behaviour of the combination wipe 'say that' which now functions both as a
report verb and a complementizer. (Note, for instance, that wherever ρέ occurs as a
complementizer, with the exception of the position after wi 'say', it can be substituted
with wi ρέ 'say that'). This suggests that ρέ must have been a verb before passing
through a similar process of de-verbalization.
8. These verbs are a sub-set of speech act verbs (Verschueren 1977), but since some of
them take objects, they are strictly speaking predicates. The only case I have encounted
so far in written Yoruba where a marked report verb plus the complementizer occurs in
a direct speech report is dähün 'answer' e.g. 6 dähün ρέ, "ό ti yd" 'He answered that,
'All is ready".
9. Note, however, that some grammarians include verbs of thinking, hoping, knowing, etc.
in report verbs. See, for example, Jespersen (1924: 290, 294).
 Reported speech in Yoruba 97

References

Bamgbose, Ayo. 1966. A Grammar of Yoruba. Cambridge. Cambridge University Press.

Bamgbose, Ayo. 1975. Relative Clauses and Nominalized Sentences in Yoruba. In: R. K.
Herbert, ed. 1975. Proceedings of the Sixth Conference on African Languages. Colum-
bus, Ohio: Department of Linguistics, Ohio State University, pp. 202-209.
Heringer, James. 1969. Indefinite Noun Phrases and Referential Opacity. In: R. Binnick.
Papers from the Fifth Regional Meeting Chicago Linguistic Society. Chicago: Depart-
ment of Linguistics, University of Chicago, Illinois, pp. 89-97.
Jespersen, Otto. 1924. The Philosophy of Grammar. London: George Allen & Unwin Ltd
Leech, Geoffrey. 1971. Meaning and the English Verb. London: Longman.
Leech, Geoffrey, 1980. Explorations in Semantics and Pragmatics. Amsterdam: John
Benjamins Β. V.
Oyelaran, Qla$Qpe. O n the Scope of the Serial Verbal Construction in Yoruba'. Studies in
African Linguistics 13, 109-146.
Perrin, Mona. 1974. Direct and Indirect Speech in Mambila. Journal of Linguistics 10,
27-37.
Quirk, Randolph et al. 1972. A Grammar of Contemporary English. London: Longman.
Verschueren, Jef. 1977. The Analysis of Speech Act Verbs: Theoretical Preliminaries.
Bloomington, Indiana: Indiana University Linguistics Club.
Zwicky, Arnold. 1971. On Reported Speech. In: C.J. Fillmore and D.T. Langendoen,
eds. Studies in Linguistic Semantics. New York: Holt, Reinhart and Winston, Inc.
pp.73-77.
Reported speech in Swahili
David P. B. Massamba

Summary

The importance of reported speech in natural languages cannot be over-

emphasized. It is an undeniable fact, for example, that while we would
like to have our public-intended messages reach everyone directly,
circumstances are such that we are almost always exposed to a limited
audience only. It is through reported speech that our utterances/messa-
ges reach hearers with whom we are not directly in touch. Yet it is
ironical to see that an important mechanism such as this is rarely given
proper attention.
In this paper we address the question of reported speech in Swahili
with an aim to examine both the descriptive and theoretical implications
that underly this phenomenon within the context of the said language. In
doing this we examine a number of issues which are crucial in Swahili
indirect speech. Such are issues like indirect speech 'indicators', catego-
ries that undergo switches and how these switches are carried out,
optionality and obligatory nature of elements within the reported clause,
the nature of indirect speech in declarative, interrogative and imperative
sentences; and the restrictions on the kind of verbs that appear in the
reporting clause.

1. Introduction
Reported speech might be defined as a message conveying utterance for
which the reporter claims no authorship. 1 It has been common practice
in the literature to distinguish three kinds of speech as described below:

1.1 Oratio recta

This is a style of speech commonly known as verbatim quotation. Within

this style reporting is made by quoting the actual words of the original
speaker as exemplified in (1) and (2) below:
100 David P. Β. Massamba

(1) "Huyu mzee ni mgonjwa," Juma alisema

this oldman is sick person Juma he-pst-say
"This old man is sick", Juma said.
(2) Amina akasema, "nitakwenda mjini kesho."
Amina she-pst-say I-fut-go town tomorrow
Amina said, "I'll go to town tomorrow".

In each of these examples the unquoted words are usually referred to

as the 'reporting clause' while the quoted form is generally known as the
'reported clause'.

1.2 Oratio obliqua

Oratio obliqua is a style of speech in which the actual words of the

original speaker are not quoted but rather are paraphrased. We might
say with Godwin (1934) that although in the reported sentence the
wording changes in order to fit the point of view and grammatical
construction of the reporter's speech the relationship (in terms of
message) between it and the original utterance is highly maintained.
Essentially what happens therefore is that the reporter uses his own
words, as it were, to say what the original speaker said. Observe the
following:
(3) Maganga alisema kwamba baba yake alikuwa mgonjwa
Maganga he-pst-say that father his he-pst-be sick
'Maganga said that his father was sick'
(4) Magesa alisema kuwa alitaka kwenda Kenya
Magesa he-pst-say that he-pst-want go Kenya
'Magesa said that he wanted to go to Kenya'

Note that in (3) and (4), unlike in (1) and (2) there are no actual words of
the original speakers (i.e. Maganga and Magesa).

1.3 Style indirect libre

The third style of speech is one which is known as 'style indirect libre.' It
is a half-way stage between direct and indirect speech (Quirk et al.: 1972
p. 789). It is the most commonly used style in modern narrative writing.
It is basically a form of indirect speech. The difference here is that in it
the reporting clause is usually omitted and all other aspects of direct
 Reported speech in Swahili 101

speech sentence structure are retained. Quirk et al. give the following
example:
"So that was their plan wasn't it? He well knew their tricks and would show a
thing or two before he was finished. Thank goodness he had been alerted, and
there were still a few honest people in the world ..."
In the above quotation reporting is indicated by the underlined words.
(Note that in the two sentences (i.e. in the quotation) there is no
reporting clause.)

2. Indirect speech 'indicators' in Swahili

By indirect speech 'indicators' we simply mean those sentential elements

that are used to either introduce or indicate a reported clause. Maw
(1969) lists the following such indicators, which she calls 'groups introdu-
cing reported speech': kuwa, kwamba, ya kuwa, kana, kana/kama
kwamba, eti, kuona. (All of these indicators are equivalent to the
English form 'that' as used in reported speech.) Unfortunately, Maw
does not give examples to show how these forms might be used as
indirect speech 'indicators'. The only form she tries to provide examples
of is kuwa; and even with this her examples are not clear enough. Many
of the problems encountered in Maw's analysis are due to the kind of
data she uses. 2 Those who speak standard Swahili will agree that most of
Maw's examples of indirect speech are derived from some dialect and
not from standard Swahili. The conclusion she draws cannot, therefore,
be taken to represent the whole of the Swahili language; and more
specifically they cannot be taken as a representative sample of data from
standard Swahili.
My own investigation of indirect speech in standard Swahili has
revealed that the following 'indicators' are the ones that are very
commonly used: kuwa, ya kuwa, kwamba, ya kwamba, kama, eti, and
sijui. It may perhaps be worthwhile at this juncture to reiterate on the
more commonly used 'indicators.' As pointed out above, these 'indica-
tors' are equivalent to the English form 'that' as used in reported speech.
From the point of view of function these 'indicators' are best regarded as
complementizer elements. In other words, they are used to introduce
complement sentences. Bearing in mind the TG-framework we might
102 David P. Β. Massamba

say that these elements are dominated by the node Comp: The indicators
etilati and sijui, however, need more elaboration. The form eti (someti-
mes ati) means something very close to the English expression Ί say!' (as
an expression of surprise). Whenever this indicator is used the implica-
tion is that there is some element of surprise involved; furthermore, the
reporter is casting doubt on what was stated by the original speaker. The
form sijui is in itself a sentence meaning Ί don't know' (i.e., si- = NEG
particle for 1st person sg.; -ju- (root) = know; -i = NEG time reference
marker). This form, just like etilati, implies that the reporter casts doubt
on what the original speaker said. However, unlike etilati, it does not
bear elements of surprise.
The less commonly used 'indicators' are kana kwamba and/or kama
kwamba.3 According to my investigation the forms kana and kuona
alone cannot be used as 'indicators' as Maw suggests.
The more commonly used indicators can be divided into two groups
depending on whether or not the reporter is casting doubt on what was
stated by the original speaker. For this reason the first five 'indicators'
(i.e. kuwa, ya kuwa, kwamba, ya kwamba and kama) fall under the first
group - they give no indication that the speaker is casting doubt on what
he/she is reporting. The last two 'indicators,' however, (i.e. eti, sijui)
imply that the reporter is casting doubt on what the original speaker said.
The following examples illustrate how these 'indicators' are used.

(5) Juma alisema kuwa angekuja kesho yake

Juma he-pst-say that he-pst-come next day
'Juma said that he would come the following day'
(6) Amina amesema ya kuwa atakwenda sokoni
Amina she-pfct-say that she-fut-go market-LOC
'Amina said that she will go to the market'
(7) Daudi alieleza kwamba mama yake alifika siku hiyo
Daudi he-pst-explain that mother his she-pst-arrive day that
'Daudi explained that his mother came that day'
(8) Magesa alieleza ya kwamba a singerudi siku hiyo
Magesa he-pst-explain that he-NEG-pst-come back day that
'Magesa explained that he wouldn't come back that day'.
(9) Mwalimu alisema kwamba atasafiri wiki ijayo
Professor he-pst-say that he-fut-travel week next
'The professor said that he would travel next week.'
(10) Fatma amesema eti hakumwona mgeni
Fatma she-pfct-say 'that' NEG/PERS-OM-see stranger
'Fatma said that she didn't see the stranger.'
 Reported speech in Swahili 103

(11) John alisema, sijui, atakuja kesho

John he-pst-say Ί don't know' he-fut-come tomorrow
'John said that he would probably come tomorrow.'
(12) Juma alisema kana kwamba asingekwenda mjini (?)
Juma he-pst-say that he-NEG-pst-go town-LOC
'Juma said that he wouldn't go to town.'
(13) Mtoto alieleza kama kwamba hakupenda kucheza (?)
Child he-explained 'that' NEG-pstneg-like to-play
'The child explained that he didn't like playing.'

The question marks in the last two constructions indicate that such
constructions are not acceptable to some speakers. This is because the
'indicators' used raise questions as to their acceptability. In fact my
research showed that kana kwamba and kama kwamba are used more
often than not to mean 'as if rather than 'that'; in which case they
wouldn't be considered as 'indicators.'
It is important to note that in some contexts these 'indicators' of
reported speech are optional while in other contexts they are obligatory.
The optionality and the obligatory nature of these 'indicators' are
dependent on the nature of the matrix verb, i.e. the verb in the reporting
clause. Observe the following examples:

(14i) Juma alisema kwamba angekuja kesho yake

Juma he-pst-say that he-pst-come next day
'Juma said that he would come the following day.'
ii) Juma alisema ( ) angekuja kesho yake
(15i) Magesa alieleza kuwa angekuja kesho yake
Magesa he-pst-explain that he-pst-come next day
'Magesa explained that he would come the following day.'
ii) *Magesa alieleza ( ) angekuja kesho yake.
(16i) Amina alinifahamisha kwamba angekuja kesho yake
Amina she-pst-me inform that she-pst-come next day
'Amina informed me that she would come the following day.'
ii) * amina alinifahamisha ( ) angekuja kesho yake
(17i) Daudi aliniambia kuwa angekuja kesho yake
Daudi he-pst-me-tell that he-pst-come next day
'Daudi told me that he would come the following day.'
ii) Daudi aliniambia ( ) angekuja kesho yake

The above examples are only a representative sample of the many cases
that one could find in Swahili. What seems clear here is that within the
104 David P. Β. Massamba

reported speech of declarative sentences the verb that occurs in the

reporting clause4 plays an important role with regard to the acceptability
of sentences of this nature. Given what we have seen above we might say
that verbs of this sort tend to fall into two groups. First, there are those
whose occurrence does not necessarily require an 'indicator'. Such are
verbs like sema (cf. 14) and ambia (cf. 17). Secondly, there are those
whose occurrence necessarily requires an 'indicator', otherwise the
sentences so formed will be unacceptable. These are verbs like eleza (cf.
15) and fahamisha (cf. 16).
The basis for the kind of grouping of verbs we are making here is
rather opaque. In literature the more common tendency for grouping or
classifying verbs has been that of transitivity, that is, examining whether
the verbs are transitive or intransitive. Clearly we cannot use this
criterion to capture the kind of generalization we are aiming at. A rather
plausible criterion would be one that hinges upon the semantic implica-
tions of the verbs used. A close examination of the verbs we have in our
data reveals that, while verbs, such as, sema and ambia tend to give mere
statements of fact, other verbs, such as, eleza and fahamisha seem to
imply more than the statement of facts. That is, they seem to go beyond
stating the facts. We would like to claim that this is a more appropriate
basis for grouping the verbs in the way we did. This criterion must,
however, be taken with precaution. There are some verbs that at first
look would tend to lean toward a particular group when in actuality they
do not belong to that group. A good example of this is a verb like shauri
(advise). A rather superficial look at this verb gives the impression that it
belongs to the eleza, fahamisha group (henceforth group B) in which
verbs do imply that something more than simply stating the facts is
actually being given. However, a more critical examination of its seman-
tic implications shows that it actually belongs to the sema, ambia group
(henceforth group A) in which verbs give statements of facts and nothing
more. This is so because the meaning of such a verb does not contain
'elaboration' or 'explanation' as integral elements; they are only given
when required. In other words 'explanations' and 'elaborations' do not
form part of the core of the meaning of the item shauri. It would mean,
therefore, that one needs to examine the semantic status of the verb
more critically before deciding under which group it should be placed.
Given this state of affairs, we can say that the grouping in some cases
cannot be taken to be straightforward. Hence the opacity of the grou-
ping.
The 'indicators' used in (5) - (13) are, as a general rule, confined to
 Reported speech in Swahili 105

reported speech of declarative sentences. With interrogative sentences a

different set of 'indicators' is used. Observe the following sentences:

(18) "unakwenda mjini?" Daudi aliuliza

You-prs-go town-LOC Daudi he-pst-ask
"Are you going to town?" Daudi asked.
i) Daudi alitaka kujua kama nilikuwa ninakwenda mjini
Daudi he-pst-want-to-know if I pst-be I-prs-go town-LOC
'Daudi wanted to know whether I was going to town.'
(19) "kwa nini Juma analia?" mwalimu aliuliza.
for what Juma he-prs-cry prof, he-pst-ask
'Why is Juma crying?" the professor asked.'
i) mwalimu aliuliza kwa nini Juma alikuwa analia
prof, he-pst-ask for what Juma he-pst-be he-prs-cry
'The professor wanted to know why Juma was crying.'
(20) "unataka nini?" baba akamwuliza mgeni.
you-prs-want what dad he-pst-OM-ask stranger
"What do you want?" dad asked the stranger.
i) baba alitaka kujua mgeni alikuwa anataka nini
dad he-pst-want to-know stranger he-pst-be he-prs-want what
'Dad wanted to know what the stranger wanted.'
(21) Amina akauliza, "mama amekwenda wapi?"
Amina she-pst-ask mom she-pfct-go where
Amina asked, "where has mom gone?"
i) Amina alitaka kujua mama yake alikuwa amekwenda wapi
Amina she-pst-want to-know mother hers she-pst-be she-pfct-go
where
'Amina wanted to know where her mom had gone'
(22) "unamtaka nani?" John aliniuliza.
you-prs-OM-want who John he-pst-me-ask
"Who do you want?" John asked me.
i) John alitaka kujua nilikuwa ninamtaka nani
John he-pst-want to-know I-pst-be I-prs-OM-want who
'John wanted to know who I wanted.'

Again the above examples are only a representative sample. We

observe from these sentences two types of interrogative sentences: the
YES-NO-questions represented by (18) and questions whose forms are
comparable to WH-questions in English.5 The latter are represented by
(19)-(22).
With reported speech of YES-NO-questions there seems to be only
one 'indicator' namely kama.6 The case of WH-questions, however, is
106 David P. Β. Massamba

rather different. Notice that while in reporting YES-NO-questions a new

item (i.e. kama) is introduced as an 'indicator' in reporting WH-
questions no new items, strictly speaking, are introduced as 'indicators'.
Instead it is the interrogative items (i.e. nani, nini, wapi kwa nani, etc.)
that serve as 'indicators'. There is, therefore, a clear difference in the
way these two types of interrogative utterances are reported. In view of
this we might say that 'indicators' of reported speech of YES-NO-
questions are explicitly expressed. Since there are no special 'indicators'
introduced in the case of WH-questions it seems reasonable to say that
'indicators' in constructions of this sort are not expressed in the surface
form. It will be noticed here that the mechanism for reporting YES-NO-
questions is, in a way, similar to that of reporting declarative sentences in
that in the latter reporting also makes use of deviant 'indicators' (cf.
(5)—(13)).
The above arguments for YES-NO-questions and WH-questions sug-
gest that in reporting interrogative utterances YES-NO-questions are
treated differently from WH-questions. That is, while for YES-NO-
questions the 'indicator' kama is used, for WH-questions 'indicators'
depend on the nature of the information presupposed in the question.
That is, for WH-questions if the question seeks information about reason
then the 'indicator' will be kwa nini; if it seeks information about place it
will be wapi/mahali ganilsehemu gani, etc.; if it seeks information about
manner it will be vipi/jinsi or namna gani; if it seeks information about
time it will be lini, etc.
With regard to verbs that may appear in the reporting clause it does
seem to be the case that in reported speech of interrogative sentences the
verbs used are uliza (ask) or taka kujua/kufahamu (want to know). 7
Recall that with declarative sentences the 'indicators' used were either
optional or obligatory depending on the nature of the verb in the
reporting clause (cf. (5)-(13)). As it turns out with reported speech of
YES-NO-questions, the 'indicator' kama is obligatory. This means an
omission of this form renders the sentence unacceptable. We exemplify
this below:

(23) Daudi alitaka kujua kama nilikuwa ninakwenda mjini

Daudi he-pst-want to know if I-pst-be I-prs-go town-LOC
'Daudi wanted to know if I was going to town.'

i) *Daudi alitaka kujua ( ) nilikuwa ninakwenda mjini.

 Reported speech in Swahili 107

(24) Amina alitaka kujua kama nilikuwa ninataka chakula

Amina she-pst-want to-know if I-pst-be I-prs-want food
'Amina wanted to know if I wanted some food.'

i) * Amina alitaka kujua ( ) nilikuwa ninataka chakula.

As for WH-questions, things are rather different. The interrogative

forms which serve as 'indicators' are optional. However, the nature of
their optionality is quite different from the one we saw in the case of
declarative utterances.
Even though the 'indicators' can be omitted here, there still has to be
some element that will make reference to the omitted form/'indicator'.
The element that usually refers to the omitted form/'indicator' is, in most
cases, a relative pronoun. That is to say a relative pronoun replaces the
omitted form/'indicator'. In order to have a clear understanding of this,
observe the following examples of the reported form of WH-questions,
In these examples sentence (i) represents cases of reported speech with
'indicators' and sentence (ii) represents cases of reported speech in
which the 'indicator' has been replaced by a relative pronoun:
(20'i) baba alitaka kujua mgeni alikuwa anataka nini
'Dad wanted to know what the stranger wanted.'
ii) baba alitaka kujua alic/iokuwa anataka mgeni
(21'i) Amina alitaka kujua mama yake alikuwa amekwenda wapi.
'Amina wanted to know where her mom had gone.'
ii) Amina alitaka kujua alifcokuwa amekwenda mama yake.
(22i) John alitaka kujua nilikuwa ninamtaka nani
'John wanted to know who I wanted.'
ii) John alitaka kujua niliyekuwa ninamtaka

We note from these sentences that the relative pronoun -cho- ((20') (ii))
refers to nini ((20') (i)); -ko- ((21') (ii)) refers to wapi ((21') (i)); and -ye-
((22') (ii)) refers to nani ((22') (i)).

3. Categories that undergo switches in indirect speech

It has been argued by Maw (1969) that Swahili reported speech has "no
other change than that of person . . . " . She claims further that declara-
tive, interrogative and imperative clauses retain their form and order.
108 David P. Β. Massamba

While we agree that in some ways Maw is correct we strongly feel that
the claim she makes is rather too bold and is, in many respects,
untenable. In order to come to grips with this fact we will examine
closely the nature of reported speech in each of the three sentence-types
she cites (i.e. declarative, interrogative and imperative).

3.1. Declarative sentences

A close examination of the form of reported speech given in (6) above

(i.e. Amina amesema ya kuwa atakwenda sokoni) reveals that its direct
form must be something like: Amina amesema, "nitakwenda sokoni".9.
Assuming that this is a correct reading of the direct form of the above
cited utterance, I think we can agree with Maw that in constructions of
this sort no other changes besides those of person take place. This fact is
also revealed in (9), (10) and (11) above. On the other hand if we
examine the reported form of speech given in (5) (i.e. Juma alisema
kuwa angekuja kesho yake) we can also say that its direct form must be:
Juma alisema, "nitakuja kesho". Here we see clearly that the direct and
the indirect forms do not tally with regard to their time references. In the
indirect form -nge- (past tense) is used while in the direct form -ta-
(future) is used. Furthermore, we also notice that in the direct form the
adverbial of time is kesho (tomorrow) while in the indirect form the
adverbial of time is kesho yake (the following day). Changes of the sort
described above are also evidenced by such examples as in (7) where the
supposed direct forms amefika (he/she has arrived) and leo (today) have
become, within the indirect form, alifika (he/she arrived) and siku hiyo
(that day) respectively; and in (8) where again the supposed direct forms
hatarudi (he/she won't come back) and leo (today) have become asinge-
rudi (he/she would not come back) and siku hiyo (that day) respectively.
Clearly these examples show that Maw's analysis is not without pro-
blems.
What seems to have misled Maw is her failure to realize the fact that in
Swahili, too, time reference plays an important role in changing forms of
direct speech to indirect speech. If the direct speech involves some kind
of action which took place, might have taken place, or was supposed to
take place some time in the past, then such changes as seen in (7) and (8)
above are bound to occur. The same applies to actions which during the
time of utterance of the original speaker were supposed to take place in
the future but are expected to have taken place at the time of the report.
 Reported speech in Swahili 109

For example, if the original speaker said: nitakuja kesho (I'll come
tomorrow) and then, two days later, if someone reported what he said,
he would say: alisema angekuja jana (he said he would come yesterday).
We note here that the time reference 'indicator -ta- (future) has been
replaced by -nge- (past) and the adverbial kesho by jana. On the other
hand if at the time of the original utterance an action was supposed to
take place in the future and if at the time of reporting the expected time
for the action to take place is yet to come then no changes other than that
of person will be effected (cf. (9), (10) and (11)).

3.2. Interrogative sentences

We have seen that interrogative utterances are divided into two groups:
the YES-NO-questions and the WH-questions. A close examination of
the YES-NO-questions (cf. (18) (i)) reveals that within the reported
clause (i.e. nilikuwa ninakwenda mjini, in the case (18) (i)) the actual
words of the original speaker are reproduced, with the exception of the
person marker: ninakwenda mjini. This seems to fall in line with Maw's
hypothesis. Notice, however, that within the reported clause there is new
information added -nilikuwa- which we may regard here as an auxiliary
verb. This new information ist crucial in two ways. First, since it is
something new, it makes both the information and the structure of the
reported clause of the indirect form rather different from the reported
clause of the direct form. Secondly, this new information tells us that the
words of the original speaker were actually uttered in the 'past'. In this
way the new information specifies both 'time' and 'tense'. Given this,
one could argue that even with YES-NO-questions we still find differen-
ces between direct and indirect speech. The two forms of speech are not
the same as Maw seems to suggest.
The arguments given above with regard to YES-NO-questions also
hold true of WH-questions. In other words in indirect speech of WH-
questions there is also new information added in the reported clause and
it also specifies 'time' and 'tense' (cf. (19)-(22)). Thus far the changes
that take place in the reported forms of YES-NO-questions are the same
as those that take place in the indirect forms of WH-questions. With the
latter, however, additional changes may take place. Recall that we
pointed out the fact that unexpressed 'indicators' in WH-questions can
be replaced by relative pronouns. Whenever this happens, additional
changes in the reported clause are bound to occur. Observe the following
sentences.
110 David P. Β. Massamba

(25i) "Juma amekwenda wapi?" baba aliuliza.

"Where has Juma gone?" dad asked.
ii) baba alitaka kujua Juma alikuwa amekwenda wapi.
'Dad wanted to know where Juma had gone.'
iii) baba alitaka kujua ali/tokuwa amekwenda Juma
(26i) "unataka nini?" Daudi aliniuliza.
"What do you want?" Daudi asked me.
ii) Daudi alitaka kujua nilikuwa ninataka nini
'Daudi wanted to know what I wanted.'
iii) Daudi alitaka kujua nilic/iokuwa ninataka

We note from the above data that the pronoun wapi in (25) (ii) has been
replaced by -ko- in (25) (iii); the pronoun nini (26) (ii) has been replaced
by -cho- in (26) (iii). These examples show additional cases in which the
reported clause of the direct speech differs from the reported clause of
the indirect speech, in the case of WH-questions.
Given this kind of evidence we can reasonably conclude that indirect
forms of interrogative constructions cannot be claimed to be indentical
to their direct counterparts; there are certain structural and/or grammati-
cal changes that take place when interrogative constructions are turned
into indirect speech. The examples also provided evidence that in Swa-
hili too the notion of 'back-shifting' (cf. Quirk et al. 1972: 786-788)
is realized, although not in exactly the same way as in English.9 In
Swahili the process of 'back-shifting' takes place in the following way:

Direct form of tense or time Indirect form of the tense or

reference time reference
-ta-
-nge-
li + kuwa -me-

Note that in Swahili 'back-shifting' depends pretty much on the exact

time reference of the original speaker's utterance. If the utterance
referred to some action in the future which has not yet come at the time
of reporting, or if it referred to an action in the recent past, then no
'back-shifting' will take place. On the other hand, if the original spea-
ker's utterance referred to an action which was expected to take place in
the distant future, or if the utterance itself was made a long time ago,
then there will be 'back-shifting'. But in such a case the 'back-shifting' is
not represented in the main verb; rather it is realized by inserting an
auxiliary verb, more specifically the auxiliary li + kuwa which will
 Reported speech in Swahili 111

necessarily precede the main verb. Finally, if the original speaker's

utterance referred to an action which was supposed to take place in the
future and if, at the time the reporting is made, that future is considered
to have passed, there will also be 'back-shifting' and -nge- will be used
instead.

3.3. Imperatives

The change from direct speech to indirect speech of imperative construc-

tions also has certain consequences. Observe what happens in the
following examples:
(27i) "soma kwa sauti", Bob alimwamuru Sue.
read with voice Bob he-pst-command Sue
"Read aloud, Bob commanded Sue.
ii) Bob alimwamuru Sue asome kwa sauti
Bob he-pst-OM-command Sue she-read with voice
'Bob commanded Sue to read aloud.'
(28i) "ondoeni takataka hii".
take away trash this
"Take this trash away."
ii) aliwaamuru waondoe takataka ile
lie/she-pst-OM-command they-take away trash that
He ordered them to take that trash away.'

A thorough examination of the facts given in (27) and (28) above

reveals that certain important changes are realized when imperative
forms in the direct form of speech are reported in the indirect form. First
we note that in indirect speech the imperative mood necessarily changes
into the subjunctive mood. This is realized by the fact that the final
vowel in the verb cluster changes from -a to -e, e.g. soma —» -some;
ondoa —* -ondoe, etc. We know that this is the case because in Swahili
the subjunctive form is realized by the change of final vowel -a into -e10.
The second change concerns the person markers. In direct speech the
2nd person singular is marked by the 0-morpheme whereas in the
indirect form it is marked by the prefix a-; the 3rd person plural is
marked, in direct speech, by the suffix -ni while in indirect speech it is
marked by the prefix wa-.
The indirect form of the imperative also requires that demonstrative
pronouns which express the idea of 'this' and 'that near' change to 'that
far', e.g.
112 David P. Β. Massamba

hii
(this (class 9))
hiyo
(that (class 9) near) - » -le
huyu (that, distant)
(this (animate))
huyo
(that (animate, near))

3.4 Pronoun transposition

In sections 3.1.-3.3. our discussion has mainly been of changes that more
or less pertain to particular or specific sentence-types. There are, how-
ever, certain changes that occur regardless of whether or not the
sentences or utterances made were declarative, interrogative, or impera-
tive. What seems to be crucial in changes of this sort are the forms of
speech themselves, i.e., from direct to indirect speech and vice versa.
Such changes are seen in the behavior of pronouns ('self-standing' as
well as possessive pronouns).

Direct speech English gloss Indirect speech

(a) mimi Τ yeye 'he'

(b) wewe 'you'sg. mimi Τ
(c) yeye 'he' ί yeye 'he'
\ wewe 'you'sg.

(d) sisi 'we' J wao 'they'

1 mimi na yeye 'he and I'

/ f mimi na Χ 'X and I'

(e) nyinyi 'you pi. j ^

(f) wao 'they' ί wao 'they'

\ nyinyi 'you'pl.

(g) -angu 'mine/my' -ake 'his/her(s)'

(h) -ako 'your(s)'sg. -angu 'mine/my'
l
(i) -ake 'his/her(s)' J -ake his/her(s)'
\ -ako 'your(s)'sg.

(j) -etu 'our(s)' ί -etu 'our/ours'

\ -ao 'their(s)'
(k) -enu 'your(s)'pl. -etu 'our(s)'
(1) -ao 'their(s)' ί -ao 'their(s)'
\ -enu 'your(s)'pl.
 Reported speech in Swahili 113

In indirect speech the self-standing and possessive pronouns are

readjusted in such a way that they can occur with the appropriate nouns
(subjects and/or objects) to which reference is made. The following are
the changes that apply to these pronouns:
Since the plural counterparts for the direct speech in the indirect speech
can be a little confusing an elaboration of the items is in order. In item
(e) mimi na χ (where χ is a variable for a noun) is used when neither the
hearer nor the original speaker is included; sisi is used when reference is
made to the reporter and the hearer. In item (f) nyinyi is used when the
hearer is included while the reporter is excluded; wao is used when the
speaker, the reporter and the hearer are excluded. In item (j) etu is used
when the speaker and reporter are included but hearer excluded; -ao is
used when the reporter and hearer are excluded. In item (k) when -etu is
used the reporter is necessarily included while the hearer may or may not
be included. Item (1) -ao is used when speaker, reporter and hearer are
excluded; while -enu is used when speaker and reporter are excluded but
the hearer is included.

3.5. Summary of categories that undergo switches

In view of the facts given in 3.1-3.4 above a number of conclusions can

be drawn with regard to the categories that either undergo or do not
undergo switches in indrect speech. Let us begin with elements within
the verb cluster. Within this group the root of the verb (e.g. som- 'read'/
'study') always remains unchanged. The final vowel, however, may
change as we saw in the case of indirect forms of imperatives. In the case
of imperative constructions we saw (cf. 3.3 (27) (ii)) that indirect forms
of utterance of this sort necessarily change the mood from imperative to
subjunctive. Thus changing the final vowel from -a to -e (e.g. soma —
- some).
The person marker prefixes appearing in the verb cluster may also
change. Switches in this category are effected in the following way:

a- 'he' (unconditionally)
ni- Τ (unconditionally)
'he' (if reference is not made) to the hearer
ί "
l u- 'you'sg. (if reference is made to the hearer)
f tu- 'we' (if speaker and reporter are included).
114 David P. Β. Massamba

tu- 'we' wa- 'they (if reporter and hearer are excluded)
(if reporter is excluded but hearer in-
m- 'you'pl. eluded)
m- 'you'pl. tu- 'we' (reporter included hearer may or may
not be included)
wa- 'they' wa- 'they' (reporter and hearer excluded)
m- 'you'pl. (reporter excluded but hearer inclu-
ded)

With regard to the so-called 'tenses' 11 switching of the tense markers is

determined by the temporal situation of the action referred to in the
verb. If the action referred to in the verb took place in the past then the
tense marker -li- will be used both in the direct and indirect forms. In
other words, there won't be any switch since -li- already refers to the
past. The -na- 'tense' (or time reference?) also never changes. However,
if the time at which the action was supposed to take place has already
passed, an auxiliary verb indicating 'past' has to be inserted (i.e. -/ikuwa
-na . . . ) ; on the other hand, if the time at which the action is supposed to
take place is yet to come, an auxiliary verb indicating 'future' has to be
inserted (i.e. -takuwa -na-...).
The -ta- tense may be 'backshifted to -nge- if the intended future has
passed; otherwise there will be no switch.
The perfective aspect -me- may also remain unchanged. Whenever
this is the case if the time reference is to a specific time in the future, the
future auxiliary verb -takuwa will have to be inserted while in cases of
distant past the auxiliary verb -likuwa will be used. Observe the follo-
wing examples.
(29i) Nimesema Juma amekwenda mjini
I-pfct-say Juma he-pfct-go town-LOC
Ί said Juma has gone to town.'
ii) Atakapofika tutakuwa tumemaliza kazi
he-fut-time-arrive we-fut-be we-pfct-finish work
'When he arrives we'll have finished (our) work.'
iii) Alipofika tulikuwa tumemaliza kazi
he-pst-time-arrive we-pst-be we-pfct-finish work
'When he arrived we had finished (our) work.'

On the other hand the perfective -me- (sometimes called the -me- tense)
switches to -li- when the perfectiveness of the action/verb is considered
part of the past (cf. (7) and (8)).
Other categories that undergo switches are the so-called 'self-stan-
ding' pronouns (i.e. mimi T , wewe 'you'sg., yeye 'he/she' sisi 'we', nyinyi
 Reported speech in Swahili 115

'you'pl., and wao 'they'), the within vicinity demonstrative pronouns

(i.e. 'this' and 'that near') and the possessive pronouns. Since these have
been discussed at length in sections 3.3. und 3.4. above we will not
reiterate on them.
Adverbials of time will undergo switches if the kind of information
given in the verb of the direct form reflects an action which is supposed
to have taken place before the time of reporting, i.e. it is already in the
past. More specifically the adverbial leo (today) will be backshifted to
jana (yesterday) if reporting is made a day after the original utterance; it
will be backshifted to juzi (the day before yesterday) if reporting is made
two days after the day of the original utterance; and it will switch to siku
hiyo (that day) if reporting is made a number of days after the day of the
original utterance. The adverbial jana will be backshifted to juzi if the
reporting is made a day after the day of the original utterance (someti-
mes a few days after the original utterance); and if quite a number of
days have passed after the original utterance it will remain jana but with
an addition of the possessive yake ('its,' which refers to the day when the
original utterance was made), i.e. jana yake. The adverbial juzi also does
not change but it is followed by the possessive yake. The adverbial kesho
(tomorrow) will be backshifted to leo if reporting is made a day after the
day of the original utterance; it will be backshifted to jana if reporting is
made two days after the original utterance; it will be backshifted to juzi if
reporting is made three days after the original utterance; and if quite a
number of days have passed, kesho will be maintained plus the posses-
sive yake, i.e. kesho yake. The adverbial kesho kutwa (the day after
tomorrow) is backshifted to leo if reporting is made two days after the
day of the original utterance; to jana if reporting is made three days after
the original utterance; and juzi if reporting is made four days after the
original utterance.

4. The unequivocal nature of reported speech in Swahili

In some languages it is sometimes difficult to tell, out of context,

whether or not a certain construction is a direct or an indirect report.
Observe the following example from English.
(30) Nick said, I ought to study hard.
In the above sentence unless we are provided with some contextual
information it is impossible to say whether I refers to the speaker or is
116 David P. Β. Massamba

coreferential with Nick. That is to say the sentence is ambigous. Since

this sentence is ambiguous it could be given either a direct or indirect
interpretation.
In Swahili (and this is true of most, if not all Bantu languages)
ambiguity of the sort discussed above does not arise. In order to come to
grips with this fact let us examine the following Swahili sentences.
(31) Juma alisema nitakwenda mjini
Juma he-pst-say I-fut-to-go town-LOC
'Juma said I will go to town.'

(32) Amina alisema nitakaa nyumbani

Amina she-pst-say I-fut-stay house-LOC
'Amina said I will stay (at) home.'

If a native speaker of Swahili is confronted with the above sentences (i.e.

(31) and (32)), even without any contextual information, he will defin-
itely say that the ni- (I) in the embedded clauses refers to the speaker of
the sentence and not to Juma/Amina. In other words to him there is no
ambiguity involved. Note, however, that although these two Swahili
constructions are not ambiguous their English equivalents are. The
question we must ask ourselves is: what is it that blocks ambiquity in the
Swahili constructions but not in their English counterparts?
A close examination of the Swahili constructions in (31) and (32)
reveals that in both cases the concordial agreement in the subject prefix
slot in the verb cluster of the reporting clause differs from the concordial
agreement in the subject prefix slot in the verb cluster of the reported
clause. More specifically, the subject prefix in the reporting clause is a-
(he/she) while in the reported clause it is ni- (I). As it turns out, it is this
discrepancy between the two verb clusters (i.e. in the reporting clause
and the reported clause) that triggers conditions which block any ambi-
guity.
In Swahili we can easily tell that I (ni-) in our two sentences above (cf.
(31) and (32)) refers to the speaker of the sentence and not to Juma/
Amina. This is so because if Τ (ni-) were to be coreferential with the
subject of the matrix sentence (i.e. Juma/Amina) then the concordial
agreement in the subject prefix slot would have to be identical in the
verb cluster of both the reporting clause and the reported clause. In
other words a situation of the sort shown below would arise:
(33) Juma alisema atakwenda mjini
Juma he-pst-say he-fut-to-go town LOC
'Juma said he will go to town.'
 Reported speech in Swahili 117

(34) Amina alisema atakaa nyumbani

Amina she-pst-say she-put-stay house-LOC
'Amina said she will stay home.'

At this point it will have become clear that in Swahili concordial

agreement plays quite an important role both in the direct and indirect
report in that it blocks ambiguity.
In English the situation is quite different. Since English is an isolating
language there are no such things as subject prefixes which would signal
information relating to the subject. Note that the kind of ambiguity
involved in (30) is triggered by the fact that there is nothing in the
sentence (or verb for that matter) that signals any information about the
pronoun T . Insomuch as there is nothing in the sentence that confines
the pronoun Τ to a specific subject its interpretation is bound to be
ambiguous; hence the ambiguity in the whole reported clause.
To recapitulate, we may conclude by saying that while in English the
interpretation of Τ in a sentence such as Nick said I should read this
paper depends entirely on contextual information (in the absence of
which the sentence becomes ambiguous) in Swahili the interpretation of
Τ does not require contextual information because concordial agree-
ment serves as a disambiguating device.
Given this state of affairs we can in fact say that in Swahili subject
prefixes, within verb clusters, also do serve as indirect speech 'indica-
tors'. These are, however, different from the 'indicators' we saw earlier
(cf. 5-11) in that, while 'indicators' like kuwa, ya kuwa, kama, eti, etc.
('that') function as introducers of reported clauses, subject prefixes
function as devices for specifying the subjects of the embedded clauses,
thereby telling us whether or not the words in the embedded clause were
uttered by the original speaker or the reporter. In this way we can say
that subject prefixes are indirect 'indicators' of reported speech.
In the above discussion we have seen that in Swahili constructions
given the different devices that are used to determine reported and direct
speech one can always tell whether an utterance is a direct or an indirect
report. In Swahili therefore it is always unequivocally clear whether an
utterance is a direct or indirect report.
118 David P. Β. Massamba

5. Conclusion

In this paper we have surveyed the different mechanisms used in Swahili

indirect speech. We have seen that indirect speech of declarative utte-
rances is introduced by 'marked' indicators. Indirect speech of interroga-
tive utterances is either introduced by an explicit 'indicator' kama (in the
case of YES-NO-questions) or is indicated by an implicit 'indicator' (in
the case of WH-questions); the implicit 'indicators' being the interroga-
tive pronouns. As for the indirect form of imperative utterances, we
noted that the mood changes from imperative to subjunctive (i.e. final -a
> -e); the 0-morpheme slot for 2nd pers. sg. is replaced by a-; (0-soma >
asome) and the 3rd person pi. suffix -ni is replaced by the 0-morpheme
while at the same time the 0-morpheme in the initial position of the verb
cluster is replaces by wa- (i.e. ondoeni > waondoe).
We have also seen that most pronouns are transposed, as it were.
These are discussed in sections 3.3 and 3.4. There is also a wide range of
categories that undergo switches. However, such switches depend on
two main things. First, they depend on the nature of utterances (i.e.
whether they are declarative, interrogative or imperative). Secondly,
they also depend on the particular 'tenses' and/or 'time references'
within which the utterances were made and during which they are being
reported. Our discussion has also demonstrated that in Swahili indirect
speech switches are not confined to the category of 'person' as some
scholars, including Maw, have suggested.
Another important aspect that has revealed itself in our discussion (cf.
4.) has to do with the unequivocal nature of reported speech. We have
seen that although in languages such as English sometimes it is impossi-
ble to tell, out of context, whether a piece of reported speech is direct or
indirect, in Swahili we can always tell, on the basis of grammtical
devices, whether a given piece of reported speech is direct or indirect.
Finally, although we did not attempt a comparative study, there seems
to be a lot in common between indirect speech in Swahili and English,
superficially at least. In this way one might venture to say that the notion
of indirect speech has indications of being a universal phenomenon.
 Reported speech in Swahili 119

Notes

1. Here we are disregarding the fact that a reporter could very well report his/her own
words.
NB: pst = past tense; fut = future; pfct = perfective; LOC = locative marker; PERS =
person; OM = Object Marker; NEG = Negative particle
2. Maw's analysis is given in her book Sentences in Swahili: a study of their internal
relationship (1969: 20-24).
3. Some speakers suggest that kwamba eti and or eti kwamba can also be used as
'indicators'. Since their acceptability is rather remote we have decided to leave them
out.
4. We are expanding the notion 'reported clause' to include the words reporting what the
original speaker said; and the notion 'reporting clause' to include every information
that precedes the reported clause.
5. We are aware of the fact that these sentential constructions are not identical per se with
English WH-questions. However, since the differences do not affect our analysis we
will use the term "WH-question" to refer to the Swahili counterparts.
6. There are a few speakers who tend to think that iwapo and ikiwa could also be used as
'indicators'. However, there is no convincing argument in support of this.
7. The two VPs (i.e. uliza and taka kujualkufahamu) are used interchangeably although
many speakers prefer the latter to the former. This is more of a stylistic question than
of grammaticalness and/or acceptability.
8. Here we are ignoring stylistic facts. That is, whether the reporting clause should come
before or after the reported clause is irrelevant at this point.
9. For a detailed analysis of 'back-shifting' in English refer to Quirk et al. 1972: 786-787.
10. We are aware that the terminology 'subjunctive' may not be the most appropriate
terminology for Swahili. We do feel, however, that whatever terminology is sought it
will not affect the facts here.
11. The question of 'tenses' has not been given proper treatment in Swahili. Whether there
are three (i.e. -Ii-, -na-, -ta-) or two (i.e. -Ii-, -ta-) major tenses in Swahili is yet to be
resolved. For the question at hand is this: Since -na- can be used both in the past and
the future, it is not clear whether it indicates 'tense' or 'time reference.'

References

Ashton, E. O. 1944. Swahili Grammar. London: Longman.

Goodwin, W. W. 1930. Greek Grammar. London.
Kabugi, C. 1974. Tujifunze Kiswahili. McGraw Hill Far Eastern Publishers.
Massamba, D. P. B. 1969. Kiswahili Fasaha. Swahili (Journal of the Institute of Swahili
Research, Dar Es Salaam) 39, 1&2.
Maw. J. 1969. Sentences in Swahili. A Study of their Internal Relationship. SO AS London.
Quirk, Randolph et al. 1972. A Grammar of Contemporary English. London: Longman.
Speech reporting in the Caucasus
B. G. Hewitt and S. R. Crisp*

Given the linguistic complexity of the Caucasus (some 38 autochtho-

nous languages divided into the five families of South, North, West,
North Central and North East Caucasian), no single article can hope to
include a comprehensive survey of all the relevant data. In what follows,
therefore, we shall concentrate on only four languages in order to
present a general idea of the techniques employed within each of the
above-mentioned families: Georgian is described for South Caucasian,
Abkhaz for North West, Avar for North East, and, for comparative
purposes between North East and North Central, reference is made to
Chechen from the North Central group. It should not, of course, be
concluded that each of the relevant sister-languages necessarily concurs
in every detail - mutatis mutandis - with the facts outlined below for the
four representative languages of this study.

1. South Caucasian (Kartvelian)

The most usual way of reporting speech in Georgian is to employ the

direct method. And three speech-particles exist in order both to mark
the direct status of the relevant word(s) and to present some (more or
less exact) information as to the person of the speaker. Where the
speaker is repeating his own words (or thoughts), he will add the enclitic
/-metki/ to the final word of the quote, which is usually the verb, e.g.
(1) net'a, 3iav, mec Senistana monadire viq'o-metki, vutxari
would-that uncle I-too like-you a-hunter I-might-be I-said-it-
to-him
bija-öems
my-uncle
Ί told my uncle that I wanted to be a hunter of his calibre//I said to
my uncle: " M a y I be a hunter of your calibre, uncle!'"
(Vaza-Psavela - Academy Dictionary)

* Sections A and Β by B. G. Hewitt, section C by S. R. Crisp.

122 Β. G. Hewitt and S. R. Crisp

The etymology of this particle is quite transparent, being a coalescence

of /me vtkvi/ Ί I-said-it'.
The second particle is /-tko/ (less commonly /-tkva/), which is similarly
attachted to the final word of a quote when the speaker is instructing the
addressee as to the words he, the addressee, is himself to repeat (usually
before some third party), e.g.
(2) tkvit, tu met'i gza ar aris, unda vimuäaot,
say-it if further way not it-is it-is-necessary-that we-work
opli vic'urot-tko
sweat we-squeeze-it-out
'Say: "If there is no other way, we must work and squeeze out the
(old) sweat'" (Barnovi - Materials for the Academy Dictionary)

Two etymologies have been proposed for this particle, and it has been
argued by Hewitt (1981 or, preferably, 1984) that the more convincing is
that of Vogt (1971: 217), who explains it as deriving from /tkva/, the 2nd
person singular Aorist subjunctive of /tkma/ 'to say' used as mild
imperative 'you should say (it)'. Note that the particle may be employed
where there is no explicit verbum dicendi,

(3) exlave ak moiq'vane, puls gaöukeb-tko

right-now here fetch-him the-money I-shall-present-it-to-you
'Fetch him here right now (telling him that) I shall present him with
the money' (C'avc'avadze - Academy Dictionary)

This second particle is regularly found in place of /-metki/ in some non-

standard dialects, particularly in western Georgia - it is also available as
an alternative to the third particle when the introductory verb has a 1st
person plural subject. Assuming the etymology given above to be
correct, we can explain this extension of usage quite easily - having
originally been appropriate exclusively in contexts where the 1st person
was instructing the 2nd person in what he was to say, the particle's
conditioning factors have been simplified by the omission of any refe-
rence to advice being offered to a 2nd person - , the trigger for the
appearance of /-tko/ is now in certain dialects merely a suitable introduc-
tory verb with 1st person subject.
The third particle ist /-ο/, of opaque etymology, which occurs in all
contexts other than those requiring either /-metki/ or /-kto/. In the
written language it behaves like the other particles in attaching to the last
word of the quote, but in actual speech it regularly follows many, if not
all, phrasal constituents as well as accompanying the final word of each
clause (or its verb), e.g.
 Speech reporting in the Caucasus 123

(4) utxra: Cerni SiSita-o cay prinvels ver

he-said-it-to-him of-me through-fear the-sky bird not
gauvlia da 3irs ö'ian£'velasa-o
it-has-been-able-to-cross-it and on-the-ground ant
da sena-o ra surieli xar-o, ro ak gadmaagde-o?
and you how foolish you-are in-that here you-threw-it
'(The dev) said to him: "Through fear of me no bird has been able to
pass across the sky and on the ground no ant, and, you, how foolish
are you to have thrown it here?'" (Gigineisvili et al. 1961: 198)

(5) vtkvit, c'igns vk'itxulobt- j

we-said-it the-book we-are-reading-it

'We said that we were reading the book' (S.A.) 1

In his Old Georgian grammar Sanidze (1976) alludes to neither of the

first two speech-particles. Similarly, in his grammatical sketch of the
oldest dated Georgian manuscript, the Mravaltavi of 864 A.D., ImnaiS-
vili quotes examples of /-of alone among the three particles, noting that it
is surprisingly frequent in this work (1975: 99-100); he also observes in
passing that in the whole of the O.G. Gospels only five instances of /-o/
occur with direct quotations - it should be noted that the speech-particles
are often absent in formal written styles within modern Georgian also.
Indirect quotation is also perfectly feasible. The complementiser /rom/
'that,' which also has wide range of other subordinating functions in the
language, introduces the quotation, whose verb is adapted to conform to
the requirements of the introductory verb, so that the indirect equivalent
of (5) will be:

(5') vtkvit, rom c'igns vk'itxulobdit (S.A.)

that we-were-reading-it

Thus a Present tense of oratio recta changes to the Imperfect in oratio

obliqua; similarly a Future becomes Conditional, whilst the Aorist (and
Perfect too) may either remain unaltered or be transformed into the Plu-
perfect. An Aorist subjunctive dependent on the modal /unda/ 'it is
necessary that' is likewise transformed into the Plu-perfect indicative
(+ /unda/).
Indirect commands, when represented by finite verb-forms rather
than by the infinitive (usually referred to by the term 'masdar' in the
relevant literature), behave as follows: the direct imperative (or Aorist
subjunctive in the case of 3rd person commands) becomes Aorist sub-
junctive after a non-past introductory verb, and either Aorist subjunc-
124 Β. G. Hewitt and S. R. Crisp

tive or Plu-perfect indicative after a past introductory verb - the comple-

m e n t e r /rom/ is optional, e.g.
(6) brjana, (rom) kuöebsi mxiaruloba Sec'q'des//
he-ordered-it in-the-streets merry-making it-should-stop//
semc'q'dariq'o
it-had-stopped
'He ordered the merry-making in the streets to stop' (E.C.)
(7) bibliotek'ars ubr3anes, (rom) k'art'otek'a da§alos//
librarian they-ordered-him file he-destroy-it
daeSala// k'art'otek'is dasla
he-had-destroyed-it// of-the-file the-destruction
'They told the librarian to destroy the file' (E.C.)

This same fluctuation between Aorist subjunctive and Plu-perfect indica-

tive is also attested in purpose-clauses in historic sequence and does not
necessarily correlate, as might be expected, with the pragmatic distinc-
tion between an order whose accomplishment is still awaited at the time
of speaking (Aorist subjunctive) and one already fulfilled (Plu-perfect)
(cf. Hewitt (1981a: Iff.) for a dicussion of the problem in purpose-
clauses).
Indirect questions are treated in one of two ways according to whether
the question is of the Yes/No- or WH-type; in either case there will, of
course, occur all appropriate changes for person and tense. To turn a
Yes/No-question into its indirect counterpart the words /tu ara/ 'or not'
are placed after the queried verb, e.g.
(8') gvek'itxeba, davesc'rebit tu ara mis korc'ils
he-asks-us shall-we-attend-it his wedding
'He is asking us whether (or not) // if we shall attend his wedding'
(E.C.)
which corresponds to either form of the direct questions:
(8') cems korc'ils daesc'rebit (tu ara)?
my wedding will-you-attend-it
'Will you attend my wedding (or not)?'

In WH-questions the interrogative word (or phrase containing it) will,

as in the direct form, continue to stand immediately before the verb (or
negative, which takes precedence in occupying the immediately prever-
bal slot); the clause may optionally begin with the particle /tu/, whose
regular role is to mark vivid (otherwise known as real or open) protases,
e.g.
(9) titoeuli sit'q'va ganmart'ebulia imis mixedvit, tu igi mititebul
each word it-is-explained that according-to it referred-to
 Speech reporting in the Caucasus 125

k'ont'ekst'Si ra mnisvnelobisaa
context-in what of-meaning-it-is
'Each word is explained according to what meaning it has in the
context referred to' (LolaSvili 1964: 6)
(10) öven ar vicit, iak'obma . . . 3eglSi sxva ra
we not we-know-it Jacob monument-in other what
Sesc'oreba seit'ana
correction(s) he-introduced-them
'We do not know what other corrections Jacob introduced into the
(literary) monument' (ibid: 37)

Combinations of complementiser and purely direct quotation (for

which parallels exist outside Kartvelian, e.g. in A n c i e n t G r e e k ) were
quite c o m m o n in O l d G e o r g i a n , where the form of the complementiser
was different f r o m modern /rom/ (cf. Hewitt Forthcoming), and is
accepted even today; although my first informant, S . A . , was at times
reluctant to accept such examples, another informant, M . D z . , stated the
construction to be very widespread, e.g.

(11) vutxari, rom es q'velaperi simonetidanaa

I-told-him//them this everything from-Simoneti-it-is
öamot'anili-tko
brought
Ί told him//them that all that was brought from Simoneti' (K'ldiaSvili
- Academy Dictionary, where /-tko/ is used for standard /-metki/)
(12) biö'ebäi k'i xma gaagdet, rom t'q'uilia-tko
among-the-lads but voice throw-it it-is-a-lie
'But raise your voice among the lads (saying that) it is a lie' (C'avö'a-
vadze - Academy Dictionary)
(13) redakciam . . . gamomicxada, rom targmani
the-editorial-board it-announced-it-to-me the-translation
ar daibeö'deba-o
not it-will-be-printed
'The editorial board announced to me that the translation would not
be printed' (K'ldiaävili 1981. II: 321)

Standing, as it w e r e , half-way between fully direct and fully indirect

quotation is what might be styled semi-indirect quotation, which differs
from the fully indirect mode by manifesting only change of person in
pronominal reference whilst preserving the original tense of the words as
spoken or of the thoughts as conceived, e.g.

(14) Joni amt'k'icebda, rom dedamic'a brt'q'elia

John he-was-claiming-it the-earth it-is-flat
'(?)John claimed that the earth is flat' (E. t.)
126 Β. G. Hewitt and S. R. Crisp

This example shows that this mode is not restricted to universal truths.
And other examples clearly confirm that retention of the direct tense-
forms is not dependent upon any requirement that those tenses should
still be appropriate at the time of reporting. To illustrate this let us take
Luke 24.23, which in the English Authorised Version reads: " . . . they
came, saying, that they had also seen a vision of angels, which said that
he was alive". In English one could hardly replace the three italicised
verbs with their direct equivalents 'have', l (?)say', 'is' respectively. And
yet what we find in a Modern Georgian edition of the New Testament is:

(15) ... tkves, rom ixiles moöveneba angelozta,

they-said they-saw//have-seen-it vision of-angels
romelnic amboben, rom is cocxalia
who they-say-PRES(ENT) he alive-he-is-PRES

Since the Georgian Aorist may correspond to an English Simple Past or

Perfect (or even Plu-perfect), the presence of /ixiles/ is not surprising,
but the same cannot be said of /amboben/ (sc. from the point of view of
an English-speaking observer). And interestingly all Old Georgian
manuscripts have here a past tense form, being Aorist in mss. DEFG-
HIK, and Imperfect in ms. C. All mss. show the Present for the final
copula, but, since it is an article of the Christian doctrine that Christ lives
still following the Resurrection, perhaps this cannot be convincingly
argued not to be a Present of universal truth 2 . Similar retention of the
direct tense but with change of pronouns is feasible for the reporting of
questions, e.g.
(16) masc'avlebelma . . . mk'itxa, meint'ereseba tu ara
the-teacher he-asked-me does-it-interest-me-PRES or not
kartuli ist'oria
Georgian history
"The teacher asked me whether I was interested in Georgian history'
(K'ldiasvili 1981. II: 92)
Thus far illustration has been offered of both direct and indirect
reporting, of co-alescence of the two (where the direct mode is intro-
duced by the typically indirect complementiser) and also of so-called
semi-indirect quotation. Finally the question arises as to what rules
might govern the employment of deictic elements when speech is being
reported in the language. To test the range of possibilities here in-
formant M. Dz. was asked to comment on the acceptability of a number
of utterances where the reference of such temporal adverbials as /xval/
'tomorrow', /dyes/ 'today' and /meore dyes/ 'on the next day' was crucial.
 Speech reporting in the Caucasus 127

Analysis of the results leads to the conclusion that, regardless of the

construction employed and as long as it is clear from the context of
conversation what day is under discussion, there seems to be a free
choice between that temporal adverbial which would have been appro-
priate at the time of direct discourse and that which is appropriate at the
moment of reporting - in other words, no grammatical restrictions apply.
However, the distinct possibility remains, of course, that, in the event of
the context not making the temporal reference quite explicit, this
freedom of choice may disappear. Since all the examples presented to
my informant were of the explicit context type, nothing further can be
said about this possibility. Firstly, then, let us exemplify a case of /xval/
appearing where it is appropriate only at the time of reporting:

(17) vtkvi, rom xval movidodnen, magram uk've ramdenime

I-said-it they-wow/i/-come but already several

dyea, rac eg gitxari, da mxolod xval gavigebt,

day(s)- since thatl-told-it-to-you and only we-shall-
it-is learn-it

martla movlen tu ara

indeed will-they-come or not
Ί said that they would come tomorrow, but it is already several days
since I told you, and only tomorrow will we learn whether or not they
will in fact come'

where the quote may equally acceptably take any of the permuations
seen below (Ν. B. that at least one of the bracketed elements will
normally be obligatory):

(17') . . . (rom) xval movlen(-metki) . . .

they-vw7/-come

On the other hand, /xval/ may be retained if either /xval/ itself or the
equivalent date formed part of the original words, e.g.

(18) p'irvel ianvars öveni saubrisas

on-the-first of-January our conversation-at-the-time-of

gitxari,
I-told-it-to-you

a) rom xval movidodnen,

they-wou/d-come
128 Β. G. Hewitt and S. R. Crisp

b) (rom) xval
movlen(-metki), da xom movidnen or ianvars
they-wi//-come and surely they-came on-the-second of-January
'During our conversation on the 1st January I told you that they
would come the next day, and they came, didn't they, on the 2nd?'
O n e final example shows the same freedom of choice when direct /xval/
is equivalent to /dyes/ at the time of reporting:
(19) guäin xom gitxari,
yesterday surely I-told-it-to-you
a) rom xval//dyes movidodnen,
they-wowW-come
b) (rom) xval//dyes movlen(-metki), da ai modian!
they-w///-come and look! they-are-coming
Ί told you, didn't I, yesterday, that they would come today, and -
look! - here they come?!'

It may, thus, be concluded not only that Georgian has a variety of

citation-methods at its disposal but also that it enjoys great freedom in its
selection of deictic elements within such quotations. O n e thing not
attested in Georgian is presence of a speech-particle in conjunction with
pronouns that have undergone a shift from their direct forms to bring
them into conformity with the requirements of the introductory verb.
We may close section A by referring to just such an anomaly in the most
archaic of Georgian's three sister-languages, Svan.
Svan seems not to have developed the fully indirect mode (Hewitt
1982: 208; though cf. Hewitt 1984a sentence [34] (14)), and the semi-
indirect method may be of recent development (Hewitt: 1982: 209). But
the most widespread device in the language combines the typically semi-
indirect phenomenon of change of person alone with the presence of a
speech-particle. Svan's equivalent to Georgian /-o/ is either /eser/ or
/}}ok(w)/, and the difference between the two languages can be seen by a
comparison of respectively the original Svan and its Georgian translation
in this LaSx dialect example (cf. Hewitt 1985, 1983):
(20a) al lavdilas lexö'oddax: may lok xäymedax
these sisters they-asked-them what? is-it-possible-for-i/iem
mins, xoxalx lok esama?
them do-i/rey-know-it anything
b) am debs hk'itxes: ra Segi3liat-o tkven,
these sisters they-asked-them what? is-it-possible-for-you you
icit-o rame?
do-yow-know-it anything
 Speech reporting in the Caucasus 129

'They asked these sisters whether they could do anything and wheth-
er they knew anything'

Clearly, if the original function of the Kartvelian speech-particles was to

mark an utterance as recapitulating exactly (sc. as far as pronominal
reference and verb-tense are concerned) the direct form of the relevant
words or thoughts, as is likely, then the Svan development represents a
considerable anomaly from a historical perspective.

2. North - West Caucasian

Abkhaz displays perhaps an even greater preference for the direct

reporting of speech than does Georgian. Unlike Georgian it possesses
just the one particle to indicate the direct status of a quote, namely /h°a/,
which historically is an archaic form of the past3 absolutive (roughly
equivalent to the English Perfect participle) of /a-h°a-ra/ 'to say', the
regular modern form of which would be /h°a-n9/, which latter is not
however employed in the role of speech-particle in the language. Quotes
usually precede their introductory verbs. When this ordering occurs,
/h°a/, which appears once at the end of the quote, will be omitted if the
introductory verb contains only one pre-radical affix (i.e. the personal
prefix correlating with the verb's subject), but where there is more than
this single pre-radical element, there seems to be a preference for /h°a/ to
be retained. Should the quote follow its introductory verb, then /h°aI will
be obligatory. E.g.

(21) S°- an+ b[a-] äa- wa- y

you when QU(ESTION) come DYN(AMIC).PRESS QU h°a4
d- ra+ z- c'aa-yt'
he them for ask FIN(ITE).AOR(IST)
'He asked them when they would come' (Sank°ba 1962: 238)
(22) ya- za+y d- r- a+ z-
that-which you happen-to (AOR) what-is-it he them it about
c'aa-yt'5
ask FIN.AOR.
'He asked them what had happened to them' (ibid.)
(23) s-ga+la- wa- m (0-)s-h°a-yt' // ya-s-h0a-yt' s-ga+la-wa-m h°a
I get-up DYN.PRES not it I say FIN.AOR it
Ί said that I would not get up' (Ζ. K.)
130 Β. G. Hewitt and S. R. Crisp

(24) y 9 -h°a ds-psa-yt' h°a

it say! he die FIN.AOR
'Say that I died' (Las°ria 1978: 39)

(in this last example the general context makes it clear that reference is
to the speaker). As in Georgian, in formal written Abkhaz the speech-
particle may simply be omitted6.
Indirect reporting is possible but restricted, because the complementi-
ser-particle /-§(a)-/ basically means 'how', and, since in indirect quota-
tion it is not always possible to dismiss the manner-interpretation, the
direct mode is preferred for its lack of ambiguity. Person-changes are as
normal, and tense-changes after a past introductory verb are: Present to
Imperfect, Future I to Conditional I, Future II to Conditional II, Perfect
to Plu-perfect - although a direct Aorist may be represented by a Past
Indefinite form, this ist not a tense-change (see Hewitt 1979: 7); because
of general grammatical requirements /-§(a)-/ may only be inserted into
non-finite verb-forms, e.g.
(25) ya-§9- q'a- y- c'a- wa- ζ
it PREV(ERB)he do DYN NON-FIN.IMPERF(ECT)
(0-)y9-h°a-yt'
it he say FIN.AOR

'He said that he was doing//would do it' (Ζ. K.)

which is the indirect representation of

(25') ya-q'a- s-c'a-wa- yt'
I FIN. PRES
Ί am doing//shall do it'
Semi-indirect reporting is quite natural and will associate the comple-
mentiser-particle with the non-finite form of the tense of direct speech,
so that a synonymous alternative to (25) is:
(25a) y9-§9-q'a-y-c'a- wa (0-)y3-h°a-yt'
DYN. NON-FIN. PRES

The fourth method described for Georgian was the association of

complementiser with speech-particle, and, where this co-alescence oc-
curs, omission of the complementiser leaves a straightforward direct
quote. Abkhaz possesses a kind of parallel construction, whereby /h°a/
may be added to a semi-indirect quotation (i.e. where there is shift of
pronouns, insertion of /-?(s)-/ but retention of the direct tense), so that
yet another alternative for (25) would be:
(25b) y9-$3-q'a-y-c'a-wa h°a (0-)ya-h°a-yt'
 Speech reporting in the Caucasus 131

Note, however, that /h°a/ is no longer combining with a structure that

could conceivably have ever formed a direct statement, even after the
removal of The result is that, whilst in both languages omission
of the speech-particle(s) leaves a semi-indirect structure, in Georgian
omission of the complementiser would produce a normal direct quote,
whereas such an omission is just not possible in Abkhaz. Note also how
/h°a/ may here precede a main verb with only one pre-radical affix.
The indirect representation of commands, when not accomplished by
means of the infinitive, place the direct verb (imperative or subjunctive)
in the so-called "purposive", produced by attaching Ι - τ + d to the stem,
regardless of the tense of the introductory verb. If the introductory verb
is past, then the verbs in any subordinate clause within the command
may either remain in the tense of the original words or change according
to the shifts listed above, e.g.
(26) s- an ya- 1- taxa-w // yä-1-taxa-z
my mother which she want NON-FIN.PRES
a- c'k'a (0-)z-3ax3-r+c
NON-FIN.PAST the dress it I sew
(0-)s9+d9- 1- c'a-yt'
it me next-to she put FIN-AOR
'My mother ordered me to sew the dress that she wanted' (Ζ. K.)

If in place of the purposive here the masdar /ä-3ax-ra/ is used, only the
non-finite Past /ya-l-taxa-z/ is permissible if this clause formed part of
the original command; if the Present /ya-l-taxa-w/ is used alongside the
masdar, this clause will be understood as a comment of the speaker
which describes a state of affairs obtaining at the time of reporting the
original command.
The indirect representation of Yes/No-questions may, if the introduc-
tory verb is past, be of the semi-indirect or the fully indirect type with the
peculiarity that /h°a/ will be present in B O T H cases - the interrogative
particle remains the same as in the original question, e.g.

(27) s-ca-wä- ma // s-ca-wä- z- ma h°a

I go DYN.NON-FIN.PRES Q U NON-FIN.IMPERF
da-s- a + z- c'aa-yt' // d-a+z-c'aa-yt'
he me it about ask FIN.AOR
'He asked (me) if I was going' (Ζ. K.) 7

For WH-questions a further possibility exists: the verb-form may stand

in its relative as opposed to its interrogative guise, being followed
nevertheless by the speech-particle /h°a/, so that the full range of
possibilities for translating the English 'He asked me how I was' will be
132 Β. G. Hewitt and S. R. Crisp

(28a) W9- § + pa-q'a-w //

you how Q U be N O N - F I N . P R E S

b) s9-§+pä-q'a-w //
I

c) se-$+pä-q'a- ζ //
N O N - F I N . Past

d) q'a-w //
I how ( R E L A T I V E )

e) s-$a-q'a-z h°a da-s-a+z-c'aa-yt' ( Ζ . K . )

The above-examples have in common that the precise form of the

actual query is clear from the context - in (28), for instance, what was
said in Abkhaz was simply the first word of (28a). However, should the
context not contain any implication as to the precise form of the direct
question, then, where the relative form of the subordinate verb is
chosen, /h°a/ will not appear, e.g.

(28f) s-$a-q'a-w // s-§9-q'a-z d-a+z-c'aa-yt'

'He enquired how I was' ( Ζ . K . )

Omission of /h°a/ from (28d, e) leaves the sentence unfinished - there is

no reference to what question was asked, since the subordinate structu-
res are interpreted as adverbial clauses of manner.
To test the problem of deixis the same range of data was offered to my
informant as was used in the investigation of Georgian. It transpired
that, regardless to the construction, it is always acceptable to use
whatever temporal adverb is suitable at the moment of reporting, so that
the reported portion of (17,17') could in Abkhaz take any of the forms

(29) . . . wac°'3
tomorrow

a) ys- aa- wä- ζ

they that come D Y N N O N - F I N . I M P E R F
( = that they would come)

b) y-aa-wa- yt' h°a

FIN. P R E S / F U T ( U R E )

c) ya-$-aa- wä
N O N - F I N . PRES/FUT

d) ya-§-aa-wa h°a . . .
'(I said some time in the past) that they would come tomorrow (, and
we shall see tomorrow if they do)'
 Speech reporting in the Caucasus 133

Compare with (29) the following example where the direct form would
have included /wac°'9/, though the reported counterpart lacks it:
(30) . . . a+y°3+m§+r3+9+n9
the-next-day

a) y3-§-aa-wa-z

b) y-aa-wa-yt' h°a,

c) ys-$-aa-wa,

d) ya-§-aa-wa h°a . . .
'(I said some time in the past) that they would come the next day
(, and they came the next day)'

On the other hand, selection of the temporal adverb that would have
been appropriate at the moment of speech or thought is tolerated only if
the construction is of the direct or co-alesced type. The example Ί said
(some time in the past) that they would come the next day (, and they
came the next day)' elicited the following judgments - the reported
words only are quoted:
(31) . . . wac°'3
tomorrow
a) y-aa-wä-yt' h°a,
b) ya-$-aa-wa h°a,
c) *y3-$-aa-wa,
d) *ya-?-aa-wa-z . . .
The same responses were produced where these four alternative sequen-
ces represented the reported portion of the English Ί said yesterday that
they would come today, and they came today'.
Thus, whilst a certain freedom characterises deictic usage in speech-
reporting in Abkhaz, there is a clear preference for temporal adverbials
at least to be selected so that the event described is temporally located in
relation to the time of reporting rather than in relation to the time of the
original utterance or thought.

3. North East Caucasian

Lack of informants, coupled with the existence of substantial lacunae in

the study of the languages of this family, inevitably impose some
restrictions on the treatment of their speech-reporting mechanisms. The
134 Β. G. Hewitt and S. R. Crisp

account of Avar which follows is based on the available linguistic sources

and on study of a corpus of written texts; material from other languages
derives entirely from secondary sources.
In common with the languages considered so far in this paper, Avar
displays a strong preference for the direct reporting of speech. The direct
status of the speech is marked by the particles /-(j)an/ and /-(j)ilan/ 8 , or
by /abun/, the past absolutive of the verb /abize/ "to say" - a formation
analogous with Abkhaz /h°a/, and with other languages from the North
East and North Central families (Lezgi /luhun/, Lak /kunu/, Chechen
/älia/). Information about the distribution of these markers is not com-
plete, but the following remarks may be made. Firstly, the use of /abun/
is uncommon with the verba dicendi proper, presumably because its use
is felt to be tautologous. It is used, however, in constructions where the
verbum dicendi precedes the direct quote, presumably to mark the end
of direct speech:
(32) Mahmudica abun bugo: hanze dunjala4ul xalgo
M. having-said is now of-the-world attention
habuge, abun
do-not-make having-said
'Mahmud said: "Do not now pay any attention to the world'"
(Sulejmanov and iabdulaev 1965: 247)

Such constructions are rare in the sources and texts consulted: they
appear to be characteristic of folklore and of the spoken language, where
the end of the direct quote cannot be marked by other means, for
example punctuation. According to Jakovlev (1940: 245), a parallel
situation obtains in Chechen with regard to the use of /ä 11a/ and /boxu§A
Secondly, there are examples of the use of /abun/ with constructions
related to speech-reporting, where there is no explicit verbum dicendi
but where an act of communication is implied:

(33) c'ale nu2, c'ale nuz, c'ale nu£ abun Leninil net^eda
study you having-said of-Lenin to-us
vasijat bugo
testament is
'Lenin's testament to us is: "Study, study, study'"
(Cikobava and Cercvadze 1962: 361)

This construction also, however, appears to be uncommon in the modern

literary language, and subject to restrictions of various kinds (direct
reporting of questions rather than statements of commands, used only
when the quote precedes the main verb of the matrix sentence). Most of
 Speech reporting in the Caucasus 135

the functions of /abun/ seem to be in the process of being taken over by

the suffix /-(j)an/, with which /abun/ is occasionally found in combina-
tion:
(34) kwen habe dir jas -jan abun harana hei
meal make my daughter begged she
'"My daughter, eat something", she begged'
(Madieva 1967: 107, quoting Muhammad XurSilov)

The next marker to be illustrated, /-(j)an/, is considered by Uslar, in the

first significant grammar of Avar (1889: 27), to be the basic means of
marking the direct status of quoted speech. He derives the marker
/-(j)ilan/ from /-(j)an/, considering it a secondary variant of /-(j)inan/
which is formed by combining /-(j)an/ with the so-called emphatic
particle /-(j)in/. For /-(j)an/ itself no etymology is proposed, either by
Uslar or by subsequent authors, and the explanation given for /-(j)ilan/,
as will be shown below, requires further comment.
In modern Avar /-(j)an/ is widely used for direct reporting of state-
ments, questions and commands, and in all permutations of ordering
between the quoted word(s) and the matrix sentence:
(35) duca haburab 4ik'4i dida kidago k'oöenaro -jan
you having-done kindness I ever will-not-forget
abuna Musacadida
said M. to-me
"Ί will never forget the kindness you have done", Musa said to me'
(Saidov and Murtazaialiev 1976: 100)

(36) hanze duca, -jan abuna dica jasalda, diq iint'amexa

now you said I to-the-girl to-me listen
'"Now", I said to the girl, "listen to me" (Madieva 1967: 107)
(37) Axkubekica hiq'ana: Muxtar arawiS -an
A. asked M. has-he-gone?
'Axkubek asked: "Has Muxtar gone?*" (ibid., 106)

The suffix /-(j)an/ may also be used in wider contexts where there is no
explicit verbum dicendi, as in the following example, which should be
compared with the Georgian sentence (3) above:
(38) dun emengi dungi kinazdago wix'a-gi-/'an wuk'ana
I father-and me-and everyone see-and-(saying) was
Ί wanted everyone to see father and me' (Hamzatov 1975: 158)

As far as can be judged from the available material, then, the suffix
/-(j)an/ is unmarked and unrestricted in its use in direct quotation of all
kinds. Hie suffix /-(j)ilan/ (the variant /-(j)inan/, despite its supposed
136 Β. G. Hewitt and S. R. Crisp

anteriority, is not found at all in contemporary sources or texts),

however, occurs only rarely as a marker of direct quotation:
(39) roq'ow δί h et'ilan Salihica ahuna? bicana
in-the-house person is-not S. by-shouting told
'Salih shouted out: "There is no-one in the house'"
(Bokarev 1949: 249)

According to the texts published by Charachidze (1981), however, this

pattern is rather common in the language of the Avars of Turkey.
Indeed, in the grammatical portion of his work (1981: 116), Charachidze
give /-(j)ilan/ as the primary form, /-(j)an/ as secondary, and illustrates
the former with examples like:
(40) dun inilan abunila Imacilas
I will-go said I.
"Ί will go", said Imaöila'

The frequency of /-(j)ilan/ for direct quotation in the Avar dialects of

Turkey raises the question of the extent to which dialectal variation may
account for the existence and use of so many mechanisms for direct
quotation in Avar (there being apparently no clear semantic or function-
al distinction between the various markers). Zirkov (1924: 59-60) for
example treats /-(j)ilan/ and /-(j)inan/ as dialectal variants. In any case,
in the modern Soviet literary language, which is to some extent normali-
sed by language planning measures, there is a clear tendency to use
/-(j)an/ as the sole marker of direct quotation and /-(j)ilan/ as that of
indirect - strictly speaking, semi-indirect - reporting (see below for
discussion and illustration of this mode.
A further suffix, of restricted use in direct quotation, is /-(j)in/,
mentioned above as an emphatic suffix but employed also in the direct
reporting of questions:
(41) kiwe-ym mun inew aht'anila asde wacal
whither yougoing shouted to-him brothers
"'Where are you going?", the brothers shouted to him'
(Borakev 1949: 251)

Any difference in function between /-(j)in/ and the other particle used
for reporting questions, /-(j)ali/, is not clear from the available material;
in any case, in recent texts both suffixes appear to be largely replaced by
/-(j)an/.
To complete the picture, mention should be made of a construction
where, instead of the addition of a special citation suffix to a finite verb
(as in all the examples so far), the suffix /-4i/, usually employed to form
 Speech reporting in the Caucasus 137

abstract nouns and subordinate direct-object clauses, is added to a

participle10. With sentence (39), for instance, should be compared the
following:

(39') roq'ow öi heö'<Ai Salihica ahunai bicana (Bokarev 1949: 248)

Bokarev's account of the difference between these two constructions is

confusing and misleading. He treats sentence (39') as "direct speech"
(prjamaja reö') and sentence (39) as indirect (kosvennaja rec'), defining
these terms however in relation to an external narrator - not to any
participant in the sentences themselves - who (it is claimed) stands in an
objective (sentence (39')) or subjective (sentence (39)) relation to the
words being reported. This distinction is found in no other source, and is
not supported by the evidence of Avar texts; it would appear that
Bokarev has drawn it artificially on the basis of an earlier distinction he
maintains (1949: 244-247) between verbs of knowing, where the verb of
a subordinate object-clause appears as participle + /4i/ ("objective"
relation), and verbs of feeling or perception, where the equivalent
construction has a finite verb + /-(j)ilan/ ("subjective" relation). This
distinction also is challenged by a mass of counter-examples from texts,
but is presented here because it appears closely connected with an
important question concerning /-(j)ilan/, namely the relationship be-
tween this speech-reporting particle and the inferential suffix /-(j)ila/.
Uslar (1889: 29) had already noted a relationship between the two
types of suffix. He expressed the distinction between them in a curious
way, saying that /-(j)ila/ replaced the citation-suffix (/-(j)an/ in his
material) "when the main sentence is only implied", i.e. to mark a
narrative statement as the words of someone other than the narrator.
Zirkov (1924: 60-61) repeated this distinction, remarking also (ibid.,
56-57) that this made the use of verb forms with the suffix /-(j)ila/ the
natural choice for the narrative of folk-tales, since there is an assumption
that the whole text is narrated from someone else's words. Bokarev
(1949: 250) goes so far as to treat /-(j)ila/ as a simple variant of /-(j)ilan/.
This equation of speech-reporting and inferentiality is repeated, on the
basis of Bokarev's data, by Friedman (1979: 344). Though this is
probably to take the relationship too far, the connection between
inferentiality (or status) and speech-reporting established by Friedman
on the basis of a range of languages does seem significant, at least to the
extent of making the etymology proposed by Uslar for /-(j)ilan/ (see
above), seem less likely than a link with the inferential particle /-(j)ila/.
138 Β. G. Hewitt and S. R. Crisp

Whatever the detailed connection, there is undoubtedly a logical rela-

tionship between inferentiality and speech-reporting in Avar.
Literary Avar has no complementiser and no mechanism of fully
indirect speech-reporting 11 . Instead, following the terminology adopted
above, Avar is characterised by the semi-indirect mode of reporting,
where the tense of the verb of the embedded speech remains unchanged
(but for the addition of the citation suffix), whilst the pronominal
reference changes to suit the relations of the matrix sentence. In this
respect the situation in Avar bears, a striking resemblance to that found
in Svan (see section A above). The normal pattern may be illustrated by
the following example, which is the semi-indirect counterpart of sen-
tence (35) above:
(35') dica haburab 4ik'4i zinda kidago k'oöenan'/a« abuna Musaca dida
I himself
'Musa said that he would never forget the kindness I had done'
(Saidov and Murtaza?aliev 1976: 100)

In this example the 2nd person pronoun of sentence (35) has become 1st
person, the 1st person pronoun has become reflexive, and the embedded
verb has the suffix /-ilan/12. It is clear from this example, moreover, that
the characteristic feature of the semi-indirect mode in Avar is not the use
of /-(j)ilan/, which may equally be used in direct quotation (see sentences
(39) and - especially - (40) above), but the changes in pronominal
reference. The use of this mode of reporting is therefore facultative to
the extent that it applies only to sentences having personal pronouns,
and the relevant pronominal switches, though commonest in combina-
tion with /-(j)ilan/, can also occur when other suffixes are employed:
(42) diq ret'el bugis-an hiq'ana dida Sawdatica
at-me clothing is-? asked me S.
'Sawdat asked me whether I had clothes' (Madieva 1967: 109)
(43) son mun i u r u / wuk'ara4i nizeda bicana
yesterday you (in)-Urux having-been us told
'(They) told us that you were in Urux yesterday' (Bokarev 1949: 248)

Lack of material and of informants makes it impossible at this stage to

give more details about the distribution of the various suffixes and
mechanisms for speech-reporting in Avar, or about which combinations
of mechanism and construction are preferred, tolerated or blocked. It
should be noted however that the situation in Avar is considerably more
complex than in most other languages of Daghestan, where the range of
citation suffixes is lacking and the only possibilities for (semi-)indirect
 Speech reporting in the Caucasus 139

reporting are pronominal switches, variations in clause order and occa-

sional changes in the form of the embedded verb (for Lezgi data, see
Gadziev 1963: 193-196; for Tabassaran, Xanmagomedov 1970: 194-195;
for Lak, Murq'ilinskij 1981: 112-113).
The question of deictic reference in Avar speech, reporting must also
await the possibility of more extensive investigation. The following
example of spatial deixis, however, indicates that for Avar, as for the
other languages considered in this paper, the point of reporting is the
controlling factor (though undoubtedly a high degree of freedom of
choice exists):
(44) jasai kin karaj yob dunjalalde bicana xeraw öijasda
girl how having-arrived that to-world told old to-man
'The girl told the old man how she came to that world'
(Charachidze 1981: 171)

Here the "remote" deictic form /yob dunjalalde/ is used to refer back to
what in the preceding direct speech was /hab dunjadalde/, "to this
world" ("immediate" deixis).
Despite the proliferation of mechanisms for reporting speech in Avar,
it is not uncommon to find examples of verbatim quotation without the
use of either suffixes or grammatical adaptations. Were such examples
only to be found in modern written texts it might be possible to ascribe
them to the influence of Russian on the literary language - as has been
claimed in the case of other languages of Daghestan (see Baskakov ed.
1969: 291, 401) - but in fact this phenomenon, though it receives scant
mention in sources on the grammar of Avar, is common in the earliest
recorded Avar texts (folk-tales from the mid-nineteenth century) and
occurs sporadically ever since, being found today both in the oral Avar
folklore of Turkey (Charachidze 1981) and in the literary works of the
most respected of Soviet Avar writers:

(45) mac'al bat'ijal ruk'a abe rak' co bugoni

languages different be say heart one if-it-is
,Say: "If there is one heart, (then) let languages be different"'
(Hamzatov 1975: 76)

This pattern of direct quotation is extremely common in Chechen, where

it is very much the preferred structure. Chechen has no citation suffixes,
so the only way of marking formally the direct status of quoted speech is
the use of the absolutives /älla/ and /boxuS/ referred to above, which - as
has already been noted - is restricted in much the same way as is /abun/
in Avar. Apart from entirely unmarked verbatim quotation and the use
140 Β. G. Hewitt and S. R. Crisp

of /älla/ or /boxuS/ withouth grammatical adaptations, then, Chechen has

a semi-indirect mode with pronominal switch, as illustrated by the
following pair of sentences taken from Jakovlev (1940: 245):
(46a) co elira: so selxana weana
he said I yesterday came
'He said: "I came yesterday'"
b) co elira, Sa selxana weana
himself
'He said that he came yesterday'

Jakovlev expresses surprise (1940: 247) that /älla/ and /boxuS/ are not
used explicitly to mark a (semi-)indirect mode; interestingly, in a recent
normative grammar of Chechen one finds examples where, in addition to
pronominal switch, the semi-indirect mode is in fact distinguished by the
addition of /älla/:
(47a) san waSas elira söga: suna höca wan läa
my brother said to-me I with-you to-go want
'My brother said to me: "I want to go with you'"
(Dzamalxanov and Maöigov 1973: 104)
b) san wasas elira söga, Sena söca wan läa älla
himself with-me
'My brother told me that he wanted to go with me' (ibid.)

Occasionally also one finds the masdar, normally characteristic of

subordinate direct-object clauses, used for the embedded verb of repor-
ted speech:
(48) ah dijcira, iz(a) war
you said he arrival
'You said that he had arrived' (Jakovlev 1940: 247)

The overlap in formal expression between speech-reporting and subordi-

nate direct-object clauses, then, is characteristic of Chechen as it is of
Avar, where /-(j)an/, /-(j)ilan/ and I Ail are all used in both types of
construction. Otherwise, while recognising the need for more detailed
study of Chechen material, one may conclude that the overwhelming
preference for verbatim quotation in Chechen means that its mecha-
nisms for speech-reporting are less complex than those of Avar.
 Speech reporting in the Caucasus 141

Notes

1. My three Georgian informants are: S. A. = Sukia Apridonidze, E. C. = Eteri Cxot'ua,

and M. Dz. = Marine DzeiranaSvili, to all of whom I am deeply grateful.
2. For comparative purposes the second of the three italicised verbs from this passage of
Luke, which is the only one to appear in finite form in all the consulted versions, is
treated as follows in other languages: Present in Greek, Latin, Russian; Imperfect in
Old Armenian, Modern Eastern Armenian; Perfect in Modern Western Armenian. In
Greek and Latin the final copula stands in the infinitive, but elsewhere it is realised by
the Present form.
3. In Hewitt (1979: 5) /h°a/ was erroneously described as an archaic form of the Present
absolutive.
4. According to my informant, my wife Zaira Khiba (= Ζ. K.), /h°a/ may be omitted.
5. Here /h°a/ may be inserted. Observe the subtle segmental difference between the verbs
of (21) and (22). In both cases the Georgian translation gives /äeek'itxa mat/ 'he asked
them' (Bakanidze 1980: 173-4).
6. Indeed the reason for the omission of /h°a1 in (22) appears to be precisely that the
example is taken from a collection of texts that has been "polished" for publication. If
(22) is read as it stands (i.e. out of written context), then /h°a/ is obligatory.
7. Alternative readings for each of the variants are 'He, asked (me) if he; was going' and
Ήβι asked (me) if he; had been going' respectively.
8. The variants /-an/ and /-ilan/ usually occur after consonants, /-jan/ and /-jilan/ after
vowels (example (34) is an exception to this rule).
9. /boxuä/ is the present absolutive of the verb /bäxa/ 'to say'. In the modern language
there appears to be no functional difference between /älla/ and/boxuS/ as they are used
for speech-reporting; quantitatively, /älla/ is more frequently found (Esxadiiev 1970:
123).
10. An exception to this rule is the case of the copula, where the suffix /-4i/ may be added
to the present tense forms /-ugo/ "is" and /hei'o/ "is not", as well as to tell correspon-
ding participles.
11. The Southern, Zakatal' dialects of Avar have borrowed the complementiser /ki/ from
Azeri (Aslanov 1977), but this has not led to the rise of an indirect mode of speech-
reporting. In Azeri itself, moreover, the use of /ki/ in speech-reporting is accompanied
only by change in pronominal reference, while the tense of the dependent verb remains
unchanged (Rüstämov et al., eds. 1959, vol. 2: 398-399). In other languages of
Southern Daghestan which have borrowed /ki/ from Azeri for speech reporting the
picture is similar: in Tabassaran there is pronominal switch but no change in the tense
of the verb (Xanmagomedov 1970: 193), while in Lezgi there are no modifications of
any kind in the form of the embedded speech (Gadziev (1963: 191).
12. The final vowel is dropped and /-ilan/ is added to the consonant. This is now the normal
pattern for such verb forms, but Zirkov (1924: 63) refers to a distinction drawn by his
informants between this pattern, which was interpreted as punctual, and the addition
of /-jilan/ to the final vowel, which was felt to give a habitual meaning.
142 Β. G. Hewitt and S. R. Crisp

References

Aslanov, A. M. 1977. 'Azerbajdianskaja model' sloinopodöinennogo predloienija ν

zakatal'skix dialektax avarskogo i caxurskogo jazykov.' In: Τ. X. Kuaseva & X. T.
Taov, eds. 1977. Voprosy sintaksiöeskogo stroja iberijsko-kavkazskix jazykov. Nalchik:
Kabardin-Balkar Research Institute, pp. 295-302.
Bakanidze, G. 1980. Ambavi nartebisa. Tbilisi: Nak'aduli.
Baskakov, Ν. Α., ed. 1969. Osnovnye processy vnutristruktumogo razvitija iranskix i
iberijsko-kavkazskix jazykov. Moscow: Nauka.
Bokarev, A. A. 1949. Sintaksis avarskogo jazyka. Moscow and Leningrad: Izdatel'stvo AN
SSSR.
Charachidze, G. 1981. Grammaire de la langue avar (Documents de linguistique quanti-
tative, 38). Saint-Sulpice de Faviöres: Editions Jean-Favard.
Cikobava, Α., I. Cercvadze. 1962. Xun$uri ena. Tbilisi: University Press.
Diamalxanov, Z. D., M. Ju. Maöigov. 1973. Noxöijn mott. 2-ya daq'a. Sintaksis. Grozny:
Chechen-Ingush Publishing House.
Esxadiiev, Ja. U. 1970. 'Sposoby vyraienija sloinopodöinennogo predloienija ν öeäens-
kom literaturnom jazyke.' Ceieno-Inguäskij Nauöno-Issledovatel'nyj Institut Istorii,
Jazyka, Literatury i Ekonomii. Izvestija 9, 2, pp. 101-154.
Friedman, V. A. 1979. 'Toward a typology of status: Georgian and other non-Slavic
languages of the Soviet Union.' In: P. R. Clyne, W. F. Hanks & C. L. Hofbauer, eds.
1969. The Elements: a parasession on linguistic units and levels. Chicago: Chicago
Linguistic Society, pp. 339-350.
Gadiiev, Μ. M. 1963. Sintaksis lezginskogo jazyka, öast' 2: sloinoe predloienie. Makhach-
kala: Dagestanskij Filial AN SSSR.
GigineiSvili, I., V. Topuria, I. Kavtaradze. 1961. Kartuli dialekt'ologia I. Tbilisi: Univer-
sity Press.
Hamzatov, R. 1975. Dande harural asaral, vol. 6. Makhachkala: Dagestanskoe Kniinoe
Izdatel'stvo.
Hewitt, B. G. 1979. Lingua Descriptive Studies 2: Abkhaz. Amsterdam: North Holland.
Hewitt, B. G. 1981. 'Eine weitere Betrachtung der georgischen Redepartikel "-tko//-
tkva'V Georgica 4, 83-85.
Hewitt. Β. G. 1981a. Contrastive study of the subordinate clause syntax of Georgian and
Abkhaz. Unpubl. doct. diss. Cambridge.
Hewitt, B. G. 1982. 'From direct to indirect speech: a South Caucasian anomaly.' Folia
Slavica 5, 206-213.
Hewitt, B. G. 1983. Ά Svan (La§x) text: part 1.' Bedi Kartlisa XLI, 278-287.
Hewitt, B. G. 1984 (to appear). 'Another look at the Georgian speech-particle '-tko//-
tkva'.' Bedi Kartlisa XLII.
Hewitt, B. G. 1985 (to appear). Ά Svan (La§x) text: part 2.' Bedi Kartlisa XLIII.
Hewitt, B. G. Forthcoming. 'Parataxis revisited (via the Causasus).' General Linguistics.
ImnaiSvili, I. 1975. Sinuri mravaltavi: gamok'vleva da leksik'oni. Tbilisi: University Press.
Jakovlev, N. F. 1940. Sintaksis öeöenskogo literaturnogo jazyka. Moscow and Leningrad:
Izdatel'stvo AN SSSR.
K'ldiaävili, D. 1981. Txzulebani (or t'omad). Tbilisi: Sabö'ota Sakartvelo.
Laä°ria, M. 1978. Sota Rustaveli: Abias-c°a zS°9w. Sukhumi: Ala$ara.
Lolaävili, I. 1964. Jve/i kartveli mexot'beni; c'igni meore. Tbilisi: Mecniereba.
 Speech reporting in the Caucasus 143

Madieva, G. I. 1967. Awar mac'. K'iabileb but'a: sintaksis. Makhachkala: Daguipedgiz.

Murq'ilinskij, G. B. 1981. Laku maz. Daniwmur Skolalul 7-8 klassirtansa uCebnik.
Makhachkala: Daguipedgiz.
Rüstämov, R. Ä. et al., eds. 1959. Azärbajdzan dilinin grammatikasy. Baku: Akademija
Nauk AzSSR.
Saidov, M. S., Μ. M. Murtaza aliev. 1967. Awar mac' grammatika. K'iabileb but'a:
sintaksis. Makhachkala; Daguöpedgiz.
Sanidze, A. 1976. Jveft kartuli enis grammat'ik'a. Tbilisi: University Press.
§3nk°ba, B. (ed.) 1962. Nart Sasr9q°'ay ps3ny°az°3y zayz°y°3k' yara yaic°ay: Apswa z°lar
rep'os. Sukhumi: Abkhazian State Publishers.
Sulejmanov, Ja., Μ. abdulaev. 1965. Awar literaturijab mac'. Makhachkala: Daguöped-
giz.
Uslar, P. K. 1889. 'Avarskij jazyk' (Etnografija Kavkaza. Jazykoznanie, 3). Tiflis:
Upravlenie Kavkazskogo Uöebnogo Okruga.
Vogt, Η. 1971. Grammaire de la langue giorgienne. Oslo: Universitetsforlaget.
Xanmagomedov, B. G.-K. 1970. Oöerki po sintaksisu tabasaranskogo jazyka. Makhach-
kala: Dagudpedgiz.
2irkov, L. I. 1924. Grammatika avarskogo jazyka. Moscow: Institut Vostokovedenija.
Reported speech in some languages of Nepal
Karen Ebert

0. Traditional grammars lead us to believe that there is a clear distinction

between what is called 'direct' and 'indirect' speech. Often this does not
even hold true for the languages those grammars intended to describe.1
The common characterization of direct speech as being an exact repro-
duction of a previous speech act - in contrast to a paraphrased indirect
report - is most misleading. What presents itself as direct speech is
usually 'constructed dialogue' (cf. Tannen, this vol.) or at most an
approximation to the original utterance. Further, thoughts, intentions
and other cognitive acts are also represented in the form of direct
speech.
The only useful criterion to distinguish direct from indirect speech is a
shift in the deictic elements. In direct speech deictics point to the
reported speech situation, in indirect speech they point to the actual
speech situation. Deictics of the original or imagined speech have
therefore to be shifted to non-deictics in an indirect report, and vice
versa if elements of the actual situation were referred to in an original
message. How the deictic elements of a given reported speech are meant
to refer is in many languages not deducible from the shape of the
sentence; i.e. direct and indirect speech are formally not distinguished.
I shall present data from three languages of Nepal, an area that has
hitherto not received much attention in general linguistic discussion.
Means for reporting speech or thought will be demonstrated in some
detail for Chamling*, a Tibeto-Burman language of Eastern Nepal. I
shall then compare the Chamling data with reported speech in Nepali,
the national language belonging to the Indo-Aryan branch of Indo-
European, and in Newari, a Tibeto-Burman language distantly related to
Chamling, spoken by the original inhabitants of the Kathmandu Valley.

1. Chamling has no morphologically or syntactically marked difference

between what might be called direct or indirect speech. Reported speech
or thought is presented as it would actually be spoken in the reported
context.
* I would like to thank the Deutsche Forschungsgemeinschaft for financial aid that
enabled me to do fieldwork in Nepal.
146 Karen Ebert

A noun phrase referring to the speaker may precede or follow the

speech; usually the speaker is not mentioned at all, but understood from
the context. In roughly 90% of the cases analysed, speech - henceforth
understood as including thought and other forms of inner speech - is
followed by a form of the verb rurjma "say" which functions as a sort of
'unquote'. It can be combined with or replaced by other report verbs. I
have found speech embedded into the following contexts:
(NP) REPORT FINAL [SPEECH] UN- REPORT (NP)
VERB PART QUOTE VERB
ruga
ruga ruga-na ludyi "said/thought"
ludyi ruga-pa senyi "told"
senyi rugma-pa dum ma "asked"
dum ma prata "speech made"
yenyi khrapa "shouted"
yenyi "cried"
"heard"
The report verbs and the unquotes are optional and more or less freely
combinable; the infinitival form rugma seems to be preferred with
questions, thoughts and the verb "hear".
-na and -pa are conjunctive markers, -no connects a series of succes-
sive actions of the same agent, -pa signals a subsequent topic shift, but is
not used quite consistently. This function of -pa explains why it is found
so frequently with questions (see (3) below). Both conjunctives corre-
spond to the Nepali 'conjunctive participle' on -era, that is rendered in
textbooks as "having (done)". Clauses ending in conjunctives are best
translated into English by when/after-clauses or coordinated by and.
The conjunctives embed speech into the matrix sentence. A construc-
tion such as
Ram-wa [SPEECH] rurja-na dum ma . . .
translates literally:
Ram [SPEECH] having-said spoke . . .
The place of embedded clauses in Chamling is before the verb. The place
of the subject may be either before the clause or between clause and
verb. The contexts of reported speech drawn out schematically above
seem to contradict this pattern. However, if a report verb precedes
speech it usually carries a sentence final marker (see 1.3, 2.2, 3.1/2
below); it may thus announce a speech, but the syntactic embedding
takes place in the sentence following. As the conjunctives are not
obligatory, speech can also be loosely integrated into a sentence with a
finite report verb.
 Reported speech in some languages of Nepal 147

The following examples 2 may illustrate different possibilities of report-

ing speech or thought in Chamling.

(1) 1. Busiri -neri pusa-ko racha. 2. pusa-na [hö-bu-ou

Β. -near went-FIN went-CONJ king-brother-VOC
dosko khana oda ta-hirja-ko thala? . . . khana
how you here 2.-settled-PART MOD you
khoda ta-pse-ko pusa] ludyi-ko racha. 3. ludyi-pana
where 2.-go-PART go! told him-FIN told him-CONJ
Busiri ludyi-ko racha, 4. [kaga-bo m-bhusi t-ug-ko
B. told him-FIN I-EMPH REL-first come-1.-PART...]
rutja-ko racha.
said-FIN.
1. (He = Wachapa) went near Busiri. 2. Having gone [brother! how
come you settled here? . . . Go wherever you'll go!] he told him. 3.
Having told him (i.e. W. to B.), Busiri spoke to him, 4. [I came
first . . . ] he said.

(I have tried to imitate the Chamling constructions as closely as possible

in the translations without any consideration to correct English.)

(2) 1. [jhara-pa-wa dhiki jhäto-da mi-ye-pa

all-CLASS-AG mortar flour mill-in 3. PL/ITR-come-CONJ
pitho camacamdei chima ise kuneu] mi-rirja-ni
flour rice any leave should not PART 3. PL/ITR-said-PART
mi-prata-ni. 2. mi-prata-pa ito-pa-na
3. PL/ITR-shouted-PART 3. PL/ITR-shouted- one-CLASS
CONJ
yenyi-ko racha. 3. [ale dhiki-da jhäto-da
heard-FIN today mortar-in flour mill-in
PART
ek mana camacam ek mana pitho chityi] yenyi-ko racha/
one mana rice one mana flour leave! heard-FIN
1. [Everybody, when coming to the mortar and flour mill, they must
not leave any flour (or) rice!] they shouted. 2. When they shouted,
one (woman) heard it. 3. [Leave today one mana rice and one mana
flour in the mortar and in the flour mill!] she understood.

ruga-no after reported speech functions grammatically as a full verb. It is

inflected for tense and it shows agreement in person and numer (cf. 2.1
mi-riga-ni) 3 . It does not count, however, as a speech act characterizing
verb. A common means of establishing text cohesion in Chamling is to
repeat the verb that describes the last action of the preceding sentence
148 Karen Ebert

followed by -pa(lo) or -n9 (1.2, 2.2). ruga cannot perform this function;
cf. 4. and 6. in the following text:

(3) 1. kic-kurugpa Saphopte-lai senyi-ko racha. 2. [a-kurugp-ou,

their-uncle S.-DAT asked-FIN my-uncle-VOC
a-nicho Khliyama paina thala?] ruga-na
my-younger sister Kh. be not MOD said-CONJ
m-kurugpa Saphopte-lai senyi-pa Saphopte-wa
her uncle S.-DAT asked-CONJ S.-AG
ruga-ko racha. 3. [kaga pani pa-chain-ug-a, kaga khro-
said-FIN I also NEG-know-l.-P I bite-
set-ug-a bo] ruga-ns dum-ma-ko.
kill-l.-P MOD said-CONJ speech-made-FIN
4. dum-ma-palo [a-nicho-na a-kurugpa
speech-made-CONJ my-y.sister-PART my-uncle
Saphopte-wa setyi-ko racha. tayä paina-ko thyo? . . . ]
5.-AG killed-FIN brains not be-FIN
5. aso [a-kurugp-ou Saphopte-ou, a-nicho-na
now my-uncle-VOC S.-VOC my-y.sister-PART
ta-setyi-ko racha; khoda ta-setyi khoda silum-holum
2.-killed-FIN where 2.-killed where burial ritual
ta-mu-ko hige thala?] ruga-na Tuwama-wa m-kurugpa
2.-made-PERF MOD said-CONJ T.-AG her-uncle
Saphopte-lai senyi-ko racha. 6. senyi-pa [odam set-ug-ko
S.-DAT asked-FIN asked-CONJ here kill-1.-
thyo, odam c-ug-ko thyo] ruga-na tyoko sya-ko tug-da
FIN here eat-l.-FIN said-CONJ that die-of place-in
pudyi-na kho-maidyi-ko racha.
took-CONJ see-made-FIN
1. (Tuwama) asked her uncle Saphopte. 2. In saying4 [uncle, is my
younger sister Khliyama not here?] having asked her uncle Saphopte,
Saphopte answered. 3. Saying [I don't know, I must have killed her
with my bite] he spoke. 4. Having spoken [my younger sister was
killed by my uncle Saphopte. Did she not have any brains? . . . ] 5.
Now: [my uncle Saphopte! you killed my younger sister; where did
you kill her, where did you make the burial ritual?] saying Tuwama
asked her uncle Saphopte. 6. Having asked him, [here I killed her,
here I ate her] saying he took her and showed her the place of death.

It is well known from African languages that the general report verb
may develop into a complementizer. Consequently it loses its illocutio-
nary force and is replaced by another speech act verb. In Kera, a
 Reported speech in some languages of Nepal 149

language of the Chadic family spoken in southern Chad, the complemen-

tizer minti (* < "say (iterative)", cf. äy mini "give message") is no longer
used as a speech act verb; its place has been taken by wdate, the iterative
of waa- "to produce, to give birth". Both direct and indirect speech are
introduced by wdate minti "said that". 5 A similar development has taken
place in other African languages. 6
In Chamling the infinitival form rugma (and possibly also ruga-no) is
currently developing into a complementizer:

(4) [khu garib hir)e] (rugma) kaga chaidü.

he/she poor is SAY I know-1.
I know that he/she is poor.
(5) [a-kururjpa] rugma pa-chait-aina, Saphopte-wa pani
my-uncle SAY NEG-know-NEG S.-AG also
[a-cyodumci] rugma pa-chait-aina.
my-niece SAY NEG-know-NEG
[(he is) my uncle] SAY did not know, Saphopte also
[(she is) my niece] SAY did not know.
She did not know that he was her uncle; S. also did not know that she
was his niece.

The status of rugma remains ambiguous, as no deictic or syntactic

changes are made in the embedded sentence. It may be complementizer
and report verb at the same time. Cf. also:

(6) m-kurugpa Saphopte-lai Khliyama-na dum pa-m-aina

her-uncle S.-DAT Kh.-PART speech NEG-make-NEG-
ko racha; [kacka oso tira-cka oso-musa phaka-cka
FIN we DL/EX so became-l.DL/EX so-doing separated-
l.DL/EX
a-nicho Khocilipa pani oso tira] rugma-pa-wa,
my-younger brother Kh. also so became SAY-CON J
tyoko-na dum pa-m-aina-ko racha.
that-PART speech NEG-make-NEG- FIN
Khliyama did not speak to her uncle Saphopte; what happened to
them both, how they separated, what happened to her brother
Khocilipa, - this she did not tell.
or: . . . "such happened to us, we separated in such way, such
happened to my brother Kh.", this she did not tell.
(7) ito nicho higa-ko, [khu pani sya-khata
one younger brother was-FIN he PART died-fell
dha-khata] rugma, [aso khuni-khuni kaici pusa-ce
SAY now where-where we DLgo-DL
150 Karen Ebert

thala] ruTjma-pa [Tuwama-cai, khana aso hui Tharu-ho

MOD SAY-CONJ T.-CONTRAST you now downTharu-king
Baju-hö-da warja . . . ] ruga-ci-na . . .
Baju-king-in go said-DL-CONJ
(sya-khata dha-khata = an idiomatic expression for "died")
There had been a brother. They thought he was dead. Where should
they go now? "Tuwama, you go down to the land of the Tharu and
Baju kings (i.e. to the Terai) . . . " they said and . . .
or: "he is dead now", they thought. "Where shall we go now?" they
thought/said . . .
or: Thinking ( = rugma-pa) about where to go now that (= rurjma) he
was dead, they said . . .

The examples considered so far were, with the exception of sentence (4),
all taken from narratives. The narrator has to take the perspective of his
characters when reporting or constructing speech or thought. In conver-
sation, however, a sentence like (5) is ambiguous without context and
can have the meaning

(5') She did not know that he is my uncle, and S. did not know that she is
my niece.

A pronominal switch may take place when something said about the
actual speaker or hearer by a third person is reported. Let me draw out
schematically the switches that are likely to occur.
A said to B: (8) Dhan Bojpur-da pusa-ko.
Dhan went to Bojpur.
Β reports to Dhan: (8') A [khana Bojpur-da ta-psa-ko7] ruga.
A said, you went to Bojpur.
rather than A [Dhan Bojpur-da pusa-ko] ruga.
Dhan has heard this (8") A [karja Bojpur-da pusug-ko] ruga,
from Β and reports to A said, I had gone to Bojpur.
F:
A said to B: (9) khana Bojpur-da ta-psa-ko?
Did you go to Bojpur?
Β reports to C: (9') A [khana Bojpur-da ta-psa-ko]
senyi.
A asked: "Did you go to Bojpur?"
or: (9") A [karja Bojpur-da pusug-ko] senyi.
A asked if I went to Bojpur.
If A reports to C what he asked B, no pronoun switch to 3rd person is
possible, thus only:
 Reported speech in some languages of Nepal 151

(9"') karja khu-lai [khana Bojpur-da ta-psa-ko]

senuga.
I asked him "Did you go to Bojpur?"
It is felt odd to talk in 3rd person about a speech act participant.
Therefore nouns and 3rd person pronouns or possessives are adapted to
the actual situation. Hie switch is possible only in one direction:
3rd 2nd/lst
Taking a deictic switch as the basic feature of indirect speech, we may
state that Chamling makes a limited use of this technique. It is, so to
speak, a concession to pragmatic practicability, even though it does in
principle lead to ambiguities. Nothing in the sentence structure of (8')
tells us whether khana refers to the actual or to the reported hearer.

2. Nepali exhibits largely the same patterns of reported speech as

Chamling does. T. W. Clark in his 'Introduction to Nepali' states:
"Speech, whether Statement, Command or Question, is reported, i.e.
conveyed to a third person, by quoting the actual words spoken, usually
without any modification" (p. 171). What Clark wants to indicate is, that
there are usually no deictic switches.
Nepali marks the end of reported speech by bhanera, the 'conjunctive
participle' of the general report verb bhannu, or by bhanne, the 'infiniti-
val participle', bhanne is used mainly before nouns such as kurä
"speech", thähä "knowledge", äs "hope" and before sunnu "hear".
Though, as in Chamling, the uses of the forms are not clearly distinct,
bhanne corresponds largely to the Chamling infinitival form rugma-pa,
bhanera to ruga-naf-pa. bhanera is used together with other speech act or
cognitive verbs, including bhannu itself8:
(10) us-le [aja sahar jänchu] bhanera ma-läi bhanyo
he-AG today town I go SAY-CONJ me-DAT said
He said to me [I'll go to town today]
(11) mero chorä-läi khup khojiniti cähincha, tyasayle
my son-DAT much curiosity requires, therefore
[tyo bätobäta kahä-kahä läncha?] bhanera sodhyo.
that roads where carries SAY-CONJ asked
may-le bhane [näni soj hay gais bhane kharsyäg
I-AG said. 1.child straight go-2. SAY Kh.
siligudhi pugchas.]
S. get-2.
My son is very curious, therefore [where do those roads lead to?]
saying he asked. I said: [Child, if you go straight, you will get to
Kharsyang and Siliguri.]
152 Karen Ebert

(12) [timi magar] bhani-thaneko ta, hoina po rahechaw

you Magar I said-thought but not PART turn out-2.
I thought [you (were) Magar], but you turn out not to be.

The second sentence in example (11) shows that in Nepali speech can be
introduced by a report verb.
bhanne, bhanne kurä (lit. saying speech"), and bhanera are also used
as complementizers. The Chamling sentence (4) was taken from the
Linguistic Survey of Nepal questionnaires 9 and is a translation of Nepali
(4') [ü gariva ho] bhanne kurä ma-läi thähä cha
he poor is SAY speech me-DAT knowledge is
(13) [timi yasto kurä garchäw] bhanne ma-läi thähä thiena
you such thing do-2. SAY me-DAT knowledge was not
I did not know that you would do such a thing
(14) [dhärä -mä päni cha ki] bhanera herna gae
tap -in water is Q SAY-CON J to see water 1.
I went to see [is there water in the tap?]
I went so see if there was water in the tap.
(15) [tapäi-läi hijo räti-ko sinemä man paryo] bhanne sune
You HON-DAT yesterday night-of cinema pleased SAY heard
I heard that you enjoyed the picture last night.
In (15) a pronominal shift has obviously taken place. The polite 2nd
person pronoun tapäi refers to the actual hearer and is a reflex of a 3rd
person noun phrase in the original speech. The meaning of (15) pre-
cludes the interpretation as an unshifted report. In other contexts 2nd
person pronouns and possessives may be ambiguous as to whether they
point to the original situation or are the result of a change in perspective.
(16) Räm-le [timro jäc bholi huncha] bhanyo
Ram-AG your exam tomorrow be-3. said
Ram said [your exam will be tomorrow.]
a) Ram said: "Your exam will be tomorrow."
b) Ram said that your exam will be tomorrow.
It should be mentioned that Nepali has an alternative mode of
marking reported speech, namely with the final particle re. With re the
source of the information is not mentioned; it may be understood from
the context or unknown.
(17) Jagbahädur eutä mämüli sipähi-bata nepäl-ko
J. one common soldier-from Nepal-of
mahäräj bhae. uni bidvän thienan.
Maharaja was/became he-HON learned was-not-Hon
 Reported speech in some languages of Nepal 153

Iekhna padhna pani ramrari jandsynathe re

write read also well knew-not-HON
taypani uni sadhay satarka bhaera basthe re
nevertheless he-HON always provident being was
Jangbahadur became from a common soldier the Maharaja of Nepal.
He was not learned. It is said, that he did not even know well how to
read and to write. Nevertheless he is said to have always been
provident.

Nepali and Chamling are not related, nor were they geographically
close before Nepali became the national language. One might therefore
conjecture that their way of reporting speech can also be found in other
languages of the area. Indeed I was told that there is no 'indirect speech'
in Nepali's close relatives, the Indian languages (including Sanskrit), and
that it is rarely used in Farsi.10 On closer interrogation I found that, as in
Chamling and Nepali, a pronominal switch is usually carried out from
3rd —» lst/2nd in Hindi and Farsi; sentences with a 2nd person pronoun
in reported speech show the same ambiguities as shown above. The
embedding contexts are of the familiar Indoeuropean type: goft ke
(Farsi), kaha ke (Hindi) "say that", with the complementizer ke intro-
ducing both direct and indirect speech. The unshifted representation of
speech is widespread also in other parts of the world, as is the embedding
by a seemingly redundant report verb, e.g.

Amharic:11
(18) [äymatum] bilo farrä
they-will-not-come saying he-feared
He was afraid they would not come.
(19) [ya-Kabbada bet yat naw] bilo tayyaqa
of-K. house where is saying he-asked
He asked: "Where is K.'s house?"
Shona:
(20) Wakamurouera pasi kuti [ndimuuraye]
He struck him down say I-kill-him
He struck him down in order to kill him.

Turkish:
(21) Cocuk-lar [vakti-miz 90k] diye yavas-yavas yürüyor-lar
children-PL time-our much saying slowly go-PL
Thinking "we have a lot of time" the children went slowly.

The fact that this type of construction is so widespread in totally

unrelated languages and areas indicates that it is a natural way of
154 Karen Ebert

reporting cognitive processes. That we sometimes find it strange is

probably due to the fact that we, as is the case with 'speech', do not have
a report word that is general enough to adequately translate bhanera,
ruga-ns and so on.

3. One of the major indigenous languages of Nepal, however, differs in

its means of reporting speech. In Newari pronouns change in indirect
speech in the way that is well-known from many languages. 12
(22) wan [jy wayaa] dhaala
he I came said
He said: "I came".
(22') wan [wa wayaa] dhaala
he he came said
He! said that hex came

(22') can only be a report of (22), but not of (23):

(23) wan [wa wala] dhaala
he he came said
He said: "He came".

which would be the same in indirect speech:

(23') wan [wa wala] dhaala
He! said that he2 came.

The ambiguous 3rd person pronoun in English indirect speech is disam-

biguated in Newari by the verbal suffix:
-yaa indicates reference identity with the speaker of the embedding
clause or situation (coreferential form),
-la indicates non-identity (non-coreferential form) 13
Nevertheless the ambiguity encountered in Chamling, Nepali and other
languages with 2nd person pronouns remains in Newari:
(24) wan [cha wala] dhaala
he you came-NONCOR said
(24) can either be equivalent to
(25) wan [jy wala] dhaala
he I came-NONCOR said
He said that I came.

or can be a report of an original speech act "A came" to A, as

exemplified for Chamling in (8) above. For (25) only the indirect
interpretation is possible (cf. 22).
 Reported speech in some languages of Nepal 155

In Newari it is only the pronoun, but not the verb, that is switched in a
non-direct report. If the reported speech were bound completely to the
actual situation one would expect the verbal suffix to mark reference
identity with the actual speaker, as it does in non-reported speech:
(26) jy wayaa
I came-COR

but
(27) cha/wa wala
you/he came-NONCOR

Interestingly in questions -yaa indicates references identity with the

hearer.
(28a) jy/wa wala laa? Did I/he come?
b) cha wayaa laa? Did you come?

For indirect speech involving 3rd persons:

(29) wan wayta [wa wa la laa] dhaala

he him he came Q say
He asked him] if he2 came.
(30) wan wayta [wa wayaa laa] dhaala
He asked him! if he! came. 14

Newari has thus for the structures examined developed greater expres-
sive power than have Nepali or Chamling (or English); for this feature it
is comparable to American Indian languages that have 'switch-reference'
markers, or to African languages which distinguish coreferential and
non-coreferential 3rd person pronouns in indirect speech.
Cf. for example Hopi: 15
(31) pam navoti:ta (pam) mo:titani -q
he thinks (he) win -NONCOR
Hei thinks that he 2 will win.
(3Γ) pam navotirta (pam) mo:titani -qa?e
he thinks (he) win -COR
He! thinks that he! will win.

Kera (Chadic):
(32) A wäate miniti a k<5or6
Shei said that she2 leave
She! said that she 2 would leave.
(32') Α wäate minti tä köor6
Shei said that she] would leave.
156 Karen Ebert

Pronominal or verbal marking of coreference is found in a great number

of languages, but I have never encountered the inverse marking with
questions as in Newari. I suppose that other languages of the Tibeto-
Burman family with coreference markers function in a similar way, but I
have no data on indirectly reported speech in any of them.

5. Conclusions. The idea that languages make a clear distinction be-

tween direct and indirect speech is for the most part a grammatical
fiction. In normal conversation, direct speech may fade into indirect and
vice versa. A construction like (33) in colloquial German16
(33) ... und dann hat er gesagt [du bist krank]
and then he said you are ill

is, without further context, just as ambiguous as are corresponding

constructions in the Nepal languages:
a) und dann hat er gesagt: "Du bist krank."
b) und dann hat er gesagt, daß du krank bist.

There is nothing in sentence (33) that indicates which interpretation is

intended.
The distinction between direct and indirect speech is probably not a
universal phenomenon. Many languages present speech or thought in a
manner that would count as 'direct', i.e. with deictics referring to the
reported situation. A deictic switch is restricted to the special case,
where a hearer is informed about a statement that was made about
himself or the actual speaker. This shift is possibly a universal in the
sense that if a language allows a partial switch to the speaker's perspec-
tive, then it will (preferably) be a switch from 3rd —» lst/2nd pronouns
rather than the other way round.
If we take a deictic switch to be the crucial criterion for indirect
speech, we might say that languages like Chamling and Nepali report
speech indirectly in very limited contexts. They prefer always to take the
perspective of the person who is speaking or thinking in their report.
This principle is outruled only by pragmatic considerations of empathy
hierarchy: The natural viewpoint is always with speech act participants.
Other languages like German, English, Newari, or Kera frequently
use indirect reports, i.e. the speaker does not change his perspective.
The embedding context remains the same in Newari and Kera (and often
also in German and English, as the complementizer is optional), but
certain elements in the reported speech may indicate indirectness, be it a
 Reported speech in some languages of Nepal 157

pronominal form (Kera), a verbal coreference marker (Newari) or the

mood/tense (German, English).
In talking about deictic switches and indirectness we usually assume
the original speech act to be basic and indirect speech to be a modifica-
tion in which the speaker has to change the perspective. In many,
probably the majority of cases, however, there is no original speech act.
Rather the speaker/narrator invents or infers what a person may have in
his mind. It is when the narrator presents another person's cognitive act
as 'direct speech' that he has to change his perspective: he has to put
himself in the other person's place. Thus from a pragmatic point of view
it is the 'direct' speech that is shifted.
The term 'direct speech' is also a misnomer because we do not
normally understand 'speech' to include 'inner speech', as do Chamling,
Nepali and many African languages. What is communicated in reported
speech is:
cognitive acts (or states)
a) presented from the speaker's/narrator's perspective (= indirect
speech)
b) presented from the perspective of the character in the narrative
(= direct speech).

Less common symbols used

CON J conjunctive suffix

COR preferential marker
EX excluding hearer
FIN sentence final particle or suffix
HON honorific
MOD modal
REL relational prefix (sometimes corresponding to a possessive pro-
noun)

Notes

1. For German see example (33) below.

2. All Chamling examples with the exception of sentence (4) are taken from folk tales I
collected during my field work in Nepal. Nothing has been published on Chamling so
far except for a few notes in Grierson's 'Linguistic Atlas of India'.
3. Agreement is always with 1st 2nd pers., either with subject or with object, and
158 Karen Ebert

sometimes with both, following a complicated pattern that is related to a split ergative
marking and a directional system.
4. A simultaneous interpretation of the conjunctive is often possible, e.g. in example (2):
When they shouted, one woman heard it.
5. Ebert (1979), Ch. 8.2.1.
6. Lord (1976); cf. also the construction i tink se found in many Pidgins and Creoles.
7. ta-ps-a-ko "you have gone"
2nd-go-PAST-PART
The adaptation of the agreement-markers is of course a consequence of the pronominal
switch.
8. Most of the Nepali examples have been adapted from introductory textbooks (mainly
from Clark, 1977). They therefore show a much simpler structure than the natural
Chamling data. The same holds for the obviously constructed Newari data in part 3.
9. I would like to thank W. Winter for making available the Chamling data collected in
the "Linguistic Survey of Nepal"-Project.
10. I owe the information on Indian languages to S. Sharma Peri, on Farsi to Farah
Sharafat. It was confirmed by a short glance at grammars of several Indian languages.
11. The Amharic examples are from Gragg (p. 78), the Shona and Turkish examples from
Roncador (p. 19).
12. The Newari data are from Hale (1971).
13. Hale uses the terms conjunctive and disjunctive instead, which I do not find very
enlightening. Moreover, I have adopted the term conjunctive for the Chamling suffixes
that correspond to the form that is called 'conjunctive participle' in Nepali grammar.
Keenan's 'same/different subject' will not do either, as in questions the Newari suffixes
are not referring to the subject in an obvious way (cf. 28-30 and footnote 14).
14. Questions can be understood as directives embedded into the situational frame YOU
TELL ME
for (28b) YOU, TELL ME [you, came?]
for (29) I TELL YOU [he asked him, [YOU, TELL ME [he2 came?]]]
for (30) I TELL YOU [he asked him, [YOU, TELL ME [you, came?]]]
15. See Keenan (1976, p. 83); for data on American Indian languages cf. also Winter
(1976). Coreferential pronouns in African languages have been described by Hagdge
(1974); for data on Kera see Ebert (1978, Ch. 8.2.1.); for Yoruba see Bamgbose (this
vol.).
16. In standard German the message would be disambiguated by using the subjunctive in
the indirect version:
(33') und dann sagte er, du seist/wärst krank.

References

Clark, T. W. (1977), Introduction to Nepali. A First-Year Language Course. London.

Ebert, Karen, H. (1979), Sprache und Tradition der Kera (Tschad): Teil III: Grammatik.
Berlin: Reimer. (Marburger Studien zur Afrika- und Asienkunde, Serie A: Afrika, Band
15).
Gragg, Gene B. (1972), Semi-Indirect Discourse and Related Nightmares. Papers from the
Eighth Regional Meeting Chicago Linguistic Society: 75-82.
 Reported speech in some languages of Nepal 159

Grierson, George (ed.) (1909), Linguistic Survey of India. Calcutta: Superintendent of

Government Printing. Vol. III.l.
Hag£ge, Claude (1974), Les pronoms logophoriques. Bulletin de la Sociiti de Linguistique
de Paris 69: 287-310.
Hale, Austin (1971), Person Markers: Conjunctive and Disjunctive Verb Forms. In:
Topics in Newari Grammar. Kathmandu: Summer Institute of Linguistics, PM1-PM12.
Keenan, Edward L. (1976), The Logical Diversity of Natural Languages. In: Stevan R.
Harnard/Horst R. Steklis/Jane Lancaster (eds.), Origins and Evolution of Language and
Speech. Annals of the New York Academy of Sciences. Vol. 280, New York, 73-91.
Lord, Carol (1976), Evidence for syntactic reanalysis: from verb to complementizer in
Kwa. In: S. B. Steever/C. A. Walker/S. S. Mufwene (eds.), Papers from the Parasession
on Diachronic Syntax, April 22, 1976. Chicago: Chicago Linguistic Society. 179-191.
Meerendonk, Major M. (1949), Basic Gurkhali Grammar. Singapore: Sen Wah Press
& Co.
Roncador, Manfred von (1980), Gibt die Redewiedergabe Rede wieder? L.A.U.T., Series
A, Paper No. 71. University of Trier.
Winter, Werner (1976), Switch-reference in Yuman languages. In: M. Langdon/S. Silver
(eds.), Hokan Studies. Papers from the First Conference on Hokan Languages, San
Diego, California, April 23-25, 1970. Den Haag: Mouton (Janua linguarum, series
practica 181.) 165-174.
Direct and indirect speech in Japanese*
Florian Coulmas

1. Introduction

Japanese is a postpositional VO-language of the agglutinative type.

Syntactic relations are expressed by case particles such as: -wa, topic,
-ga, subject, -o, direct object, -ni, indirect object or locative/dative, and
-de, instrumental. Verbs are marked according to whether they occur in
sentence final or non-final position. Person, number, and gender are not
marked. There are two tenses, past and non-past, but it is arguable that
Japanese is not a tense language at all, but rather an aspect language
where past tense at verbs and adjectives indicates a perfective or factive
aspect. Sentence mood is expressed by suffixes and sentence final
particles such as: -yo, emphatic declarative; -ka, interrogative; -tai,
desiderative; -yo, probability; etc. Sentence complementation for the
purpose of speech reporting is quite straightforward as the word order is
the same of main clause and embedded clause. Yet reported speech in
Japanese has been a controversial topic for some time.
It has been claimed repeatedly that no distinction is made in Japanese
between direct and indirect discourse (see for instance McClain 1981:
75).1 While this is hardly true, it is interesting to examine the reasons
that might give rise to such a claim.

2. Narrative style

Japan looks back on a long literary tradition. In early writings, neither

words nor sentences were clearly demarcated. Diacritic marks for quota-
tion were unkown. In classical Japanese literature a narrative style was
quite common where the author of what was being related was not
always clearly indentified.2 As R. A. Miller (1982: 89) puts it, "reported

* This paper overlaps in part with Coulmas 1985. Some sections have been taken from the
earlier paper, others have been revised, and yet others are new.
162 Florian Coulmas

conversation - what one character is saying to another, according to the

author - is linguistically and grammatically intermingled with narrative
per se - what the author is saying about the characters, the scene, and
the situation - in such a complex way that these two strands of the text -
conversation on the one hand and narration on the other - more often
than not fuse into a single amorphous entity."
Similarly, linguist Shigehiko Toyama explains: "Generally, in the
classical literature of our country, a mixed form of direct and indirect
speech was used which contributed to the pecularity of its style" (1981:
152).3 As an example he cites the following two sentences from the
Manyöshü collection of poems:

(la) Otori no hagai no yama ni na-ga kouru imo wa

mountain on you Sub love woman Top
imasu to hito no ieba
is Com man Sub say-Cond4
If someone says, "on Mount Hagai in Otori is the woman you
love," . . .
b) Otori no hagai no yama ni wa-ga kouru imo wa
mountain on I Sub love woman Top
imasu to hito no ieba
is Com man Sub say-Cond
If someone says that the woman I love is on Mount Hagai in Otori

According to Toyama (la) is more like a direct quote while (lb) is

indirect speech. "However," he continues, "to make a distinction on the
basis of the linguistic form is difficult."3 In other words, an indirect
reading for (la) and a direct reading for (lb) cannot be ruled out. This
may be one reason why it has been questioned that there is a clear
distinction between direct and indirect speech in Japanese. Still, the
evidence is not compelling.
Part of the problem stems from the fact that without further contex-
tual information it is impossible to determine whether the reporter of the
two sentences (la) and (lb), hito (man), is a particular person or a
general, unspecified source, as the difference between definite and
indefinite noun phrases is not necessarily marked in Japanese. In the
former case the subject of the matrix verb ieba (if say) and the subject of
the verb kouru (love) in the embedded sentence in (lb) could be
referentially identical: "if someone says: O n Mount Hagai in Otori is the
woman I love.'" Wa-ga and hito cannot be coreferential in the indirect
reading of (lb) which is more likely. Accordingly, the official English
translation of the Manyöshü chose an indirect interpretation for (lb) and
 Direct and indirect speech in Japanese 163

an unspecific reading for hito: "Men tell me that my wife is in the

mountains of Hagai . . . " In order to determine whether (la) and (lb)
should be interpreted as direct or indirect speech, the referential identifi-
cation of the personal pronouns wa (I) and na (you) is crucial. The
decision for an indirect reading of (lb) in the translation is based on
contextual rather than linguistic information: The verse in question is
part of a song whose narrator mourns his wife's death. The problem at
hand is thus not really peculiar to Japanese.
(2) He says the woman I love is on Mount Hagai.
An indirect reading - assuming a suppressed complementizer - seems to
be more likely for (2), but a direct reading cannot be ruled out. A
decision for either one presupposes that the reference of the first person
pronoun can be established, that is, that contextual and situational
information is available. This kind of direct-indirect ambiguity can be
observed in many languages, and its occurrence does not allow the
conclusion that indirect speech is not coded in the grammar of a given
language, let alone that it cannot be expressed.
Modern written texts are usually self-sufficient, in the sense that they
can be understood if read. This, however, was not always the case. It is a
typical feature of literary works of cultures in transition from orality to
literacy that they presuppose more of a common context of situation
than can be reconstructed from the text alone. This is so because literary
works were then communicated to the public orally, rather than visually
by being read by individuals not sharing the Zeigefeld with the author or
reciter. This is also true for early Japanese literature, a fact which may be
partly responsible for its vagueness and for referential ambiguities of the
kind mentioned above.As a translator of classical as well as modern
Japanese literature, Edward Seidensticker has been concerned with
these problems vere directly. "Because the monogatari [tale] was being
read aloud," he explains, "meaning could be conveyed by non-verbal
means, such as emphasis and tone of voice. If the author was the reader,
then there could be explanation along the way, in response perhaps to
questions, which did not then find its way into the text" (Seidensticker
1976: 55). Similar observations have been made about early Greek
Literature, 5 where situationally bound expressions were used pervasively
despite the lack of background information necessary for their appro-
priate interpretation. Thus the problem with sentences such as (la) and
(lb) appears to be one of genre rather than Japanese, and Toyama's
conclusion that quotation in Japanese makes use of a kind of reported
164 Florian Coulmas

speech which is neither indirect nor direct but "medium" (chükan

waho), while being correct for some instances - just as it is in other
languages - is hardly valid as a general statement about quotation in
Japanese. Let us now consider some of the characteristics of reported
speech in Japanese in more detail.

3. Complementation

In Japanese (as in any other language), sentences can be quoted directly

or indirectly. In both cases they are marked with the complementizer
particle to and embedded in a matrix sentence with a verb of saying,
usually following the quote. The particle to was originally reserved for
reporting another speaker's statement. In modern Japanese it has be-
come a general complementizer with a much wider application. As a
quotative particle to occurs in combination with verbs of saying and
thinking, such as iu (say), kataru (relate), kiku (ask), omou (think), etc.
Some examples illustrate this use.
(3) Tarö wa Harumi ο nikundeiru to itta.
Taro Top Harumi Akk hate Com said
Taro said that he hated Harumi.
(4) Yamada-san wa kyö byöki da to kikimashita.
Mr. YamadaTop today sick is Com heard
+hon
I heard that Mr. Yamada is sick today.
(5) ashita wa ii tenki ni naru to omou.
tomorrow Top good weather Dat become Com think
The weather will be good tomorrow, I think.

The quotative particle to in conjunction with the verb iu (say) can also
be used as a noun complementizer as in Kimiko to iu onnanoko (a (the)
girl called Kimiko). Any expression that is metalinguistically mentioned
rather than used can occur in this context. The basic pattern is as follows,
(6) x toV

where X can be a sentence, a clause, a word, or an onomatopoetic

expression. The verb can be finite or non-finite. It is usually a verb of
communication or thought, but when X is an onomatopoetic expression
other verbs can also be used.
 Direct and indirect speech in Japanese 165

(7) Dotsun-dotsun to sumoo-san wa kaidan ο nobotte kita.

Com wrestler Top stairs Akk climbing came
The sumo wrestler climbed up the stairs with heavy steps.

In (7) the "matrix" verb is noboru (climb) and the complement is the
sound-imitating word dotsun-dotsun. Words of this sort are quite com-
mon in Japanese. They can be used as adverbs as they are, but someti-
mes the complementizer to ist added. This latter usage underscores the
special character of these words. They are self-referential inasmuch as
they mean by virtue of their form and this makes them quotation-like
expressions.
X in (6) can also have the form " x to V". Thus, to can also serve
as a complementizer for recursive embedding.

(8) John wa [Mary wa [Pete ga kuru] to yü] to itta.

John said (that) Mary said (that) Pete will come.

An important question about to is whether there is any structural

justification for distinguishing the complementizer to from the quotative
particle to. To put it differently, is the verbatim quotation occurring
within the scope of the quotative particle to a part of the matrix sentence
in the same sense as the complement clause is a part of the sentence in
which it is embedded, or is its status relative to the matrix sentence
different? It has been argued for other languages, for instance by Partee
(1973) for English, that the direct quote is not part of the reporting
sentence. Rather, a sentence such as (9) has to be analyzed as two
sentences, (9a) and (9b).
(9) Mary said: "Pete will come."
a) Mary said b. Pete will come

This analysis is not too attractive, because Mary said is not a sentence.
A complement is necessary that replaces the quoted sentence: Mary
said But this dummy symbol approach is quite counterintuitive in
many instances. Yes in (10), for example, would have to be considered as
not being a part of the sentence.
(10) Mary said yes.

For Japanese, such an analysis that recognizes the complement clause as

part of the matrix sentence while claiming that the quoted sentence does
not belong to the sentence in whose context it occurs is also very
counterintuitive, because many sentences occurring as a complement
clause in the context of to verbum dicendi can be used
166 Florian Coulmas

independently as direct speech. This is not to say that indirect speech

cannot be distinguished from direct speech. Direct speech does have a
number of characteristic features, but stripping a sentence of these
features does not necessarily turn it into a dependent clause which can
occur only as a complement of a matrix sentence. Conversely, not every
complement sentence is clearly marked as a dependent clause. Some,
however, are. Consider the following examples.
(11a) Ashita made ni kono shigoto ο yatte kudasai
tomorrow until this work Acc do please
to kare wa iimashita.
Com he Top said
He said, "Please finish this work by tomorrow".
b) Ashita made ni kono shigoto ο yarn yöni
tomorrow until this work Accdo thus
to kare wa iimashita
Com he Top said
He told me to finish this work by tomorrow.
The word order (in 11a and b) is the same. This is not coincidental but
rather a regular feature typical of Japanese and other Altaic languages
such as, e.g., Korean. (11a) is marked as a direct quote as the form
kudasai requires direct address. It is one of a set of directional verbs
which is inherently deictic. As the honorific form of the verb kureru
(give) its conditions of use require the addressee to be the subject. Thus,
the "giver" cannot be the speaker, but is always the addressee while the
former is the recipient.
In indirect speech, therefore, this form is replaced by yöni, which
literally means "thus". Yöni is a shortening of yöni shite kudasai ("please
do in this manner") now chiefly used as an idiomatized marker of the
imperative and hortative moods in indirect speech. Unlike kudasai, yöni
cannot occur in sentence final position, and hence the embedded clause
in (lib) is clearly marked as indirect speech. Ashita made ni kono
shigoto ο yarn yöni could not have been the original utterance that is
reported in (lib), nor could it function as an independent utterance in its
own right. But many embedded clauses can. The complement clauses in
(3)-(5) can all be used as they are as independent declaratives. Other
moods are somewhat more difficult.
Moods are most commonly indicated in Japanese by sentence final
particles. To some extent they serve a function as illocutionary force
indicating devices and as hedges for intensifying or downtoning the
speaker's degree of certainty about the statement. These sentence final
particles are thus characteristic of direct speech.
 Direct and indirect speech in Japanese 167

(12a) Hoteru no heya kara nigeta hitobito wa 'matatakuma

hotel Gen room from escaped people Top immediately
ni kuroi kemuri ga jümanshita' to katatte iru.
black smoke Sub filled Com telling are
The people who escaped from the hotel are saying that the rooms
filled with black smoke immediately.
b) Hoteru no heya kara nigeta hitobito wa 'matatakuma
hotel Gen room from escaped people Top immediately
ni kuroi kemuri ga jümanshita yo' to katatte iru.
black smoke Sub filled ! Com telling are
The people who escaped from the hotel said: "The roomsfilledwith
black smoke immediately!"
The embedding in (12a and b) is identical except for the emphasis
particle yo attached to the quote in (12b) but not in (12a). The former is
thus marked as a direct quote. In colloquial speech sentence final
particles are extremely common, but their absence from a quote cannot
be taken as evidence for its indirectness. (12a) is ambiguous as to
whether it is direct or indirect speech.

4. Speaker-addressee relationship

As there are no changes in word order or tense, the distinction between

direct and indirect speech in Japanese is based on deictic switches, on the
one hand, and switches concerning speaker-addressee relationship and
discourse organization, on the other hand. These two kinds of switches
interact since the honorific systems serves both relational and deictic
functions (cf. Coulmas 1982).
Japanese has a highly differentiated honorific system. The use of
honorific terms is determined by and to some extent influences the
relationship between interlocutors; they are bound to the speech situa-
tion, their function being pragmatic rather than semantic. They cannot,
therefore, easily be taken out of their original context of utterance and
put into another, unless the new context is structured in such a way that
it evokes the original speech situation. It follows that honorifics are not
likely to occur in indirect discourse.
Take, for instance, the polite auxiliary verb masu. It is attached to the
infinitive -i- of other verbs (e.g., hanasu (speak) 4- masu > hanashi +
masu, hanashimasu) and usually occurs at the end of a sentence, rather
168 Florian Coulmas

than in embedded clauses. Its function is to mark one's style as polite.

Thus, -masu may be regarded as a stylistic marker operating on linguistic
units that contain a verbal element and can function as complete
utterances. It seems plausible, therefore, that -masu should not occur in
subordinate clauses which could not constitute functionally complete
utterances. Yet the use of -masu is less restricted than would be expected
from these considerations. Actually, it quite commonly occurs in subor-
dinate clauses of conditionals such as in (13), where the conditional is
like an implicit piece of advice.

(13) Kono ronbun ο yomimasu to mondai no muzukashisa ga

this article Acc read + hon when problem Gen difficulty Sub
wakaru yöni narimasu
understand will 4- hon
When you read this article you will understand the difficulty of the
problem.

Yomimasu in (13) could be replaced by the plain form, yomu, and the
sentence would still be polite, as the matrix verb, narimasu, is so
marked; in the present context, however, yomimasu is just as natural.
The politeness function of the auxiliary-masu is sufficiently carried out
when it is attached to the final verb of a sentence, but it is syntactically
possible and in certain contexts quite common to use it in non-finite
positions.
Whether or not the masu - form is used at all depends on the relation-
ship between the interlocutors and on whether the speaker wants his
speech to be polite. It is never grammatically nessary. Clearly, then, this
form is a good candidate for an element that should not occur in indirect
speech, because (a) its every occurrence in an utterance belongs to the
speech of the actual speaker, and (b) in indirect speech it would occur in
a subordinate clause. Nakau (1973: 88) in his dissertation on sentential
complementation in Japanese therefore states that embedded sentences
containing a masu-iorm are direct quotes. This claim is quite plausible,
for the reasons give above, but it seems to be only a rule of thumb.
Consider the following example.

(14) Kare wa watashi wa mata machigaimashita to iimashita.

he Top I Top again wrong was Com said

In this sentence, the embedded clause watashi wa mata machigaimashita

("I was wrong again") contains the past tense form of masu, mashita.
Thus, if Nakau's claim was correct, it should be a direct quote, in which
 Direct and indirect speech in Japanese 169

case the two pronouns kare (he) and watashi (I) would have the same
referent. The question now is whether watashi could also refer to the
speaker of (14), in which case the embedded clause could not be a direct
quote, because the speaker of the original utterance who is referred to by
the third person pronoun kare, the subject of the matrix sentence, could
not have used the first person pronoun in referring to the person who
"again made a mistake." If this reading, where watashi refers to the
speaker of (14), were possible, (14) would be ambiguous with both (14a)
and (14b) as possible translations.
(14a) He said: "I was wrong again",

b) He said that I was wrong again.

Sentence (14) is not a very nice sentence, but it is grammatical and

acceptable. The reason why it is not a very nice sentence is not so much
the occurrence of a polite auxiliary in the embedded clause as the use of
the two pronouns which make it sound somewhat outlandish. Even
though person is not marked at the verb in Japanese, personal pronouns
are used relatively infrequently, and sentences leaving the subject or
object position unfilled are not felt to be incomplete, provided the
necessary information is easily inferred from context or situation. Given
the appropriate circumstances where the agents of the saying and the
erring can be inferred, (15) would be more natural as an utterance than
(14).

(15) Mata machigaimashita to (kare wa) iimashita.

again wrong war + hon Com he Top said + hon

The preferred reading of (15) is clearly direct quotation, but for some
speakers an indirect reading becomes possible when iimashita is replaced
by a respectful verb such as ossharu (say 4- honorific, where the subject
is the addressee). It seems that the politeness marker -masu can then be
used in the subordinate clause without enforcing a direct interpretation.

(16) Mata machigaimashita to osshaimashita.

again wrong was + hon Com said (respectful)

The masu-iorm in the embedded clause could be taken as making the

whole utterance yet more polite rather than indicating that the embed-
ded clause, mata machigaimashita, is direct discourse. In this reading,
the speaker of (16) is the subject of machigaimashita, and the addressee
is the subject of osshaimashita: "You said that I was wrong again." It
170 Florian Coulmas

seems, therefore, that under certain circumstances -masu can occur in

the complement clause of indirect speech, although it cannot be denied
that the direct reading is usually the preferred one in cases where -masu
is used in such a context.

5. Personal pronouns

Notice that there is no pronoun in (16). As stated above, this is very

common in the spoken language, since directional verbs, such as kuda-
saru, and honorific verbs, such as ossharu, are restricted with respect to
the subject and its role in the speech situation: It cannot be the speaker.
There are other cases, however, where the subject can only be inferred
from context. Let us consider the following example:
(17) Ganbareba wakaru yöni naru to kare wa itta.
try hard if understand become Com he Top said

Sentences (17a-f) are possible translations of (17).

(17a) He said if I'd try hard I would understand.
b) He said that if you'd try hard you would understand.
c) He said that if she tried hard she would unterstand.
d) He! said that if hej tried hard hej would unterstand.
e) He! said that if het tried hard they would understand.
f) He! said that if he 2 tried hard they would understand.

The above renditions of (17) presuppose different speaker-addressee

relations and different relations between the original speaker, the re-
porting speaker, and the respective agents of the two embedded verbs.
The speaker of (17a), for instance, must have been the agent of both of
the verbs in the embedded clause in (17) and is likely to have been the
addressee of that sentence. Similar changes in speaker perspective are
presupposed in (17b-f). All of the above sentences are phrased in
indirect speech, however; direct interpretations as in (17g and h) are also
possible.
(17g) He said: "If I try hard I will understand."
h) He said: "If you try hard you will understand."
 Direct and indirect speech in Japanese 171

The direct interpretation is possible because the embedded clause of

(17), ganbareba wakaru yöni naru, would make a natural utterance by
itself, although some sentence-final particle such as, e.g., wa, yo, or ne,
or the presumptive form of the copula, desho, would then be likely to be
attached.

6. Sentence-final particles

As pointed out above, sentence-final particles are common in colloquial

speech, but they are not always grammatically necessary or stylistically
required. Hence, while their presence distinguishes direct quotes, their
absence does not distinguish indirect speech. To put it differently, most
sentence-final particles are indicative not only of the sentence mood but
also of the speaker's cognitive and emotional state or attitude towards
the proposition and/or his interlocutor. They are bound to the speech
situation and have nothing to do with the meaning of a sentence, which is
what is typically reported, or partially reported, when another's utter-
ance is conveyed in indirect speech. They are phatic rather than refer-
ential elements of speech and therefore cannot occur in indirect speech.
The embedded sentence in (18) is therefore unequivocally direct speech.
(18) Otokonohito wa kono koto wa naisho yo to
man Top this matter Top secret ! Com
itte so iu tokoro de hanashi ga owarimasu
saying such place at story Sub end
The story comes to an end where the man says: "This matter is a
secret!"
The only exception to the general principle that sentence-final partic-
les imply direct speech is the particle ka indicating the interrogative
mood. It may occur in indirect speech as well as in direct speech, a fact
which can be attributed to the absence of changes in word order in both
questions and complement clauses. An example of the interrogative
particle ka in indirect speech is provided in the second sentence of
example (19) discussed below.
172 Florian Coulmas

7. Speaker perspective

Before turning to examples (19)-(21), consider again the main points

emerging from the above observations:
(i) Grammatically, the embedding of complement clauses and direct
quotations within the scope of the quotation particle to is identical,
because complement clauses do not undergo any changes in tense or
word order. Verbatim quotes, insofar as they are grammatically
well-formed, are constituents of the matrix sentences into which
they are embedded. There are no grammatical reasons to think
otherwise.
(ii) In all sentences lacking honorific forms in both matrix and comple-
ment clause, honorific switches cannot be relied upon for distin-
guishing direct from indirect speech.
(iii) The presence of honorific verb forms is not always a reliable feature
identifying direct quotation, since they can also occur, under certain
circumstances, in subordinate clauses in indirect speech.
(iv) If no deictic pronouns for person or object reference are used, no
deictic switches indicating speaker perspective are necessary in
indirect speech. Subjectless sentences are common, and use of
personal pronouns is often avoided.
Taken together, these conditions contribute to blurring the distinction
between direct and indirect quotation or, rather, to creating relatively
many Japanese sentences that could be given a direct as well as an
indirect reading.6 Especially in colloquial speech, lack of honorifics and
scarce usage of pronouns often coincide. Hence, when colloquial speech
is reported in colloquial speech, the distinction between direct and
indirect speech is often lost or, more precisely, fails to be formally
marked. However, there is a great variety of means indicating speaker
perspective: directional and respectful or humble verbs, other lexical and
morphological honorifics, personal pronouns, deictic demonstratives.
For illustration of the latter kind consider the following data. (19) and
(20) are reports of two different speakers who were shown the same
segment of a video-taped TV-drama and were then instructed to relate as
accurately as possible what they had seen and heard to a third person.
They were not aware of the purpose of their report. (21) is the transcrip-
tion of the actual dialogue in the film.7
(19) Otokonohito ateni futö ga todokerareru no
man addressed to envelope Sub send Pass
 Direct and indirect speech in Japanese 173

desu ga, sore ο onnanohito ga uketoru wake desu ne . . .

is but that Acc woman Sub receive case is
Sono fütö ο mottekita hito wa dare ka to kikimasu.
that envelope Acc brought person Top who Inter Com ask
Soshite onnanohito wa kotaete . . .
then woman Top answering
(20) Otokonohito ga onnanohito ni Yuki-chan kono fütö ο
man Nom woman Dat this envelope Acc
mottekita no wa donna otoko datta to kö kikimasu.
brought Top what kind man was Com thus ask
(21) Yuri-chan sä, kono fütö ο mottekita yatsu donna
this envelope Acc brought guy what kind
otoko datta.
man was
Hey, Yuri, what kind of a man was the guy who brought the
envelope?

Taken by itself, the second sentence in (19) could be either direct or

indirect speech, but given the context in which it occurs, it can only be
indirect. The demonstrative sono (that) is anaphoric, its use being
justified by the prior mention of fütö (envelope) to which it refers. This
reference reflects the reporting speaker's perspective rather than that of
the original speaker, whose utterance is given in (21). In (21) the original
speaker uses kono instead of sono since he is acutally holding the
envelope in his hand while asking the question. Kono is an adnominal
demonstrative used for referring to something close to the speaker.
Sono, in contrast, is either anaphoric or used deictically to refer to
something close to the hearer. In (20), where the embedded sentence is
clearly marked as a direct quote by the direct address Yuki chan, the
original kono is thus reproduced as kono, while a change of the speaker
perspective leads to its replacement by sono in (19).
Yet another feature marking the embedded clause in (19) as indirect
speech is the interrogative particle ka. As pointed out above, this
particle may occur both in direct quotes and in indirect speech. How-
ever, its use without the copula - as, for instance, in ano hito wa dare ka,
"Who's that?" - is stylistically very harsh and not likely to occur in finite
position in spoken discourse. Rather, in colloquial speech, in order to
mark an utterance as a question, ka is attached to the copula desu or
replaced by rising intonation as in (21), donna otoko datjtä. The use of ka
without the copula in the second sentence of (19), therefore, suggests
that the question reported there is meant to be indirect speech.
The speaker of (20), on the other hand, takes pains to make it clear
174 Florian Coulmas

that his report is a direct quote. Not only does he repeat the name of the
addressee, Yuki, which is stylistically somewhat awkward and unneces-
sary as it is referentially identical with the descriptive term onnanohito ni
reflecting his, i.e., the reporter's perspective; he also emphasizes the
faithfulness of this report by modifying the matrix verb kikimasu (ask)
with kö (thus, in this manner). Several other adverbial expressions can
be used in a similar way, indicating relative closeness to the original
utterances and/or confidence in one's own report. In (22), for example,
which is taken from the same data, iufüni (like this) is used to indicate a
somewhat lesser degree of faithfulness and confidence.
(22) Onnanohito wa sono shashin ο mite 'kore wa
woman Top that photograh Acc seeing this Top
köhökachö ga itte ita suküpu no shashin de wa
publicity manager Sub said scoop Gen photograph is
nai ka' to iu funi iimashita.
not Inter Com like this said.
When the woman saw the photograph, she said (something like)
"Isn't this the scoop photograph the publicity manager was talking
about?"

Notice that in this sentence, as in the case of (20), the speaker perspec-
tive and the directness of the quote are most clearly indicated by deictic
switches. In (22) the same photograph is referred to once as sono shashin
(that photograph) and then as kore (this). Sono is anaphoric, referring to
something recently mentioned; thus, if kore, the individual demonstra-
tive referring to something close to the speaker, is used by the same
speaker to refer to the same object, there must have been a change in
speaker perspective.
Deictic switches are the most reliable indicators of speaker perspec-
tive. This is generally true, and it is true of Japanese. However, in the
absence of grammatical switches that unequivocally determine in all
cases whether a passage of reported discourse represents direct or
indirect speech, their function is particularly important. On the basis of
their knowledge of stylistic levels and socially conditioned honorific
speech, speakers of Japanese can make reasonable guesses about wheth-
er or not an embedded sentence ascribed to another speaker is likely to
be a verbatim rendition of his actual utterance. Such guesses rely more
on an understanding of the social factors determining the choice among
stylistic registers than on grammatical rules. Yet the fact that this kind of
information is not always available and the infrequent use of pronouns
often combine to leave the question unresolved whether some other
 Direct and indirect speech in Japanese 175

speaker's speech is reported as a direct quote or in indirect speech.

Presumably, this is one reason why the distinction between direct and
indirect speech in Japanese is said to be vague and gradual by some
Japanese linguists. Traditionally, indirect speech was not an important
concept of Japanese grammar and, (it might be added without indulging
in too much hazardous speculation) it has never been an important
cultural notion to distinguish verbatim quote and indirect report.
It should be noted, however, that referential ambiguities of pronouns
and the ensuing failure to identify decontextualized sentences as direct
or indirect speech is a very common problem of reported speech not
restricted to Japanese. For example, (23) is just as ambiguous as (24).
(23) Mr Tanaka said that he will go to Kyoto.

(24) Tanaka- san ga Kyoto ni iku to iimashita.

Mr. Sub to go Com said + hon

In one interpretation he is coreferential with Mr. Tanaka, in which case

the embedded clause is phrased in indirect speech, whereas a direct
reading is possible only if he refers to a person other than Mr. Tanaka.
The situation is quite similar in Japanese. Kyoto ni iku is the elliptical
embedded sentence with the subject left out. An interpretation where
the subject of iku and the subject of iimashita are the same, namely
Tanaka-san, is likely, but, given an appropriate context, an interpreta-
tion with different subjects for matrix and embedded verb is also
possible. This kind of ambiguity in decontextualized sentences does, of
course, not mean that in normal conversation it would be unclear
whether it is Mr. Tanaka or somebody else who goes to Kyoto.
Usually, unequivocal interpretation of deictic pronouns is possible on
the basis of contextual information, and, accordingly, it can be determi-
ned whether a given embedded clause should be interpreted as direct or
indirect speech.

8. Conclusion

While, structurally, many sentences are ambiguous in Japanese with

respect to the interpretation of the embedded clause as direct or indirect
speech, such ambiguities are, for the most part, automatically resolved in
176 Florian Coulmas

natural discourse. Notice that this kind of ambiguity is characteristic only

for complement constructions of the form discussed here.
(25) s to verb of communication

Many honorific lexical items of the relational kind often compensate

the absence of pronominal or nominal subjects and thus serve a disambi-
guating function. Still, it is quite common for sentences of this pattern
that they do not allow a definite classification as direct or indirect
discourse.
Other complement constructions used for conveying another spea-
ker's message or for marking an utterance as being based on second-
hand information are always indirect. Particularly frequent are the
following three:
(26) s to yü koto da
Com say thing is
Kanojo wa daigaku no sensei datta to iu koto da
she Top university professor was
I understand that she was a university professor.

(27) s to no koto da
Com Gen thing is
A n o h i t o wa Nihone iku to no koto da
that man Top Japan to go
I am told that he is going to Japan.

(28) S sö da
thus is
Kyö uchi ni kuru sö da
today my house to come
I hear he'll come to my house today.

They can all be translated as "I hear/understand that . . . , I am told

that . . . " In (26), koto can be replaced by other nouns such as hanashi
(story), uwasa (rumor), etc., and da by other forms of the copula. The
generalized pattern is thus

(29) s to iu Ν + copula

Another pattern with to iu koto ο as an object-sentence marker is

discussed at length by Maynard (in this volume). (28) has a hearsay
quality, that is to say, the speaker relates second-hand information.
However, in the other patterns, too, the embedded clause is always
understood as being phrased from the reporter's point of view. The
 Direct and indirect speech in Japanese 1T7

utterance reported in (28), for instance, may have been something like
(30) in its original phrasing, depending on when, where, and by whom it
was made.
(30) Ashita otaku ni ikimasu
tomorrow your house to go
I'll come to your house tomorrow.
Lexical switches, such as otaku (your house) —> uchi (my house), have
the quality of deictic switches necessitated by many socio-relational
words. The general tendency in everyday speech is to adapt these words
to the point of view of the reporter whenever they occur in an utterance
to be communicated to a third person. Hence, it seems that there is a
general preference for indirect reports over verbatim quotation in Japa-
nese.

Notes

1. Lewin, for instance, explicitly denies the existence of indirect speech in Japanese:
Das Japanische kennt keine Sätze mit indirekter Rede und auch keine abhängigen
Fragesätze. Jeder Ausspruch wird in der Form der direkten Rede als Zitatsatz gegeben
und unterscheidet sich syntaktisch nicht von den üblichen Satzformen. (Lewin 1975:
213)
Lewin's statement is correct inasmuch as indirect speech is not marked by a subjunctive
mood, a special word order, or any other overt syntactic feature. Nonetheless, distinc-
tions can be made between direct and indirect speech, as is demonstrated in this paper
and is also shown by Maynard (in this volume). See also Endo 1982. Lewin's argument,
it seems, is based on a very narrow notion of indirect speech that is based on its
grammatical encoding in Indo-European languages.
2. Similar observations have been made about other languages, such as, for instance,
classical Greek (Rosier 1983).
3. Ippan ni wa-ga kuni no jodai no bungaku ni wa chokusetsu wahö to kansetsu wahö no
konzai shita keishiki ga okonowarete ite, sore ga buntai no tokushoku ο nashite ita.
Keredomo, katachi no ue kara kubetsusuru no wa konnan dearu. (My translation.)
4. The following abbreviations are used in interlinear glosses:
Top = topic Gen = genitive/possessive
Sub = subject Cond = conditional
Acc — accusative Com = complementizer
Dat = dative hon = honorific
5. See Rosier 1983 for a stimulating treatment of the use of deictic words in oral and
written poetry in classical Greek. Cf. also Eideneier 1982.
6. Given this state of affairs, the assumption that direct and indirect speech are transforma-
tionally related seems to be particularly dubious. Okutsu 1970 tries to analyze the
quotative structure of Japanese along the lines of this assumption. He proposes seven
"transformations" which supposedly "indirectify" the direct quotation. Unless he uses
178 Florian Coulmas

the term transformation in a loose, metaphorical sense it is very hard to understand how
he can subsume such a variety of switches as, for instance, "time adjustment," "deletion
of initial and final," or "first and second pronoun adjustment" under the same heading
of this grammatical notion. The assumption that direct and indirect speech are transfor-
mationally related has been convincingly refuted by Banfield (1973) arguing against
Ross' "performative analysis."
7. This explains the different names, Yuki and Yuri, in (20) and (21).

References

Banfield, A. 1973. Narrative Style and the Grammar of Direct and Indirect Speech.
Foundations of Language 10: 1-39.
Coulmas, F. 1982. Some Remarks on Japanese Deictics. In: J. Weissenborn, W. Klein,
eds., Here and There. Cross-linguistic Studies on Deixis and Demonstration. Amster-
dam: John Benjamins, 209-21.
Coulmas, F. 1985. Direct and Indirect Speech: General Problems and Problems of
Japanese. Journal of Pragmatics 9: 41-63.
Eideneier, Η. 1982. Leser- oder Hörerkreis? Zur byzantinischen Dichtung in der Volks-
sprache. Anatypon apo ta "Ellinika". Tomos 34 os, Tessaloniki, 119-50.
Endo, Hiroko. 1982. Nihongo no waho. Gengo 19, 3: 86-94.
Lewin, B. 1975. Abriss der japanischen Grammatik: Wiesbaden: Harrassowitz.
The Manyöshü. The Nippon Gakujutsu Shinkokai Translation. UNESCO Collection of
Representative Works - Japanese Series. New York: Columbia University Press.
Maynard, S. K. 1985. The Particle -ο and Content-Oriented Indirect Speech in Japanese
Written Discourse. In this volume, 179-200.
McClain, Υ. M. 1981. Handbook of Modern Japanese Grammar. Tokyo: Hokuseido Press.
Miller, R. A. 1982. Japan's Modern Myth. New York, Tokyo: Weatherhill.
Nakau, M. 1973. Sentence Complementation in Japanese. Tokyo: Kaitakusha,
Okutsu, K. 1970. Inyo közö to kansetsuka tenkei (Quotative Structure and Indirectifica-
tion). Gengo kenkyü 56: 1-26.
Partee, Β. H. 1973. The Syntax and Semantics of Quotation. In: S. Anderson, P. Kiparsky,
eds., A Festschrift for Morris Halle. New York: Holt, 410-18.
Rosier, W. 1983. Über Deixis und einige Aspekte mündlichen und schriftlichen Stils in
antiker Lyrik. Würzburger Jahrbücher für die Altertumswissenschaft N. F. Bd. 9).
Toyama, S. 1981. Nihongo no Sugao. Tokyo: Chüo köronsha.
The particle -ο and content-oriented indirect speech
in Japanese written discourse*
Senko K. Maynard

1. Introduction

Traditionally in Japanese, the particle -to and its variation -tte are
considered major linguistic devices for speech and thought representa-
tion. As it is generally assumed that -to marks both "direct" and
"indirect" quotation, little attention has been paid to any other linguistic
device. Contrary to this frequently held assumption which regards -to as
the only device for speech representation, this paper argues that the
object marker -o along with the complementizer koto in the form of
koto-o and -to yuu koto in the form of -to yuu koto-o are also used for
reporting speech and thought in Japanese discourse. Specifically, it will
be argued that the -o clause plays a significant role in representing what
we call content-oriented indirect style. Closely related to the issue of
linguistic devices is the more fundamental question of how direct and
indirect speech and thought representation should be defined and what
functions these two devices play in discourse.
In this paper then, we attempt to clarify the notions of direct and
indirect speech in Japanese but with more attention given to two sub-
groups of the so-called indirect quotation, namely, what we call "direct-
oriented indirect" and "content-oriented indirect representation." In
other words, we are interested in investigating this usually neglected
linguistic device, the particle -o, in association with the traditional direct/
indirect dichotomy and in contrast with indirect quotation marked with
-to. What are the syntactic constraints of the koto-o clause? What are its
semantic implications? Are the verbs that can co-occur with the koto-o
strategy limited to certain types? And ultimately, what motivates a
writer to choose a particular mode of speech and thought representation
in Japanese? In what follows, we attempt to answer these questions.

* This is a revised version of my earlier paper titled "Functions of -to and koto-o in Speech
and Thought Representation in Japanese Written Discourse," 1984.
180 Senko Κ. Maynard

2. Direct, direct-oriented indirect and content-oriented

indirect styles
One of the "common" views in regard to distinguishing different styles
of quotations, and without doubt, one of the simplest is to use grapholo-
gical marks - such as quotation markers r~ _i in Japanese - as its
criterion (see for example Ross (1976-77)). While at first this criterion
may seem feasible, such a distinction in reality serves little in terms of
capturing syntactic characterizations of direct quotation in contrast with
those of indirect quotation. Graphological markings are a useful conven-
tion in delineating the quoted clause from the narrative body and are
frequently used in Japanese written discourse. However, quotations
which carry all of the features of the directly reported clause frequently
appear without graphological markings in fictional writings.1 For ex-
ample, observe the following data.

(1) Sono hanashi no oshimai ni wa daredemo, maa

that talk of end at Theme everyone well
ichido Otsugi-san ο mite-oinaae, sore wa uttui
once Miss Otsugi Obj. go and see that Theme beautiful
ohito yae, to it-te musubu no
person be Quot. say-and conclude Comp(lementizer)
ga tune deari soosuruto kikite wa
Subj. ordinary be-and then listener Theme
tachimachi kyoomi ο tsuriage-rare-te, Hirayama made
immediately interest Obj. increase-Pass.-and Hirayama to
dekakete-shimai, Kaeto onajiyooni, jitsubutsu ga
end up going-andKaeand same actual person Subj.
soozooijooni utsukushii no ni ikkyoosuru
more than imagined beautiful Comp, at be surprised
no deat-ta.
Comp. be-Past
(Ariyoshi 1970: 6)
'Whoever discussed the couple always concluded, "You must see
Otsugi, she's incredibly beautiful." The listener whose curiosity was
aroused to go to Hirayama was generally surprised to find that
Otsugi's beauty truly surpassed the imagination - as it had Kae's.'
(Hironaka and Kostant 1978: 4)

As evidenced in the English translation, the quoted clause invokes the

familiar features of immediacy and authenticity one finds in direct
 Content-oriented indirect speech in Japanese 181

speech - such as regional dialect in mite-oinaae 'go and see' and ohito yae
'(she) is a person.' And yet, in the original Japanese, no graphological
marks are used.
In fact, contemporary fictional writings containing quotations with all
utterance features appear frequently without graphological markings,
especially when the quotation itself constitutes a part of the matrix
sentence, as in (1). Defining direct/indirect quotation in terms of quota-
tion markers, then, fails to capture the general characteristics of the
Japanese language in two ways. One, it is not necessarily a useful
criterion for all instances, and two, syntactic and semantic characteristics
of the quoted clause are not distributed in such a way as to justify a
division based on the criterion.
Unlike Ross, Inoue (1982), following Nitta (1979), proposes a set of
criteria based on the constraints that are imposed on direct and indirect
quotation.2 According to Inoue, linguistic forms such as non-past tense
of psychological and perception verbs, modals of intention, imperative
endings, polite forms etc. are indicators of direct speech, while change of
deictic elements in terms of pronouns, temporal and spatial expressions
along with the tense agreement of the quoted phrase are considered
indicators of indirect speech.
Another set of syntactic criteria to distinguish direct and indirect
quotation is proposed by Kamada (1981, 1983). Kamada argues that
Japanese quotation may be represented by direct quotation, indirect
quotation and semi indirect quotation, and proposes the following
characteristics of indirect quotation (see Kamada, 1981: 98).
1) Pure IQ (Indirect Quotation) is a style of quotation in which all the
viewpoint elements such as pronouns, demonstratives, time/place
adverbs and viewpoint verbs take a form which conforms to the
speaker's point of view.
2) The QC (Quoted Clause) of an IQ (Indirect Quotation) is syntacti-
cally less free than that of a DQ (Direct Quotation) in regard to its
capacity to accomodate modal auxiliaries desu/masu, sentence final
particles like yo and transformations.
As for the semi indirect quotation, Kamada proposes that it can use
these features, and yet grammatically represent not the speaker's but the
narrator's point of view. Kamada's major contribution lies, as in the case
of Inoue (1982), in identifying the general syntactic behavior of direct
and indirect quotations marked by the particle -to.3
Although the criteria proposed by Inoue and Kamada are useful, we
find them inadequate in two important aspects. One, despite these
182 Senko Κ. Maynard

criteria, there are cases where the distinction between direct and indirect
speech in Japnese becomes obscure. And two, in their conception of
direct/indirect speech, they exclude speech representation marked by -o.
With these points in mind, let us now observe sentences (2) through (7).

(2) Otoko wa kenmeini "Kirishitan nanka ja nai-n-da"

man Theme eagerly Christian such Neg.-Comp.-be
to kyoochooshite-i-ru.
Quot. insist-be-Nonpast
'The man insists strongly (by saying), "I'm not a Christian!"'
(3) Otoko wa kenmeini kirishitan nanka ja nai-n-da
man Theme eagerly Christian such Neg.-Comp.-be
to kyoochooshite-i-ru.
Quot. insist-be-Nonpast
'The man insists strongly (by saying): I'm not a Christian!'
(4) Otoko wa kenmeini "Kirishitan dewa-nai" to kyoochooshite-
man Theme eagerly Christian Neg.-be Quot. insist-
i-ru.
be-Nonpast
'The man insists strongly (by saying): "I'm not a Christian.'"
(5) Otoko wa kenmeini kirishitan dewa-nai to kyoochooshite-i-ru
man Theme eagerly Christian Neg.-be Quot. insist-be-Nonpast
'The man insists strongly (by saying), "I'm not a Christian.'"
'The man insists strongly (by saying) that he is not a Christian.'
(6) Otoko wa kenmeini jibun wa kirishitan dewa-nai
man Theme eagerly self Theme Christian Neg.-be
to kyoochooshite-i-ru.
Quot. insist-be-Nonpast
'The man insists strongly (by saying) that he is not a Christian.'
(7) Otoko wa kenmeini jibun ga kirishitan de-nai
man Theme eagerly self Subj. Christian Neg.-be
koto ο kyoochooshite-i-ru.
Comp. Obj. insist-be-Nonpast
'The man insists strongly (by saying) (the fact) that he is not a
Christian.'
(Endo 1981: 113)
Traditionally, sentences (2) and (4) are classified as direct speech with
the quoted section being graphologically marked. However, sentence (3)
which is not signalled by graphological marks is also direct speech by
virtue of the fact that the extended predicate nai-n-da (rather than nai) is
used to reflect the utterance aspect of the quoted clause, and the quoted
 Content-oriented indirect speech in Japanese 183

clause is assumed to carry all of the utterance features that accompanied

the actual deliverance.
Sentence (5) is ambiguous in that it can be interpreted either as a
direct or as an indirect quotation as shown by two English translations.
Thus, wereas in English different surface forms must be chosen, in
Japanese an identical form can serve to present two modes of represen-
tation. If we assume (5) to be an accurate reproduction of (4), (5) is
considered direct speech, whereas if we assume (5) to be associated with
(2), we can interpret (5), as an indirect quotation of (2). The mere
absence of graphological marks does not designate (5) as an indirect
quotation as pointed out earlier.
It should be noted that the ambiguity of (5) is caused in Japanese (but
not in English) primarily because of three syntactic characteristics. First,
there is no verbal tense agreement in Japanese indirect quotation, and
therefore the tense in direct speech remains as it is in indirect speech.
Second, the same quotative marker -to is used for both direct and
indirect marker; and third, the personal pronouns which are often
ellipted in Japanese do no show the deictic shift as required in English.
As shown by this example, the Japanese language presents some cases
where we cannot tell whether a particular sentence represents speech (or
thought) directly on a purely formal basis.
While defining the direct/indirect quotation is not an easy task, for the
sake of argumentation presented in this paper, we propose the following
definition which incorporates Inoue (1982) and Kamada (1981, 1983).4

Direct quotation:
is the quotation marked by -to where the quoted clause is assumed to
carry all the utterance and deictic features of the actual utterance and
where no shift in point of view is involved.

Indirect quotation:
is the quotation marked by -o, koto-o and to yuu koto-o where the
quoted clause conveys the propositional meaning of the quotation and is
not assumed to carry any of the utterance features as well as deictic
elements of the actual utterance. In indirect quotation, the point of view
shifts from the speaker to the person who reports the speech event.

Sentence (6) exemplifies indirect quotation. In the quoted portion, the

reflexive jibun 'self is used instead of watashi Ί . ' While it is possible for
184 Senko Κ. Maynard

the speaker of the quoted clause to refer to himself as jibun 'self during
the actual deliverance of speech, it is also possible to interpret jibun 'self'
as the reporter's relexicalization of the original watashi Ί . ' In this
situation jibun 'self' is used to refer to the man as 'he' in line with the
reporter's viewpoint. Note also that (6) contains nai 'negative non-past
be' in the root form without further utterance features. Again while it is
possible to assume nai was utterred in an actual speech event, in the
context of the novel, it is less likely that nai is actually uttered. 5 We
assume here that the actual utterance contained an extended predicate
such as n(o)-da 'it is that' as in sentence (2).
Sentence (7) which appears in Shuusaku Endo's Silence, also repre-
sents indirect discourse - in this case with the particle -o preceded by the
complimentizer koto. Some may argue that the koto-o expression is not a
quotation by pointing out that the embedded clause is not intended to
report speech per se. However, the embedded clause clearly conveys the
propositional meaning of the assumed verbal action committed by the
speaker and clearly meets the definition of indirect quotation given
earlier. In this broad sense we can consider koto-o as a device used for
speech and thought representation in Japanese. 6
As we have seen then, we have identified (and I do not intend this to
be exhaustive) at least four ways speech may be represented in Japanese.
These are, direct speech with graphological signals, direct speech that is
not graphologically marked, indirect speech with -to and lastly, indirect
speech with koto-o.
This last reporting strategy has been excluded from previous studies,
and it is this strategy which we attempt to investigate in this study. It is
perhaps most appropriate to consider the indirect speech marked with
-to and the indirect speech marked with -o as two sub-classes of the direct
quotation defined earlier. And although -to may mark indirect quota-
tion, since it also marks direct quotation, and based on the evidence
discussed in the following two sections, we identify it as direct-oriented
indirect representation. We also identify indirect quotation marked with
-o as content-orientied indirect representation for reasons which will be
presented in the following sections.
It is significant at this point to contrast this speech style taxonomy in
Japanese with that of English. Note that in English a clear distinction is
available between direct quotation and indirect quotation both from
formal and semantic perspectives. In English, indirect quotation takes
the subordinate clause with all characteristics of embedded clauses such
as tense and deictic shifts and is often marked by the complimentizer
 Content-oriented indirect speech in Japanese 185

that, while direct quotation, often accompanied by graphological marks,

will contain a verbatim expression expressed as an independent sen-
tence. In Japanese, on the other hand, there are two delineating factors
involved in characterizing speech and thought representation: one,
between direct -to and indirect -to on the basis of changes that take place
in the quoted clause and two, between direct-oriented indirect and
content-oriented indirect speech on the basis of the particles marking the
distinction, namely -to and -o respectively, and also on the basis of
various syntactic and semantic characteristics that are discussed in the
following sections.

3. Syntactic characteristics of koto-o clause

The most significant syntactic characteristics of koto-o clauses in contrast

with -to quotation is the fact that -o must invariably mark a subordinate
clause, while in -to quotation, -to may not do so.7 As mentioned earlier,
-to marks direct quotation which consists of independent sentences with
all sentential features as well as direct oriented indirect quotation. Hie -o
clause on the other hand, constitutes an object NP of reporting verbs and
must bear the features of subordinate clauses. This is clearly shown by
the fact that whenever -o is used, the complementizer koto 'fact' must
accompany it.8 Koto designates the preceding clause to be subordinate in
the sense that it modifies koto. This is evidenced in the following data,
where (8b), unless a contrastive reading is assigned, is unacceptable.
This is due to the constraint in Japanese that -wa in subordinate clauses
must be realized as -ga unless a contrastive reading of -wa is implied (see
Kuno 1973).
(8) Otoko wa kenmeini Tanaka-san a. ga Amerika
man Theme eagerly Mr. Tanaka Subj. America
b. *wa
Theme
e it-ta koto-o kyoochooshite-i-ru.
to go-Past Comp.-Obj. insist-be-Nonpast
'The man is strongly insisting on the fact that Mr. Tanaka went to the
United States.'
(9) Otoko wa kenmeini Tanaka-san a. ga Amerika
man Theme eagerly Mr. Tanaka Subj. . America
b. wa
Theme
186 Senko Κ. Maynard

e it-ta to kyoochooshite-i-ru.
to go-Past Quot. insist-be-Nonpast
'The man is strongly insisting that Mr. Tanaka went to the United
States.'

Thus, the unacceptability of (8b) versus the acceptability of (9b) stems

from the fact that (8b) contains a subordinate clause while (9b), the
sentence with -to quotation, does not.
Additional syntactic evidence to support the view that -o invariably
marks the subordinate clause while -to does not is found in the phenome-
non that follows. Whereas the -o clause must be accompanied by a
matrix verb of speech and thought in order to syntactically assign the
subordinate clause marked by -o, quotations marked with -to, even in
indirect mode, may occur without matrix verbs of communication.
Observe, for example, sentence (10).
(10) Bannin-tachi wa shinto-tachi ni meijite,
guards Theme Christians to order and
shitai ο yakootoshi-ta ga, shisai to
dead body Obj. be about to burn-Past but priest and
shinto-tachi wa, sore wa kirishitan no oshie
Christians Theme that Theme Christianity of teaching
to chigau to gankoni kyozetsushi-ta.
with differ Quot. firmly refuse-Past
(Endo 1981: 144)
'The guards gave orders to the Christians to burn the body; but they
all protested that such a course of action was contrary of Christian
teaching
(Johnston 1980: 172)

Sentence (10) which carries no utterance features is considered to be an

indirect quotation. Note that if it were a direct quotation, the Christians'
speech should carry dialectal features as is always the case in the novel.
In case of priests, their verbatim quotation is only available through the
translator in the novel. Note that although priests is not translated in
Johnston, in the original as is shown in (10) it constitutes a part of the
subject of the verb kyozetsusuru 'to refuse.'
The point of significance here is that direct-orientied-indirect speech
which is marked by -to can appear without accompanying verbs of
communication. Whether one assumes that the verb of reporting is
ellipted or not, it is significant that the quotation marked by -to can
appear on the surface without a verb of communication, while the to
clause cannot, as will be shown in (13). In (10) the object of the verb
 Content-oriented indirect speech in Japanese 187

kyozetsushi-ta 'refused' must be understood as 'to burn the body,' not

the clause marked by -to, since Christians and priests did not 'refuse' the
fact that 'such a course of action was contrary to Christian teaching'.
A similar example is given in (11).
(11) Kyoo watashino uchi e kite uchiake-ru
today my house to come-and confess-Nonpast
to tameiki ο tsuite-i-ta.
Quot. sigh Obj. release-be-Past
'He sighed (saying) that he would come to my house to make his
confessions today.'

In sentence (11), the matrix verb tsuku 'release' is accompanied by the

direct object NP, tameiki 'sigh' and therefore the indirect quotation plus
-to cannot be the direct object of the matrix verb. The quotation
modifies the verb 'released a sigh' and does not supply an object clause
for the verb tsuku 'to release.' Note that in the English translation, we
must provide the verb of reporting to point out that the subordinate
clause is not an object clause but rather a quoted clause.9
The inherent difference between English and Japanese quotation
strategies is exemplified by the fact that while in English it is natural to
provide verbs of communication, in Japanese, -to quotation can fre-
quently occur without specific verbs of communication. That is, English
quoted clauses tend to be part of the matrix sentence especially when it is
indirectly quoted. In Japanese, however, -to quotation even when it is
indirect, tends to be syntactically indepentend of the matrix sentence and
is not marked as a subordinate clause on the surface.
Observe, however, that content-oriented indirect quotation with -o
clause must be accompanied by the verb that immediately governs the
clause as shown by sentences (12) and (13).
(12) Kyoo watashino uchi e kite uchiake-ru
today my house to come-and confess-Nonpast
to yuu koto ο itte tameiki ο
Quot. say Comp.Obj.say-and sigh Obj.
tsui-ta.
release-Past
'He sighed (saying) (the fact) that he would come to my house to
make his confession today.'
(13) *Kyoo watashino uchi e kite uchiake-ru to
today my house to come-and confess- Nonpast Quot.
yuu koto ο tameiki ο tsui-ta.
say Comp. Obj.sigh Obj.release-Past
188 Senko Κ. Maynard

*'He sighed the fact that he would come to my house to make his
confession today.'

Sentences (10) through (13) exemplify the fact that -to quotation does
not necessarily constitute an object clause of the matrix verb, and
therefore does not necessarily constitute a subordinate clause. Koto-o
and to yuu koto-o constitute subordinate clauses and must always be
accompanied by verbs of communication.
Another point of interest in regard to content-oriented quotation in
contrast with direct-orientied indirect quotation is the dominance rela-
tion of -to and -o. In -to quotation in Japanese we find what Martin
(1975) calls 'transitive quotation,' as exemplified by (14), reproduced
here from Martin (1975: 997).

(14) Kimi wa boku ο bocchan da to kangaete-i-ru

you Theme I Obj. boy be Quot. think-be-Nonpast
rashii ga . . .
seem but
'You still think of me as "Sonny-boy", it seems, but . . . '

Note that while it is possible to raise boku Τ of the quoted clause boku
wa bocchan da Ί am "Sonny-boy"' to an object NP position of the
higher verb kangaete-iru 'be thinking' in data (14), this manipulation will
produce an ungrammatical sentence when accompanied by -o as shown
by data (15).
(15) *Kimi wa boku ο bocchan da to yuu koto ο
you Theme I Obj. be Quot. say Comp. Obj.
kangaete-i-ru rashii ga . . .
think-be-Nonpast seem but
'You still think that I am a "Sonny-boy," it seems, but . . . '

While the dominance relation of -o, then, must extend over the whole
quoted clause, the dominance relation of -to may cover only a portion of
the quoted clause. Here, -to does not dominate boku-o but rather it
constitutes a sister node to boku-o. Strictly speaking, the verb kangaeru
'to think' does not take two object NP's, and therefore, in this structure
we must assume that the second NP is not an object NP, but something
else.
In characterizing what this NP might be, it is helpful to examine
another phenomenon, the anaphoric reference to a quotation. Observe
that when referring to a quotation in sentence (16), an adverb soo 'in
such a way' may occur only when it is not accompanied by -to as shown
 Content-oriented indirect speech in Japanese 189

by (17), while soo can occur alone as well as co-occur with koto-o, but
not with -to yuu koto-o as shown by (18).

(16) Tanaka-san wa "Ashita iki-masu yo" to

Mr. Tanaka Theme tomorrow go-Polite-Nonpast Emph. Quot.
it-ta
say-Past
'Mr. Tanaka said, "Tomorrow I'll go.'"

(17) Tanaka-san wa a. soo

Mr. Tanaka Theme so it-ta.
b. *soo to say-Past
so Quot.
'Mr. Tanaka said so.'
(18) Tanaka-san wa a.soo
Mr. Tanaka Theme so
b. soo yuu koto ο it-ta.
so say Comp. Obj. say-Past
c. *soo to yuu koto ο
so Quot. say Comp. Obj
'Mr. Tanaka said so (such a thing).'

This seems to indicate that -to quotation itself functions as something

like an adverbial phrase to describe the manner, while the -o clause can
be used to mark a subordinate object clause which functions as an
anaphoric reference. The case of transitive quotation and the anaphoric
device both support the view that -to does not necessarily mark the
subordinate clause while -o must.

4. Semantic implications of koto-o clause

The -to quotation and the koto-o clause also differ semantically. -To but
not -o is more likely to co-occur with verbs that strongly imply reference
to the actual utterance. Examine the following data.
(19) Otoko wa "Nihon wa madamada mazushii kuni
man Theme Japan Theme yet poor country
na-n-da yo" to tsubuyai-ta.
be-Comp.-be Emph. Quot. mutter-Past
'The man muttered, "Japan is still a very poor country.'"
190 Senko Κ. Maynard

(20) Otoko wa nihon wa madamada mazushii kuni da to

man Theme Japan Theme yet poor country be Quot.

tsubuyai-ta.
mutter-Past
'The man muttered that Japan was still a very poor country.'

(21) * Otoko wa nihon wa madamada mazushii kuni da to

man Theme Japan Theme yet poor country be Quot.

yuu koto ο tsubuyai-ta.

say Comp Obj.mutter-Past
'The man muttered the fact that Japan is still a very poor country.'

(22) *Otoko wa "Nihon wa madamada mazushii kuni

man Theme Japan Theme yet poor country
na-n-da yo" to kuwashiku nobe-ta.
be-Comp.-be Emph. Quot. in detail state-Past
'The man stated in detail, "Japan is still a very poor country.'"

(23) ?Otoko wa nihon wa madamada mazushii kuni da

man Theme Japan Thema yet poor country be

to kuwashiku nobe-ta.
Quot. in detail state-Past
'The man stated in detail that Japan is still a very poor country.'

(24) Otoko wa nihon wa madamada mazushii kuni da

man Theme Japan Theme yet poor country be

to yuu koto ο kuwashiku nobe-ta.

Quot. say Comp. Obj. in detail state-Past
'The man stated in detail the fact that Japan is still a very poor
country.'

The verb tsubuyaku 'to mutter' is semantically characterized by imme-

diacy of actual verbalization and the koto-o clause cannot co-occur as
shown in (21). On the other hand, the verb noberu 'to state' which can
take both -to and -o, when accompanied by an adverb kuwashiku 'in
detail,' cannot co-occur with a direct quotation as shown in (22).
Semantically, the inappropriateness of sentence (22) is caused by the fact
that the adverb 'in detail' explains the manner in which the statement
was made and thus shifts the focus away from the verbatim of the speech.
Sentence (24) is intended to do more than merely report, it is meant to
qualify the content of what was spoken.
Sentence (23) is acceptable but somewhat awkward. This again is
 Content-oriented indirect speech in Japanese 191

perhaps due to the conflict between the orientation toward direct and
immediate reporting realized by -to in (23) and the adverb 'in detail'
which is associated with a description of the style of delivery rather than
with the actual speech.
Semantically, then these examples indicate that -to is direct report
oriented with immediate access and with the reporter's commitment to
actual verbalization of the quote, whether it occurs with direct quotation
as in (19) or with indirect quotation as in (20). Conversely, the koto-o
clause is suited exclusively for indirect reporting - with no commitment
to quoting verbatim words but with the emphasis on the content itself -
and therefore is less likely to co-occur with verbs that refer to actual
verbalization.
Related to the semantic characteristics of the -to quotation and koto-o
clause is their functional difference as evidenced in the following data.
(25) "Senor, Gracia."
Otoko wa porutogarugo de koeokake-ta. Kimyoona
man Theme Portuguese in speak-Past strange
tadotadoshii hatsuon dat-ta ga, sore wa
halting pronunciation be-Past but that Theme
tomokakumo porutogarugo ni chigainakat-ta.
at any rate Portuguese as must be-Past
"Senor."
"Palazera ä Dios nuestro Senor."
(Endo 1981: 112)

(26) "Senor, Gracia,' he said.

So he was speaking Portuguese. The pronunciation was strange and
halting, but it was certainly Portuguese.
'Senor.'
'Palazera ä Dios nuestro Senor."
(Johnston 1980: 132)
(27) Palazera ä Dios nuestro Senor to it-ta.
Quot. say-Past
'He said: Palazera ä Dios nuestro Senor.'
(28) *Palazera ä Dios nuestro Senor to yuu koto ο it-ta.
Quot. say Comp. Obj. say-Past
'*He said (the fact) that palazera ä Dios nuestro Senor.'

It is both possible and in fact useful to quote foreign speech in direct

quotation with graphological marks as in sentence (25) and its English
translation, sentence (26), as well as quotation marked with -to without
graphological marks as in (27). However, toyuu koto-o cannot comfort-
192 Senko Κ. Maynard

ably co-occur with a foreign language quote as shown in (28). This is

created by the conflict between an orientation toward immediate access
and the expression of the actual wording, as in the case of foreign
language quotation and the tendency toward content reporting as in the
case of to yuu koto-o. Moreover, since to yuu koto-o syntactically
constitutes a subordinate clause and expresses a narrative description
representing the narrator's view, it is expected to be consistent with the
language and style used by the narrator in the novel. To convey meaning
successfully, it is usually expected that the narrator continues in the same
language when he indirectly reports what was actually said - even when
the speech is delivered in a foreign language.
The semantic distinction between direct and content orientation ob-
served in verbs of speech is also found in thought representation as
shown by data (29) and (30).

(29) Kae ga akogarete-i-ru yooni Otsugi mo mata Kae ni

Kae Subj. adore-be-Nonpast as Otsugi also Kay by

horekonde-ite, dakara Kae no su-ru koto

be charmed-and therefore Kae Subj. do-Nonpast thing

na-ra donna sasaina koto demo kinii-ru

be-Conditional any small thing even be-liked-Nonpast

no dearoo to omou to, Kae wa issoo seiodashite

Comp, be Quot. think when Kae Theme more diligently

hataoora-zuniwa irare-nakat-ta.
weave-Neg. stay-Neg.-Past

(Ariyoshi 1970: 43)

'Yet she was extremely pleased to see Otsugi contented with her.
Every day she tried to increase the production because she felt her
feelings were reciprocated by her mother-in-law.'
(Hironaka and Kostant 1981: 41)

(30) Kae wa shu ga hurubite saeta iro ο

Kae Theme vermilion Subj. get old-and clear color Obj.

shite-iru norisakazuki no nakani moriagu-ru

do-be-Nonpast sake cup of inside rise-Nonpast

yooni tatae-rarete-i-ru koi sake ο mitsumete.

as till-Passive-be-Nonpast strong sake Obj. see-and
 Content-oriented indirect speech in Japanese 193

koko kara atarashii ikikata ga hajima-ru koto

here from new way of life Subj. gegin-Nonpast Comp.

ο omot-ta.
Obj. think-Past
(Ariyoshi 1970: 22)
'As Kae watched the sake reach the brim, thoughts of her new life
raced through her mind.'
(Hironaka and Kostant 1981: 31)

In sentence (29) -to is chosen over the koto-o clause while in sentence
(30), koto-o is chosen. Both are followed by the verb omou 'to think.' In
interpreting (29), the thought expressed carries an instantaneous flash of
thought rather than a deep careful thought process. In (30), however,
Kae's thought is not a burst of inspiration but rather a carefully thought
out one. The expression koto-o omou is semantically similar to omoi-o
megurasu which may be translated as 'to be in deep thought - thinking of
various facts and possibilities that are related . . . ' Interestingly, the
English translation by Hironaka and Kostant reflects this difference to
some degree.
The difference in semantic interpretation observed here may be
understood from the perspective stated earlier, namely, the direct vs.
content oriented orientational difference between -to and koto-o. Since
-to has an orientation toward direct, immediate reporting, we interpret
(29) to represent exactly how Kae thought, namely, Kae ga akogareteiru
yooni Otsugi mo mata Kae ni horekondeite, dakara Kae no suru koto
nara donna sasaina demo kiniiru no de aroo 'just as Kae adored Otsugi,
Otsugi adores Kae and therefore whatever small things Kae did, Otsugi
liked her anyway.' On the other hand, in (30), the interpretation would
have to be that Kae's thought was not limited to her realization that koko
kare atarashii ikikata ga hajimaru 'new life will begin now.'
In fact, the direct versus content orientation of -to versus koto-o
clauses may be used for categorizing verbs of communication. 10 The
more direct the semantic interpretation of the verbs of communication,
the more likely the occurrence of -to. Conversely, the more indirect and
the more content-oriented the verb of communcation is, the greater the
likelihood of koto-o to occur.
Although many verbs can take both -to and koto-o, the -to quotation
in both the direct and indirect mode seems to be the dominant strategy
and the use of the koto-o clause seems to be more restricted. Indeed, the
fact that the Japanese language provides -to to mark both direct and
194 Senko Κ. Maynard

indirect quotation may be supporting evidence that in Japanese the

directness oriented mode of speech and thought representation is prefer-
red.11

5. Scope relations of -o clause

Another significant difference between direct-oriented indirect and con-

tent-oriented indirect speech and thought representation involves scope
relations. Inoue (1982) explains that the scope of to yuu spans across
sentences. The cross-sentential scope of -to, however, is not limited only
to -to yuu. Observe date (31).
(31a) Tomogimura to onajiyooni koko ni mo Fereira-shi
Tomogimura village with as here at also Mr. Fereira
wa ki-ta koto wa nai no ka.
Theme come-Past Comp. Theme Neg. Comp. Ques.
b) Soretomo Fereira to yuu na wa nihon no
or Fereira Ouot. say name Theme Japan of
shinto noaidade wa kuchinidashite wa naranu
Christian among Theme mention Neg. must not
kinku ni natte-i-ru no de wa nai ka
taboo as become-be-Nonpast Comp, be Neg. Ques.
to sae kare wa omot-ta.
Quot. even he Theme think-Past
(Endo 1981: 106)
b) 'Was the very name of Ferreira a word that was not even to be
mentioned by the Christians?, he asked himself.'
(Johnston 1980: 126)

Since sentence (31a) is not translated at all in Johnston's Silence, I am

adding here my translation as follows: 'Is it possible that Ferreira has
never been here just as he has never been to Tomogi village?' In order to
fully appreciate (31), we must assume that (31a) is under the scope of 'he
asked himself.' As this actual case from modern fictional writing exem-
plifies, the scope of -to transcends the sentential boundary, which
contributes to discourse cohesion.
On the other hand, in koto-o quotations the scope of koto-o does not
expand beyond the subordinate clause thus marked. This is because
when the preceding sentence is a fully independent sentence, and not an
 Content-oriented indirect speech in Japanese 195

NP. -o, which requires an NP as a preceding element, cannot syntacti-

cally co-occur and therefore it does not extend its scope beyond the
object NP.
Let us now contrast scope relations of-to used in direct-oriented
indirect as in (32) and koto-o used in content-oriented indirect speech as
in (33).
(32a) Gooruden ueek ga yatteku-ru.
golden week Subj. come-Nonpast
b) Kondono kinyoobi wa hannichi ni su-ru to
next Friday Theme half day as make-Nonpast Quot.
shachoo ga koogenshi-ta.
president Subj. declare-Past
a) 'Golden week is coming.'
b) 'The company president declared that he will make this coming
Friday a half day.'
(33a) Gooruden ueek ga yatteku-ru.
golden week Subj. come-Nonpast
b) Kondono kinyoobi wa hannichi ni su-ru koto
next Friday Theme half day as make-Nonpast Comp.
ο schachoo ga koogenshi-ta.
Obj. president Subj. declare-Past
a) 'Golden week is coming.'
b) 'The company president declared (the fact) that he will make this
coming Friday a half day.'

In (32), we can assume that what the president said includes (32a). For
example (32) can easily be interpreted as the indirect quotation of
Gooruden ueek ga yattekuru naa. Kondono kinyoobi wa hannichi ni
shiyoo, >Golden week is coming. Let's make this coming Friday a half
day.'
However, the most natural interpretation of sentence (33) is, as indicat-
ed in translation, that the scope of koto-o spans only over (33). To
interpret (33a) to be under the scope of 'declare' is most difficult, if not
impossible.
196 Senko Κ. Maynard

6. Concluding remarks: literary functions and implications

As discussed in the preceding sections, we have observed that in Japa-

nese both the -to quotation and koto-ol-to yuu koto-o clauses are used for
indirect speech and thought representation and we identified them as
direct-oriented indirect and content-oriented indirect, respectively.
The syntactics, semantic and discoursal differences that are investigat-
ed in this study between the two aforementioned strategies are signifi-
cant not only from a linguistic standpoint but also from the perspective of
literary style. The direct-oriented quotation marker -to in Japanese, even
when it clearly marks an indirect quotation, reflects the actual speech
utterance more realistically than the -o clause. The -o clause, while
resembling the so-called English indirect quotation because in both
English and Japanese the quoted clause constitutes a subordinate clause,
is restricted in its usage. As mentioned earlier, the verbs of communica-
tion that can take koto-o, to yuu koto-o are limited in number. This
linguistic constraint may very well be directly correlated to the pre-
ference of direct-orientation over content-orientation in Japanese speech
and thought representation.
In fact, although based on only cursory observation, a preference for
direct style over indirect style is observed in modern Japanese fictional
writings examined in this study. And in more than a few instances where
the quotation of the English translation or original is indirect, the
quotation of the Japanese original or translation is direct. This strategic
difference has a profound influence on literary effect because the direct
mode represents the speaker's view allowing the reader access to utter-
ance aspects of speech deliverance, while the indirect mode represents
the narrator's view more strongly and as a result increases the distance
between the reader and the person whose speech is being quoted.
Another significant functional difference observed in the Japanese
quotation system involves the shift and maintenance of point of view.
First, note that as shown in (1), direct quotation without graphological
markings and -to plus a verb of communication can constitute a subordi-
nate clause of a higher sentence. This sentence then represents dual
points of view - direct speech representing the speaker's, while narrative
description representing the narrator's view(s). Second, it is significant
that in Japanese, direct-oriented indirect quotation marked with -to as
shown in (32) can easily cross over the sentential boundary. Observe that
in (32) a particular point of view can be successfully maintained across a
 Content-oriented indirect speech in Japanese 197

sentential boundary without specific signals. In this situation, the reader

interprets the point of view on the basis of the scope relations of -to.
Both situations (1) and (32) seem to point to the fact that in Japanese
shifts and maintenance of points of view are less specifically manifested
by linguistic devices. The fluidity of viewpoints seems to be made more
accessible to the writer since the language itself allows for it and perhaps
encourages it. Indeed, the frequently occurring mixture of direct quota-
tion without graphological marks and narrative description as well as the
mixture of direct-oriented indirect and content-oriented indirect repres-
entations provide an extra communication tool to the narrator since
representing multiple voices can be accomplished without disturbing the
flow of discourse. And since -to has the power to cross sentential
boundaries, it provides the narrator with expanded manipulative flexibil-
ity in dramatically sustaining the point of view through cross-sentential
signals that only -to can offer. The narrator's choice of speech mode,
then, is based on his/her manipulation of the point of view and aesthetic
distance that he/she wishes to create between speech initiator, reader
and the author himself/herself.
Although the existance of both direct and indirect modes of speech
and thought expression is most likely to be universal, the devices
available in each language differ. In Japanese the availability of direct-
oriented indirect speech and the limited access to content-oriented
indirect style constitute the idiosyncratic characteristics of the speech
and thought representation mode. The difference in devices and manipu-
lations and the preference for specific strategies imposed by the language
system itself, then, constitute a critical focus in examining stylistic gaps
that may occur among various literary traditions.
The study reported in this paper is by no means intended to be
exhaustive. The issues surrounding speech and thought representation
are obviously too many and varied to be studied under one heading.
Moreover the questions raised at the beginning of this paper are only
answered in part. However, investigation of literary discourse from a
linguistic perspective, I believe, constitutes a vital and important area
and I hope that the interdisciplinary approach attempted in this study
will contribute to a better understanding of communicative and artistic
functions of language and its usage.
198 Senko Κ. Maynard

Notes

1. This observation is made based on data examined for this project which includes
Shuusaku Endo's Silence, Sawako Ariyoshi's The Doctor's Wife, Yukio Mishima's
Spring Snow and John Updike's Couples and their translations.
2. One point to be raised regarding Inoue's criteria is the tense agreement constraint in
indirect quotation. Inoue does not provide examples in which 'agreement of the quoted
phrase tense with that of the main sentence or the time of the event' is satisfied.
Contrary to this statement, the agreement of tense in indirect speech in Japanese
cannot be accomplished if the intended tense is to be interpreted correctly. For
example, in a sentence such as (i), the only interpretation is that the child actually said
that he 'went' to the store in his direct speech.
(i) Kodomo wa hahaoya ni sono hi mo mise e
child Theme mother to that day also store to
it-ta to it-ta.
go-Past Quot. say-Past
'The child said to his mother that he went to the store again that day.'
3. For other categories or subcategories of direct and indirect speech and their syntactic
constraints, see Endo (1982).
4. There are cases not covered by either direct or indirect quotation here. The quoted
clause that carries some of the utterance features, must be categorized as semidirect. A
discussion of other possible modes of speech and thought representation that may be
located in various distances between and beyond two polar cases, direct and indirect
quotation, however, is beyond the scope of this paper.
5. In Silence, the farmer's speech in direct quotation is consistently marked by the
performative feature, and thus, ending a sentence with nai without other interpersonal
communication devices attached would be unusual, if not improbable.
6. See sentence (30) for an example of koto-o used for thought representation. Yooni that
co-occurs with omou 'to think' and yuu 'to say' is another device that is used for speech
and thought representation, but is excluded from the present study.
7. Note that we are defining 'subordinate clause' in a broad sense. Considering the fact
that an indirectly quoted clause bearing features of subordination can be marked by -to
and that direct quotation is also marked by -to, we state that -to may or may not mark
the subordinate clause.
8. Although wherever -o is used koto must precede it, it is not necessarily the rule that in
all cases where koto is used, -o must follow. See, for example, data (ii).
(ii) Minasama ogenkide bengaku ni isoshinde-i-ru koto
everyone healthy study at engage in-be-Nonpast Comp,
to omoi-ma-su.
Quot. think-Polite-Nonpast
Ί think (that) all of you are in good health and studying hard.'
9. Direct quotation such as (iii) frequently appears without the verb of reporting.
Interestingly in English translation, the translator often provides verbs of communica-
tion as shown by Galagher's translation.
(iii) "Sore ga ore ni wa ichibanno nazo na-no-da
that Subj. I to Theme first riddle be-Comp.-be
 Content-oriented indirect speech in Japanese 199

to Honda wa hotondo kokkeini mie-ru shinshina

Quot. Honda Theme almost comical be seen-Nonpast earnest
tameiki ο morashi-ta ga . . .
sigh Obj. release-Past but
(Mishima 1977: 114)
'"And that for me is the greatest riddle," said Honda, sighing so earnestly
that it seemed almost comical.'
(Gallagher 1972: 97)
10. The semantic distinction we observe here between -to and koto-o is closely related to
some of the notions developed regarding complimentizers koto, no and -to in Japanese.
Kuno (1973), for example, using the concepts of presupposition and factivity, makes
the distinction between factive kono/no and nonfactive -to. Moreover, Kuno proposes
that koto is used to nominalize an 'abstract concept' while no nominalizes a 'concrete
event.' McCawley (1978) makes a claim that complimentizers no, koto and to are
distributed in such a way that truthfulness is presupposed in the order of no > koto >
to. Josephs (1976) introduces another set of notions, namely the features 'direct' and
'indirect.' Josephs writes (1976: 325): ... No means something like 'directly perceived,
simultaneously occurring, or imminent action, event, etc.,' while koto means 'non-
simultaneous, nonrealized, or abstractly perceived action, event, state etc.' We can
summarize this inherent meaning difference by characterizing no with a semantic
feature like 'direct' and koto with a semantic feature 'indirect'. The fact that koto has
an indirect feature is similar to what I am proposing in this paper. However, in our
study of reported and non-reported speech, we are concerned with two devices (-to and
koto-o) that co-occur with verbs of communication. Although Josephs' notions are
useful, his analysis is restricted to koto versus no in applying notions such as 'direct' and
'indirect.'
11. The fact that in Japanese, direct quotation is preferred over indirect quotation has been
pointed out by Ross (1976-77), Anzai (1983) and others.

References
Ariyoshi, Sawako. 1970. Hanaoka seishuu no tsuma. Tokyo: Shinchoosha.
Anzai, Tetsuo. 1983. Eigo no hassoo. Koodansha Gendai Shinsho. Tokyo: Kodansha.
Banfield, Ann. 1973. Narrative style and the grammar of direct and indirect speech.
Foundations in language, 10: 1-39.
- . 1982. Unspeakable sentences. Boston: Routledge and Kegan Paul.
Endo, Hiroko. 1982. Nihongo no wahoo. Gengo, 19, 3: 86-94.
Endo, Shuusaku. 1981. Chinmoku. Shinchoo Bunko. Tokyo: Shinchoosha.
Gallagher, Michael. 1975: Spring snow, The sea of fertility 1, translated from Haru no yuki,
Hoojoo no umi 1 by Y. Mishima. New York: A Washington Square Press.
Hironaka, Wakako and Ann Siller Kostant. 1978. The doctor's wife, translated from
Hanaoka seishuu no tsuma by S. Ariyoshi. Tokyo: Kodansha International.
Inoue, Kazuko. 1982. An interface of syntax, semantics and discourse structure. Lingua
57: 259-300.
Johnston, William. 1980. Silence, translated from Chinmoku by S. Endo. New York.
Taplinger Publishing Company.
200 Senko Κ. Maynard

Josephs, Lewis. 1976. Complementation. In: M. Shibatani (ed.), Syntax and semantics,
Vol. 5: 307-369. New York: Academic Press.
Kamada, Osamu. 1981. Indirect quotation in Japanese. In: S. Makino (ed.), Papers from
the Middlebury Symposium on Japanese discourse analysis, 89-124.
- . 1983. Nihongo no kansetsu wahoo. Gengo 20, 9: 108-117.
Kuno, Susumu. 1973. The structure of the Japanese language. Cambridge: The M.I.T.
Press.
Kuroda, S.-Y. 1973. Where epistemology, style, and grammar meet: a case study from
Japanese. In: P. Kiparsky, S. Anderson (eds.), Festschrift for Morris Halle, 337-391.
New York: Holt.
- . 1976. Reflections on foundations of narrative theory from a linguistic point of view. In:
T. van Dijk (ed), Pragmatics of language and literature, 108-140. Amsterdam and New
York: North-Holland Publishing Company.
Martin, Samuel. 1975. A reference grammar of Japanese. New Haven, London: Yale
University Press.
McCawley, Noriko. 1978. Epistemology and Japanese syntax: complementizer choice.
Papers from the Fourteenth Regional Meeting: 272-84. Chicago Linguistics Society.
Mishima, Yukio. 1977. Haru no yuki. Hoojoo no umi 1. Shinchoo Bunko. Tokyo:
Shinchoosha.
Miyamito, Yookichi. 1975. Kapuruzu, translated from Couples by J. Updike. Shinchoo
Bunko. Tokyo: Shinchoosha.
Nitta, Yoshio. 1979. Types of expressive potentiality of Japanese sentences: on the notion
of the relation between restrictions of the persons of the subject and sentence structures.
In: Committee for Publication of the festschrift for professor Eiichi Hayashi on his 60th
birthday, English and Japanese, 287-306. Tokyo: Kuroshio Shuppan.
Okutsu, Keiichiro. 1968. Inyoo koozoo to kansetsuka tenkei. Gengo kenkyuu 56: 1-26.
Page, Norman. 1973. Speech in the English novel. London: Longman Group.
Partee, Barbara. 1973. The syntax and semantics of quotation. In: P. Kiparsky and S.
Anderson (eds), Festschrift for Morris Halle, 410-418. New York: Holt.
Ross, Claudia. 1976-77. Reporting style as discourse strategies - a study in Japanese and
English. Papers in Japanese Linguistics, 243-59.
Updike, John. 1968. Couples. New York: Knopf.
Some semantic aspects of indirect speech in
Hungarian
Ferenc Kiefer

1. In this paper I am going to discuss a number of problems concerning

indirect speech in Hungarian. To be sure, indirect speech in Hungarian
shares a lot of features with other languages. Thus, for example,
personal, temporal or locative shifts work by and large in the same way
as in well-known languages such as English, French, German, etc.
However, it is a distinctive feature of Hungarian that it lacks the
sequence of tenses, i.e. the tense of the embedded verb is independent of
the tense of the matrix verb. An utterance such as
(1) Jänos elmegy.
'John is leaving'

is reported by
(2) Peter mondta, hogy Jänos elmegy.
lit. 'Peter said that John is leaving'

The fact that the matrix verb is in Past Tense does not affect the tense of
the embedded verb. That is, the tense of the original utterance is always
maintained in reported speech.
Another feature of indirect speech in Hungarian is that quotation and
indirect speech can be kept apart by means of demonstratives. Consider

(3a) Peter azt mondta, hogy Jänos elmegy.

lit. Peter that said that John is-leaving
b) Peter ezt mondta: Jänos elmegy.
lit. Peter this said John is-leaving
'Peter said this: "John is leaving'"

If the demonstrative ezt 'this' is used, indirect speech is not possible, if,
on the other hand, azt 'that' is used, indirect speech is the only possibil-
ity.
A further property of indirect speech in Hungarian is that the connec-
tive hogy 'that' is used to introduce indirect speech not only in the case of
indirect statements but also in the case of indirect questions. For
example,
202 Ferenc Kiefer

(4) P6ter megmondta, hogy hoi volt.

lit. Peter said-perf. that where was he
'Peter told us where he had been'

However, as far as I can see, the most interesting aspects of indirect

speech in Hungarian are related to the semantics of the matrix verb.
Therefore, the problems which I am going to discuss will all have to do
with properties of the reporting verb. The first problem concerns the
relationship between verbs of saying and factivity, the second problem
that between factives and emotives. The third problem leads us to a
different field: it has to do with the behavior of verbs of saying in
connection with the nonpropositional elements of the utterance.

2. In Hungarian certain verbs of saying have a factive and a nonfactive

variant. For example,

(5a) Peter mondta, hogy Anna beteg.

'Peter said that Ann is sick'
b) Peter megmondta, hogy Anna beteg.
(6a) Peter irta, hogy Anna beteg.
'Peter wrote that Ann was sick'
b) Peter megirta, hogy Anna beteg.

The verb mond 'say' is not factive, the speaker need not commit himself
to the truth of the proposition 'Ann is sick'. In contrast, the verb
megmond is factive: the speaker takes the proposition 'Ann is sick' for
granted. The relationship between the two verbs is quite transparent: the
second verb contains the perfective prefix meg, i.e., it is a perfective
verb, whereas the first one is imperfective. The same difference can be
observed in the case of (6a) and (b) as well. The verb ir 'write' is
imperfective and nonfactive whereas the verb megir is perfective and
factive.1
Next, let us consider indirect questions.

(7a) Peter mondta, hogy hol järt.

'Peter said where he had been'
b) Peter megmondta, hogy hol järt.
(8a) Peter irta, hogy hol järt.
'Peter wrote where he had been'
b) Peter megirta, hogy hol järt.
 Indirect speech in Hungarian 203

Again, the difference between (7a) and (b), on the one hand, and
between (8a) and (b), on the other, is similar to that noted in connection
with (5a-b) and (6a-b), respectively. The (b)-sentences are factive in
some sense whereas the (a)-sentences are definitely nonfactive.
The quasi-factivity of certain matrix verbs with embedded questions
has often been noted. 2 It is most apparent in the case of such verbs as
know, discover, realize, etc. For example,
(9a) Bill knows where Peter was.
b) Bill has discovered where Peter was.
c) Bill has realized where Peter was.
Notice that for Bill the embedded sentence 'where Peter was' does not
contain any unknown variable and he could easily replace the question
word where by a locative expression. The speaker, however, need not
share this knowledge.
Let us return to the sentences (7a-b) and (8a-b). When Peter tells us
where he was, we need not believe him. Though in all the sentences
(7a-b) and (8a-b) the question word where can be replaced by a locative
expression known to everybody who has witnessed Peter's utterance, the
proposition obtained in this way need not be accepted by the speaker as
a true proposition in the case of (7a) and (8a). In the case of (7b) and
(8b), on the other hand, such propositions are considered to be true.
Examples such as (9a-c) show that factivity with embedded questions
has to be restricted to the mental state of the person referred to by the
subject of the matrix sentence. The speaker need not commit himself to
the truth of the proposition expressed by the that-clause. Examples such
as (7a-b) and (8a-b), on the other hand, bear witness to the fact that
communicated knowledge and accepted knowledge are two different
things.
It is generally known that perfective verbs are often used to report on
facts whereas imperfective verbs are used to describe ongoing processes
or activities. This may explain why megmond 'say (perfective)' and megir
'write (perfective)' are factive whereas mond 'say (imperfective)' and ir
'write (imperfective)' are not. This does not mean, however, that all
perfective verbs are automatically factive (of course, they are not). The
majority of factives seem to be statives, they describe states and are
consequently neither perfective nor imperfective. This may mean, how-
ever, that whenever we have an imperfective-perfective pair such as
mond-megmond, ir-megir which may take an embedded that-clause, in
addition to the difference concerning perfectivity there might also be a
204 Ferenc Kiefer

difference concerning factivity between the two verbs, the perfective

verb being factive and the imperfective one nonfactive.
Semantically embedded questions can roughly be characterized as
consisting of an expression containing one or more variables (which can
be referred to as the propositional content of the question) and an
attitudinal operator.3 Stating, asking, wondering, wishing, etc. are such
operators. That is, one can ask a question, one can state that one does
not know the answer to the question, one can express one's wish to know
the answer, one can wonder what the answer may be, one can look for an
answer, etc. One can also state that one knows the answer, that one has
discovered the answer or that somebody has told us the answer, etc. In
other words, one may have various attitudes toward the propositional
content of a question. These attitudes seem to fall into two major classes.
One class is characterized by the fact that one does not know the answer,
the other one by the fact that one does know the answer. The verbs of
saying discussed above and cognitive verbs such as know, discover,
realize (cf. examples (9a-c) above) belong to the latter type. There is,
however, as pointed out above, an essential difference between the
cognitive verbs representing the result of cognitive search (for an ans-
wer) and the verbs of saying. Evidently, the answer must be known to
the person referred to by the subject of the matrix sentence in both
cases. But the speaker need not share this knowledge in the case of
cognitive verbs. Cognitive search is a private matter and the result need
not be communicated to others. Cognitive verbs do not presuppose a
speech situation with speaker and hearer. On the other hand, verbs of
saying normally require such a speech situation. One can report on
somebody's cognitive operations without knowing anything about the
result but when we report on something we were told then we must know
something. It is at this point that the distinction between factive and
nonfactive verbs of saying comes into the play. If the speaker uses a
factive verb of saying in his report, he accepts the reported state-of-
affairs as a fact; if, on the other hand, he uses a nonfactive verb of
saying, he does not commit himself to the truth of the state-of-affairs at
hand.
Notice that in the case of states the perfective prefix may be used to
describe the beginning of a state. For example, tud 'know' and megtud
'become to know'. Of course, both verbs are factive. The first one
describes a cognitive state whereas the second one refers to the coming
about of a cognitive state. It does not come as a surprise, then, that a
sentence such as
 Indirect speech in Hungarian 205

(10) Jänos megtudta, hogy Peter hol järt.

lit. John became-to-know that Peter where was
'John found out where Peter had been'

is interpreted in a similar fashion as (9a-c), i.e., the speaker need not

know anything about where Peter was.
We may thus conclude that only factive verbs of saying induce factive
presuppositions, cognitive factives do not, or at least not in the same
sense as the factive verbs of saying mentioned above.

3. Our next problem leads us to an interesting relationship between

emotive factives and emotive verbs of saying. In Hungarian one encoun-
ters pairs of verbs such as sajnäl 'regret' - sajnälkozik 'utter one's regret,
lament', restell 'be ashamed' - restelkedik '(verbally) express one's
embarrassment', etc., where the first verb is a true factive and the
second verb is nonfactive. Notice that the existence of such pairs may
have important implications concerning the status of presuppositions in
linguistic theory. According to a widely held view, sentences such as
(11a) imply, rather than presuppose, the truth of the embedded proposi-
tion 'Ann has left': ( l i b ) contains a contradiction. The negation of (11a),
on the other hand, does not have this implication since ( l i d ) is fully
acceptable. 4
(11a) Peter sajnälja, hogy Anna elment.
'Peter regrets that Ann has left'
b) *Peter sajnälja, hogy Anna elment, bar Anna nem ment el'
'Peter regrets that Ann has left though Ann has not left'
c) Peter nem sajnälja, hogy Anna elment.
'Peter does not regret that Ann has left'
d) *Peter nem sajnälja, hogy Anna elment, mert Anna nem ment el.
'Peter does not regret that Ann has left because Ann has not left'

Unfortunately, however, ( l i d ) is as bad as ( l i b ) in Hungarian. That is to

say, both (11a) and (c) entail the truth of 'Ann has left'. This speaks
definitely in favor of a non-implicational account of presupposition. The
Hungarian verbs sajnäl 'regret' and restell 'be ashamed' are true factives:
in whatever context they occur they always induce the 'factive' presuppo-
sition.
Consider next the following sentences:
(12a) Peter sajnälkozik, hogy Anna elment.
'Peter is lamenting that Ann has left'
206 Ferenc Kiefer

b) Peter sajnälkozik, hogy Anna elment, pedig Anna nem ment el.
'Peter is lamenting that Ann has left but Ann has not left'
c) Peter nem sajnälkozik, hogy Anna elment, pedig Anna nem ment el.
'Peter is not lamenting that Ann has left though Ann has not left'

The sentence (12b) is completely well-formed, that is, (12a) neither

implies nor presupposes the truth of 'Ann has left'. On the other hand,
though the sentence (12c) is rather awkward, it does not contain any
contradiction. Notice that the verb sajnälkozik 'utter one's regret,
lament' is a verb of saying, it can be used to report somebody's
utterance. This is not the case with sajnal 'regret'. The same holds true
for the pair restell 'be ashamed' and restelkedik '(verbally) express one's
embarrassment'. Only the latter but not the former can be used as a verb
of saying. The verb sajnälkozik is paraphrasable by sajnälkozva mondja
'say with regret' and restelkedik by restelkedve mondja 'say with embar-
rassment'. That is, (13a-b) and (14a-b) are close paraphrases.
(13a) Peter sajnälkozik, hogy Anna elment.
'Peter is lamenting that Ann has left'
b) Peter sajnälkozva mondja, hogy Anna elment.
lit. 'Peter says with regret that Ann has left'
(14a) Piter restelkedik, hogy eltörte a poharat.
approx. 'Peter apologizes that he has broken the glass'
b) P6ter restelkedve mondja, hogy eltörte a poharat.
lit. 'Peter says with embarrassment that he has broken the glass'

In the sentences (13b) and (14b) the forms sajnälkozva 'with regret' and
restelkedve 'with embarrassment' function as manner adverbials. One
may thus argue that verbs such as sajnälkozik 'lament' and restelkedik
'apologize' contain - at least in one of their readings - the following
semantic elements:

(15a) the speaker utters ρ

b) the speaker says that he feels regret/embarrassment because of ρ

We saw above that the emotive verbs of saying are not factive.
Furthermore, we also saw that there are factive verbs of saying in
Hungarian. The question may thus arise whether the lack of factive
emotive verbs of saying in Hungarian is an accidental gap or whether it is
due to some semantic regularity. The fact that the sentences (16a-b) with
the factive megmond 'say' are quite acceptable seems to speak in favor of
the 'accidental gap'-explanation.
 Indirect speech in Hungarian 207

(16a) Peter sajnälkozva mondta meg, hogy Anna elment.

lit. 'Peter said with regret that Ann had left'
b) Peter restelkedve mondta meg, hogy eltörte a poharat.
lit. 'Peter said with embarrassment that he had broken the glass'

Emotive verbs such as sajnälkozik, restelkedik, etc. are not always

interpreted as verbs of saying, however. In fact, they are ambiguous
between the verb of saying-reading and the pure emotive reading. The
latter reading is the only possible one in the following sentences:
(17a) P6ter sajnälkozik a fönöken.
'Peter feels pity for his boss'
b) Peter restelkedik a pohär miatt.
'Peter is embarrassed because of the glass'

The verb of saying reading of emotives is only possible if the verb is

followed by a that-clause. This is not at all surprising since hogy 'that' is
the connective generally used with reporting verbs.
Consider next (18a-b).
(18a) Peter restelkedik amiatt, hogy eltörte a poharat.
'Peter is embarrassed because he has broken the glass'
b) P6ter amiatt restelkedik, hogy eltörte a poharat.
'It is because he has broken the glass that Peter is embarrassed'
or 'It is because he has broken the glass that Peter is apologizing'
Sentence (18a) is connected with the presupposition that 'Peter has
broken the glass' in spite of the fact that the verb restelkedik 'be
ashamed/embarrassed' is not factive. This presupposition has to do with
the topic-comment structure of the sentence. The demonstrative amiatt
'because of that' points to the embedded that-clause, it is its 'forerun-
ner', as it were. If we encounter amiatt in the main clause, we will know
that the connective hogy introduces a clause of cause. The position of
demonstratives such as amiatt in the sentence indicates the topic-com-
ment structure of the sentence. If it follows the verb (in which case it is
always unstressed) it expresses unambiguously the fact that the that-
clause belongs to the topic of the sentence.
Sentence (18a) can only be used in a context in which one wants to
know something about Peter's reactions to the event described by the
that-clause. If, on the other hand, the demonstrative amiatt precedes the
verb and carries main stress, the presupposition mentioned above need
not hold. Sentence (18b) is used in a context in which it is known that
Peter is embarrassed and we are inquiring about the reason of his
208 Ferenc Kiefer

embarrassment. That is, the that-clause belongs to the comment of the

sentence. 5
We noted above that emotive verbs of saying are not factive. This is
not at variance with the facts presented in (18a-b). Notice that the verb
restelkedik 'be ashamed/embarrassed' would not normally be interpreted
as a verb of saying in (18a). It merely refers to Peter's emotional state.
The sentence (18b), on the other hand, admits the verb of saying
reading, though it may also express an emotional state. In other words,
factivity brought about by the topic-comment structure of the sentence
seems to exclude the verb of saying reading of emotives whereas this
reading is possible if the topic-comment structure of the sentence makes
it clear that the that-clause is not a presupposition.
Similar differences can also be observed in the case of (19a-b).
(19a) Peter sajnälkozik azon, hogy Anna elment.
'Peter regrets that Ann has left'
b) Peter azon sajnälkozik, hogy Anna elment.
'It is because Ann has left that Peter feels regret'
or 'It is because Ann has left that Peter is lamenting'

In these sentences we encounter another demonstrative, namely azon

'on that'. The form of the demonstrative depends on the rection of the
verb. Again, the position of this demonstrative indicates the topic-
comment structure of the sentence. In (19a) the that-clause belongs to
the topic of the sentence, in (19b), on the other hand, it is the comment.
It is not surprising, then, that the verb sajnälkozik admits the verb of
saying interpretation in the case of (19b) only.
In addition to sajnälkozik 'regret or lament', restelkedik 'be ashamed/
embarrassed or apologize' there is a whole class of emotives in Hunga-
rian which behave in a similar fashion (kesereg 'lament', zsembel
'grumble', zsörtölödik 'nag', örvendezik 'rejoice', etc.). In the verb of
saying reading these verbs are never factive. If, however, they refer to
emotional states, they may induce the 'factive presupposition'. This
presupposition appears if the that-clause belongs to the topic of the
sentence. In such cases the verb of saying reading is excluded as we saw
above. The combination of the elements 'verb of saying', 'emotive' and
'factive' seems to be impossible even in cases where factivity is not
determined lexically.
Notice that emotive verbs of saying admit embedded statements only.

(20a) *Peter sajnälkozik, hogy ki ment el.

'Peter is lamenting that who has left'
 Indirect speech in Hungarian 209

b) * Peter restelkedik, hogy ki ment el.

'Peter is apologizing that who has left'

The ungrammaticality of (20a-b) can be explained along the following

lines. It would seem that emotive predicates which can function as matrix
verbs for embedded that-clauses can only be used if the that-clause is
fully specified, i.e., if it expresses a proposition. Notice that sentences
such as (21a-b), too, are ungrammatical.
(21a) *Peter sajnälja, hogy ki ment el.
'Peter regrets that who has left'
b) *Peter restelli, hogy ki ment el.
'Peter is embarrassed that who has left'

In contrast, both (22a-b) and (23a-b) are possible.

(22a) Peter sajnälkozva mondja, hogy ki ment el.
'Peter says with regret who has left'
b) Peter restelkedve mondja, hogy ki ment el.
'Peter says with embarrassment who has left'
(23a) Peter sajnälkozva mondta meg, hogy ki ment el.
'Peter said with regret who had left'
b) Peter restelkedve mondta meg, hogy ki ment el.
'Peter said with embarrassment who had left'

We saw above that sajnälkozva mondja is synonymous with sajnälkozik

and restelkedve mondja with restelkedik if sajnälkozik and restelkedik are
used as verbs of saying. There is, however, an important syntactic
difference between the two types of predicate. The verb mond 'say' and
its factive variant megmond can be used to report practically any kind of
utterance. Moreover, in (22a-b) and (23a-b) the question sentence is
embedded under mond and megmond, respectively, which are modified
by manner adverbials denoting emotions. The verbs sajnälkozik and
restelkedik are emotives and as emotives they do not tolerate unspecified
embedded that-clauses.

4. Verbs of saying can be divided into four major groups: (a) Verbs of
uttering (e.g. utter, speak, sing, whisper, shout, etc.), (b) verbs of
asserting (e.g. assert, state, say, announce, declare, report, deny, etc.),
(c) verbs of requesting and questioning (e.g. order, request, command,
invite, ask, answer, etc.), and (d) conversational verbs (e.g. talk, discuss,
converse, chat, gossip, etc.). Conversational verbs cannot be used as
matrix verbs in indirect quotations. In general, verbs of requesting and
210 Ferenc Kiefer

questioning are used with questions and requests, verbs of asserting with
statements. Finally, verbs of uttering can be used with statements,
questions and requests alike.
It is a generally held view that any verb that allows direct quotation
also allows indirect quotation but the converse is not true: verbs like
state, assert are not used to make direct quotations.
The problem to be addressed in this section pertains to the difference
between verbs of uttering and verbs of asserting with respect to embed-
ded statements. It has been pointed out that verbs of asserting "appear
to change their meaning as a function of the objects or complements with
which they occur. For example, say, state and assert are virtually synony-
mous when used to make indirect quotations; there are no obvious
differences between the meanings of
He said that the room was bugged.
He stated that the room was bugged.
He asserted that the room was bugged.
"With other constructions, however, this virtual synonymy disappears"
(Miller!Johnson-Laird: 644-645). Notice, however, that apart from the
fact that only say but not state!assert can be used with indirect questions,
there are other important differences between say and state/assert. The
verb say can be used to report nonpropositional elements of an utterance
whereas state!assert admit embedded propositions only. Consider

(24a) Bill said that John would probably arrive tonight,

b) *Bill stated that John would probably arrive tonight.

Sentence (24b) is clearly ungrammatical: one can only state propositions

but not propositions which are modified by an expression of speaker
attitude. In fact, one can use the frame 'X stated t h a t . . . ' in order to find
out what is propositional in an utterance. Of course, there are things
which have to be dropped when say is used as the reporting verb. For
example,

(25a) 'You see,' Henry Cooper said, Ί have a dead baby in my bag.'
b) 'Ah well,' the driver said, 'he won't feel the heat, will he?'

Let's now transform these sentences into indirect quotations:

(26a) Henry Cooper said that he had a dead baby in his bag.
b) The driver said that he wouldn't feel the heat.
 Indirect speech in Hungarian 211

This seems to indicate that interjections and expressions with purely

conversational functions ('you see', 'ah well', tag-questions, etc.) have to
be omitted in any indirect quotation.
In what follows we will further explore the differences between what
can be said and what can be stated.
One thing seems to be quite evident: whatever can be stated can also
be said but the converse is not true. But what are the things that can be
said but not stated?
This is the point where we have to return to Hungarian. Hungarian is
extremely rich in modal particles. The modifier 'modal' should be taken
in its broadest sense: it is meant to cover all kinds of speaker attitudes
(i.e. it is not restricted to expressions of possibility and necessity).
Consider now the following pairs of sentences:
(27a) Peter azt mondta, hogy Jänos bizony reszeg volt.
'Peter said that John had really been drunk'
b) * Peter azt ällitotta, hogy Jänos bizony reszeg volt.
'Peter stated that John had really been drunk'
(28a) Peter azt mondta, hogy dehogyis ärulja el a barätjät.
'Peter said that he would by no means betray his friend'
b) *Peter azt allitotta, hogy dehogyis ärulja el a barätjät.
'Peter stated that he would by no means betray his friend'
(29a) Peter azt mondta, hogy Jänos joformän csak a het vegen van otthon.
'Peter said that John is virtually only on week-ends at home'
b) *Peter azt ällitotta, hogy Jänos jöformän csak a het vegen van
otthon.
'Peter stated that John is virtually only on week-ends at home'
(30a) Peter azt mondta, hogy Jänos okvetlenül eljön.
'Peter said that John would come under any circumstances'
b) *Peter azt ällitotta, hogy Jänos okvetlenül eljön.
'Peter stated that John would come under any circumstances'

It can be seen that whereas the sentences with mond 'say' are
acceptable, the corresponding sentences with ällit 'state/assert' are all
ungrammatical. The attitudes expressed by bizony 'really', dehogyis 'by
no means', joformän 'virtually' and okvetlenül 'under any circumstances'
are speaker attitudes, they are attributed to the speaker. In the case of
mond 'say' these attitudes can be interpreted as being the attitudes of the
person referred to by the subject of the matrix sentence. This interpreta-
tion is out of the question in the case of ällit 'state/assert'. What Peter
states or asserts about the world cannot contain his attitudes. This means
212 Ferenc Kiefer

that the attitudes expressed by the modal particles must be attributed to

the speaker, i.e., to the person who utters the sentences in question. But
his attitudes cannot be part of what the person referred to by the subject
of the sentence states or asserts. They can modify the matrix sentence
only. Sentences such as (31a-b) are fully acceptable.

(31a) Peter bizony azt ällitotta, hogy Jänos reszeg volt.

'To be sure, Peter stated that John had been drunk'

b) Peter dehogyis ällitotta, hogy nem ärulja el a barätjät.

'Peter by no means stated that he would not betray his friend'

One states things about the world: whatever is stated can be true or false.
In other words, one states propositions. Propositions do not contain
expressions of speaker attitudes. Speaker attitudes are, in fact, never
stated: they are expressed or indicated.6 As we saw above, one can say
things which cannot be stated. But one cannot say everything. Interjec-
tions, expressions with purely communicative functions cannot occur in
clauses embedded under mond 'say'. Expressions of speaker attitudes,
on the other hand, are not excluded from these clauses. This seems to
suggest that mond 'say' occupies an intermediate position between the
verbs of uttering and the verbs of asserting. Recall that the grammatica-
lity of the (a)-sentences in (27a-b) - (30a-b) is due to the fact that the
corresponding attitudes can be attributed to the person referred to by the
subject of the matrix sentence.
On the basis of the above observations I would like to put forward the
following hypothesis:
A is an expression of speaker attitude if it can occur in clauses
embedded under mond 'say' but not in clauses embedded under ällit
state/assert'.
Let us now test this hypothesis on further examples. Take, for
instance, the particles i'me 'see, here you are!', ja 'oh, ah' and läm
'indeed, well'.
(32a) ?*Peter azt mondta, hogy ime nehäny pelda.
'Peter said that here are some examples'

b) *Peter azt mondta, hogy ja igaz, holnap nem er rä.

'Peter said that, by the way, tomorrow he would be busy'

c) *Peter azt mondta, hogy, läm igaza volt.

'Peter said that, well, he had been right'

Sentences (32b) and (c) are ungrammatical. (32a) may be acceptable if

the embedded clause is interpreted as a direct quotation, that is if ίme
 indirect speech in Hungarian 213

'see!, here you are! here' is uttered by Peter, the person referred to by
the subject of the main clause. Our hypothesis predicts, then, that none
of the particles ime, ja and läm express speaker attitudes. This squares
perfectly well with our intuitions. These particles seem to have purely
communicative functions which may change according to the speech
situation. Speaker attitudes, on the other hand, have well-defined con-
ventional meanings.
Speaker attitudes are not only expressed by what we called modal
particles but also by sentence adverbials. This has been demonstrated in
independent arguments.7 Let us now see whether this is correctly predic-
ted by our hypothesis.

(33a) Peter azt mondta, hogy Jänos sajnos elment.

'Peter said that, unfortunately, John had left'

b) *Peter azt ällitotta, hogy Jänos sajnos elment.

'Peter stated that, unfortunately, John had left'

(34a) Peter azt mondta, hogy Jänos valöszinüleg elment.

'Peter said that John had probably left'

b) *Peter azt ällitotta, hogy Jänos valoszinüleg elment.

'Peter stated that John had probably left'

(35a) Peter azt mondta, hogy Jänos föltetlenül eljön.

'Peter said that John would definitely come'

b) *Peter azt ällitotta, hogy Jänos föltetlenül eljön.

'Peter stated that John would definitely come'

Thus, our hypothesis seems to be borne out by the facts. The sentence
adverbials sajnos 'unfortunately', valoszinüleg 'probably' and föltetlenül
'definitely, by all means' do not belong to what it asserted about the
world, they are not 'propositional'. Rather they are used to express
various speaker attitudes. These sentence adverbials, as shown by
(33a-b) - (35a-b), can occur in clauses embedded under mond 'say' but
not in clauses embedded under ällit 'state/assert'.
Consider next (36a-b).
(36a) Peter azt mondta, hogy Jänos biztosan elment.
'Peter said that John had certainly left'
or 'Peter said that it was sure that John had left'

b) Peter azt ällitotta, hogy Jänos biztosan elment.

'Peter stated that it was sure that John had left'

This example brings out an interesting point. As indicated by the English

translations, sentence (36a) is ambiguous. The ambiguity is due to the
214 Ferenc Kiefer

polysemy of biztosan which may mean either 'certainly, surely' or 'for

certain, for sure'. In the first case it expresses a speaker attitude and is
not 'propositional', in the second case, on the other hand, it is part of
what is stated about the world, i.e. it is propositional. In contrast to
(36a), sentence (36b) is not ambiguous. This sentence can only be
interpreted as stating that a certain state of affairs (= John left) holds for
sure.
Notice that in certain cases expressions of speaker attitudes may occur
even in the frame 'X states that '. Consider

(37a) Peter azt ällitotta, hogy akär tizszer is megpröbälja.

'Peter stated that he would try it even ten times'
b) Peter azt ällitotta, hogy jöformän egesz nap otthon volt.
'Peter stated that he had been at home almost the whole day'
c) Peter azt ällitotta, hogy koräntsem erzi jol magät.
'Peter stated that he didn't feel well at all'

In the above sentences the particles akär 'even, as well', joformän

'practically, virtually, as good as, almost' koräntsem 'by no means, not at
all, far from it' may be interpreted as being the attitudes of the person
referred to by the subject of the main clause. This interpretation is
possible only if the subject of the embedded clause is identical with the
subject of the main clause. The following sentences in which this is not
the case are all ungrammatical:
(38a) *Peter azt ällitotta, hogy Jänos akär tizszer is megprobälja.
'Peter stated that John would try it even ten times'
b) *Peter azt ällitotta, hogy Jänos joformän egesz nap otthon volt.
'Peter stated that John had been at home practically the whole day'
c) *Peter azt ällitotta, hogy Jänos koräntsem erzi jol magät.
'Peter stated that John didn't feel well at all'

Notice that the interpretation exemplified in (37a-c) is restricted to a

small number of modal particles and it is out of the question in the case
of sentence adverbials. The fact that (38a-c) are ungrammatical suggests
that if the person referred to by the subject of the main clause (i.e. if he
or she is not the subject or agent of the that-clause) the attitudes
indicated in the embedded clause cannot be attributed to the person
referred to by the subject of the main clause. This seems to be a rather
plausible explanation. It should also be noted that in cases such as
(37a-c) the meaning of the verb ällit 'state/assert' gets slightly shifted
toward the meaning of the verb mond 'say'. Unfortunately, however, I
 Indirect speech in Hungarian 215

cannot offer any explanation for the fact that only a few modal particles
admit the interpretation exemplified in (37a-c).
Before concluding this section an important remark must be made
with respect to the English renderings of the Hungarian particles. It is
well-known that particles are extremely difficult to translate. They do
not have a logically clear meaning and they may fulfil a lot of different
functions depending on the context. Hungarian is very rich in modal
particles (over 70), English, on the other hand, does not seem to have so
many of them. It follows almost automatically from this disparity that
what is expressed in Hungarian by means of a modal particle (i.e.
speaker attitude) may have to be rendered in English by something
which belongs to what is stated about the world (i.e. which is part of the
proposition). It may suffice to compare the Hungarian sentences in
(38a-c) with their English equivalents. The Hungarian sentences contain
expressions of speaker attitudes in their embedded clauses whereas it can
be claimed that the corresponding English expressions are 'propositio-
nal.' In fact, in certain cases it seems to be possible to 'propositionalize'
expressions of speaker attitudes. That is, very often when reporting
somebody's utterances one may have the choice of either omitting from
the reported utterance everything that is not propositional or to 'proposi-
tionalize' attitudinal expressions. Compare, for example, the following
sentences:
(39a) * Peter azt ällitotta, hogy Jänos jöformän egesz nap otthon volt.
'Peter stated that John had been at home practically the whole day'
b) Peter azt ällitotta, hogy Jänos majdnem egesz nap otthon volt.
'Peter stated that John had been at home almost the whole day'

The first sentence, (39a), is identical with (38b). If we replace the

attitudinal expression jöformän 'practically, virtually' by the logical
element majdnem 'almost' we get a well-formed sentence (cf. (39b)).
What happens here is that the meaning of jöformän gets 'propositionali-
zed' by majdnem. In other words, majdnem expresses the same thing
propositionally what jöformän expresses nonpropositionally. This is, of
course, true for the given context only. In other contexts, this relation-
ship between jöformän and majdnem may no longer be valid. It is
possible that jöformän has to be 'propositionalized' by another logical
element or that it cannot be 'propositionalized' at all.8

5. To summarize, then, we saw that the reporting verbs in Hungarian

exhibit some properties which the corresponding verbs in other langua-
216 Ferenc Kiefer

ges need not share. Consequently, certain distinctions which are drawn
in Hungarian in an unambiguous way may shed some light on a number
of controversial issues in linguistic theory. It should be clear by now,
among other things, that the factivizing role of question embedding
verbs is more complex than assumed in previous work, that the behavior
of factives in Hungarian provides a clear argument in favor of a semantic
notion of presupposition (it does not allow for an implicational account),
and finally that the difference between the embedding properties of
mond 'say' and ällit 'state/assert' forces us to refine the classification of
verbs of saying and that, at the same time, it helps us to clarify the notion
of speaker attitude.

Notes

1. Cf. with respect to a detailed discussion of factivity in Hungarian Kiefer (1978). An

outline of the aspectual system of Hungarian is given in Kiefer (1982).
2. Cf., for example, Karttunen (1977).
3. I discussed question embedding verbs in terms of attitudes elsewhere (Kiefer 1981).
4. Such an account was advocated, for example, in Kempson (1975) and Wilson (1975).
5. For a more detailed account of the relationship between topic-comment structure and
factivity cf. Kiefer (1978).
6. We cannot go into the details of this difference in the present paper. Cf., however, Lang
(1983).
7. Cf. Lang (1979).
8. Notice that the difference between mond 'say' and ällit 'state/assert' discussed above
divides the class of verbs of asserting into two distinct groups. Verbs such as bevall
'admit', hangoztat 'insist on', emleget 'mention', fejteget 'explain', etc. belong to the ällit-
group whereas verbs such as kijelent 'declare', välaszol 'answer', közöl 'announce', ir
'write', etc. belong to the mond-group.

References

Karttunen, Lauri. 1977. 'Syntax and semantics of questions.' Linguistics and Philosophy 1/
1, 3-44.
Kempson, Ruth M. 1975. Presupposition and the delimitation of semantics. Cambridge:
Cambridge University Press.
Kiefer, Ferenc. 1978. 'Factivity in Hungarian.' Studies in Language 2/2, 165-197.
Kiefer, Ferenc. 1981. 'Questions and attitudes.' In: W. Klein and W. J. M. Levelt, eds.
1981. Crossing the Boundaries in Linguistics. Dordrecht: D. Reidel, pp. 159-176.
Kiefer, Ferenc. 1982. 'The aspectual system of Hungarian.' In: F. Kiefer, ed. 1982.
Hungarian General Linguistics. Amsterdam: J. Benjamins, pp. 293-329.
Lang, Ewald. 1979. 'Zum Status der Satzadverbiale.' Slovo a Slovesnost XL. 3, 200-213.
 Indirect speech in Hungarian 217

Lang, Ewald. 1983. 'Einstellungsausdrücke und ausgedrückte Einstellungen.' Studia

Grammatica XXII, 305-342.
Miller, George A. and Johnson Laird, Philip N. 1976. Language and perception. Cam-
bridge: Harvard University Press.
Wilson, Deirdre. 1975. Presuppositions and non-truth-conditional semantics. New York:
Academic Press.
Reported speech in Danish
Hartmut Haberland

1. Reported speech: General remarks

Reporting another utterance (either one's own or another person's) must

come close to a universal of linguistic action. It requires quite an effort to
imagine a language where this would not be possible. To begin with, we
cannot exclude repetition of that other utterance (which in the following
I shall call the model or model utterance), and repeating the model
utterance would be a way of reporting it. One would only have to make
sure that in repeating the utterance one did not repeat the original
speech act (with all that this would entail), but rather that one displayed
it. Still, there would be the problem of identifying that repetition: as of
what it is a repetition of - whose utterance, made when, where and
directed to whom. If a language has grammatical devices for those two
functions, we would say that it was equipped for reporting speech. Most
languages we know of seem to have these devices in one form or other,
and even languages which are not too well-equipped, somehow seem to
make do with whatever devices they have, mostly relying on context
features. Of course there is still the possibility (unlikely though it may
be) of a culture where one simply does not report people's utterances,
not even one's own. Until such a culture is discovered, we may maintain
the idea that reported speech is a universal of human action.
We are on far less safe ground when we look at direct and indirect
speech as two different ways of carrying out this reporting. In many
languages a more or less clear-cut distinction between direct and indirect
speech seems to exist. Still, we cannot be sure whether this distinction is
a universal1. Investigating this question would in the first place require
making up a definition of direct and indirect speech which is independent
of the grammar of any particular language. A definition of indirect
speech in Classical Greek, for example, would probably have to refer to
the fact that indirect speech in Classical Greek sometimes contains verb
forms in the optative. Such a definition would refer to a category (the
optative) which many, if not most, languages do not have and can
therefore simply be dismissed as a candidate for a general definition.
220 Hartmut Haberland

I would propose the following definition: any report of a model speech

act by another speech act which reenacts the model speech act, is called
direct reported speech (or in short, direct speech). Any report of a model
speech act which merely reports the contents of the model, is called
indirect reported speech (or in short, indirect speech).
In the case of direct speech, we would therefore expect the same
linguistic form as, or a linguistic form similar to the one which carried the
original speech act, that is, often a full sentence. The propositional
content and the illocutionary force of the report are the same as the
model's; the difference is that the illocutionary force of the model
utterance is only indicated or displayed in the report, not performed or
enforced. This is what I mean by 'reenacting the model speech act'. It is a
question of repeating the original (model) utterance without repeating
the original (model) speech act. If a witness in court says, "And then he
said, 'You're a bloody fascist'", the witness in reporting repeats the
utterance of the defendant, but she certainly does not repeat the original
insult (and therefore cannot be sued for it). In reenacting the model
speech act, one can avoid repeating the illocutionary act, but hardly what
Austin (1962: 92) calls the 'phatic' act: as we all have experienced as
children, it is very difficult to tell on somebody for having said a 'dirty'
word without actually using the 'dirty' word again - with all the sanctions
which this implies.
In the case of indirect speech, the standard form would be a subordi-
nate clause having the same propositional content as the model, whereas
the model's illocutionary force either is made explicit by the speech act
verb which the subordinate clause is a complement of, or has to be
deduced from the context.
Direct speech reproduces the deictic references of the original utter-
ance. This is a consequence of the fact that in direct speech, the
perspective of the speaker of the model is assumed, whereas in indirect
speech the perspective of the model is abandoned and the perspective is
that of the reporting speech act. This shift in perspective means that in
Bühler's (1934) terms, the 'here, now and I' of the model speech act (its
'origin', as Bühler calls it in analogy to the origin of a Cartesian co-
ordinate system) is reproduced in direct speech, although the origin of
the matrix speech act is maintained for everything outside the direct
speech. Speakers speak in direct speech as if they were the speaker of the
model. This means that they disregard their own here, now and I inside
direct speech and take the role of the speaker of the model. In direct
speech, speakers refer to the participants in the model speech situation
 Reported speech in Danish 221

in the same way as the model speakers did, using expressions like I2 (for
the speaker of the model) and you (for the hearer of the model). They
refer to third persons in that situation as he, she etc., even if these
persons happen to be participants in the actual speech situation. They
point out objects in the spatio-temporal context of the model speech act
as this etc. They refer to the point of time of the model speech act as now
and its point in space as here. In order to refer to actions, states etc.
occurring or obtaining at the speaking time of the model they employ the
tense used to refer to actions and states occurring or obtaining at the
point of time of speaking (like the present tense in English) and another
appropriate tense (if available) in order to refer to actions, states etc.
occurring or obtaining at other points of time.
In indirect speech, this doubling of the origin will not occur. The
perspective of the model disappears into the perspective of the reporting
speech act. Now, there is only one origin left, that of the reporting
speech act. If the speaker of the model and the report speech act are
identical, the I will still be the same, the now will have changed (and,
with it, often also the tense of the utterance), unless it originally referred
to a time-span encompassing both the point-of-time of the model speech
act and of the report, like if somebody had told me yesterday "Now you
can travel to America faster than 100 years ago" and I reported it today -
now would still be now today. If the speaker of the model and the
reporting speech act are not identical, changes will occur in the personal
pronouns, etc.
If Alice tells me on Sunday that she's going to shoot pheasants on
Monday and will come by with one for my wife and me on Tuesday, then
she might say to me:

(1) /"m going to come by on Tuesday with a pheasant for you.

I could report this to my wife on Monday by saying.

(2) Alice told me yesterday, she's coming by with a pheasant for us
tomorrow.

where I was transformed to she, Tuesday to tomorrow, and you to us.3

These are properties of the reporting utterance which stem from the
fact that its perspective (and, with it, its deictic origin) has changed.
Another set of changes stems from the fact that the form of the report
has changed from an independent main clause which can carry an
illocutionary force to a complement clause which cannot. In many
languages this will imply certain grammatical changes, such as changes in
222 Hartmut Haberland

the sentence mood, in word order, in the possibility of the occurrence of

sentence adverbials and modal particles, etc. These changes are mostly
idiosyncratic, and our earlier decision to dismiss grammatical distinctions
between direct and indirect reported speech as a basis to define their
difference is actually based on the idiosyncratic nature of these distinc-
tions. This distinction might be very striking in languages like Classical
Greek or German where an optional difference of mood exists between
direct and indirect report. But apart from the fact that this difference
usually only is facultative, it is not of much use when we want to define
the difference between direct and indirect reports language-independ-
ently. This difference marks the distinction between the two kinds of
reports more or less clearly, but it cannot be essential for the distinction.

2. Models, direct and indirect reports, and just statements

2.1 Model vs. report

What we have to do with here, is a three-fold distinction between what

we called the model (M) and two types of reports: direct reports (DR.)
and indirect reports (IR):

DR (direct report)

model (M)

IR (indirect report)
One might think that either the direct report or the indirect report
might stand in a closer relationship to Μ than the other. If (1) is a model
and (2) contains an indirect report of (1), then one might be tempted to
see the direct report of (1) contained in (3),
(3) Alice said to me, "I'm going to come by on Tuesday with a pheasant
for you."
 Reported speech in Danish 223

as something much closer to (1) than the indirect report contained in (2).
To make it even clearer: if we compare (1) with the report parts of (2)
(here called (2')) and (3) (here (3')), we get the following picture:
(1 ) I'm going to come by on Tuesday with a pheasant for you.
(3') I'm going to come by on Tuesday with a pheasant for you.
(2') She's coming by with a pheasant for us tomorrow.

Here, (3') and (1), the direct report and the model, are identical.
Therefore, it might seem as if direct reports are somehow more prim-
itive, or less derived than indirect reports.4 Accordingly, Coulmas (1985)
characterizes direct speech as being closer to what is reported than
indirect speech reports, since speakers commit themselves in two ways
with direct reports, but only in one with indirect reports:
"Both direct and indirect speech serve the function of marking a statement as
that of someone else than the speaker. The main difference between them can
be seen to lie in the speaker's attitude towards the reported speech. In
marking an utterance as a direct quotation, he commits himself to faithfully
rendering form and content of what the original speaker said; this is what
direct quotation suggests. An utterance marked as indirect speech, on the
other hand, implies a commitment about the contents but not about the
form." (Coulmas 1985: 42)

I agree with Coulmas that one should talk about 'commitment to being
faithful' here rather than considering 'faithfulness' as a convention or the
like. This is in keeping with Wunderlich's proposal (1977: 16) that
truthfulness is not a convention for statements. Rather it is a kind of
safeguard-rule for speakers that they can avoid problems, if they speak
truthfully, which follows from the fact that speakers commit themselves
to whatever they have said. It is wiser to commit oneself to something
true than to something false. In the same way, it is wiser to commit
oneself to somebody else having said something which that person
actually has said than to something else (at least on a great number of
occasions). This does not rule out that one might commit oneself to
somebody (or oneself) having said something which actually has not
been said, and an utterance implying such a commitment would still be a
quotation. But is it actually the case that speakers of direct reports
commit themselves to rendering the form of the model? I would main-
tain that the identity of (3') and (1) is by no means accidental, but that
this identity in form between model and direct report is not a necessary
property of the relationship between model and direct report. The
material this presentation is based on5 contains a number of cases of
224 Hartmut Haberland

direct speech, where it is either impossible or unlikely that they are cases
of direct quotation. It is quite clear that no actual commitment to the
form of the model is involved where variables like such-and-such or so-
and-so are employed in direct speech:
(4) [Dagny: 110]
Sä kommer man til en d0dssyg generalforsamling, og de forstär ikke
en b0nne af, hvad formanden stär og siger til dem . . . " . . . og der og
dir har de faktisk lavet det og det..de forstär det faktisk ikke.
'Then you go to a dead boring general meeting, and they do not
understand a word of what the shop steward is talking to them
about . . . "... and there-and-there they have got done such-and-
such ...", actually, they do not understand it.'6

This cannot be called quotation. Rather it is sketchy reenactment of the

original scene.7
Sketchy enactment is also found in other languages, like in Greek.
The following example is from Tannen (1983: 365):
(5) . . . απλώς, »ήρθα από τον τάδε«, και τα λοιπά
'[I] simply [said], "I come from such-and-such", and so on.'

Here από τον τάδε 'from such-and-such' is a variable, obviously not used
in the model, where the actual name of the person would have been
used.
This is similar to cases where direct speech is employed to characterize
a whole class of models. Here we do not have a literal quote, rather an
enactment of a typical or average scene:

(6) [Lizzie: 33]

Det er jo sädan, at när man kommer hen et sted, bliver man spurgt
om . . . er du alene, har du born, hvor mange, hvornär blev du alene,
hvorfor blev du alene - alle de sp0rgsmäl der, de gär igen, ligemeget
hvad for en arbejdsplads, man kommer pä.
'Well, it's like this, when you come to a place, you're asked about...
are you on your own, do you have children, how many, since when
have you been on your own [lit. when did you become on your own],
why are you [lit. did you become] on your own - all these questions,
they come again, no matter what kind of a place of work it is you
come to.'8

Another example is the following:

(7) [RP: 4a: 169]
Jeg havde en hvor moren sagde til ham, "Gä op og ring pä, hos
Reinhard der, inden du gär i skole", og sä har han ringet pä, jeg har
sagt "ja", og sä er jeg faldet i s0vn igen.
 Reported speech in Danish 225

Ί had a [friend], whose mother said to him, "Go up and ring the bell,
at Reinhard's there, before you go to school", so he rang the bell, I
said "yes", and then I fell asleep again.'

It must be mentioned in passing that there is also another phenomenon

which could be called 'hypothetical direct speech'. The speaker reports
something which some other speaker might have said, but which he or
she actually did not say, like the following:

(8) [Karen: 13]

for sä kunne han jo bare ha' sagt til mig, "Ja, D e kan godt gä hjem -
glaedelig jul".
'because, in this case, he might as well have said to me, "Well, you
might as well go home - Merry Christmas!".'

Also in this case it is problematic to talk about a 'direct quotation'. The

model is just imagined by the narrator.
These examples show that the use of direct speech only implies a
commitment to the content of the model speech act, viz. to its intention
(illocutionary force) and its propositional content. No commitment to its
wording (its form) is implied. It is only true that if one wants to commit
oneself to the form of the model as well, then one has to avail oneself of
direct speech as a means of reporting the model speech act, since this is
the only way of quoting directly. One might restrict the term 'direct
quotation' used by Coulmas to those direct reports which show this
identity. The connection would then be: if one wants to quote directly,
one has to use the form of direct report, but not every case of direct
speech is a direct quotation.
On the other hand, one has to question how far commitment to the
faithfulness of the contents of the report applies to both direct and
indirect reports. This question could be answered empirically in a very
simple way if we had access to sufficiently many pairs of models and
direct reports, and of models and indirect reports, respectively, in the
same corpus. Unfortunately, I have not been able to find a single case in
my material where both the model utterance and a report in direct
speech were represented. This is not surprising in the case of narratives,
which usually are not self-referential. In interviews, it is possible to find
references to earlier parts, or earlier sessions, of the same interview, but
for some reason these tend to be in indirect speech:

(9) [RP: 3a: 263]

D u fortalte han blev afskediget efter en strejke.
'You said he was fired after a strike.'
226 Hartmut Haberland

In some cases, it is actually possible to trace an instance of reported

speech back to its model. In my corpus, I have the following question in
an interview:
(10) [RP: 3a: 512]
Jörgen: Du siger, da du var omkring tre är, da flyttede I i
Hj0rringgade.
'You say, when you (sg.) were about three years old,
you (pi.) moved to Hj0rringgade.' 9
Now,what J0rgen says is certainly an example of reported speech, and it
is equally certain that it is not an example of direct speech, which can be
seen from the use of du 'you' (sg., informal) for reference to the listener.
(In direct reported speech, one would expect jeg T . ) But what is the
model speech act to which J0rgen's speech act refers? He has talked with
the interviewee twice about his family moving from Gernersgade to
Hj0rringgade. The first time in the following bit of dialogue:
(11) [RP: 3a: 211]
J0rgen: Hvornär flyttede I:: . . . eller hvornär flyttede
r de
Reinhard: L til til til til Hj0rringgade Hj0rringgade? Ja, ja det er
jo altsä inden, der mä- der der der mä jeg have vaeret
der mä jeg have vaeret en fire, fem är . . .
J0rgen: da I flyttede til Hj0rringgade?
Reinhard: ja, og sä flyttede vi til Koldinggade
'J0rgen: When did you:: (pi.) . . . or when did they
Reinhard: to to to
to Hj0rringgade Hj0rringgade? Well, well now that's
before, there mu- then then then I must have been . . .
then I must have been some four, five years . . ,
J0rgen: when you (pi.) moved to Hj0rringgade?
Reinhard: yes, and then we moved to Koldinggade.'

Roughly twenty minutes later the topic is brought up again, this time in
connection with the question of how long the interviewee's family lived
in the flat in Gernersgade, that is, the place where they lived before
moving to Hj0rringgade:
(12) [RP: 3a: 503]
J0rgen: Hvor laenge var det I boede i Gernersgade?
Reinhard: Ja, der har jeg vaeret, Gernersgade, hvad har jeg
vaeret, jeg . . . har vaeret en tre, fi-, tre är, tre fire är, da
vi flyttede, det er altsä . . . det er ikke meget for- og sä
flyttede vi jo i Koldinggade, der gik jeg i skole. Sä det,
der kan ikke vaere mange är der i, der fra fire til
syv, de tre är der, vel.
 Reported speech in Danish 227

J0rgen: Du siger da du var omkring tre är, da flyttede I i

Hj0rringgade
'J0rgen: How long time did you (pi.) live [lit. was it that you
lived] in Gernersgade?
Reinhard: Well, then I was, Gernersgade, how old [lit. what] was
I, I . . . was three, fou-, three years [old] when we
moved, that's now . . . that's not very wro- and then
we moved to Koldinggade, there I went to school. So
that, it can't be many years there in, there from
four to seven, those three years there, can it?
J0rgen: You say when you (sg.) were about three years old,
you (pi.) moved to Hj0rringgade'

What was quoted as (10) above, occurs immediately after (12) and has
been repeated here for convenience.
Looking for the model, one realises that there isn't any, at least not in
the form of something like Da jeg var omkring tre är, da flyttede vi i
Hj0rringgade 'When I was about three years old, we moved to Hj0rring-
gade'. What is reported here is the contents of the two bits of dialogue
quoted, not any particular utterance appearing in them. So the model
must consist of the entire two bits of dialogue. This means stretching the
concept of a 'model' quite far - possibly beyond what some people might
think is advisable - but the alternative seems to be not to call cases like
(10) reported speech at all. This seems counterintuitive to me, since
cases like (10) obviously are experienced as cases of indirect speech,
independently of whether we recognize or remember the underlying
model or not. Even if a reply like "I did not say that" followed (10), (10)
would still be a case of indirect reported speech.
But if we accept a model-report relationship between (12) (and (11))
and (10), then we also have to face the fact that the interviewee's
statements as to when his parents moved from Gernersgade to Hj0rring-
gade are not quite consistent in (11) and (12). If we require that indirect
speech be faithful to the contents of its model (which is different from
stating that the speaker has committed him- or herself to having deliver-
ed a faithful report), then (10) would qualify only as an indirect speech
report vis-ä-vis (12), not vis-ä-vis (11). This is clearly undesirable: we
want to have both (11) and (12) as models of (10).
The result of this is that neither model and indirect report, nor model
and direct report stand in a straightforward relationship to each other.
Often we can reconstruct a 'possible model' - and that is obviously what
that literature at the same time has drawn heavily on what one could find
of reporting styles in the vernacular and the verbal folk arts.
228 Hartmut Haberland

2.2 Reported speech vs. simple speech acts

Sometimes it is difficult to see whether some utterance is to be interpre-

ted (by the analyser) as reported speech or not. One cannot always tell
from the form of an utterance which is which. This means that reported
speech is a pragmatic phenomenon, although it has certain syntactic
manifestations. This can be made clear by the following example (which
is adapted from Stubbs (1983: 22)10):
The utterance

(13) There was money there.

looks as innocent as can be and seems to be a statement. But in the

following context it becomes clear that it also can function as an example
of reported speech, more precisely, as an example of indirect speech:

(14) A: you told me when I was about five - r or six

B: L there was money there
A: there was m- money hidden there

It is clear that the sentence uttered by Β and formally identical to (13)

could be 'catalysed' (to use a glossematic term) or 'expanded' (which to
some people might be a more familiar term) into

(15) Β: (I told you that) there was money there.

That would mean that Β states that he told A that there was money in the
'haunted house'. This is no statement to the effect that there was any
money, actually, and there is no reason to assume that there was any
money there at all. But the utterance could alternatively be a statement
made by Β to the effect that there actually was money there; a statement
triggered off by B's remembering that he had told A so. In the first case
we would have a case of reported speech, that is, here, indirect speech.
In the second case we would just have a simple statement.
The actual decision as to how to interpret B's utterance can only be
made by inspecting the rest of the story and finding out that there
actually was no money in the haunted house. (This applies to the
analyser, not the participants, who know it all the time.) So B's utte-
rance, which in the first place looked like an innocent statement, turns
out to be a case of indirect speech - without, in this case, any syntactic or
semantic marking.
The following is a similar case.
 Reported speech in Danish 229

(16) [RP: 4a: 086]

Og sä havde vi jo ogsä spisning pä skolen dengang, jo det der med
varm mad men det ville min far ikke have, det var for de fattige og sä
videre, det. Han var en af de gamle stolte socialdemokrater.
'And then we had school meals, too at that time, well, [that there
with] hot meals, but my father didn't like it at all [lit. thät my father
did not like at all], that was for the poor and so on, [that]. He was one
of the old proud Social Democrats.'

The clause det var for de fattige 'that was for the poor' can either be
interpreted as a report of a statement by the interviewee's father, who
had said on some occasion that school meals were for the poor, 11 or
alternatively, it could be the interviewee's explanation of why his father
did not like him taking part in school meals. 12 In the first case the
interviewee would not necessarily identify with his father's attitude to
school meals; in the second, he would use his own attitude to school
meals (that is, that they are for the poor) as an explanation of his father's
attitude towards them. In this case, there is a formal means of arguing
for the interpretation as reported speech: it is the formula og sä videre
'and so on', which is similar to variables like 'such and such' etc. used in
what I called 'sketchy reenactment': the interviewee's father used to say
that school meals were for the poor, and he used to say a lot more, which
the interviewee does not quote here.
A related case is the following:

(17) [Karen: 11]

og det var altid os, der skulle rydde op efter dem, det var kutyme
derude, et var det kvinderne var der til, sagde de.
'and it was always us who had to clean up after them, that was a
custom there, that was what the women were there for, they said'

Det var kutyme derovre could either be part of the narrator's description
of her place of work, where it was part of 'the code' that the women
cleaned the place after the men had left. It could also be part of the
men's explanation of 'the code' to the woman. In the latter case sagde de
'they said' would have det var kutyme derude, det var det kvinderne var
der til as its scope, in the former only det var det kvinderne var der til. The
difference would be the degree to which the narrator accepts 'the code':
either she is saying "that's the way it was" or she is telling that the men
told her at that time "that's the way it is", without accepting this as her
own conception of the code. (The formal difference is that the past tense
in the first case is part of the narrative, whereas in the second case,
230 Hartmut Haberland

reported speech, it comes in by deictic shift.) In this case, it is very

difficult to decide between one of the two equally possible interpreta-
tions.

2.3 Direct speech vs. indirect speech

Another problem is: how can we distinguish between what is a direct

report and what an indirect report, in any specific case? Our defining
criterion - the assumption of two deictic origins vs. one - has the
disadvantage that it does not directly correspond to an element or
elements on the level of expression (or form). We will have to identify
what is direct speech and what is indirect speech by reconstructing the
deictic origin(s) involved, and from observing the secondary grammati-
cal distinctions associated with the two forms of report. Of course, it will
depend on the grammar of the language in question whether the two sets
of characteristics that differentiate between direct and indirect speech
(viz. deictic and grammatical) clearly mark off the two forms of reported
speech from each other. To take an extreme case: in a language that does
not have obligatory surface subjects, and where subject pronouns are
relatively rare, which furthermore has an aspect system rather than a
tense system, which does not have basic word order differences between
main and subordinate clauses, and which has no special grammatical
mood to mark either complement or quotative clauses, the difference
between direct and indirect speech as it is defined here might hinge on
the deictic shift of temporal and local adverbials alone. We can assume
that adverbials of this kind are not nearly as frequent as all the other
expressions mentioned before. This situation would not make for a very
clear distinction between the two types of reported speech, simply
because direct and indirect speech in a majority of cases would look
exactly alike. In fact, one would expect that speakers of such a language
would not pay much attention to any difference of this kind, and that the
difference between the two concepts never would have developed into a
received distinction in narrative or literary style. In a language and a
tradition like that it would not make much sense to distinguish between
direct and indirect speech at all.
Not even in a language like English, where most of the distinctions
mentioned above usually can be made, is it always easy to find out what
is direct and what is indirect speech. Take the following passage from
Paul Theroux' Old Patagonian Express:
 Reported speech in Danish 231

'"My daddy had two guns,' said the preacher. One here and one here.' He
spanked two pockets.
The lady said that one day her daddy had tried to take a gun into a Dallas
department store. He was just a stranger in town, from San Tone. Woke up
that morning and strapped on his gun, like he always done. Nothing funny
about that. Done the same thing every day of his life. Went in the store,
packing his old gun. He was a huge man, way over six feet tall. The
department store girls figured it was a holdup as soon as they seen him. They
stomped on the alarm. All hell busted loose, but daddy didn't mind one bit.
He pulled out his gun, and when the police come along, daddy said, 'Okay,
boy, let's git 'im!'
The lady's husband said that daddy had been eighty-four years old at the
time." (Theroux 1979: 29)

The story reported here clearly starts off in indirect speech (The lady said
that...) and ends in the lady's direct speech (He pulled out his gun ...),
even to the extent that it contains an extra layer of direct speech inside
the lady's direct speech, where she reports her father's utterance. But
where precisely is the transition point? Where does indirect speech stop
and direct speech take over? Moreover: even if it could be determined
analytically where the breaking point is, would it be relevant for the
analysis of the whole passage? Is it not rather the case that what is going
on here is a smooth transition from direct to indirect speech?
On top of this, the last sentence quoted here contains a mixture of
direct speech and indirect speech elements as well. The sentence in itself
is in indirect speech as is made clear by the that introducer of the
complement and the pluperfect tense; but daddy is certainly a non-
shifted (person) deictic element: Theroux would not call that Texan man
"daddy", but the lady's husband would.
In Danish, very often features of direct speech and indirect speech are
mixed in one utterance, as in the following example:
(18) [Karen: 12]
Han spurgte, hvorfor jeg ikke var der mere. Nej, jeg var gäet, fordi
jeg ikke kunne fä mere i l0n.
'He asked, why I was not there any more. No, I had left, because I
could not get more pay.'

Here jeg var gäet, . . . Ί had l e f t . . . ' is clearly not direct speech (a direct
report would have been jeg er gäet Ί left'). But nej 'no' is clearly a direct
speech feature. 13 It almost looks as if direct and indirect speech were
distinguished by degree rather than categorically, at least in Danish, and
at least in the verbal style investigated here. Ulis is because the features
constitutive for indirect speech do no always occur in a bundle.
232 Hartmut Haberland

Literary theory has known this for a long time, but has rather
complicated the matter by assuming a third, distinct category, which is
variously called style indirect libre (Bally 1912), erlebte Rede (Lorck
1921), or oratio tecta (Br0ndum-Nielsen 1953) (Neubert (1957: 8-9)
gives a long, though not complete, list of the various terms used).
However, the style indirect libre as a category of literary stylistics must be
distinguished from the same phenomenon as a structural-grammatical
category. (One could call this structural-grammatical category 'quasi-
direct discourse', following the English translation of VoloSinov (1929)
(Volosinov 1973)). As a stylistic feature, it is a historical category, as is
said quite clearly by Neubert:
"ER [erlebte Rede] ist ein stilistisches Mittel zur Darstellung von Bewußt-
seinsvorgängen. Die Möglichkeit ihres Auftretens in der Literatur wird also
zunächst davon abhängen, inwieweit ein vertiefter Einblick in das Innenleben
seiner Gestalten vom Schriftsteller als bedeutsam und zur Verwirklichung
seiner künstlerischen Absichten notwendig erachtet wird. Andererseits
kommt in einer immer verfeinerteren Anwendung der ER die steigende
Wichtigkeit zum Ausdruck, die die Autoren der seelischen Introspektion
beimessen. Es nimmt deshalb nicht wunder, dass die ersten bewußten Ansätze
im Gebrauch der ER erst zu Anfang des 19. Jahrhunderts festgestellt werden
können, nämlich bei Jane Austen." (Neubert 1957: 11)
It is also clear why literature has used this category at a certain historical
stage rather than at another: whereas direct speech as a grammatical
phenomenon is suitable for reporting speech, it is less well suited to
reporting thoughts. Quasi-direct discourse, on the other hand, although
well suited for reporting speech as well, is equally suited for reporting
thoughts. At the moment where literature becomes interested in report-
ing 'states of consciousness' or 'mind', it would have to develop an
interest in these forms of quasi-direct discourse. But, as M0ller Kristen-
sen (1938: 35) has pointed out, one should not get confused by the fact
that the same grammatical device (which I prefer to call quasi-direct
discourse, if it is to be considered as a category in its own right at all) can
be used with two completely different functions: reporting other peoples'
thoughts (as in the literary inner monologue) and reporting other
peoples' (or one's own) speech acts. As to the latter function, the
phenomenon is much older in Danish literature than often assumed
(J. P. Jacobsen (1847-1885) is often considered the first writer in Danish
literature to employ the inner monologue extensively). As Br0ndum-
Nielsen (1953) has pointed out, it can not only be found in H. C.
Andersen (1801-1875), but also in letters and historical literature of the
18th and even the 17th century, like in Princess Leonora Christina's
 Reported speech in Danish 233

(1621-1698) "Jammersminde". His conclusion is that quasi-indirect dis-

course, or, as he calls it, oratio tecta, goes back to a fairly old feature of
the vernacular. At least in Danish, quasi-direct discourse is therefore not
a stylistic means mainly to be found in literature, as VoloSinov seems to
imply (1973: 147ff., 158ff.): as we shall see in the following, it is still very
much alive in colloquial Danish. Br0ndum-Nielsen also points out that
Jacob Baden is the first Danish grammarian to have described it. In his
lectures on the Danish language from 1792, he mentions a 'historical
subjunctive', which is a use of the past tense to report what another
person has said:
"Hvad A. har sagt, fortacller B. saaledes: 'Han (nemlig A.) kiendte nok disse
ski0desl0se Folk; men dersom de ikke passede paa deres Sager, vilde det gaae
dem ilde.' Men Or dene har A. saaledes udf0rt: 'Jeg kiender nok disse
ski0desl0se Folk; men dersom de ikke passe paa deres Sager, vil det gaae dem
ilde.' Dette kan kaldes en historisk Coniunctivus i Dansken, som beretter en
andens Ord, uden at indf0re dem selv med uforandrede Former, men gemeen-
lig omvender Praesens af den andens Tale til Imperfectum, Perfectum til
Plusquamperfectum, den f0rste Person til den tredie, og den anden underti-
den til den f0rste." (1792: 235)14

B's utterance in the 'historical subjunctive' is characterised both by main

clause word order (kiendte nok, cf. (23) and (24) below), viz. a direct
speech characteristic, and by deictic shifts of tense and personal pro-
nouns, viz. indirect speech characteristics. So it is neither direct reported
speech nor indirect reported speech, and since a matrix verb of saying is
lacking as well, we have a case of quasi-direct speech.
In the following we shall look more systematically at Danish direct
and indirect speech reports, and investigate whether the distinction
between direct and indirect speech which we have defined on the basis of
their different deictic relationships to the utterance they are embedded
in, is always paralleled by the grammatical distinctions to be expected.
This requires a very short outline of Danish grammar first.

3. A brief outline of Danish grammar

Danish is a Germanic language of the Northern Germanic or Scandina-

vian sub-group. Typologically, it belongs to what Perlmutter calls Type
A languages (Perlmutter 1971: 115; Haiman 1974: 90f.), that is, in every
clause or sentence (except main clauses in the imperative), the subject of
234 Hartmut Haberland

the finite verb is expressed either by a full NP or at least by a pronoun or

subject dummy. There are no person or number markers in the verb.
The verb's moods are indicative, imperative (and, if you call it a mood,
infinitive)15. The Danish verb has two non-compound tenses, present
and past, as in the following verbs.
present past
flytter 'move(s)' flyttede 'moved'
bor 'live (in a flat)' boede 'lived'
falder 'fall(s)' faldt 'fell'
fylder 'fill(s)' fyldte 'filled'
bliver 'become(s); stay(s)' blev 'became; stayed'
kommer 'come(s)' kom 'came'
fär 'get(s)' fik 'got'
kan 'can' kunne 'could'
er 'am; are; is' var 'was; were'

The past tense also functions as a conditional in counterfactual condi-

tional clauses pretty much like in English. There is no future tense;
future time reference is usually made by the present tense (like in
German) or, more rarely, with the help of modal auxiliaries (like in
English), as skal 'shall', vil 'will'. There are two compound 'tenses',
present perfect and pluperfect. The present perfect is formed from the
present tense of one of the auxiliaries at have 'to have' or at vcere 'to be'
and the past participle, and we can describe the pluperfect as the past of
the present perfect.

present perfect pluperfect

jeg har lavet Ί have done' jeg havde lavet Ί had done'
jeg er kommet Ί have come' jeg var kommet Ί had come'
jeg har vaeret Ί have been' jeg havde vaeret Ί had been'

(Roughly speaking, perfective intransitive verbs are combined with at

vcere, others with at have).
As far as word order is concerned, Danish is basically a SVO
language, or, rather: a finite verb-second (V/2, TVX)-language. In a
main clause, there is one 'position' in front of the verb, which usually is
filled with the subject, but which most other constitutents of the sentence
can occupy as well, if they become fronted for either topicalization or
emphasis. Non-fronted subjects appear immediately after the verb.

(19) Ham kendte jeg i 100 är

him knew I for years
V
 Reported speech in Danish 235

(20) Sä d0de han.

then died he.
V

The fronted element can be a subordinate clause:

(21) Da jeg var tre är, flyttede vi.
when I was three years (old) moved we
V
(22) Hvis du kommer, gär jeg.
If you come go I
V
'If you come, I'll go.'

In subordinate clauses, word order is basically the same as in main

clauses, except that the negation ikke and a class of adverbials, which
occupy the same positions in a sentence as ikke, precede the finite verb
in subordinate clauses whereas they follow it in main clauses (with
possibly a non-topic subject in between). This class of adverbials in-
cludes a number of words like sgu 'by God!', da (roughly like German
'doch'), and nok 'probably', which are extremely frequent in the spoken
language. Therefore, the placement of these adverbials is a highly
operative test for determining whether a given verb is the verb of a main
or subordinate clause:
(23) Hun ringede ikke hjem
She called not home
V Neg
(24) Jeg blev urolig, da hun ikke ringede hjem.
I got nervous because she not called home.
Neg V
In (23) the verb precedes the negation (main clause), in the subordinate
clause of (24), the negation precedes the verb.
In indirect questions, the question word is considered as standing
outside the 'finite verb second' pattern, which gives indirect questions a
different shape from direct questions:
(25) Jeg ved ikke, hvornär han d0de.
I know not when he died
V
(26) Hvornär d0de han?
When died he?
V
In the main clause (26), the Λν-word occupies a position analogous to a
topic, whereas it is placed like a conjunction in (25).16
236 Hartmut Haberland

Fronting inside a subordinate clause is possible, but subject to strong,

not too well unterstood restrictions. On the other hand, it is possible to
extract an element from a subordinate clause and make it the topic of the
main clause. In this case, no trace of the topicalized element is left in the
subordinate clause17:

(27) s[J e g ve
d s[du elsker ham]]
I know you love him
(28) s [Hami ved jeg s [du elsker J
Him know I you love

If we look for criteria for calling something 'direct' rather than 'indirect
speech', or vice versa, we can use the difference in the treatment of
personal pronouns. If any deictic shift occurs here, we have a piece of
indirect speech, otherwise not. This will not work in all cases, since not
all pronouns (especially not all third person pronouns) are affected by a
change of the deictic origin. On the other hand there will be quite a lot of
them since subject NPs are obligatory, and NPs are minimally realized
by pronouns. There is also the possibility of distinguishing between
direct and indirect speech by the deictic shifts of temporal and local
adverbials, although these shifts are not so obvious since the elements
involved do not occur as frequently as subject and object pronouns. In
the verb, a grammatical marking of indirect speech by a special mood is
not possible, but direct speech imperatives will, of course, be matched by
a modal auxiliary construction in indirect speech. We could, however,
expect deictic switches in tenses.
These criteria will be useful; on the other hand, they cannot be
applied mechanically: reconstructing the deictic mechanism presupposes
understanding of the utterance in question and at least a knowledge of
context. But there are other distinctions such as word order: in a
construction with a speech act verb V like sige 'say', we could distinguish
between two cases of NP V S: either S has main clause word order or it
has subordinate clause word order. We would expect the first if S is a
case of direct reported speech, and the latter if it is indirect reported
speech. Furthermore, any extraction of an element from an embedded
clause into a matrix clause would be unlikely in the case where the
embedded clause is not a subordinate clause (that is, direct reported
speech); extractions of that type would therefore indicate indirect
speech. Indirect reports should also be characterized by a complementi-
zer at 'that', and the occurrence of a verb of saying. And finally, we
would except that elements which cannot be transformed into comple-
 Reported speech in Danish 237

ment clauses or parts of such clauses due to their lack of a propositional

content, would only appear in direct speech.

4. Deictic expressions in Danish reported speech

If one considers the difference in deictic orientation as basic to the

distinction between the two kinds of reported speech, one should first
look for deictic expressions in the utterances analysed.

4.1. Time and place deixis

In my material, there are a couple of utterances with such deictic

expressions.
(27) [Karen: 15]
"Ja, naturligvis," siger jeg, "jeg kom da hjem i fredags."
"Yes, of course', I say, Ί came home last Friday after all."
(28) [Karen: 24]
. . . og sä sagde jeg: "Ja, nu skal vi se, om vi fär det lavet i dertne her
uge, ellers laver vi det samme nummer p& mandag".
' . . . and then I said, 'Well, now we'll see, if we can get it done
[during] this week, otherwise we'll play the same trick on Monday.'.'

In these cases it is clear that the temporal expressions refer to the point
of speech of the model: i denne her uge 'during this week' refers to the
moment of speech of the model situation, not of the report situation.
Likewise, pä mandag 'on Monday' refers to the Monday following the
model speech situation, not the Monday after the report situation (the
interview). In principle, there should be no problem in transforming
these two adverbials into expressions suitable for indirect report: i
fredags 'last Friday' would become fredagen f0r 'Friday before', denne
her uge 'this week' would become denne der uge 'that week', pä mandag
'on Monday' would become den f0lgende mandag 'the following Mon-
day' or just om mandagen 'on Monday'. 18 Expressions like these appear
in the sections reporting events in the same texts, but they are conspic-
uously missing from sections reporting speech. This might be accidental
or a bias in my corpus; it could also hint at a general tendency, namely
that the occurrence of time adverbials of this kind (which are not all that
238 Hartmut Haberland

frequent to begin with) trigger off direct speech, in an attempt to keep

deictic complexity in the text low. Expressions denoting 'relayed' tempo-
ral relationships tend to be slightly complex (in Danish as well as in other
languages).

4.2. Deixis into the spatio-temporal context

Similarly, deictic expressions connected with pointing gestures would

require quite elaborate circumlocutions in indirect reports and tend to
trigger off enactments of the model scene rather than reports containing
indirect speech:
(31) [Tove: 68]
Da de havde gennems0gt hele huset og kom tilbage, fik de pludselig
0je pä et billede af os allesammen pä skrivebordet. "Hvem af dem er
det?" Jah, det var Hans, og det var Erik.
'When they had searched the whole house and came back, they
suddenly saw a picture of all of us on the desk. 'Which of them is it?'
Well, this was Hans, and that was Erik.'

On the other hand, this example contains deictic shifts for the very
deictic element where such a shift can be accomplished easily, viz. tense:
var 'was' is past tense. This means that the deictic origin relevant for the
local deictic elements is the origin of the model scene, whereas the
deictic origin for the tense of the verb is the origin of the reporting
utterance. Again, we have a case of two deictic origins involved inside
one report, which makes the report something in between direct and
indirect speech. We also have a case (like with time adverbials, see
section 4.1) where the choice between a shifted element and a non-
shifted element seems to depend on the ease with which the deictic shift
can be accomplished.

4.3. Person deixis

It is almost not necessary to mention examples of shifted personal

pronouns. They function in the most obvious way:
(32) [Karen: 12]
og han havde ovenik0bet spurgt mig [ . . . ] om jeg kunne trykke
'and he had asked me on top of that [ . . . ] if / knew how to print'
 Reported speech in Danish 239

The model corresponding to this case of indirect reported speech can

be reconstructed as something like Kan du trykke? 'Do you know how to
print?'. An interesting case is also the following, which involves personal
pronouns:
(33) [RP: 4a: 187]
Det er hende der kom hjem og sagde hurt havde vaeret sygeplejerske
for min mor.
'That's her who came home to us and said she'd nursed my mother
[lit. she had been a nurse for my mother].'
Hie scene which is referred to here we must imagine as follows: after the
interviewee's mother has died, this woman (her name is never mention-
ed) appears at the door and claims she has nursed the interviewee's
mother (in order to get a job as a house-keeper). Now what she would
have said might have been something like Jeg har veeret sygeplejerske for
fru P. Ί have nursed Mrs. P.' The shift from jeg Ύ to hurt 'she' is an
obvious case of deictic pronoun shift of the type we find described in
even the most simple-minded accounts of reported speech; but the shift
from fru P. 'Mrs. P.' to min mor 'my mother' is different. It involves two
things: on the one hand, the way a person is identified in speaking. For
the interviewee, what is 'the interviewee's mother' or 'Mrs. P.' in the
context of this article, is min mor 'my mother'. This is a case, where not
just the origin of Bühler's Zeigfeld, but the whole cognitive map of the
different speakers is shifted between the two speech acts. Here, person
deixis involves also social deixis: even if to the 'woman' mentioned the
person she is speaking about is one fixed individual, she can refer to her
in different ways, and is even obliged to do so according to whom she is
speaking to. Talking to the interviewee's father she would have had to
refer to Deres afd0de kone 'your late wife', whereas she actually could
have said something like din mor 'your mother' to the interviewee
himself who was eleven years old at that time.

4.4. Tense deixis

Sentence (31) also contains a verb form havde vaeret which is in a tense
different from what we would expect for the corresponding verb form in
the model har vceret. Since we decided to consider havde vaeret (the
'pluperfect') as just the past of har vceret (the 'present perfect'), this
seems to be a case of a deictic shift where a present becomes a past in
indirect speech. This fits nicely in with cases like
240 Hartmut Haberland

(34) [RP: 4a: 104]

jeg h0rte at hun skulle til hospitalet
Ί heard that she had to go to the hospital'

where we can imagine a model like "Hun skal til hospitalet" '"She has to
go to the hospital'", where the model has the present tense and the
indirect report has the past. But we have to be careful not to see this
'shift one tense back' or 'shift to the past' in Danish indirect speech as a
direct analogue to the subjunctive in Standard German and the optative
in Classical Greek 19 . This shift occurs only if the relationship of the
point-of-event of the reported speech act to the point-of-speech of the
reported speech act is different from the relationship of the point-of-
event of the reported speech act to the point-of-speech of the reporting
speech act, that is, different in a way that would affect the expression of
this relationship through the tense system of the Danish verb. If this is
not the case, no shift occurs, as in the following (made-up) example:
(35) A: (on the telephone): "Ved du hvorfor Peter ikke kom til
m0det?"
B: (Peter's sister): "Han er i Amerika."
Ά: "Do you know why Peter didn't come to the meeting?"
B: "He's in America.'"
Now, if A is to report what Β told him on the following day, he could
say something like Jeg ved det ikke, men da jeg snakkede med hans s0ster
i gär, sagde hun til mig at han var i Amerika Ί don't know, but when I
talked to his sister yesterday, she said that he was in America.' But
America is quite far away from Denmark. So A can assume that when
Peter was in America yesterday, there is a chance that he is still there
today. So A's report can as well take the form . . . sagde hun til mig at han
er i Amerika ' . . . she said that he is in America'. The difference between
reporting that Peter's sister had said that Peter was in America at the
time Peter said it, and reporting that Peter's sister had said that Peter is
in America these days, is simply a difference in A's interpretation of the
present tense in B's utterance. In both cases, the present tense has
present time reference, but its scope is different: in the narrow scope
reading, it only refers to B's point of speech (and requires tense shift in
the report), in the wide scope reading, it still encompasses A's point of
speech, and therefore does not require tense shift.
The use of the past tense in reports of a present tense model is
sometimes described as expressing a certain reservation on the side of
the speaker (Diderichsen 1962: 124). I do not think that this is a quite
fitting description. It is not the use of the past tense as such that
 Reported speech in Danish 241

expresses the reservation. It is the narrow scope reading of the present

tense of the model that, by way of a Gricean implicature (similar to, but
distinct from the implicature at work with the German subjunctive in
indirect speech, see footnote 19), indicates that the speaker does not
want to go into a discussion of how far he or she subscribes to the claims
of the speaker of the model. Reporting the model with the narrowest
reading possible minimizes the report speaker's commitment. So if the
speaker does not want to commit her- or himself to the truth of what the
original speaker said, he or she has to use a past tense for a present tense
of the model, but this being the unmarked case, the past tense in itself
does not express any reservation on behalf of the speaker.
A case like the following brings this out beautifully:
(36) [RP: 5a: 015]
Reinhard: Jeg har nu altid vaeret noget genert og sädan noget
'Reinhard: Now I have always been a bit shy and that sort of thing'
[RP: 5a: 019]
Anne: Du siger du var genert
'Anne: You say you were shy'

Anne's remark does not question the interviewee's statement that he has
always been shy, nor does she express a reservation about it. She just
doesn't want to embark on a discussion as to whether this is true and
whether this still applies - she is only interested in his statement in
connection with the story the interviewee is about to tell. Therefore, her
use of the past tense.
Indirect reports in the present tense with a verb of saying in the past
seem to be possible, but relatively rare. I could not find a single one in
my material. The only cases I could find were examples where the verb
of saying was in the historical present (with past time reference in both
matrix and complement clause), or where the time reference of both the
matrix sentence and the complement was to the point-of-speaking of the
present speaker, like in:
(37) [Else: 62]
det er sä rutinerede og gode kranf0rere, de har derude, siger de
'they are [lit. it is] such experienced and good crane drivers they have
out there, they say'
242 Hartmut Haberland

5. Grammatical distinctions between direct and indirect

reports in Danish

Realizing whether the report and the embedding text share a deictic
origin presupposes that one understands what is talked about. Gramma-
tical differences between direct and indirect reports, on the other hand,
should be recognizable independently of a previous understanding, since
they are formal properties of the utterance. The question is whether they
always are there. There are two possible reasons for their non-appea-
rance: first, that the elements potentially affected do not appear in the
sentence in question at all, second, that they are not affected in the way
expected. A sub-case of the latter will be the case where grammatical
elements commonly associated with either direct or indirect speech occur
together in one sentence.

5.1 The at complementizer

Subordinate clauses containing indirect reports of statements or orders

(in contradistinction to reports of questions) often start with the comple-
mentizer at 'that'. This complementizer is not obligatory, however. It is
lacking in (38), which is not an unusual case at all:
(38) [RP: 5a: 019]
Du siger du var genert
'You say you were shy'

It is clear from the context that this is an indirect report, since du 'you'
in both instances refers to the hearer.
On the other hand, at can appear at the beginning of what otherwise
looks exactly like direct speech, like in the following:
(39) [Lillian: 91]
sä siger vaerkf0reren bare at der er sgu altid sä meget i vejen med
Dem
'so the foreman just says that God! there is always so much the
matter with you'

In (39), the report has a non-shifted deictic element (Dem '(polite) you')
and main clause word order (sgu before finite verb). So in (40) all
elements characteristic for direct speech are present, but the report is
introduced with the at complementizer. Here, at seems to function as a
 Reported speech in Danish 243

general report marker. This phenomenon is not restricted to modern

colloquial Danish. It is found in Old Icelandic (Han svarar, at "ek skal
rida til Heljar at leita Baldr." 'He answers, that "I shall ride to Heljar in
order to look for Baldur"', cf. Mikkelsen (1911: 563)), in Classical
Greek after the conjunction ότι 'that' (like in Xenophon's Anabasis, 2,
4, 16 "Proxenos said that Ί am the man you are looking for'", and in
Russian (VoloSinov, see reference in footnote 1, has an example from
Gogol's "Hie Inspector General" (Revizor), where Osip says "The
innkeeper said that (öto) Ί won't give you anything to eat until you paid
for what you've had'").

5.2 Word order

It was mentioned earlier that the placement of a number of words like

the negation ikke and others like da (= Ger. doch), nok 'probably', godt
'well' before or after the finite verb is a test for the main/subordinate
clause distinction. This sometimes works with reports, as in the following
example:
(40) [Karen: 14]
Sä sagde jeg til ham, at jeg ikke ville have, at han byttede om p i det
lager.
'Then I told him that I didn't like him changing around that storage-
room.'

This is an indirect report, since main clause word order would be: jeg
ville ikke have at...
But there are a lot of cases, where these words occur in the place
where one would expect them in a main clause, both in reports with an at
complementizer (like in (39) above), and without (such as (41)):

(41) [RP: 3b: 100]

og du har selv sagt du kunne godt drikke bajere
'and you said yourself you could well drink a pint'

(41) is different from (39) in that it has deictic shift in the subject
pronoun du 'you' (the past tense of the verb is not a result of a deictic
shift). Diderichsen (1964: 71) mentions that subordinate clause word
order rules do not always apply after at, and this seems also to apply to
cases where the optional at is dropped as in (41). This slackening of the
subordinate clause word order rule is not restricted to reported speech,
but it is in reported speech where the rule is most easily flouted. Since
244 Hartmut Haberland

this means that the main/subordinate clause distinction is less clearly

developed in reports, the consequence is another blurring of the border-
line between direct and indirect reports in Danish.

5.3 Fronting

Fronting of a constituent other than the subject in Danish subordinate

clauses has definitely something murky about it. Diderichsen's grammar
(1962) ignores the possibility of other constituents than the subject in
front of the verb completely. Nevertheless, fronting occurs, although
other strange things happen with it, like a placement of ikke and related
adverbs way back in the clause, where we would expect them in a main
clause. In indirect speech, the restrictions for fronting in subordinate
clauses do not seem to apply: elements can be fronted (e.g., topicalized)
more or less as would be possible in direct speech reports as well. An
example is (42), where the topicalized element det is part of the predicate
expression:
(42) [Lizzie: 33]
Vaerkf0reren var ked af at jeg holdt op, men jeg sagde, at det skulle
han ikke vaere
'The foreman was sorry that I left but I told him that he shouldn't be
[lit. that thät he shouldn't be]'

Again, we find the mixture of direct and indirect report elements we

have encountered several times so far. The report has deictic shift of
tenses and personal pronouns, but it also shows fronting (topicalization)
and as a consequence, main clause word order. (The at complementizer
is so ambivalent that one cannot really consider it as a marker of direct or
indirect speech, although it, of course, should introduce a subordinate
clause, being a complementizer). Since subordinate clauses with a fron-
ted non-subject constituent do not differ in word order from main
clauses, det skulle han ikke ν cere in itself would qualify as quasi-direct
speech, but here it is introduced both by a verb of saying and a
complementizer (at). Rather than assuming a fourth category here,
'quasi-direct speech after verbs of saying' (which one could, of course), I
suggest that one allow for mixtures of features of direct and indirect
speech to varying degrees.
That the det topicalized in (42) is a pronoun and therefore has
discourse deictic functions is not really relevant for the direct/indirect
distinction here. If we imagine (43) as the corresponding model,
 Reported speech in Danish 245

(43) Vaerkf0reren: Jeg er ked af at du holder op.

Lizzie: Det skal du ikke vaere.
'Foreman: I'm sorry that you quit.
Lizzie: [That] you shouldn't be.'

the same process (traditionally called pronominalization) would occur

here too, just not intra-sententially, but across speech acts, and this
would also be the case in a direct report version, as one easily can figure
out.
Clefting, which is often heavily stylistically marked in Danish subordi-
nate clauses, is quite often found in indirect speech, like in the following
example:
(44) [Jette: 52]
Men sä havde han engang f0r min tid spurgt dem eller sagt til dem at
det var da nogle frygtelige stole, de sad pä, när de fyldte is i.
'But then he had asked them once before my time, or told them that
it were [some] awful chairs they sat on when they put the ice in.'

5.4 Extraction phenomena

The following example shows a curious extraction phenomenon in

connection with a so-called der-cleft.20
(45) [Jette: 52]
Der var flere ting de sagde han havde lovet
'There were quite a few things they said he had promised'

It is not quite clear what constituent structure one should give (45),
but (46) might be a candidate:
( 4 6 ) s [Der var NP [flere tingj s [de sagde s [han havde lovet _ j ] ] ] ]

(46) shows that an element inside the report introduced by de sagde has
been extracted. The model of this report would either be Han har lovet
flere ting 'He has promised quite a few things' or Der er flere ting han har
lovet 'There were quite a few things he had promised'. (46) is also clearly
different from a possible
(47) Der var flere ting, sagde de, han havde lovet

which has a completely different structure. The base of (47) is something

like (48),
(48) s[Der var NP [flere ting; s [han havde lovet _ ; ]]]
246 Hartmut Haberland

with sagde de put in between (the inversion is probably due to the verb-
second rule).
This complex interaction between the report and the introducing
clause (at is missing here again) shows that the whole construction of (45)
is treated as one whole complex syntactic domain which would not be the
case with elements of direct speech. There, syntactic processes don't
operate across the borders between syntactic elements contained in
different speech acts.

5.5 Expressive elements

Finally, I would like to point out two phenomena which have to do with
differences between direct and indirect reports. They are lexical rather
than grammatical phenomena, but as such they belong to the level of
form and are therefore easily observable.
It has sometimes been claimed that 'expressive elements' like curses
etc. cannot actually form a part of indirect speech reports (Coulmas
1985: 45). These 'psycho-ostensive expressions' which stand outside the
syntactic construction of the sentence, have actually the value of a whole
sentence themselves, but they do not have any propositional content and
cannot therefore be paraphrased by a subordinate clause.
This is obviously not true for fossilized elements like sgu, which
originally was an oath but which now has become an ordinary particle
functioning as a clause adverbial. But it is neither true for elements like
Helvede 'Hell!', of which one hardly can say that they have been
grammaticalized to the same extent:
(49) [Jette: 49]
jeg tudbr0lede og sagde, hvorfor Helvede hun var sädan en skide-
satans-lorte-tarvelig kaelling
Ί burst into tears and asked her why the hell she was such a bloody
awful low-down bitch'

5.6 Verbs of saying

The other phenomenon is the occurrence of a verb of saying. The lack of

a verb of saying is usually attributed to quasi-indirect discourse, but in
my corpus there are also cases of direct speech without any introductory
verb. In some cases we have the usual deictic shifts, like in (50) and (51),
 Reported speech in Danish 247

(50) [RP: 3a: 098]

Hun ville have at jeg skulle vaere barber. Det var sädan et renligt og
paent arbejde.
'She would have liked me to become a barber. It was such a clean and
neat occupation.'
(51) [Karen: 15]
Else, hun for derop og var helt hysterisk. "Ja, sä ku' vi bare gä -
allesammen. Det bet0d ikke noget." Og sä gik vi.
'Else, she charged up there and was completely hysterical. "Well,
then we could go - all of us. That didn't matter." So we all left.'

and we can therefore attribute the phenomenon to the category of quasi-

indirect discourse. In other cases, we have 'pure' direct reports, as in
(52):

(52) [RP: 4a: 248]

Og ligesom da han skulle til at laere latin og det, kan jeg huske,
derovre pä efterslaegten, ikke, der kommer jeg til skolelaereren, "ja
han er da frisk"
'and like when he was to learn Latin and that, I remember, over at
Efterslaegten21, well, there I come to the schoolteacher, "Well, he's
OK'"

where what is reported is what the teacher said and han 'he' refers to the
inverviewee's son. The lack of a need of a speech-act describing verb is
obviously connected with the fact that, in spoken language, direct speech
is always displayed. An ever so slight change in the tone of voice, a body
movement, or simply short pauses can make clear who's speaking. It is
only in 'texts' 22 that a verb of saying achieves its real importance in
structuring the dialogue reported. This is not the place to speculate
about the historical development of speech-act reports and the role of
speech-act describing verbs in reporting speech. But there is a certain
reason to suspect that a rich system of several possibilities for reporting
speech, as it is exhibited in colloquial spoken Danish, has to be seen in its
development as interdependent with the development of literacy.23 The
'text' creates a deictic problem by superimposing a third deictic origin on
report and reported model: that of the listener or reader. At the same
time, visual text (writing) compensates for the lack of intonation,
gestural and other paralinguistic features by the new graphic media for
structuring discourse such as quotation marks. It is small wonder that
literacy has contributed to the development of oral reporting styles and
that literature at the same time has drawn heavily on what one could find
of reporting styles in the vernacular and the verbal folk arts.
248 Hartmut Haberland

6. Concluding remark

We started off by asking whether reported speech was an universal of

linguistic action. We still hardly have the right to doubt this. We also
asked, whether the direct/indirect report distinction has any claim to be a
universal as well. Our material from spoken Danish suggests that there is
a clear basis for assuming that there are two basic ways of orientation in
reporting speech which could be called direct and indirect speech. It is
doubtful whether they make it possible to classify reports into two
mutually exclusive and clearly distinguishable sets. Rather, they can be
interpreted as two tendencies. This makes us ask if the clear-cut division
between direct and indirect speech, mainly developed on the model of
Latin and Classical Greek, and in different ways applied to a handful
(but not even all) of the European literary languages (of which Danish is
one), ever really was as clear as it appeared from the point of view of the
study of the literature of those languages in certain historical periods.

Acknowledgement

I want to thank Anne Cornelius, Erik Aamand Knudsen, Jörgen Paagaard Christensen and
Poul Lyk S0rensen for the permission to use their tapes (see footnote 5). I have had
discussions about reported speech with Jens Balslev, Florian Coulmas, Sysse Engberg,
Anne Marie Heltoft, Jesper Hermann, Margaret Malone, Jacob Mey, Gunter Senft, Karen
Sonne Jakobsen, Ole Nedergaard Thomsen and Johannes Wagner, and I got comments on
an earlier draft of this paper by Karen Ebert, Lars Heltoft Niels Haastrup and Johan Van
der Auwera. All these people not only gave me a lot of good ideas, they also prevented me
from getting carried away with some of the fancier ideas of my own and falling into the
most obvious traps, and some less obvious ones. If not for them, this paper would look
considerably different from what it looks now.

Notes

1. Volosinov (1973: 126, and fn. 2) mentions Russian as a language where the distinction
is not too clearly developed. See also Ebert (this volume) for Chamling.
2. I am using the English word I as an illustration only. The concepts of speaker, hearer,
point-of-time of speaking etc. do not have to be straightforward concepts in every
language. Some languages (Danish is one of them) make a polite vs. familiar distinc-
tion for addressee reference (in Danish, du (singular) and I (plural) are used for
familiar address, and De (both singular and plural) for formal address.)
 Reported speech in Danish 249

3. Some grammars describe this switch as pronominal shift ("Pronominalverschiebung",

Helbig and Buscha 1975: 167ff.). As should be clear by now, this is nothing which
affects pronouns in any special way. Rather the phenomenon is one of deictic switch or
deictic shift, which affects all deictic elements in the model utterance. Tense is a deictic
element as well.
One might wonder by the way, if aspect is too. Between (1) and (2) above, there is a
change of the verbal form I'm going to come by to she's coming by. As far as tense is
concerned, no shift is to be expected, since Tuesday, when Alice is expected to come, is
ahead of both Sunday, the time of the model speech act, and Monday, when I talk to
my wife:

ahead of

ahead of

time-axis
•Alice
to

tQ speaking point-of-time of speaker s

ti point of event for Alice's coming by

Both the relationship between t^ and t,, and t^J16 and t! require future time reference
in the verb referring to the event taking place at tj. So no deictic shift should be
involved for that reason. On the other hand, there is a difference between the two ways
of future time reference 'be going to + Infinitive' and 'Present progessive'. In the first
case, we have what Quirk et al. call "future of present intention" (1972: 87), in the
second "fixed arrangement, plan, or programme" (1972: 88). This is certainly not a
difference in tense, rather in aspect. But no matter what it is otherwise, it is a deictic
difference: what Alice stated as her intention, has become an arrangement (with her)
for me. So it looks as if even the difference in aspect between (1) and (2) could be seen
as a deictic difference. I'll leave this problem to students of indirect speech in English.
4. Nevertheless, Banfield (1982: 26-28) gives a good many reasons for not deriving
indirect speech from direct speech transformationally, just as she does for not doing the
opposite, deriving direct speech from indirect speech (1982: 28-34).
5. The material is taken from two corpora of spoken Standard Danish I am working with.
One consists of a collection of interviews from a book called Kvinder pä fabrik
(Women in the factory) (Jespersen 1971). Material from this book is quoted by a first
name, which is the name of the woman interviewed in the relevant chapter of the book,
and a page number, like the following: [Karen: 16], [Paula: 44], I have not had access
to the original tapes of that corpus and I use the transcriptions as they appear in the
book. This seems valid to me, since it is not my intention to show any differences
between spoken and written Danish; therefore it does not matter whether a piece of
250 Hartmut Haberland

material is 'fully authentic', or rather 'edited' spoken Danish as long as native speakers
accept it as a rendering of spoken Danish (this has been checked with native speakers).
The other corpus consists of tape-recorded interviews with a retired Copenhagen
worker born in 1902, collected by Anne Cornelius, Erik Aamand Knudsen, J0rgen
Paagaard Christensen and Poul Lyk S0rensen. (I have used my own transcription of
this corpus). Material from this corpus is identified like this: [RP: xx: yy:], where xx is
the section of the interview, and yy the tape counter number of the passage quoted.
6. Translations of material from my corpus are rough translations. They are as close to the
Danish original as possible without impairing understandability, even at the expense of
English idiom.
7. Another example is the following:
[Jette: 52]
Men sä siger de bare: Det er sgu underligt, at han ikke siger det og det, det er
sgu underligt at han ikke g0r sädan og sädan, at vi ikke fär nogen besked i det
mindste . . .
'But then they just say: It's funny that he doesn't say such-and-such, it's, God!
funny that he doesn't do such-and-such, that we don't hear anything at the
least.'
8. It will be clear in the light of the discussion of some properties of Danish syntax in
section 3 (sentences (25) and (26)), that Hvornär blev du alene is a main clause and
therefore a direct question, not an indirect one.
9. Hj0rringgade, Koldinggade and Gernersgade are all streets in Copenhagen. (Danish
gade means 'street'.)
10. The example is adapted in the sense that Stubbs' original example also contains a very
good case of 'split recipient design', which has been left out here as irrelevant to the
present discussion.
11. Or who just held the opinion that school meals were for the poor, and was known for
holding this opinion (possibly without himself actually knowing that he was known to
hold this opinion). This brings us to the difficult question of indirect 'speech' as report
of a thought or attitude which I cannot go into here further.
12. Syntactically, this case is different from (14). In (14), we had an ambiguity between
statement and indirect reported speech. In (16), the relevant part (det varfor defattige
'that was for the poor') is ambiguous between statement and quasi-indirect speech.
13. A different possibility would be to consider Nej as direct speech and what follows as a
simple statement, like in the cases discussed above, (14) to (17). I don't think that this
interpretation, although formally possible, is very likely here.
14. Also in the first edition of 1785, quoted by Br0ndum-Nielsen (1953: 17).
English translation: What A. has said, is told by B. like this: He (that is, A.) knew
indeed those negligent people, but since they didn't take care of their business, they
would come to a bad end. But A. has put his words like this: I know indeed those
negligent people, but since they don't take care of their business, they will come to a
bad end. This can be called a historical subjunctive in Danish, which relates another
person's words, without introducing them with unchanged forms, but generally chang-
ing the present tense of the other's speech into past tense, past perfect into pluperfect,
the first person into the third, and the second person sometimes into the first.
15. "The Danish infinitive has a narrowly restricted use as a sort of 'optative', expressing a
wish or a polite request." (Diderichsen 1964: 61) In most cases, this use has to be
considered formulaic, if not obsolete, like in the beginning of a recipe Man tage ...
 Reported speech in Danish 251

'Let one take . . . ' , or in Leve Köngen! '[Long] live the King!' The obsolescence of the
latter formula, usually used in grammars to illustrate the optative, is aptly underlined
by the fact that Denmark hasn't had a king since 1972 (the present monarch is a
woman). Cf. also Diderichsen (1949: 147).

16. (25) is completely analogous to:

Jeg tror ikke, at han d0de
I believe not that he died
V

If the Λν-word represents the subject of the sentence, no other constituent can fill the
place between the hv-word and the verb, and a subject dummy is inserted:

Jeg ved ikke, hvem der d0de

I know not who Δ died
V
17. As Danish is a type A language in Perlmutter's sense, the subject of a subordinate
clause cannot be fronted in this way (which would remove it from that clause). For
some general information on Danish extraction phenomena, see Erteschik-Shir (1982)
(Erteschik-Shir's acceptability judgements of Danish sentences are to be taken with a
certain caution, though).
18. This presupposes, of course, that the report is not uttered in the same week as the
model.
19. That these phenomena are not analogues, is also strongly hinted at by the fact that the
German subjunctive and the Greek optative are optional stylistic features. In German,
it is very often a device marking reported speech when no other features marking
reported speech are present or only very weak ones. Furthermore, in German indirect
speech the subjunctive can by way of a Gricean implicature ("why else would the
speaker have to lay such emphasis on the fact that this is indirect speech?") hint at a
disagreement of the speaker with the content of the reported speech act, although this
'distancing function' by no means is a part of the (semantic) meaning of the German
subjunctive. None of this occurs in Danish. As we shall see, lack of tense shift between
model and report is not an optional stylistic feature, but can always be interpreted
semantically.
20. Cf. Br0ndum-Nielsen (1930). For a closely related phenomenon in Norwegian, cf.
Fretheim (1970: 62f.) The phenomenon is not too well understood, but whereas real
clefts in Danish (and Norwegian) have something to do with emphasis, this type, which
always has an indefinite NP in its focus (here: flere ting 'quite a few things'), seems to
have the function of moving as much material as possible to a non-given, non-topic
position in a sentence. Norwegian clefts of this type are usually formed with det (like
real clefts) and therefore formally even less distinguishable from real clefts than their
Danish counterparts.
21. A secondary school in Copenhagen.
22. In the sense of Malinowski. For Malinowski, a 'text' is a piece of discourse "divorced
from its context of action and situation" (1935: 8). Similarly Ehlich (1979: 195 and
426).
23. This does not mean that cultures where literacy and writing play, or until recently have
played, a minor role will have less rich systems of reporting speech. Hutchins'
252 Hartmut Haberland

transcript of a land ligitation on the Trobriand islands (Papua New Guinea) shows that
in Kilivila it is possible to use a number of different ways of introducing reported
speech both with and without a speech act verb (Hutchins 1980: 71-109).

References

Austin, John L. 1962. How to do things with words. London: Oxford University Press.
Baden, Jacob 1792. Forelcesninger over det danske Sprog. Copenhagen: P. Horrebouws
Enke.
Bally, Charles 1912. Le style indirect libre en fran9ais moderne. Germanisch-Romanische
Monatsschrift 4: 544-56 and 597-606.
Banfield, Ann 1982. Unspeakable sentences. Narration and representation in the language
of fiction. Boston: Routledge and Kegan Paul.
Br0ndum-Nielsen, Johs. 1930. Om nogle "episke love". In: Grammatical miscellany
Jespersen. Copenhagen: Levin & Munksgaard pp. 373-377.
Br0ndum-Nielsen, Johs. 1953. Dcekning- Oratio tecta i dansk litteraturf0r 1870. Saertryk af
K0benhavns Universitets festskrift. Copenhagen: Schultz.
Bühler, Karl 1934. Sprachtheorie. Jena: Fischer.
Coulmas, Florian 1985. Direct and indirect speech. General problems and problems of
Japanese. Journal of Pragmatics 9: 41-63.
Diderichsen, Paul 1949. Morpheme categories in modern Danish. TCLC 5: 134-155.
Diderichsen, Paul 1962. Elementar dansk grammatik. Copenhagen: Gyldendal.
Diderichsen, Paul 1964. Essentials of Danish grammar. Copenhagen: Akademisk Forlag.
Ebert, Karen H. 1985. Reported speech in some languages of Nepal. This volume,
145-159.
Ehlich, Konrad 1979. Verwendungen der Deixis beim sprachlichen Handeln. Bern: Lang.
Erteschik-Shir, Nomi 1982. Extractability in Danish and the pragmatic principle of
dominance. In: Elisabet Engdahl and Eva Ejerhed, eds. Readings on unbounded
dependencies in Scandinavian languages. Acta Universitatis Umensis 43: 175-191.
Fretheim, Thorstein 1970. "Utbrytning av setningsledd" sett fra transformasjonsgramma-
tisk synspunkt. In: Eskil Hanssen, ed. Studier i norsk spräkstruktur. Oslo: Universitets-
forlaget. pp. 53-85.
Haiman, John 1974. Targets and syntactic change. The Hague: Mouton.
Helbig, Gerhard and Joachim Buscha 1975. Deutsche Grammatik. Ein Handbuch für den
Ausländerunterricht. Leipzig: Enzyklopädie.
Hutchins, Edward 1980. Culture and inference. Cambridge, Mass.: Harvard University
Press.
Jespersen, Karen et al. eds. 1971. Kvinder pä fabrik. Copenhagen: Hans Reitzel.
Lorck, Jean Etienne 1921. Die erlebte Rede. Heidelberg: Winter.
Malinowski, Bronislaw 1935. Coral gardens and their magic. Vol. II. London: George
Allen and Unwin.
Mikkelsen, Kr. 1911. Dansk Ordföjningslcere. Copenhagen: Lehmann & Stage (Reprint
Copenhagen: Hans Reitzel 1975).
M0ller-Kristensen, Sven 1938. /Estetiske studier i dansk fiktionsprosa 1870-1900. Copenha-
gen: Munksgaard.
 Reported speech in Danish 253

Neubert, Albrecht 1957. Die Stilformen der "Erlebten Rede" im neueren englischen
Roman. Halle: Niemeyer.
Perlmutter, David M. 1971. Deep and surface structure constraints in syntax. New York:
Holt, Rinehart and Winston.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1972. A grammar
of contemporary English. London: Longman.
Stubbs, Michael 1983. Discourse analysis. Oxford: Blackwell.
Tannen, Deborah 1983. "I take out the rock - dok!": How Greek women tell about being
molested (and create involvement). Anthropological Linguistics 25: 359-374.
Theroux, Paul 1979. The Old Patagonian Express. Boston: Houghton Mifflin.
Voloäinov, V. N. 1973. Marxism and the philosophy of language. Translated by Ladislav
Matejka and I. R. Titunik. New York: Seminar Press [1929].
Wunderlich, Dieter 1977. Assertions, conditional speech acts, and practical inferences.
Journal of Pragmatics 1: 13—46.
Reported speech in French and Hungarian
Ivan Fonagy

1. Direct (mimetic) quotation

Limitations

It has been pointed out repeatedly that direct quotations are not necessa-
rily literal and authentic reports (see: Kalik-Teljatnicova 1965-1966.
Strauch 1974, Sternberg 1982). The distance between real speech events
and their stylized literary representations becomes conspicuous if we
compare the reported speech acts as presented in a naturalistic novel
with the 206 pages of the 'First five minutes' of an interview in the
'sound-perfect' reproduction of Pittenger, Hockett and Danehey (1960).
Even the Russian skaz which is intended as a precise graphic representa-
tion of live speech (Vinogradov 1926, Bakhtin 1971) lags far behind this
'sample of microscopic interview-analysis' (o.e. title-page).
Direct quotation, by virtue of its formal features, however, can be
considered as a means for restaging a verbal performance, and as such it
creates the illusion to witness the scene evoked by the narrator. Plato
clearly refers to the limitations inherent in direct quotation as he
opposes, for the first time in European literature, mimetic direct report
(mimesis) to narration (diegesis):
But when the poet, in fact, reproduces spoken words as if he were somebody
else, would not we say that he adjusts his speech as well as he can [my italics]
to the individual manner of the person who's words he is supposed to report
(The Republic III, 393-394, c, d, e)?
The character's manner of speaking is reflected most of the time in an
indirect way, thus, by means of similes, conventional or genuine meta-
phors.
( . . . ) hangjänak orgonaviräg szaga van ( . . . )
['her voice exhales a fragrance of lilacs']
(Kriidy, Puder p. 10)
Ce n'etait meme pas une voix imperieuse. C'etait une voix molle, grasse, une
voix dont chaque mot faisait plof, qui tombait sur vous comme une crepe, une
voix dans laquelle il semblait qu'on aurait pu enfoncer le doigt.
(Felicien Marceau, Bergere 16gere p. 116)
256 Ivan Fönagy

Voice and mimetics of the character are mirrored with scientific

precision in Proust's novels. 1 Intonation is characterized, most of the
time, through the message conveyed, that is by describing the speaker's
attitude and intentions.
- Mais pour etre heureuse! (et avec quelle vivacite de ton! un 'pardi!' est sous-
entendu ( . . . )
(Montherlant, Jeunes filles p. 164)

The phonological limitations of the alphabetic writing-systems can be

got around in different ways. Emphatic lengthening may be suggested by
repeating the letters representing the lengthened consonant or vowel.
According to Malraux's transcriptions the lengthening of hard conso-
nants seems to be an idiosynchratic feature of baron Clappique in the
novel 'La condition humaine':
De nouveau confidentiel: " . . . une pp'etit Martini. Sevfcre: "trres sec" (p. 25)

The principle of isomorphism allows for the representation of extra-

stress by means of capital letters, boldface, etc.
Je suis malade,
JE VEUX MOURIR.
(Verlaine, D6ception)

Halting speech, hesitation, irregular speech-pauses are replaced by

their visual equivalents such as the separation of letter sequences by
inserting blanks, hyphens, suspension-points. On the other hand, the
suppression of interspace may suggest precipitation, the sweeping away
of junctures:
"R6pete un peu voir", qu'il dit Gabriel.
"Repeter un peu quoi?"
"Skeutadittaleur . . . " [= Ce que tu as dit tout ä l'heure].
(Queneau, Zazie dans le Metro p. 8)

Expressive phonetic shifts may be reflected by means of phonemic

(graphemic) substitutions. Thus, the grapheme e representing the pho-
neme /e:/ in Hungarian may allude in Aläszolgeja, bäro urf (instead of:
Aläszolgäja 'Your humble servant') to a very tense and fronted articula-
tion (Borszem Janko, June 1921: 10). Frigyes Karinthy characterizes the
symbolist poet, author of "The afflicted hearts', incidentaly a robust giant
with a stentorian voice, by means of a repeated graphic vowel-shift e a
(corresponding to the phonetic substitution /ε/ —> lal) rattanatas ambar
instead of rettenetes ember 'a terrible man' (IJj görbe tükör p. 161).
At the same time, the lack of prosodic features has to be compensated
 Reported speech in French and Hungarian 257

in writing by a shaping of the original delivery at the syntactic level.

Speech acts reported in novels are regularly filtered.2 The term stylized
direct speech assigned by Werner Günther (1928: 11) to French and
German epic poetry of the 11—12th centuries can also be applied to the
representation of speech acts in most novels of the 19th or 20th centu-
ries. Thus, the reported speech, characterized as hesitant, halting in the
writer's comment, is, in fact, mostly regular, explicit and succinct in
Camus' novel (see e.g. Peste p. 29). The occasional reproduction of
syntactic noises in literary prose is all the more effective.
- Anormalement?
- Heu, dit Richard, le normal, vous savez . . .
(Camus, La peste p. 31)

Quotation marks in writing and speech

Attempts at isolating quotations, essentially citations of the Scriptures,

occur already in early Latin manuscripts (Gurk 1961, Catach 1968:
299-300, Ouy 1979, Vezin 1979, Libera 1984). Since the 16th century,
two horizontal ( . . ) or vertical (:) dots served either to lend prominence
to the subsequent words, or to mark the beginning of a citation. The
colon was and still remains functionally ambiguous. In the 'Cancionale'
(1574) of the Hungarian writer Gäspär Heltai the colon was used (a) to
introduce an adversative clause, (b) to point out a causal relationship, (c)
to introduce a comment, (d) an address; but most of the time (e) a
quotation. An indexical gesture, Void! pointing at the subsequent
utterance, could be the basic meaning of (:) in most instances.
In contradistinction to graphic homonymies and synonymies of the
16th century, apparant inconsequences in modern novels correspond to a
high degree of differentiation. Thus, writers, who mark usually direct
reports by means of hyphens, may use the quotation marks "..." for
other purposes: (a) to oppose inner speech to vocalized verbal messages:

- Mit keresel itt?

Mözes kicsit gondolkozott, s egyszeriien, szeliden azt akarta mondani:
"A fiatalsägomat."
[ - What are you searching here for?
Moses pondered for a moment, and would have liked to reply simply and
gently:
"For my youth."]
(Kosztolänyi, Elbeszelesek p. 219)
258 Ivan Fonagy

The "monologues interieurs" are marked in the same way in certain

novels of Balzac (Ursule Mirouet p. 93), Jules Romain (Les amours
enfantines pp. 22,63,113 etc.), Sartre (L'äge de la raison 33, 39, 53 etc.);
(b) to distinguish written quotation from spoken words;3 (c) to refer to
words pronounced in the past;4 (d) to mark quotations embedded into
the referred speech act or inner speech of the character (Jules Romains,
Les amours enfantines p. 108, Läszlo Nemeth, Irgalom vol. 1: 195, 232
etc. vol. 2: 50, 68 etc.). 5 At the same time, according to other graphic
conventions observed in other French novels, inner speech introduced
by a colon may remain unmarked. It is often put into parentheses
(sordino) in Romain Rolland's novel "Annette et Sylvie" (1922), espe-
cially if they make the counter-point to the utterance spoken aloud:6
- Et toi, demandait Annette. Qu'est-ce que tu en penses, Sylvie?
(Flüte!' pensait Sylvie).
- Je pense comme toi, cherie.
(p. 147)

In some cases lack of quotation marks means that the speech act
remains unnoticed or not recognized as such. Thus, in Queneau's novel
'Zazie dans le Metro' the words of the mysterious pseudo-policeman
addressing Gridoux, the shoemaker, are reported in direct style, first
person singular, but without any kind of graphic indication - as long as
the addressee does not seem to notice his partner's utterances.7 Cumula-
tive quotations are repeatedly unmarked in Anatole France's novels:

Durant ces jours, avait-il coutume de dire ä ses freres, je bouillais dans la
chaudiere des fausses delices.
(Thai's p. 11)

The absence of quotation-marks lends to cumulative quotation an inter-

mediate status between direct report and narrative report.
The intuition that prosodic features are necessary and sufficient means
to distinguish quotations from descriptive narration (Verschoor 1959: 9)8
is only partly supported by acoustic analysis and semantic tests. Subjects
confronted with the laryngographic ("filtered") version of a narration
containing reported speech acts are frequently hesitant when distinguish-
ing quotation from descriptive narration. They tend to interpret changes
of tonal level and/or ironic intonation patterns as quotations even if such
sequences do not contain reported speech. 9 In fact, both features are
highly frequent in direct as well as indirect reports.10 The speaker may
adopt in mildly ironic self-quotations a recurrent singsong tune - rise of a
fifth followed by a slow fall of a minor third - corresponding to one of the
 Reported speech in French and Hungarian 259

French enumerative intonation patterns, proper to lyrical or bored

narrative (Fonagy I. and J. 1983: 168-169). The reference to a previous
speech act is, however, always clearly marked by different types of
interrogative intonation patterns. Thus, de dicto questions expressing
misbelief keep closely to a melodic pattern consisting of a rise of a fifth
and a down step of a minor third, with rising tone in the last and the
penultimate syllable (Fonagy I. and J. o.e. 156-157). Queneau explici-
tely refers to melodic quotation marks in connection with ironic de dicto
questions.
- Pourquoi "cette fois-ci"? demanda Marceline en roulant les derniers mots de
sa question entre des guillemets (Zazie dans le Metro p. 148).

Since intonation is an optional mark, live speech cannot do without

constant repetition of some few verbs of saying. In a recorded French
conversation a young employee repeated i'm dit et j'lui dis eight times in
30 seconds. In popular French qu'i'm'dit / que j'dis are the always
postposed quotation indicators (see: Bauche 1946: 134).
"Comment que vous pouvez croire, que je lui dis, comment que je lui dis,
comment vous pouvez croire que vous allez trouver lä-dans l'oiseau reve".
(My italics)
(Queneau, Zazie p. 71)

In Hungarian, the phonetically and semantically reduced form

['asonda] for azt mondja 'he/she says [it]' is used most of the time as a
quotation mark.

Some views on the syntax of reporting

The nature of the syntactic relation between reporting and reported

utterances is still a controversial issue. They are regarded as independent
utterances by some authors (Davidson 1968-1969, Hall-Partee 1973);
unconnected at the level of syntax, although related at the level of
meaning (Denes Szabo 1958 vol. 2: 367-369); they cannot be classified
either with subordinate or with coordinate clauses, although they have a
vague similarity with object clauses (Räcz, in Benczedy, Fabian and
Räcz 1976: 378). Meir Sternberg carefully distinguished the embedding
of indirect citation and the framing of the quoted 'inset'. He insists,
however, on the integrative and subordinating effects of the frame (1982:
70).
The Grammar of the Hungarian Academy of Sciences (Tompa ed.
260 Ivan Fonagy

1961-1962) takes an implicit but nevertheless clear position in listing

'quotation' under the heading of 'object clauses' (vol. 2: 359-360).Mar-
guerite Lips (1926) and Adrien Verschoor (1959) consider reported
sentences in French either as object clauses or as appositions. The
syntactic and semantic analysis of 1266 quotations of a contemporary
Hungarian novelist, Gabor Thurzö, leads Adrianne Dömötör (1983) to
the conclusion that there are as many categories of reported sentences as
subordinate clauses in Hungarian. Reported sentences may function as
an object clause, a subject clause, a predicate clause, an attributive
clause, an adverbial clause or as an apposition; object clause equivalents,
however, largely predominate.
According to Annette Saban (1978) the relationship between the
transitive introductory verb such as to ask and the referred question is
analogous to the relation between to ask and its nominal object: a
question.u Fred D. Cram (1978) goes even a step further: he considers
the reported sentence as a part of the reporting sentence. The quotation
is integrated into the basic sentence by means of lexical insertion. It is a
lexical unit, whether it is a word, a phrase, or a sentence (o.e. 46). Their
grammatical role is a function of their individual lexical meaning: this
property is to be specified in the lexicon (I.e.).12 Any constituent
representing a judgement has to be integrated into the semantic repres-
entation of a sentence by means of an operator such as X ENOUNCE
THAT, argues Marc Plenat (1979), where X is a variable, and its value is
determined by the linguistic or extra-linguistic context. This principle
applies both to global and to partial quotations, to words and phrases
(o.e. 114-115).

The reporting sentence in weak and strong position

We can hardly formulate general statements concerning the syntactic ties

between the reported sentence and the reporting clause, without taking
into account the position of the latter. Syntactic as well as prosodic
indices clearly reflect the dependency and the relative weakness of
postposed reporting clauses. The main sentence shows much more
independence in initial position, especially in Hungarian. Postposed
reporting sentences are weakly stressed, and pronounced on a mid-low
level. They have inversed word-order in Hungarian as well as in French
and other Indo-European languages (see Kieckers 1912: 145 ff.) Such
inversions are, generally, the consequence of an emphatic stress placed
 Reported speech in French and Hungarian 261

on a preceding word. In the case of reported speech, inversion lends

emphasis to the reported sentence as a whole.
The syntactic dependency of the postponed reporting clause is still
more apparent in Hungarian. Verbal prefixes are regularly separated
from, and put behind, the verb: this implies a preceding topicalized
verbal unit. The potentially transitive verbs13 of the reporting sentence
always follow the objective conjugation: mondta 'he said it', olvasta 'he
read it', gondolta 'he thought it' (see also Dömötör 1983: 474). In
Hungarian the verb follows, broadly speaking, the objective conjugation
if it points to a definite object (Tompa ed. 1961-1962 vol. 1: 207). The
verb in the objective conjugation implies, in the absence of an object in
the main clause, that the reported sentence is conceived as an object.
Even originally nontransitive verbs may follow the objective conjuga-
tion, on the model of mondta. Thus sohajt 'he sighes' follows the
objective conjugation on p. 218 of the collected short stories of Dezsö
Kosztolänyi:
- sohajtotta Daniel 'sighed [it] D.

but not on p. 201:

- sohajtott a fiu 'sighed the boy'.
(Öregurak, in: Elbeszelesek)
Though both reporting sentences show inverse word-order, weak
stress and mid-low pitch, the verb in objective conjugation creates closer
links between the reporting and the reported sentence, and suggests
indeed a similarity with object clauses (Räcz 1975, Dömötör 1983:
474).14
Transitive verbs of saying and thinking such as mondta 'he said [it]'
require a pronominal object in initial position: W erre ezt mondta [N
thereupon said this]7. Here the subsequent quotation cannot satisfy the
syntactic need for completion of the reporting sentence as is possible in
median and final reporting clauses (see also Kiefer in this volume).
This statement, however, needs some qualification. The demonstra-
tive pronoun ezt 'this [acc.] (nearby) might be substituted by azt 'that
[acc.]' (distant). In the introductory sentence azt points beyond the
reporting clause towards the reported one; the tone only slightly des-
cends, in contrast to ezt mondta followed by a pause or juncture marked
by a low fall in the last syllable. In both cases un-weakend stress and
direct word order mark the syntactic independence of the introductory
main clause. In fact, azt mondta usually introduces indirect quotations
(see Kiefer p. 201).
262 Ivan Fönagy

The reduced and frozen form of azt mondja (see: p. 282) reflects,
however, semantic reduction, and functions as a deictic morpheme
pointing to the reported utterance. The deictic morpheme /asonda/ can
be opposed to azt mondja which maintains its literal meaning ('this is
what he says/pretends'). In literary text of the 18th and 19th centuries we
meet a frozen and reduced form of ügy mondja 'he says/it/ like that':
ugymond, a deictic which always follows the reported clause. The
Hungarian ügy mond, /asonda/ together with French [kimdi], short form
of qu' il me dit have hardly any referential function. This is equally true
for the deictics of the classical languages, Latin inquit, Greek emi (see:
Kieckers 1912) or the Hopi /ay'w/ immediately following the reported
speech (Sternberg 1982:108). There is a marked tendency to confine non
referential quotation-deictics to the weak position. The contrast between
full forms and weak forms leads us to the distinction of the deictic and
the referential function of the verba dicendi and cogitandi.15
There is a clear distinction between deictic and descriptive use of the
verbs of saying in Hungarian. In medial or final position deictic use
implies, in the case of originally intransitive verbs such as sir 'he cries',
zokog 'he sobs', the objective conjugation and inverse word-order:
( . . . ) , zokogta a sziiz meg-megcsuklo hangon.
[sobbed the virgin in a faltering voice.]
(Tibor Dery, Α Kiközösitö, p. 65)

The same verbs follow the subjective conjugation and the reporting
clause the Subject + Predicate (direct) word-order, if the reporting
sentence is not used in its deictic capacity.
- Jaj istenem - nöi hang sirt fel a vonal mäsik oldalän jöjjenek hamar . . .
['Good heavens - a woman's voice cried out at the other end of the line -
please, come immediately . . . ]
(Thurzo, Belväros p. 154, cit. Dömötör 1983: 476)

The reported clause can be more intimately connected with the main
clause by enframing, a procedure frequently used both in French and
Hungarian literary texts. The quotation is inserted into the reporting
clause, by splitting it.
M. de Cleves reprenant la parole avec un ton qui marquait son affliction: - Et
Μ. de Nemours, lui dit-il, ne l'avez-vous point vu ( . . . )
(Madame de Lafayette, La Princesse de Cleves p. 141)

Splitting gives the semblance that the character suddenly interrupted the
author to put his own words in. We meet this syntactic figure already in
Medieval French literature.16
 Reported speech in French and Hungarian 263

Le bo£u la teste desouz:

"Alez, que honis soiez-vous,
Dist-il, se vous ne revenez."
[The hunchback his head upside down:
"Now go, and be accursed,
Said he, if you don't come back."]
(Des trois bogus, Fabliaux vol. 13:
Integration of the reported utterance into the main clause creates still
closer links. The reported clause figures as the object in active sentences.
Mult haltement, 'Munjoie!' renuvelent.
(Chanson de Roland v. 3300)
Ce faisant, elle croit devoir jeter au lieutenant un "£a va, dans le placard?"
(Colette, Mitsou p. 10)

It appeares as the subject of the main clause in sentences in the passive

voice or in adverbial phrases introduced by par.
De tutes parz est 'Munjoie!' escriee.
(Chanson de Roland v. 107)
Le percepteur ( . . . ) foudroya par un "Quand je vous le disais!"
(Balzac, Ursule Mirouet p. 18-19)
It figures as an apposition, immediately following the subject.
Un matin, la depeche: "Etat stationnaire", fut interpr6tee dans un sens
favorable.
(Mauriac, Le myst£re Frontenac p. 91)

Some divergencies between French and Hungarian

In contrast to Hungarian, French transitive introductory verbs of saying

and thinking may dispense with a noun or pronoun in object case: they
are completed by the quotation itself. In French epic poetry of the 12th
century introductory verbs without pronoun alternate with transitive
verbs with pronominal object.
Qo dist Rollant: ' ζ ο ert Guenes, mis parastre.'
Dient Franceis: 'Car il le poet ben faire . . . '
[R. said this: 'It will be G. my stepfather.'
The French said: 'Sure, he is the right man for this task]
(Chanson de Rolland w . 277-278)
The pronoun does not 'surface' in the Modern French translation of
Joseph Bedier.
The introductory sentences eztlazt mondta / suttogta / javasolta etc. 'he
264 Ivan Fönagy

said, whispered, proposed this/that' of the Hungarian novel Ά kiätkozo

[The excommunicator]' of Tibor Deri are regularly translated il dit / il
chuchota / il proposa etc. (Kassai - Appercell Paris 1967). Similarly, no
trace of the adverbial pillars, used to lend some independence to the
Hungarian introductory clause, is found in the French version: igy
folytatta 'and he went on like this' - il continua. The reported French
utterance may function as an object:
Et il poursuivait son chemin, songeant:
- Je suis en veneration ä cette vieille femme . . .
(Anatole France, Thais p. 48)

The divergence between French and Hungarian reporting sentences is

still more conspicuous in medial and final position. In Hungarian in
addition to the genuine quotation-deictics of saying and thinking, many
other verbs are assimilated to the primary ones. Thus, in a novel by a
contemporary Hungarian writer, Läszlo Nemeth, 'Irgalom [Mercy]' 97
different items are used in inquit-position (see: Sipka 1966: 262), in
deictic capacity, with inverse word-order, the transitive verbs following
the objective conjugation. The use of 'verba non dicendi', especially of
'verba agendi', became increasingly frequent in Hungarian literature
during the 20th century (see: Szabolcsi 1976).
The semantic field of Hungarian reporting verbs and nouns can be
divided into two parts in accordance with the basic meaning of the
words. I. The expressions belonging to the first half imply verbalization:
production and percepting of oral and written messages, on the one
hand; expressions denoting internalized verbal processing, on the other.
- II. In the second field we find a variety of terms denoting non-verbal
activities. In spite of their impressive diversity, all of these expressions
are genuinly related to communication. 17 We shall be concerned here
with this second (and secondary) semantic field. 18
1. Non-verbal human sound-making: nevetett 'laughed', nevette (obj.)
'laughed [it]', kuncogott 'tittered', kacagott 'laughed out aloud', sirt
'cried', sirtfel 'cried out', sirta (obj. = objective conjugation) 'cried [it]',
szipogta (obj.) 'whimpered [it]', nyüszitette 'whined [it]', zokogott 'he
sobbed', zokogta (obj.) 'sobbed [it]', lihegte 'wheezed [it]', dsitott 'yaw-
ned', äsitotta (obj.), 'yawned [it]', nyögte (obj.) 'groaned [it]', nyögdic-
selt 'moaned gently', hörögte (obj.) 'rattled [it]', lehelte (obj.) 'breathed
[it]'
- Gyäva szajhos . . . - hörögte
[ - You bragging coward . . . - he rattled.]
(Karinthy, Gyilkosok p. 29)
 Reported speech in French and Hungarian 265

A number of these reporting verbs are untranslatable onomatopes ( - J0

- bakkantott az öreg 'Well, yapped(?) the old man', Möricz, Erdely
p. 16).
2. Verbs denoting (originally) natural or mechanic sound phenomena:
dörögte, mennydörögte (obj.) 'thundered [it]'.
- Hovä? - dörgi vissza elkäbultan. (My italics)
[ - Where? - he thundered back dazedly.]
(Szomory, Harry Russel-Dorsan p. 92)
fuvoldzta (obj.) 'tootled [it] (= as on a flute), 'csengett a bätor kisläny-
hang ['rang the voice of the fearless litte girl]' (Thurzo, Belväros p. 107),
'nyekergi a verkli [grinds the street organ]'
3. Facial mimetics: mosolygott 'smiled', mosolyodott el 'break into a
smile', vigyorgott 'he grinned' rdndult meg az ajka 'his lips contracted',
villant fellcsillant fel a szeme 'his eyes sparkled', hunyoritott 'he squinted
his eyes', räncolta össze a homlokat 'he knitted his eye brows'
- Miert olyan? - fintorodott rä Fusztulan. (My italics)
[- Why is it like that? (lit.:) grimaced at him F.]
(Tamäsi, Abel p. 249)
4. Bodily movements: bolintott 'nodded his head (as a sign of approv-
al)', csovdlta a fejet 'shook his head (to indicate disapproval)', kaptafel
a fejet 'jerked up his head'; mutatott oda 'pointed at sy/sg', intett neki
'beckoned to him', nyujtotta a kezit 'held his hand out (for a hand-
shake)'; dörzsölte a kezit 'rubbed his hands (to express his satisfaction)',
csapta össze a kez0t 'clapped his hands (to express his amazement)',
legyintett 'waved his hand (to express discouragement)', cirogatta meg
arcät 'caressed her face', tette vällära a kezet 'put his hands on his
shoulders (as a sign of confidence and goodwill)'; karolta ät, ölelte meg
'put an arm around her', ölelte meg; vont vällat, vonta a vdlldt (obj.),
rdntotta a vdlldt (obj.) 'shrugged (his shoulders)'; rngta felre 'kicked it
aside', rügott bele 'gave it a kick', riigta ki maga alöl a szeket 'kicked
aside the chair'.
- Ilyen szerencsetlen vagyok! - kapkodott a fejehez anyäm. (My italics)
[ - Ο good Lord! - stroke my mother (repeatedly) her head -]
(Margit Kaffka, Szinek es 6vek p. 17)
fordult feleje 'turned towards him', hiizodott vissza 'drew back from
him'; lepett hozzd 'walked up to him\futott utdna 'ran after him', erte be
'reached him\jött vissza 'came back', rohant vissza 'came back running',
rontott elö 'rushed forward'; iiltfel 'sat up (in his bed), iilt le 'sat down',
dllt fel 'got up', ugrottfel 'jumped up'.
266 Ivan Fönagy

- Szegöny az ördög, - pattant fei Adameczne ( . . . )

[ - Poor is the devil - jumped up Mrs Adamecz ( . . . ) ]
(Mikszäth, Szent Peter esernyöje p. 26)

ugrott egyet 'jumped up'; räzkodott össze 'shivered (feeling chilly)',

reszketett 'shivered'.
5. Verbs denoting social behaviour: videlmezte (obj.), vette (obj.)
vedelmebe 'took his defence', 'said in his defence', segitette 'he helped
him', ällt αpärtjära 'took his part', alkalmazkodott hozzä 'complied with
his ideas', tämadt rä 'attacked him', harcolt 'he struggled', avatkozott be
'came in between', kergette el (obj.) 'expelled him', szalojtotta fel 'made
him hurry up (the staircase)', hasznälta ki az alkalmat 'took advantage of
the opportunity (to intervene)'.
- De a szäkalluk istenes - kerestem valami dicserni valöt is
[ - Why, their beard is not so bad - I tried to find something praiseworthy.]
(Tamäsi, Abel p. 249)

6. Verbs referring to conversational strategies (without being verbs of

saying): terelte el a szot 'shunted the conversation', kerülte el a kifejezest
'circumvent the (ominous) expression', fogalmazta ät 'changed the word-
ing',
( . . . ) , tette ät Agnes kerdöjelbe az öhajät.
[, Agnes lent an interrogation mark to her wish.]
(Läszlö Nemeth, Irgalom vol. 2: 93, cit. Sipka 1966: 267)

nyelte le a kerdest 'swallowed the question (which was on the tip of the
tongue)'.
7. Verbs expressing emotive attitudes without implying a speech act:
csodälkozott 'wondered', kepedt el 'staggered', döbbent meg 'stood
aghast', ütödött meg 'became indignant at sth', remiilt meg 'took fright at
sth', örült meg 'rejoiced at sth', vidult fel brightened up', mutatta ki a
jokedvet 'showed his good humour'; 'cherished hopes'; makacskodott
'showed stubborn obstinacy'; bosszankodott 'he was annoyed by sth',
legyürte haragjät 'surmounted his anger'; bätorkodott neki 'plucked up
his courage'.
It seems that only verbs expressing extrovert attitudes are used as
verbs of saying, since these tend towards verbal or non-verbal expres-
sion. Verbs denoting introvert emotive states, such as felt 'he was afraid',
szomoru volt 'he was sad', haragudott 'he was angry' have no such
speech-deictic power.19
The inquit-function of verbs denoting an activity related only by
accident with the reported speech act is, as far as I can see, a more recent
 Reported speech in French and Hungarian 267

development: an extension of the original rule stipulating natural ties

between the reported speech and the meaning of the reporting verb.
Generally speaking, the speaker's simultaneous gestures carry informa-
tion completing the speech act.
- TJgy hozatta - mondta Eszti.
- Gondolt rä? - meregette nagymamuska keseruen a csomagot. (My italics)
[ - He ordered it for me (= the cake), said Eszti.
- He provided for it? eyed Granny with bitterness the parcel.]
(Läszlo Nemeth, Egetö Eszter p. 63)

Looking so suspiciously at the parcel, a present of Eszti's father, who is

held in contempt by the grandparents, is a counterpoint to the words of
acknowledgement.
This is also true for metaphorical paraphrases substituted for a verb of
saying.
( . . . ) , libikokäzott ugyanebben a hangban a bekes szep este felkinäläsa s egy
pici harcikeszseg.
[, a gentle offer of a peaceful evening and a bit of fighting mood see-sawed in
the same voice.]
(Läszlö Nemeth, Irgalom vi. 1: 311, cit. Sipka 1966: 265)

Such images are somewhat far-fetched and certainly unique (hapax

legomena). Some other literary metaphors or metonymies are lively,
perhaps individual, creations but much closer to conversational meta-
phors.
- Vege mindennek, tördelte. (My italics)
[ - The end, this is the end, (lit.:) she broke it (i.e. the utterance, speaking with
a 'broken voice'.]
(Kaffka, Elbeszelesek p. 59)
- Es mi az? - villant fel a mäsik. (My italics)
[ - And what is it? - flashed suddenly the other man.]
(Ferenc Molnär, Az aruvimi erdö p. 20)

The metaphor felvillant 'he flashed' could be a short form oifelvillant a

szeme 'his eyes sparkled'.
Conventional semantic transfers play an essential role in the develop-
ment of the system of reporting verbs. Verbs denoting non-verbal sound-
phenomena, such as the above quoted verbs dörögte, mennydörögte
(obj.) 'he thundered' gradually shift from the category of verbs denoting
sounds generated by non-human (natural or mechanic) source, to the
category of 'verbs of saying conveying additional information at the level
of vocal expression'. The verb sohajt 'he sighes', in inquit-position, is a
verb of saying maintaining its original meaning, that of a non-verbal
268 Ivan Fonagy

sound emitted by humans. If it follows the subjective conjugation

(sohajtott 'he/she sighed') its original meaning is brought into promi-
nence; but if the verb follows the objective conjugation (sohajtotta 'he
sighed [it]'), it is assimilated to other verbs of saying. The verbs lelkese-
dett 'he became enthusiastic about it', mergeskedett 'he was angry, he
behaved angrily' originally denoted attitudes which did not imply verbal
expression. In their role of reporting verb they took up a new meaning
and became 'verbs denoting verbalized attitudes'.
In general, secondary verbs of saying are metaphors. The verbs setting
out from the most diverse domains of verbal and non-verbal communica-
tion converge towards a centre, as if attracted by an ideal, completely
unmarked verb of saying. In contrast to individual lexical or grammatical
metaphors, it is the underlying categorial transfer which preconditions
the generation of secondary verbs of saying.

Verba non-dicendi Verba dicendi

Meaning

Function

The meaning of the underlying basic transfer could be this: 'this non-
verbal action is to be interpreted as a speech act.'
The distinction between primary and secondary verbs of saying goes
back to the first linguistic theories of reported speech. Otto Behaghel
(1877,1899,1923-1928 vol. 3: 694-711) pointed out that verbs describing
the circumstances of the speech act are substituted for the erased
genuine verb of saying. Hermann Paul gave a similar interpretation of
verba agendi in inquit-position (1916-1920 vol. 4: 172f.). More recently,
Miklos Szabolcsi (1976) derives the secondary verb of saying ugrott fel
'jumped up' from mondta es felugrott 'he said and jumped up'; sugge-
sting that the change took place in four stages: (1) mondta 'he said'; (2)
mondta is felugrott 'he said and jumped up'; (3) mondta felugorva 'he
said jumping up'; (4) ugrott fel 'he jumped up'. Similarly, Annette
Sabban (1978) derives secondary verbs of saying trembla (Β. N.) from dit
tout en tremblant (o.e. 57) 18 postulating the deletion of the verb of saying
dit, and the transformation of the adverbial complement into a verb. In
contrast to Behaghel, Paul, and Sabban, Szabolcsi considers ugrott fel as
the result of a transfer, and not as an elliptic utterance. Deletion and
 Reported speech in French and Hungarian 269

metaphor are, indeed, two basically different kinds of expressive trans-

formation. The meaning conveyed by means of deletion is a pragmatic
(indexical) one; it reflects an attitude. Metaphor, however, creates new
meaning through the synthesis of two widely differing concepts, e.g. that
of 'jumping' and 'saying'.
The fact that facial mimetics, non-verbal human noise-making, sud-
den movements, voluntary or unintended, may serve as verbs of saying is
of considerable semiotic importance.
- Toväbb! - dobogott a kirälyfi szive.
[ - Onwards! - beat the Prince's heart.]
(Kosztolänyi, Elbeszelesek p. 270)

The Prince's heart-beat is equated with a self-directive speech-act.

Similarly, the character's violent bodily movement
( . . . ) - rugta ki maga alol a szeket.
[ - he jumped up and kicked aside the chair]
(Mikszäth, Gavallerok p. 38)

is raised to the level of verbal communication. We meet sometimes with

highly condensed metaphors in every-day conversations. Miss Ε. K., 21
years old, relates a recent event to her girl-friend (in May 1941).
- Elmesz? csillogtam.
[ - You are already leaving us, I sparkled]

The verb csillog 'to sparkle and glitter' refers to the eyes shining
suddenly with joy, and to the face radiant with pleasure as the guest
seems to be about to leave.
In all these cases an instantaneous semantic transfer spans the distance
of hundreds of thousands of years of semiotic evolution.
The possibility of expressing the meaning of gestures in words is
explicitely stated in one of the inquit-equivalents of Läszlo Nemeth:
( . . . ) , adta vissza a legyintes ertelmet szavakban is.
[, he rendered also in words the content of his gesture.]
(Irgalom vol. 1: 254)

The semantic variety of French verbs of saying is much more limited.

Simplicity and restraint characterize the use of speech-deictics in the
novels of Balzac, Flaubert, Maupassant, Zola, Anatole France, Proust,
Romain Rolland, Jules Romains, Gide, Mauriac, Aragon, Camus, or
Rene Bazin. 20
The best represented semantic category is probably that of the verbs
denoting the manner of speaking: (a) intensity markers: murmurer,
270 Ivan Fönagy

chuchoter, souffler; crier, s'eerier, clamer, s'exclamer, tonner, hurler,

vociferer; and (b) markers of voice-quality: susurrer, beugter, bougon-
ner, brailler, gazouiller, glapir, mugir, rugir; zezayer, bredouiller, balbu-
tier; chanter; (c) voice quality and attitude: grommeler, ronchonner. The
various aspects of the speech turn are clearly distinguished by the verbs
of reporting: commencer, continuer, poursuivre, repeter, ajouter, termi-
ner; interrompre; repartir, reprendre, repondre, repliquer21. The verbs
conclure, trancher are halfway between conversational verbs (organi-
zers) and verbs of thinking.
Directives are also used as matrix-verbs: ordonner, commander, de-
mander, supplier. Some mediate directives could complete the list:
recommander, proposer, insister, reprocher, protester. I have only one
example of a commissive inquit verb: promettre. Questioning figures in
inquit-position already in the Chanson de Roland (demantv. 1525), the
verbs questionner, interroger are less frequent. Declarative verbs (or
assertives, Miller and Johnson-Laird 1976, specifying the mode of say-
ing, the degree of certainty of the statement) tend to fade in inquit-
position: affirmer, assurer, commenter, declarer, expliquer, exprimer,
observer, proferer, professer, remarquer, souligner some gradually closer
to the neutral basic verb, dit-il.
Secondary verbs of saying denoting originally non-verbal glottal noise
are less diversified in French than in Hungarian: gemir and soupirer are
the most typical instances. In a similar way, emotive verbs which do not
imply verbal expression, are represented by some types, such as s'emer-
veiller, s'alarmer, s'empörter, se revolter, s'indigner and a few isolated
tokens (qu'il me taquine, Celine, Voyage au bout de la nuit, p. 11-12).
Crying, weeping or loud laughter may occasionally yield matrix-verbs22,
though sanglota-t-il is quoted as an example of bad French by Albalat
(1921, cit. Spitzer 1926 p. 242).Tesniere quotes reprovingly rit-il (1976:
549). Verbs belonging to the semantic fields of facial mimetics, gestures,
bodily movements are not used as speech deictics, apart from the
traditional fait-il, fit-il (see: Spitzer 1926: 223-257), and some few verbs:
lancer, jeter, lächer.23
The answers given by 20 lecturers in linguistics (of both sexes, age-
group: 30-40) reflect a more liberal attitude.24 The matrix-verb pleurait-
il was rated as normal by 8 subjects, as acceptable by 4, and rejected as
unacceptable by 8; pleurnicher proved to be more suited to play the role
of a speech-deictic (normal: 14, acceptable: 3, unacceptable: 3). The
verbs sanglotait-il, gloussait-il were both qualified as normal by 15
subjects, as acceptable by 3, and as unacceptable by 2. The expressive
 Reported speech in French and Hungarian 271

bodily movement sursautat-il was declared normal by 9 subjects, as

acceptable by 6 and as unacceptable by 5.
Extroverted attitudes were more easily accepted as reporting verbs in
postponed position: s'indigna-t-il was rated as normal by 13 subjects, as
acceptable by 5 and as unacceptable by 4; insistait-il was normal for 13
subjects, acceptable for 6 and unacceptable for 1; se resigner, however,
was rejected as speech-deictic by 14 subjects, qualified as acceptable by
3, and as normal by 3. Interestingly, the egocentric se defendit-il proved
to be normal or acceptable for 18 subjects and unacceptable for only 2; le
defendit-il was normal or acceptable for 7 and rejected by 13 subjects. In
a similar way: menaga-t-il was qualified as normal by 11 subjects, as
acceptable by 6, and as inacceptable by 2 (one subject did not answer the
question); reprimanda-t-il was rated as normal by 12 subjects, as accepta-
ble by 5, as unacceptable by 3; whereas le louait-il was marked as normal
by 3 subjects, as acceptable by 8 subjects, as unacceptable by 9; le calma-
t-il was sensed as normal by 3 subjects, as acceptable by 5 and rejected by
12. The acceptability of the verb seems to increase with the increase in
directivity: encourager was more readily accepted, than louer (see:
above), the former was qualified as normal by 11 subjects, as acceptable
by 5 and rejected by 4. In a similar way, the verb gourmander was felt
'normal' by 13 subjects and accepted by another 4; see also: riprimander
quoted above; on the other hand, quereller was unanimously rejected as
a 'normal' reporting verb.
Verbal paraphrases, most liberally admitted in Hungarian texts, as of
verbs of saying in weak positions (see p. 261 f.), were categorically
refused by our French informants. Two periphrases, demeura ahuri and
prit son courage ά deux mains were only barely acceptable for two
subjects. All the other examples, translated from Hungarian novels,
such as prit la parole, leva les yeux, prit sa defense, trouva le mot juste
etc., were unanimously rejected. 25
The lack of concensus in most cases is in itself significant. It might
reflect a change in progress. The hesitations could be due partly to the
fact that the acceptance or refusal of a verb as a quotation-deictic is not
entirely a matter of linguistic competence. It also depends on the
informant's literary preferences.
The contrast between deictic and descriptive use is less explicit in
introductory position. This may account for the fact that verbs not
admitted in weak position may nontheless introduce a citation in strong,
initial position (see also Sabban 1978: 57). In the Chanson de Roland
there are but four verbs of saying figuring in weak position: s'escrier (vv.
272 Ivan Fonagy

1234, 2151, 3091, 3620), respundre (vv. 633, 1073, 1375), demander (v.
1525), renuveler (only in "Munjoie!" renuvelent v. 1375). In initial
position, however, other verbs may occur dispensing the utterance with a
quotation-deictic: apelat 'he called, summoned' (vv. 14, 643, 1020 etc.),
araisunet 'he rouses' (v. 3536), despersunent 'they insult' (v. 2580),pluret
'he cries' (v. 1853), ad sun piz batud 'he tapped his breast' (v. 2368),
saluz lifirent 'they welcomed (him)' (v. 2710), Les escheles Charlun li ad
mustret 'he shows him the army corps of Charlemagne' (v. 3314), cleimet
sa culpe 'he cries peccavi' (v. 2383). Thus, the verbal expressions,
denoting gestures; banished from reporting clauses in weak position, are
freely admitted in initial position as early as the 12th century.

2. A few remarks on indirect reporting

Indirect report in Hungarian is covered in some detail by Ferenc Kiefer's

chapter in this volume. The papers of Fischern (1899) Bally (1912, 1914)
Lips (1926), Günther (1928), Verschoor (1959), Strauch (1970, 1972),
Rey-Debove (1971, 1976), Gauvenet and Moirand (1974), Authier and
Meunier (1977), Authier (1978, 1979), Sabban (1978), Plenat (1979)
throw light on the syntactic and semantic problems of indirect quotation
in French. I shall deal with some selected aspects of this topic here.

Semantic forward drift

The theory of secondary verbs of saying which introduce referred speech

through the omission of the primary ones, was outlined for the first time
by Zsigmond Simonyi in the context of a classical study on Hungarian
complex sentences (1881-1883). He interprets structures such as gya-
korta külde, hogy 'he frequently sent [= messangers] to him that' as an
elliptic figure of thought where the second predicate 5 azt izene 'and he
let them tell' has been omitted (o.e. vol. 2: 156). The theory of the
double role played by verba non-dicendi in indirect quotations was
worked out in more detail much later, and independently from Simonyi,
by Läszlo Hadrovits (1969: 204-211). Hadrovits interprets in a similar
way expressions like feleje fordultam, hogy mit ohajt lit.: Ί turned to him
what he wanted', or Örökke sir, hogy nines ρέηζβ 'He always cries that
 Reported speech in French and Hungarian 273

he has no money', as elliptic forms for 'he turned to him and said', and
'he said crying'. The verb of saying is deleted, and the verb referring to a
simultaneous event absorbs its meaning (o.e. 204).
Syntactic figures of this sort and the use of verbs of acting (or verbal
expressions denoting emotive states and social attitudes) as reporting
verbs introducing indirect citation, overlap to a large extent with those
admissible as verbs of saying in direct quotation (see: p. 264f.). In both
cases recurrent syntactic distortions are gradually incorporated into the
grammar, and are regularly generated in live speech. In the dictionary of
the Hungarian Academy of Science (A magyar nyelv ertelmezö szötära,
Budapest: Akademiai kiado 1961, vol. 5: 1212) the verb sir is still listed
as intransitive. In fact, systematic rule transgressions altered the seman-
tic and morphosyntactic character of the word enabling it, for instance,
to take in certain cases a pronominal object (Mitsirsz? lit.: 'What do you
cry?')
We meet with similar transgressions in French where intransitive
verbs denoting originally non-verbal human soundmaking may occur in
the matrix clause of indirect reports:

et nous soupirons que vous n'ayez point le meme plaisir.

(Madame de Sevigne, Lettres vol. 4: 92, cit. Littre)

The verb soupirer 'to sigh' is used here as a verb of saying.

Not all verba non-dicendi admissible in inquit-position function alike
in governing a reporting clause. The Hungarian verb enyelgett 'he flirted'
focuses attention on the pleasant verbal and non-verbal activity itself. It
can be used as speech-deictic commenting on the preceding speech act
(pronounced flirtingly), but not as a matrix verb of an indirect quotation.
In a similar way, the French verb quereller, concentrates on the emotive
performance and its external features and diverts the attention away
from the actual communicative content. This could account for the
negative reaction of our informants 26 who rejected quereller as a repor-
ting verb of direct and indirect quotation (14 subjects out of 20 declared
the verb unfit for introducing an indirect quotation). 27 The same subjects
reacted in an opposite way in the case of s'emerveiller (normal for 15,
acceptable for 5 subjects) and s'indigner (normal for 4 and acceptable for
15 subjects).
Conversational verbs, too, were differently evaluated: poursuivit
(que) was admitted by all subjects (normal for 14); termina (que) has
been rejected, however, by 16 subjects out of 20, and interpella (que) by
19 out of 19.
274 Ivan Fönagy

Generally speaking, the verbs governing indirect quotation are cha-

racterized by a semantic right or forward drift, in contrast to the
postposed speech-deictics in direct questions, characterized by a left or
backward drift. The semantic forward drift is inherent in directives, in
verbs, of inquiry which are generally accepted by our French informants
as verbs governing indirect quotation. 28 The French verbs of thinking
apprendre, concevoir, se douter, inventer, savoir attested as matrix verbs
in indirect quotation, 29 cannot figure as postposed speech-deictics in
direct quotation.
Verbs expressing a reaction provoked by a speech act are necessarily
confined to the inquit-position. Verbs lacking semantic forward drift are
the Hungarian instantaneous verbs, such as, felfortyant 'he burst out in
sudden anger'; and so much more the verbs signifying facial expressions,
sudden gestures, bodily movements, highly frequent in inquit-position
(see p. 267f.). Verbal forms of politeness, such as üdvözölte 'he greeted
him', lacking anticipatory power, are also limited to the weak position in
direct quotation.
In Hungarian the objective conjugation seems to favour backward
drift, unless it is counterbalanced by a deictic pronoun or adverb
pointing forward:
(...), zokogta (obj.) vs. azt zokogta (obj.), hogy . . .
'he sobbed [it]' lit.: 'he sobbed that
In some verbs the backward drift of the objective conjugation contrasts
with the forward drift of the subjective conjugation:
(...), sirta (obj.) 'he cried [it] vs. sirt, hogy . . . 'he cried that . . .
(...), jätsza (obj.) a magnö vs. a magnö azt jätsza, hogy . . .
'plays the tape-recorder' 'the tape recorder plays that . . .
Most of the Hungarian as well as the French primary and secondary
verbs of saying can point in both directions: they may figure both as a
short comment on the preceding speech act (in direct quotation), or
introduce a subsequent report (in indirect quotation). This is true for
most assertive verbs, 30 and the verbs characterizing the manner of
speaking. 31
The acceptability of French and Hungarian verbs as matrix verbs of
indirect quotations is generally predictable on the basis of semantic and
morpho-syntactic features. Historical, i.e. unpredictable, factors, play
an important role nontheless. Verbs denoting facial mimetics or gestures
are limited to direct quotation in Hungarian. Some of these verbs are,
however, freely admitted in the same role: intett, hogy 'he beckoned
 Reported speech in French and Hungarian 275

that', bolintott, hogy 'he nodded that'. The French secondary verb of
saying lancer was admitted as matrix verb governing indirect quotation
by all 20 informants. The opinions varied in the case of the parallel term
il jeta (que), which was judged as normal by 7 subjects, as acceptable by
6, and rejected by the remaining 7.

Subjectivity in indirect reporting

In Hungarian direct quotation, the reporting verb can be considerably

extended even in week position, and offers wide possibilities for subjec-
tive characterization of the reported verbal process (see p. 268f.). Mar-
guerite Lips' contention (1926: 31) concerning the objectivity of direct
reporting is not confirmed by Hungarian literary examples. The matrix
verbs cannot be extended in a similar way in indirect quotation. The verb
itself offers, however, much variety, enabling the speaker or writer to
express his confidence or misbelief, his admiration or his contempt,
concerning the reported utterance.
The most essential vehicle of the reporter's attitude in live speech is
intonation. The semantic relevance of prosodic means is the most
conspicuous in artistic performances. Thus, the sentence A ζ uram rdm-
pamncsolt, hogy täncoljak neki 'My lord ordered me to dance for him' is
pronounced in Ferenc Herczeg's play ' Violante' by the lady of the manor
during a medieval trial (the role was performed by Ilona Torday). The
unmarked intonation pattern of the sentence should be falling gradually
(see: Figure 1). The melodic line of the actual utterance is quite
unexpected: its sharp rise has to be interpreted as the expression of the
underlying emphatic message: the categorical refusal of the lady: 'Hat

= cco 102 104

vm
^ ^L

0 Β ρ ρ; ψ
Az u -rem rött po - ran-csolt,hogy tan -col-jak ne-ki.
J =cco 31
> ?
k): Β . J iJ J J Λ J * J J j j j j

J = cco 104 >

Hdt m£g mit nem ? ! Figure 1

276 Ivan Fönagy

meg mit nem! (What do you take me for!)' (Figure 1 c).32 The melodic
pattern of the speaker's non-verbal answer is grafted on the intonation of
the reported utterance.
The speaker's feelings concerning the reported utterance are expres-
sed in a less sophisticated but nontheless effective way in every-day
conversations. In a recorded dialogue, Madame LH, a senior lecturer in
linguistics, reports with parody the behaviour of a French film critic.
There is no pause between narration and quotation, the juncture,
however, is clearly expressed by a much lower average pitch-level - a fall
of about 7 half-tones. The interval is strongly reduced in the reported
utterance, it rarely exceeds a third. The lack of pauses, the low intensity,
the chest register, the level tone in the unstressed syllables combined
with the monotonous repetition of the same rising-falling pitch pattern in
stressed syllables expresses boredom by vocal means, and the implicit
judgment of mechanical criticism based on preconceived ideas:
Et les critiques intellectuels disent que \ [deep fall:] bien >-v sür c'est u n ^
film extraordi- naire c'est une documenta- tion puis on en fait une
interpretation mar- ^ xiste ou catholique etcaetera etcaetere- / ra.33

Detachment, mixtures, neutralization

As in the above quoted conversation, sentences reported in indirect style

become increasingly independent from the main clause in lively reports.
The narrator regularly switches from indirect into direct style (Spitzer
1939: 81).
Combinations of the two basic forms of reporting are also common in
French and Hungarian, both in literary prose and conversation. In child
language they could be considered as grammatical slips.
'Le gar$on il dit ä sa maman que c'est pas moi qui Va casse' said a four years
old French boy to his mother (Lentin 1972: 145) [My italics].

Similar inconsistencies occur, however, in stories told by adults where

direct speech breaks through the frame of indirect quotation: 'Alors il a
dit que moi je suis contre' (a 38 old employee) (see also Verschoor 1959:
142). We meet with the same switching in strategy in Hungarian conver-
sation, and even in literary prose:
Azt kiältozzäk, hogy: Ambrus legyen a mi püspökünk. (My italics)
[They shouted that: Ambrus should be our bishop.]
(Tibor Deri, Α kiatkozö)
 Reported speech in French and Hungarian 277

De akkor jött az, hogy Dezsö, igazän ne legyen olyan fäd.

[But then came that D. don't be so tedious.]
(Dezsö Keller, Pest, az Pest p. 245)
The opposition between direct and indirect reporting is regularly soften-
ed and sometimes completely neutralized in assyndetic sentences.
Non, j'aime autant vous l'avouer, je ne me sens plus de force ä me tracasser
davantage.
(Celine, Voyage au bout de la nuit p. 354)
Kerdem, hoi van az a vagon?
(Karinthy, Krisztus es Barabäs p. 109)
This last sentence could mean either Ί asked where the carriage was,' or
Ί asked: where is the carriage?' A similar neutralization took place in
reported questions in Popular French: Je me demande qu'est-ce qu'elle
fait? (see Henri Frei 1929: 213-214).
An interpretative ambiguity is inherent in Popular French qu'il dit as a
speech deictic. The epenthetic que can be interpreted as an elliptic
anaphore: qu'est-ce que - que. But it could also be conceived of as a
displaced conjunction of a corresponding indirect quotation.
En Amerique, j'allais pas assez vite, que tu disais . . .
(Celine, Voyage au bout de la nuit p. 369)
This sentence could be related to (a) Έ η Amerique, j'allais pas assez
vite, c'est ce que tu disais', or (b) 'Tu disais qu'en Amerique, j'allais pas
assez vite.' (The disjunction 'either/or' is not an exclusive one.)

The status of elliptic indirect quotation

Repetitive de dicto questions may appear at the surface as simple Yes/No

or Wh-questions. The rising interrogative intonation pattern in repetitive
Wh-questions clearly alludes to an underlying 'Did you say that pT 'Did
you ask whether pT
- Comment? \ (Wh-Question)
- Comment?? / (Secondary, de dicto question)
In her dialectic monologue, Cliges repeatedly questions her former
statement:
Bien s'en avra aparcevoir,
S'il onques d'amors s'antremist,
Ou s'il par parole en aprist:
Aprist? Or ai je dit oiseuse. (My italics)
278 Ivan Fönagy

[He will become aware of it ( = of her love)

If he ever was in love.
Or if he learned about love from hearsay.
From hearsay? I talked nonsense.]
(Chretien de Troyes, Clig£s w . 1010-1014)

The conjunction of the underlying subordinate reported clause implies a

preceding principal clause in French as well as in Hungarian.
Que sans chemise je m'y mette?
Fi, ma soeur, vous n'y pensez pas!
(Anonymous poet of the 18th century)
Majd bolond vagyok, hallja kend! Hat megveszett kend? Hogy en a kollekciö-
mat szetszörjam, amit olyan färadsägosan gyujtöttem?
[Do you take me for a fool! Are you crazy? That I should waste my collection I
painstakingly gathered?]
(Mikszäth, A szelistyei asszonyok p. 58)

Some authors (Bally 1912,1914, Lips 1926) consider such echo-questions

as instances of free indirect speech (see p. 283 ff.), as the verb of saying is
deleted, and the speaker reports the words of another. In contrast to all
other examples of free indirect style the quotation is explicit and it is not
integrated into a narrative (see also Verschoor 1959: 18, and Schürr
1926: 305).

3. Forms of fictitious quotation

Information given in quotation marks may differ in several ways of

reported speech events.
Non-verbal communication by means of facial mimicry may be verba-
lized. Anatole France paraphrases a smile in nine lines
lis me regarderent avec un sourire de pitie qui semblait dire: 'Est-ce que nous
ecrivons 1'histoire, nous? Est-ce que nous essayons d'extraire d'un texte, d'un
document, la moindre parcelle de vie ou de v6rite? ( . . . ) II faut etre bien vain
pour ecrire l'histoire: il faut avoir de l'imagination.'
(L'ile des Pingouins p. iii)

'For-example'-quotations are on the border-line between reality and

fiction.
II repond des trues pas serieux comme: 'Et ton oiseau ä toi, tu te l'es fait
denicher souvent?' Des blagues, quoi ( . . . )
(Queneau, Zazie dans le Metro p. 72)
 Reported speech in French and Hungarian 279

Fictitious quotations of the type 'what might he think now' or 'what

will he say' try to come close to a potential verbal reality.
Un soir, nous sommes attendus ( . . . ) On vient au-devant de nous; nous
n'arrivons point. Que sont-ils devenus? Quel malheur leur est-il arrive? Per-
sonne de leur part! La soiree secoule ä nous attendre.
(Rousseau, Emile, book 5 p. 544)

The reader is supposed to react verbally in Diderot's 'Jacques le

Fataliste'
( . . . ) ils furent acueillis par un orage qui les contraignit de s'acheminer
Oil? - Oü? lecteur, vous etes d'une curiosite incommode! ( . . . )
(o.e. vol. 1:39)

More extensive but no less fictitious dialogues with the reader or the
opponent are frequent in scientific literature. 34 Philosophical or moral
dialogues from Plato's Hippias to Alfred Renyi's "Letters on probabi-
lity" are extreme cases of fictitious quotation.
The speaker may give, with polemic intention, an exaggerated expres-
sion to the imputed ideas of his opponent.
Isabelle:
Oh j'sais c'que tu vas m'dire: 'que j'ai 18 ans que j'ai pas d'metier, Marc non
plus, euh, on n'sait meme pas oü on va euh, j'ai l'examin la s'maine prochaine,
e'est patati patalere.'
Jean (the father of Isabelle): J'aurais dit patati patalire?
('La gifle', film by Claude Pinoteau 1974)

This type of fictitious quotation is raised to the level of literary genre

in parodies. Such condensed false quotations, omitting contingent ele-
ments, come closer to the essence of the original than verbatim repro-
duction of the text (see Meyer 1961: 109-112).
The verbal projection and impersonation of the 'inner voice', reminis-
cent of schizophrenic hallucination, constitutes another literary form of
false quotation. 35
Negated quotation ('he didn't say p') is a special case of fictitious
report.
( . . . ) il s'interessait tant ä eile que cette affection rendait encore plus etroits
ses liens avec le docteur qui n'osa jamais dire a' ce vieux garqon:
- Et vous aussi, vous avez done perdu des enfants.
(Balzac, Ursule Mirouet p. 34)

Self-quotations may also be fictitious. The verbum dicendi is used to

lend emphasis to the subsequent or preceding statement.
280 Ivan Fonagy

Kritikus: En mondom neked . . . az ilyen tökeletes ϊΐΐύζίόί adö maszkok . . . az

legenda.
[The Critic: I tell you . . . such masks which create a perfect illusion . . . belong
to fiction.]
"J'aime Gertrud, et je le respecte, vous dis-je, ( . . . )
(Gide, Symphonie pastorale p. 79)

The use of inverted commas transgresses the domain of reported

speech (see: Rey-Debove 1971, 1976; Milner 1978; Fonagy 1980, Gaul-
myn 1981, Bergheaud 1984). The semantic extension of quotation marks
is based on the principles governing lexical or grammatic transfers. The
detachment of quotation marks from their basic function is partly due to
selective or partial quotation. Selective quotations are generally partial
in the evaluative sense of the term. Quotation-marks expressing a
deprecative judgment easily turn into marks of irony; especially so, if
they point to an unidentified or vaguely defined source. 36
Ghil est un imbecile. Moreas
N'en est foutre pas un lui, mais, helas!
II tourne, ainsi que ce Ghil, "chef d'ecole"
Et cela fait que de lui Ton rigole.
(Verlaine, Conseils)

Ironic quotation marks, allowing for saying ρ without meaning it,

became widely used in Hungarian newspapers and political leaflets of the
Stalinian period. Thus, the word 'philosophy' was thoroughly put into
quotation marks if applied to non-marxist doctrines, suggesting that such
ideas 'have nothing to do with philosophy'. Quotation-marks may thus
serve the purpose of self-defence, either by throwing the reposibility
upon some other person (Bergheaud 1984), disconnecting the words
enframed from the speaker (Milner 1978: 301);or by invalidating the
words put into inverted commas. In the latter case, the quotation marks
disconnect the isolated sequence from its basic meaning: Ί didn't mean
it, I was only joking'. They function as a hedge.

4. Forms of hidden quotation: the hidden text

Allusive quotation, as a figure of thought, or as a guessing game is

necessarily unmarked (cryptical, according to Hermann Meyer 1961:
229). The reader may take pleasure and satisfaction in discovering the
well-known in the still unknown. The aesthetic impact of hidden quota-
 Reported speech in French and Hungarian 281

tion is due to a succesful synthesis of redundancy and condensation, just

as in the case of other figures. The author echoes the words of another
author: redundancy; in the new context his utterance is at the same time
personal and novel, and the evoked words appear as a kind of second
voice or as an incompletely erased line in a palimpsest: condensation.
The underlying text is supposed to be present in the reader's me-
mory. 37 The Roman reader easily detected a verse by Alkaios in Hora-
ce's ode I, 9 ('vides ut alta stet nives candidum / Soracte'). In modern
times some texts of Shakespeare and Goethe may rival the bible. Sändor
Petöfi (1820-1849), shortly before his death on the battlefield, in the last
three lines of his last poem 'Szörnyü idö [Horrible times]' clearly alludes
to the well-known verses of Macbeth (act V. scene v).38
In his poem Ά ket növer [The two sisters]' Mihäly Babits (1883-1941)
alludes to Baudelaire's 'Les deux bonnes soeurs' who himself quotes
implicitely Swinburn's 'Death and Birth' and 'Birth and Death.' Mal-
larme evokes Verlaine's words and style in his Epitaph (Tombeau) of
Verlaine. Verlaine evokes Villon in the 'Ballade de la vie en rouge'; and
Villon does the same for the authors of the past in the 'Ballade ä ce
propos en vieil langage franQois.'
Mihäly Babits intentionally and openly uses the collage-technique of
quotations (see: Räba 1960), just as does T. S. Eliot who alludes to 35
authors in "The Waste Land," from Ezekiel up to Verlaine. This is in
full agreement with Babits's vision of world-literature: 'Only the greatest
writers belong to wored-literature, in the true sense of the term. The
history of world-literature is the history of these giants who shake hands
over the people's heads' (History of European literature, Budapest:
Nyugat 1936: 11).
Tendencious modifications of the original text are a further source of
information. A contemporary Hungarian writer, Peter Esterhazy combi-
ned successfully a speech given be Kälmän Tisza, the reactionary Hunga-
rian Prime Minister of the First World War, with that of the Stalinist
leader, Mätyäs Räkosi. In his still unsurpassed monograph on the forms
and functions of literary quotation, Hermann Meyer (1961) gives en-
lightening examples of similar twisted quotations (o.e. 73, see also Ch.
Jones 1968).
The rules generating allusive twisted quotations have not yet been
formulated. There are certainly limits that the modifications of the
original text cannot go beyond without completely erasing the underly-
ing text of the 'palimpsest'. How far can the syntactic and semantic
structure of the sentence be modified? Transformation by negation
282 Ivan Fönagy

seems to be more easily admitted than active—>passive, or passive^ ac-

tive transformation (see: Calbris 1982, Gresillon 1984). Similarly , the
substitution by antonyms is more readily accepted than other changes in
semantic quality of the lexical elements (Freud, Jokes [1905] in Standard
Edition vol.2; Fonagy 1982). The rules governing parallelism (see
Driver 1896, Steinitz 1934, Hightower 1959, Jakobson 1966) could offer
a sound basis for the study of twisted quotations. In contrast to twisted
quotation where the modified text is superimposed to the hidden origi-
nal, in the case of parallelism the modified version follows the original
one. In both forms of partial repetition the structure of the text is
essentially preserved (principle of synthesis, according to Robert Lowth
1778). Lexical substitutions are based on analogy and antonymy (Lowth
o.e. § X-XI, Driver o.e. 385ff., Steinitz o.e.; Gresillon 1984: 112ff.).
There are, nevertheless, transformations typical for parallelism - such
as gradation, comparison, question and answer - but not for twisted
quotation. And, conversely ironic distortions are typical of twisted
quotation, and do not occur in parallel verses. In twisted quotations a
humoristic effect may be achieved by means of remotivation, i.e., by
restoring the original, concrete meaning of an abstract term. 39
Parody of style and genre constitutes an extreme case of twisted
quotation. The features belonging to the level of expressions are strictly
maintained or even condensed (exaggerated). The content, however, is
completely changed or reversed. 40
Common places, routine formulae (Coulmas 1979), are the best
hidden, and linguistically the most relevant forms of quotation. When
using proverbs we are aware of quoting texts of anonymous authors.
(The Hungarian speaker may refer in such cases to a fictitious man who
lived once upon a time: mondta az egyszeri ember 'said the man of
yore'.) Prefabricated jokes, moral commonplaces occur, however, with-
out quotation marks and without reference to 'the man of yore'. In a
previous work (1982) I attempted to show the frequency and variety of
prefabricated sentences automatically elicited by typical (recurrent)
situations. If we consider such mechanically reproduced, holistic utteran-
ces as hidden quotations, the statement that quotations are in banal
every-day conversations at least as frequent as individual statements,
may not be excessive. Human speech is interspersed with the speech of
others, as Bakhtin (1978: 157 ff.) puts it. Riffaterre considers citation a
special case of cliche (1979: 48).
 Reported speech in French and Hungarian 283

5. The hidden speaker

Free indirect quotation

Otto Behaghel mentioned in 1877, for the first time, a special kind of
reported speech which could be interpreted as a synthesis of quotation
and narration. By a variety of verbal techniques, the narrator points to a
speech act or verbalized mental process of another person, the 'subject
of consciousness' of A n n Banfield (1973: 29), the enunciator, according
to Marc Plenat (1979: 116). The narrator can refer in the same way to
public opinion (Lerch 1914: 486, 1930: 11), Herczeg 1963, 218, Beyerle
1972: 361) or to his own statement formulated in the past. In contrast to
hidden literary quotations, his speech act or his written text is our only
source of knowledge. Nevertheless, the underlying message must be
retrievable without an overt act of quotation.
The some twenty terms denoting this type of implicit reference
represent not only divergent conceptions, but reflect at the same time
the complexity and diversity of the investigated phenomena. 4 1
The procedures of encoding may operate both at the level of grammar
(Bally 1912, 1914, Lips 1926, Kalik-Teljatnicova 1965-1966), and at the
level of discourse ( R e y - D e b o v e 1971,1976, Strauch 1974, Authier 1979).
Since recurrent parole-phenomena may be integrated into the grammar,
the two procedures are largely overlapping. The choice of the level of
encoding also depends on the possibilities offered by a given language.

Grammatical devices of encoding

Verbal expressions of saying, communicating or thinking may hint at a

verbal or mental act without governing an explicit or implicit quotation.
les chevaliers ki la menoent
cumanda tost a arester:
descendre vuelt e reposer.
[she ordered the knights
who escorted her to make a halt:
she wished to alight and have a rest.]
(Marie de France, Le lai du chevreiueil)
A jo tanulo udvariasan vegtelen megertessel köhög. Termeszetesen, valami
erdekes peldät (My italics).
[The first class pupil cleared his throat politely, with great comprehension. Of
course [one has to look for] an interesting problem].
(Karinthy, Tanär ur kerem p. 31)
284 Ivan Fonagy

In the last example, the little cough is clearly an act of communication: it

punctuates the pupil's silent answer.
In both languages, French and Hungarian, the conjunction which
outlived the deletion of the reporting clause, figures as an unequivocal
indicator of a speech act: TO SAY that ρ -» 0 that p.
D'abord c'etait des folies de contracter encore des dettes . . . Que dejä
rembourser Gorloge, $a serait un travail d'Hercule! (My italics)
(Celine, Mort ä credit p. 238, cit. Spitzer 1939: 81)
Egy kisebbik fiu is ott volt. Hogy 6 majd bedugja, ha alszik, a boszorkänyt a
kemencejebe (My italics).
[The younger boy stood by, too, That he will put, as soon as she will fall
asleep, he will push the witch into her stove.]
(in an improvised narrative of a 5 years old boy)
Free indirect style can be syntactically marked by tense-shift (Bally 1912:
597)42, present —» imperfect, future —» present or conditional, see: Lips
1926: 187-189, Auerbach 1967; 451^53), shift of moods (imperative - »
subjunctive, see: Lips o.e. 73), just as in indirect quotations, with the
same restrictions (Damourette and Pichon 1911-1936, vol.5: 175-245;
Cohen (1967: 389); Plenat 1979:112-125). In indirect quotation the verb
of the main clause governs the shift taking place in the subordinate
clause. In the absence of such verbs in free indirect style, the reader may
conclude that the shift is due to the virtual presence of a verb of saying or
thinking, and he will attribute the words to the hidden speaker.
Such indicators are not available for the Hungarian writer since there
is no tense agreement, in Hungarian, and the tense-system itself is much
less complex. He will have to compensate for this deficiency of the
language by exploiting the possibilities offered at the level of discourse.
For instance, a piece of free indirect speech marked by a tense shift in
French will be put in Hungarian into direct speech or indicated at the
level of discourse.43
Let us add that shifts of mood and tense, while they are conspicuous to
the writer or speaker, do not always guarantee unambiguous decoding.
In order to interpret the encoded message correctly, we have to be aware
of the respective shifts in person, tense and mood. Such an awareness is
enforced only in the case of syntactic incompatibility: when the use of
moods and tenses clashes with the grammar of narrative prose within a
given context.
Du coup, Bordenave comme^a ä se fächer. II repeta le mot idiot, chercha
quelque chose de plus fort, trouva imbecile et cretin. On siffleraif, l'acte
finiraif pas (my italics).
(Zola, Nana vol. 1: 54-55)
 Reported speech in French and Hungarian 285

In the above, nothing justifies the conditional form which is hence to be

understood as a shifted future tense: 'People will hiss, one will have to
stop the act, said Bordenave'. In other instances the conditional or the
imperfect may seem acceptable at first glance, and we might be tempted
to attribute the statement to the author; a wider context, however, may
force the reader to change his mind, and attribute the utterance to the
character (see: Wagner 1972: 87). The ambiguity of the conditional or
the imperfect is often unresolved in the novels of Camus.
Justement, le concierge etait devore par la soif.
(La peste p. 31)

The imperfect of έίαϊί could be a transformation of the present of direct

speech ('le concierge s'est plaint d'etre devore par la soif', 'le concierge
se plaignait: "Je suis devore par la soif"), as well as a genuine narrative
imperfect. Considering the grave condition of the concierge, it would
seem unlikely, though perhaps not impossible, that he would have
uttered such a long and elaborate sentence.
An unambiguous device is the use of quotations marks contrasting
with the continuity of the narration.
U l'interrompit presque grossierement: "Pourquoi jouait-elle cette comedie?
Elle savait bien ce qu'il redoutait"
(Mauriac, Fin de la nuit p. 172, cit. Landry 1953: 17)

Colons are widely used in French as well as in Hungarian to indicate

that the following sequence has to be attributed to the speaker (see: Lips
1926: 187-188).
Le bazardeur souffle: eile se croit chez Fior, cette petite conasse. II fait un joli
sourir ä pleines dents pour dire: "Pas la peine, regardez-moi 9ui-la."
(Queneau, Zazie dans le Metro p. 44)

Encoding at the level of discourse

Any feature - phonetic, syntactic or semantic - confined at the level of

discourse to live-speech, may allude to a speech act which is not overtly
represented in the narrative.
(1) The text can reflect with the help of graphic means some articula-
tory or prosodic particularities of the speaker just as in direct quotation
(see: p.256). 44
Via essmefie (= Viola eile se mefie)
(Queneau, Zazie dans le Metro p. 54)
286 Ivan Fönagy

In Maupassant's novel 'Notre coeur' (p. 22) a capital letter is used to

ironically reproduce the emphasis given to a word (Elle):
( . . . ) il se sentit humilie dans sa vanite masculine par le sentiment de
classement qu'Elles font de nous ( . . . )

Jänos Arany systematically markes by means of italics the words attri-

buted to the character. Usually, these references are somewhat ironic.

Ez, mikor nem halljäk, es mikor nem lätjäk,

Pirulva kerdezi töle . . . testverbätyjät.
[And she, when heard and seen by nobody,
inquired bashfully after her . . . brother]
(Csalädi kör - Family circle)

"Free indirect style, before becoming a grammatical form ( . . . ) is an

intonation pattern" according to Albert Thibaudet (1922: 280). Implicit
references are indeed more often indicated by prosodic means than overt
quotation. Prosodic allusions generally reflect a negative attitude: the
speaker dissociates himself from the reported statement or opinion,
expressing disapproaval, reproach or irony. Ironic intonation may make
a reported reproach seem completely unfounded. A 46-year-old Hunga-
rian house-wife relates a discussion she had with her mother-in-law:
Persze megint az έη hibäm volt O f course, this was again my fault'. The
speaker seems to make the rejected criticism her own as she reproduces
it in the first person singular. Such forms of ironic inversion of roles are
typical in Hungarian as well as in French (Bally 1914, Lips 1926: 82) or
German conversations (Lerch 1914, 1930: 99, Günther 1928: 109). They
are based on rhetorical figures (irony, chiasm) and more or less integra-
ted into the grammar.
On the artistic level, the intonation pattern may indicate the implicitly
quoted speaker's attitude with great precision. The Hungarian actor
Oszkär Ascher, for instance, uttered the sentence Zajcsek ordit, hogy ot
nem hagyjak tanulni [Z. cries that one doesn't let him study] with a
superimposed melodic pattern generally associated with utterances such
as '/fagyjatok engem bikeben [Don't bother me]' with repeated rise-falls
in the stressed syllables: I ' h a j j a t o k ,-v 'engem ^ 'be:ke:ben /.
(2) Morpho-syntactic particularities of informal or popular style are
widely-used indicators of hidden speech acts. The possessive structure
SUBSTANTIVE! ä SUBSTANTIVE 2 is such a signal of free indirect
speech in French novels ("la femme a Bourron", Zola, Travail p. 382;
cit. Lerch 1930: 111, see also Jenny Simonin 1984: 50).
(3) The syntactic structure of spoken utterances necessarily differs
 Reported speech in French and Hungarian 287

from that of written prose. In conversations or discussions we need to

react instantaneously to verbal or non-verbal stimuli. The Latin proverb
verba volant, scripta manent also implies that the words, once uttered,
escape our grasp. The stylized reproduction of such 'noises' regularly
allude to an underlying speech act. Thus, for instance, corrective reset-
ting
Mais comment . . . Mais comment cela etait-il possible!
(Romain Rolland, L'äme enchantee 1: 115)

sloppy repetition:
M.le cure n'a pu vaincre sa resistance; car M.le cure trouve qu'il ne doit pas
refuser une honnete fille, sous le pretexte qu'elle a ete femme de chambre.
(Stendhal, Le rouge et le noir, Oeuvres 1: 259)

dislocation (Bally 1941):

C'etat pas de sa faute, ä lui, Gabriel . . .
(Queneau, Zazie dans le Metro p. 280)

omission of more or less essential constituents (ellipsis, aposiopesis); (see:

Lips (1926: 80), Herczeg (1963: 38, 49)).
(4) Words proper to the spoken language, such as, argot-terms con-
trasting with the style of the narrator point to speech acts extraneous to
the narration (see: Lips 1926: 69, Lerch 1930: 109).
Loute etait stupefaite. Manquant d'aisance, de genie, eile trouvait sa soeur
tres forte (my italics). (Cocteau, Le grand ecart, p. 110)

The expression tres fortltres forte referring to some outstanding intellec-

tual quality, is still, and certainly was in 1924, a colloquialism.
(5) The balance of linguistic functions changes as we pass from one
type of discourse to another. The conative and expressive functions
(Jakobson 1960: 355-357) have a greater impact on direct speech acts
than on written scientific or literary communication. The dominance of
the expressive function may result in non-referential emotive utterances
and interjections in live speech. Consequently, interjections are most
suitable for evoking direct speech acts (see: Tobler 1889, Lerch 1914,
Ullmann 1964: 111). Similarly, colloquial hyperboles may signify every-
day conversations. There have been examples of this type of indirect
free quotation since the Old French roman courtois (Günther 1928: 55,
111).
(6) Sentences routinely elicited by recurrent situations are subject to
semantic erosion. Such situation-specific expressions allude to a speech
act outside the narrative frame-work. The sentence Le colonel est trop
288 Ivan Fdnagy

bon as a semantically unrestricted statement may occur in narration as

well as in direct speech. As a semantically reduced bound sentence it
evokes live-speech, and will be transformed by the reader from the third
to the second person.
Leurtillois, vous mangez? Le colonel etait trop bon. Leurtillois avait mange.
(Aragon, Les communistes Mai 1940 p. 16)
Commonplaces, far beneath the information level of the narration, will
be assigned to the character or to a social group (Spitzer 1928 a: 174).
( . . . ) il leur appartenait ä toutes deux, et elles le jugeaient d'un sourir
indulgent, ironique, comme un grand gösse amüsant, epuisant, pas serieux,
pas tres sage . . . - (Tous les hommes sont de meme!) On ne lui en voulait pas
(my italics).
Romain Rolland, L'äme enchantee 1:69)
(7) Semantic erosion is generally compensated by the precise pragma-
tic contours of the semantically weak expression, as is the case with
French adverbs such as franchement, decidement (Ducrot et al. 1980),
peut-etre. Such adverbs, used in their situation-specific pragmatic capa-
city always refer to an underlying speech act.
Le Midi enchanteur par son climat, mais incommode vu les moustiques, et le
Centre, franchement, n'avait rien de curieux (My italics)
(Flaubert, Bouvard et Pecuchet p. 17)
II piquait decidement un soleil (my italics).
(Aragon, Les communistes, Mai 1940 p. 35)
II n'allait pas faire les dindes, peut-etre (my italics).
(Zola, L'Assomoir p. 299, cit. Lips 1926: 299)
The comma that separates peut-etre from the rest of the sentence
represents a juncture and counterbalances the pragmatic function of the
sentence adverb by suggesting something like 'Let's hope they won't
behave like silly geese.'
The Hungarian adverb is (also) may be semantically reduced in live
speech to merely serve the expression of various attitudes, such as
nonchalance, sulkiness, grumbling.
Panaszkodott, hogy färadt, hogy a läbän is verholyag nött, ö hajnalig nem
megy mär toväbb.
(Läszlö Nemeth, Bün p. 9)
In a non-colloquial narrative the clause α läbän is verholyag nött would
correspond to 'he had blisters also on his feet (i.e. and on other part of
his body). In the present context, however, is has a pragmatic function
only, similar to that of an intonation pattern expressing annoyance or
complaint. The sentence alludes unmistakably to a speech act:
 Reported speech in French and Hungarian 289

He complained that he was tired, and that he had blisters on his feet, he for his
part was not willing to go any further before daybreak.

Ferenc Kiefer's examples of ungrammatical forms of indirect report on

p. 214 - such as Peter azt ällitotta, hogy Jänos akär tizszer is megprobälja
'Peter stated that Jänos would try it even ten times' - could be instances
of expressive rule transgression, reflecting a hidden speech act - the
utterance of Jänos in the example quoted above (38 a in Kiefer's paper).
The loss of information is often reflected on the phonetic level (see:
Horn 1923, Fonagy 1973). The full form azutän of the Hungarian adverb
of time 'after, afterwards' is reduced to aztän if it is used in its weak
meaning, 'well then', or like a menacing gesture. Similarly, mär megint
'once again' is reduced to mär meg as a pragmatic modal 'as it happens'.
The weak forms unmistakably indicate direct speech.

Meg aztän, amikor az ember vagy egy hetre szabadulni akar hazurol, milyen jo
elöhozakodni awa, az asszony elött, hogy nini, mär meg Aradra kell menni.
[Well then, whenever one would like to escape for a week, it is rather helpful
to tell your wife, that why! as it happens, one has to got to Arad.]
(Tömörkeny, Hat szäl gyertya - Six candles p. 31)

(8) In Hungarian narrative texts the unmotivated occurrence of a third

person pronoun may indicate free indirect speech.
Egy mozi elött kepek voltak ( . . . ) Ö meg sohasem volt moziban (My italics).
[There were pictures in front of the cinema. He had never been to a movie.]
Läszlö Nemeth, Biin p. 44)

As the third person is marked by suffixes, the personal pronoun is

redundant, unless it marks contrastive emphasis which is then also
expressed by heavy stress and high pitch. This, however, is not the case
in the above sentence. Such silent (unstressed) emphasis is reserved to
the first person pronoun.
The Hungarian reader will automatically relate the narrative sentence
to a directive speech act such as: £n meg sohasem voltam moziban Ί
have never been to a movie'. The exceptional presence of the unstressed
pronoun of the first person singular, en, could be explained on the
ground of an ego-centered interpretation of the situation, and interpret-
ed in terms of 'narcissistic emphasis', as opposed to contrastive empha-
sis.
French lui qui / eile qui (or: lui / eile + Past Participle) in free indirect
reports are comparable with the ego-centered use of Hungarian en T . In
contrast to Hungarian, the French pronouns are always stressed if used
emphatically.
290 Ivan Fonagy

comment voulait-on qu'elle vecut au fond de ce trou, eile habituee ä sa

chambre si claire, ( . . . ) (my italics)
(Zola, La bete humaine p. 283)
II s'etonnait lui-meme: lui qui avait toujours affect6 le plus grand mepris des
femmes.
(Aragon, Les beaux quartiere p. 290)

(A typical instance of narcissistic emphasis is the formulaic sentence

C'est moi qui vous le dis! implying that the referred statement is certainly
true. Its counterpart, the equally situation-specific C'est toi qui le dis!
means that the preceding statement is probably false.)
There are other sentences bound to the first person singular. Ou ai-je
la tete! - Qa, je m'en fiche pas mal! - Je ne sais pas ce qui me retient
(de . . . ) . All these sentences, transposed, arbitrarily to the third person
in the frame of a narration, will always evoke direct speech.
Elle ( = Nana) ne savait pas ce qui la retenait de flanquer ce sale monde ä la
porte.
(Zola, Nana p. 127)
Mais Couteau, ricanant, le hua. lis s'en fichait pas mal de son rapport!
(Zola, Le Debacle p. 290)

C'est ma folie! in the first person singular is a formulaic sentence

characterized by narcissistic indulgence. The transposition into the third
person strips the utterance of its empathy and indulgence. The original
narcissistic attitude is, however, maintained, if the transposition takes
place in the frame of an indirect implicit quotation:
Elle ( = Madame Bovary) declarait adorer les enfants; c'etait sa consolation, sa
joie, sa folie (my italics)
(Flaubert, Madame Bovary, p. 90)

Sentences regularly used in the second person, such as Cela vous fait
honneur, transposed to the third person in the narrative, function
similarly.45
M. Rateau etait plein d'estime pour le chocolatier Barrel. Non seulement ä
cause des scrupules de conscience qu'il avait vis-a-vis de ses ouvriers et qui lui
faisait honneur, mais aussi pour tout ce qui aureolait cette personnalite
industrielle; (my italics)
(Aragon, Les beaux quartiers p. 28)

The French pronoun on functioning as a general subject, is often used

like a first person pronoun.
Due to its semantic variability, on plays an important part in pseudo-
hidden quotation.
 Reported speech in French and Hungarian 291

. . . Apres quoi, eile s'amusait de plus belle. A mesure que le jour s'avan5ait
elle redevenait plus jeune. Les soir on allait [isi person plur.] au theatre; et
c'etait l'eternel plaisir de reconnaitre dans la salle des memes eternelles
figures; - le plaisir, non de la piece qu'on jouait [general subject], mais des
acteurs qu 'on connaissait [twisted 1st person singular], et dont on relevait une
fois de plus les travers ( . . . ) On trouvait que l'ingenue avait un filet de voix
"comme une mayonnaise tournee" . . . (my italics)
(Romain Rolland, Jean-Christophe vol.5: 169)

Conversely, some formulaic sentences in the first or the second person

are not interpreted in terms of direct speech by the reader (see: Bally
1914: 410, 419).
Si Jean revenait devant Cecile, je ne saispas, moi, avec la croix de guerre? (my
italics)
(Aragon, Les communistes, vol. 4: 2)
Pas tous les mois evidemment qu'on a la chance d'un de ces enterrements
comme celui de la vieille dame Co tin, de la rue Longue, pour laquelle un
eveque s'etait derange, pensez done, ( . . . ) (my italics)
(Aragon, Les beaux quartiers, p. 12)

In the first example, je ne sais pas, moi could be considered as a variant

of par example. In the second, pensez done is a non-referential emphasi-
z e s Such pseudo-direct speech acts could be regarded as mirror-images
of free indirect report. In the latter, the point of departure is a pseudo-
narrative text inviting the reader to look for a hidden speaker. In the
case of pseudo-direct speech the presence of the speaker is veiled
through the demotivation of the surface structure (the first or second
person).46

narrative text free indirect report speech act

>• <

dramatization demotivation

In contradistinction to the narrator, the speaker engaged in actual

conversation makes use of the deictic field (Bühler 1934: 386 ff.) to point
to the object in space and time. The use of verbal gestures - deictics,
anaphoras, shifters (Jakobson 1971: 130-147) - in a narration reflects a
direct speech act.
Nem is az a kis kave kellett volna neki, csak a jo szo.
[What he wanted was not so much this little coffee, rather a few nice words].
(My italics)
(Läszlo Nemeth, Bun p. 286)
292 Ivan Fonagy

( . . . ) eile voulait parier du parfum qui emanait de ce meussieu.

(Queneau, Zazie dans le Metro p. 7)
(10) Sentences narrowly linked with conversational situations point
beyond the narrative frame. Thus, directive or commissive speech acts
are necessarily dramatic, involving the participation of at least two
speakers. Karl Vossler (1921: 74) correctly interprets in terms of free
indirect discourse the verses 233-236 of the Pelerinage de Charlemagne'
where the emperor's vow is related in the third person:
L'emperere de France i out tant demoret
De sa mollier Ii membret qu'il olt parier:
Ore irat le rei querre qu'ele li out loet,
Ja n'en prendrat mais fin, tresqu'il l'avrat trovet.
[The emperor of France has whiled there long;
He remembered the words of this wife.
He will return now as he promised her;
He will not rest till he will find her.]
(ed. Karl Voretsch)

(11) Verbal strategies of politeness or impoliteness also imply a direct

verbal interaction. In such cases, the reader is compelled to transform
the narration into direct speech.
Une vieille qui ne part pas leur a prepare un jus bien chaud, Ce n'est pas de
refus avec ce crachin qui vous penetre. (My italics)
(Aragon, Les communistes, Mai 1940: 60)
Rhetorical questions have often been considered as markers of free
indirect quotation (Kalik-Teljatnicova 1965: 133; Herczeg 1963: 87).
Very often they do, indeed, allude to direct speech, unless they can be
attributed to the author. Questions which cannot be considered as purely
formal, are more reliable indicators; for instance elliptic questions intro-
duced by et (Hungarian es) reminding the partner of circumstances which
escaped him:
Et puis, ses affaires, ä l'hötel? ( . . . ) Et le linge, chez la blanchisseuse?
(Roger Martin du Gard, Les Thibault III, La belle saison 2: 181)
or repetitive de dicto questions:
Az aszottkepü bennfentesen mosolygott a bajusza alatt . . . Hogy nehez-e oda
bejutni? Bizony sok fät elfureszelt ö a meltosagos lireknal ( . . . ) a meltösägos
ür is häny barätjät összetelefonälta . . . (My italics)
[The one with the emasticated face smiled knowingly . . . If it is hard to gain
admittance? Well he (= as for him) chopped up a lot of wood for his lordship
( . . . ) his lordship called up quite a few friends of his on the phone.]
(Läszlo Nemeth, Bun p. 34)
 Reported speech in French and Hungarian 293

(12) Giulio Herczeg attributes a great part to para-phrastic criteria,

such as loose organization of verbal and mental processes (1963: 36).
(13) Terms of address are determined by the social relations between
addresser and addressee. In French narrative, ces Messieurs impersona-
tes a waiter addressing customers.
On a fait des oeufs ä ces Messieurs ( . . . )
(Aragon, Les communistes, vol.4: 309)

Terms of address are especially revealing in a language such as

Hungarian, carrying the verbal traces of a feudal system. LAST NAME
+ PROFESSION, for example, is a respectful form of address appro-
priate for a subordinate addressing his superior.
Arra gondolt, megkeri aζ epitesz urat, hogy egy mäsik epületre helyezze.
[He thought he would ask the engineer (lit.: engineer + mister) to transfer
him to another building]
(Läszlö Nemeth, Bün p. 10).

Addresses as markers of free indirect speech were mentioned both by

Bally (1914: 420) and Marguerite Lips (1926: 68).
(14) Leo Spitzer devoted a paper to the use of illogical reasoning as
indication of free indirect speech in Louis Philippe's novel 'Bubu de
Montparnasse' (1928 (a): 166-207). The delusive ideas of the heroine of
Jänos Arany's (1817-1882) ballad repeatedly figure as quotation marks.
The Good-wife Agnes who assisted her lover in killing her husband sinks
gradually deeper and deeper into dementia. 47
Szöghajät is megsimitja
Nehogy azt higyek: megbomlott.
[She smoothes her fair hair
so they will not believe she has gone mad]
Hallja a hangot, erti a szot,
S mig azt erti: "meg nem oriil."
[She hears the sounds, she understands the words,
And whilst she understands them: "she can't be mad."]
(Agnes asszony - The Good-wife Agnes)

Several scholars consider the way of thinking a major cue for identify-
ing free indirect speech (Lips 1926: 66, 105, 155; Lerch 1930: 170;
Bakhtin 1968: 128; Wagner 1972: 180). The importance of markers at the
content level was highlighted recently by Brian McHale (1978: 272-273).
The argument concerning discursive markers of free indirect report,
based on the transfer of speech characteristics to narration, needs some
qualification. All the discursive-level markers above from 1 to 12 are
294 Ivan Fonagy

based on the opposition between live-speech and narrative. The narra-

tive style is, however, an unsteady reference-level. The increasing liveli-
ness of the narration gradually reduces the distance between direct
speech and narration: thus the discursive markers based on feature-
transfer become less and less effective.
In Kaiman Mikszäth's novels interjections (Kutyateremtette 'by
George!' lit.: 'dog's creation' Gavallerok p. 13) and emphatic modals
{De szep 'How beautiful!' Gavallerok p. 43) are clearly attributable to
the author. He often makes use of the imaginary deictic field which he
shares with the reader.
Ahoi jön ni a csalitos aljän a ket szürken Bogozy Miklos bäcsi.
[Look, there comes at the lower part of the brush-wood on two grey horses
uncle Μ. B.]

He often lends to a description the form of a dialogue ('De hol van

Endre? Nini hiszen epen Ott all mär elötte [But where is Endre? Look,
there he stands, just in front of her]) Resettings, ellipses, interrupted
utterances are quite as common in the narratives of some other Hunga-
rian writers at the turn of the century.

De hat megmondhatnä ezt elöszöval - ki csak no? igen, - Gubos Menyhärt, a

szabo.
[This could be stated viva voce by - by whom then? why! by M. G. the tailor.]
(Tömörkeny, Hat szäl gyertya p. 31)

There is no clear distinction between narration and dialogues in

Celine's novel 'Voyage au bout de la nuit', neither at the lexical, nor at
the syntactic level. Celine always follows the rules of Popular French.
Direct quotations are framed by qu'il dit instead of dit-il; there is no trace
of standard forms of negation: ne is omitted regularly. The ambiguity of
free indirect quotations in La Fontaine's Fables (see: Krantz 1970) is
partly due to the lively style of his narration. It has to be kept in mind
that the loosening of syntactic structures, the oratio soluta is also a
rhetorical figure deliberately used by writers and poets. The potential
discursive markers of free indirect style must be evaluated in view of an
unstable reference-level, varying in function of literary genres, literary
trends and individual style.48
Lively, dramatic narrative style does not coincide with free indirect
self-quotations of the author.
Dühösen 6s gesztikulälva magyaräztam, hogy milyen nagy es ritka tünemeny-
röl van szo, es hogy mindjärt vege lesz, hiszen csak arröl van sz0, hogy
ätjöjjenek az utcai szobäba ( . . . )
 Reported speech in French and Hungarian 295

[Shouting and gesticulating I told them what an important and exceptional

event is was [= a solar eclipse], and that it would be over soon, why, they only
had to take the trouble to come to the front-room ( . . . ) ]
(Karinthy, Tanär ύΓ k6rem p. 78)

Complexity, ambiguity, plurivocity

Bally declared that the complexity of the process of identification was

not to be confused with indecision (1914: 459). Cases of indecision are,
however, not infrequent in the interpretation of literary texts (see: Lips
1926: 47, 58, 102; Landry 1953: 10, 29-30; Spitzer 1928: 201; Herczeg
1983). Some passages of Flaubert's novel 'Madame Bovary' which were
presented as examples of free indirect reporting by Stephen Ullmann
(1964: 96) were considered as the author's statement by Beyerle (1972:
358, 363). Günther (1928) refuses to accept as instances of free indirect
quotation the examples quoted by Marguerite Lips from Old French
texts. According to Beyerle, the ambiguity becomes increasingly fre-
quent as the intellectual standing of the figure comes closer to that of the
author (1972: 356).
The degree of ambiguity inherent in free indirect reports, in a context
of η lines, could be determined by means of semantic tests. In such a
pilot-test 30 samples of 8-10 lines taken from Flaubert's novel 'Educa-
tion sentimentale' were presented to 10 post-graduate students in lingu-
istics, who were asked to mark the words which might be attributed to
one of the novel's characters. (All of the samples were interpreted by at
least one author as instances of free indirect report.) 10 samples out of 30
were attributed by the majority of the subjects (from 8 to 10) partly or
wholly to the character. In 11 cases a thin majority (6-7 students)
suspected a hidden speaker; whilst the remaining 9 samples were attribu-
ted by a majority to the author. According to a χ 2 test the probability of a
random distribution of the answers was ρ < 0.001.
The term 'ambiguity' may cover two distinct concepts. The first is
based on an exclusive disjunction: It permits a choice between two
interpretations, but excludes the simultaneous validity of both of them.
The second is based on an inclusive disjunction, which allows for the
possibility of the simultaneous presence of the two voices: the words of
the character interfering with the author's reflection. The two messages
may, in fact, be interwoven. This seems to be the case in a number of
passages of Karinthy's satirical masterpiece, 'Tanär ur kerem [Please,
teacher]'. He succeeds in presenting in a highly condensed form, with
296 Ivan Fonagy

ironic precision the multiple facets of school-life. In the chapter on the

'bad pupil', the latter remains standing by the blackbord petrified,
following a disgraceful performance.
He is standing alone in the crowd as if on an island. But still he does not return
to his desk. No one told him that he could go - he stands, his heart hollow, and
depraved, made an outcast - no one said anything, no one. He is still in the
middle of his viva. Should he walk back again between the rows of the desks?
No, he would rather stand stupidly: his hands haltering, he fidgets on the
blackboard amidst the ruins of the equation, as a pilot, who has crash-landed,
and wanders aimlessly around the broken pistons of this engine.
(Tanär ur kerem p. 38)

The text in Roman type is simply narrative. The sequences in italics can
be attributed on the basis of syntactic and lexical criteria to the 'bad
pupil'. 49 The sequences in spaced-out type are less easy to interpret. He
would rather stand stupidly is seen simultaneously from the inside and
the outside. The metaphors 'odvas es züllött, kitaszitott sziwel' rendered
by 'his heart hollow and depraved, made an outcast' reflect precisely the
pupil's feeling. Such poetic precision, however, would be far beyond his
verbal skills. The attributes züllött 'depraved' and kitaszitott 'outcast'
may figure in ready made utterances more likely to come into his mind
( Ί am an outcast of society'). His feelings and verbal associations are
'received' by the author, and retransmitted in a condensed literary form.
The symbiosis of the author's and his character's verbal creation is
even more apparent and more complex in a preceding sequence reflec-
ting the state of mind of the 'bad pupil' during his inadequate perfor-
mance.
Irja, irja, mint Agnes asszony, tudja, miröl van szo, lätja a tetelt, 'eppen ugy
mint akkor ejjel', mikor elaludt mellette es fogalma se volt rola, mit jelent az
egesz.
[He writes, and writes it again, as the good-wife Agnes did, he knows what is
being referred to, he saw the formula 'just as it was that night', when he fell
asleep over it, and he had no idea what all this was about.]
(Karinthy o.e. p. 36)
Goodwife Agnes is the murderous heroine of Arany's above quoted
ballad that even 'bad pupils' are supposed to know by heart, irja, irja
recalls the verse vires leplet mossa, mossa 'she washes, and washes again
her sheet covered with blood'. 'Just as it was that night' is another hidden
quotation from the ballad, simultaneously alluding to the tragic night
when the murder took place, and the night of the total failure of the 'bad
pupil' predicting and preparing his fall, the capital sentence he is
awaiting.
 Reported speech in French and Hungarian 297

Notes

1. 'En parlant, Albertine gardait la tete immobile, les narines serrdes, ne faisait remuer
que le bout des levres. II en rdsultait un son trainard et nasal dans la composition
duquel entraient peut-etre des hdrdditds provenciales, une affectation juv6nile de
flegme britanique, les legons d'une institutrice etrangfere et une hypertrophic conge-
stive de la muqueuse du nez' (A l'ombre des jeunes filles en fleurs. A la recherche du
temps perdu, ed. Pldiade: Paris 1954, vol. 1: 877).
2. Speakers of a symposium, confronted with the dactylographic version of their impro-
vised oral delivery, often hesitate before recognizing in the type-script their original
wording. The more precise the transcription of the tape, the stranger they find it. The
sharp contrast between the recorded speech and the wording of the written version,
edited by the author shows the gap separating the unpolished event and its literary
transposition.
3. A letter is read, silently, put between quotation-marks, in Balzac's 'Ursule Mirouet'
(p. 64), and in Läszlö Ndmeth's '£getö Eszter' (p. 55). Jerphanion, in 'Les amours
enfantines' by Jules Romains, reads aloud passages from an old newspaper article he
just came across. The quoted sentences are put into quotation-marks (p. 36), as well as
the quoted verses of Baudelaire (p. 51).
4. Läszlö Nemeth, systematically puts into quotation marks sentences remembered; and
introduces by hyphens sentences spoken in the present (£get<5 Eszter p. 114).
5. Läszlö Nemeth uses another type of quotation-mark for citations of the second degree:
[ . . . ] as opposed to [ " . . . " ] for non-embedded quotations.
6. Speech and inner-speech, are conceived as a polyphonic composition of point and
counter-point, both presented simultaneously, side-by-side, in the novel 'Heinrich von
Walheim' (Frankfurt-Leipzig 1785, vol. 1: 301-316) by the Austrian writer Johann
Friedel (quoted by Leslie Bodi 1977: 216-222).
7. ( . . . ) il eteint le megot et le range soigneusement dans une boite de Valdas, une
habitude de l'occupation. Puis quelqu'un lui demande vous n'auriez pas un lacet de
Soulier par hasard je viens de p6ter le mien. Gridoux Ιένε les yeux et il l'aurait parie.
C'est le type et qui continue de la sorte:
"Y a rien de plus agagant, pas vrai?
- Je ne sais pas, rdpond Gridoux.
(p. 72-73)
8. One of the first announcers of the French Radio incurred reproaches for reporting a
subversive statement without oral quotation marks (see P. Duprd, Encyclop6die du
bon f r a ^ a i s , Paris 1972: 1174, cit. Anne Bergheaux 1984).
9. The results of the experiments are to be published in vol. 5 of the Travaux de l'lnstitut
de Linguistique et Phondtique Generates et Appliqu6es.
10. Mr. H. G. ironically quotes the apologies of the representative of the Hungarian Home
Office by means of repeated final rise-falls, varying between a sixth and an octave:
'le bonhomme du service d'etrangers qui . . . qui . . . enfin elle lui demandait rien, euh,
elle / lui dit/proprio motu / r* Ah votre ^v pere ah dvidem- /-v ment c'est pas
d'chance / euh hmm vot' p6re c'est pas d'chance / ( . . . ) '
^-v symbolizes a rise-fall taking place in the subsequent syllable
/ marks pauses.
11. Intransitive verbs may introduce a quotation only if they can be paraphrased by a
298 Ivan Fönagy

verbal phrase where the nominal derivative of the verb figures as the object, e.g.
witzeln 'to joke' —> einen Witz machen, mentir —> faire un mensonge (op. cit. 34).
12. In the examples given by the author quotations appear, however, either as objects and
subjects, or as attributive appositions. The grammatical role of the quotation does not
seem to depend on its 'individual lexical meaning'. It is hard to see, in any way, how the
lexicon could contain and specify an infinite number of utterances as possible candi-
dates of reported speech.
13. The semantic conditions of 'potential transitivity' have not yet been fully elucidated
(Hadrovits 1969: 63).
14. I am less confident in the strength of other formal criteria advocated by Adrienne
Dömötör (1983) in defence of the proposed adverbial, attributive and predicative
quotations. She considers, for instance, the reported sentence adverbial, if an adverbial
pronoun of the main sentence 'shadows' the quotation.
(...), igy kiältott fei vidäman, hangosan: (My italics)
- Itt vagyok!
[, she cried out joyfully, loudly, like this:
- Here I am!]
(Thurzö, Belväros es videke p. 24, Dömötör 1983: 475)
The presence of an adverbial or attributive shifter in the reporting clause does not
warrant the adverbial or attributive character of the reported utterance. According to
the understated rationale of the author, reported sentences could be integrated into the
reporting clause as object, subject, adverb, attribute or predicate; in the same way as
subordinate clauses do in the frame-work of a type of classical grammar (see: ed.
Tompa 1961-1962, vol.2: 330ff., Wagner and Pinchon 1962: 545ff., Sanfeld 1977:
7 ff.). Such a contention is not supported in the case of reported sentences supposed to
be attributive, adverbial or predicative.
15. The reporting clause is a device of naming by demonstration, "as we may identify a
referent by using the deictic word this' as pointed out by Leech (1974: 352).
16. Otto Behaghel reports analogous figures from Germanic languages (1928 vol. 3: 696),
and E. Kieckers quotes Greek and Latin instances of splitting (Spaltung) (1912:
154 ff.).
17. Dieter Wunderlich rightly assigns to such verbal expressions the feature + COMMU-
NICATIVE (1969: 99). We shall be concerned here with the second, and secondary,
semantic field.
18. Our analysis is based on Hungarian novels, short-stories and humoresques listed on
p. 303 f.
19. The question - why s'emporter may figure as a verb of saying but not s'apaiser - raised
by Annette Sabban (1978: 30) could be answered perhaps in terms of the more
extrovert character of s'emporter.
20. Some other French writers make a much more liberal use of deictics. Thus, Boris Vian
in his novel 'L'ecume du jour' admits some 50 verbs in mgMif-position; some of them
are rarely used by other authors, such as prevint, s'acquit, acquiesga. Queneau goes
even further creating reporting verbs of his own: - Ah! oui vuvurre Zazie (Zazie dans le
Metro p. 44), - Tapage nocturne, surhurlirent ä ce moment les nouveaux flics compe-
tes, eux, par un panier ä salade (Zazie p. 166).
21. Gougenheim analyzed in detail the use of conversational verbs in the novels of French
authors of the seventeenth century (1938: 313-319).
22. ricana (Cocteau, Grand ecart p. 107; Paul Morand, Magie noire p. 174); pleurnicher
 Reported speech in French and Hungarian 299

(Arland Marcel, L'ordre p. 482, Triolet, Premier accroc p. 375), sangloter (Paul
Morand, Magie noire p. 227).
23. lancer (in the French version of Tibor Dery's novel Ά kiközösitö' L'excommunicateur
p. 13, 52; 'jetta Brahms' a commentator of France Musique 5th June, 1983, jeta
(Colette, Mitsou p. 10; Queneau, Zazie dans le M6tro p. 31; Fichtre! lächa simplement
D. (Zola, L'argent p. 358).
24. The informants had to fill in a questionaire containing in random order 49 verbs,
presented as speech deictics in weak position, e.g. ( . . . ) , lut-il; approuva-t-il etc. They
had to assign them to one of the following three evaluative categories: (1) normal, (2)
acceptable, (3) unacceptable.
25. According to Annette Sabban (1978: 32 ff.) only resultative transitive verbs are
admitted in direct quotation. This contention is not supported either by Hungarian or
by French verbs admitted as verbs of saying in weak position. The verbs bredouiller,
zezayer, crier, clamer, observer, taquiner etc. may allow but do not imply a definite
object.
26. The same subjects who tested the French verbs of reporting in direct quotations (see
p. 270 f.) were asked to consider the acceptability of 50 verbs as matrix-verbs governing
indirect quotation (a) either subordinate clauses, or (b) infinitival clauses (de/ά + Infini-
tive). The list contained: (1) assertives, (2) conversational verbs, (3) directives, (4) me-
diate-directives, (5) verbs of protest, (6) other emotive verbs, (7) verbs characterizing
the manner of saying, (8) expressing non-verbal human soundmaking, (9) denoting fa-
cial mimetics or (10) gestures. The subjects had a triple choice, considering the verbs in
their capacity of matrix-verbs as (a) normal, (b) acceptable, (c) inacceptable.
27. The corresponding Hungarian verb veszekedett 'he quarreled' is freely admitted in the
role of matrix verb of indirect report: veszekedett, hogy . . . 'he quarreled that . . . ' .
28. The forward drift of primary and secondary verbs of saying in indirect quotation has
been clearly formulated by Läszlö Hadrovits (1969): 'The verbs introducing indirect
quotation (mond 'he says', beszel 'he speaks', mesel 'he relates') ( . . . ) suppose that the
source of their emotive kernel should be unrevealed. The continuation of the report is,
thus, unavoidable' (o.e. 202).
29. 'Je ne pouvais concevoir qu'on püt sortir impunöment des bras de la padoana'
(Rousseau, Confessions p. 295) - 'Elle inventa qu'elle avait horreur du mariage'
(Mauriac, Le baiser du 16preux p. 34) - Ί1 £tait tout au souci de garder sa vitesse,
sachant bien que la vraie qualit6 d'un mecanicien ( . . . ) consistait ä marcher d'une fa$on
rdguliere ( . . . ) . ' (Zola, La bete humaine p. 206).
30. Thus, approuver was accepted by 19 out of 20 subjects as an inquit-vcrb in direct
quotation, and by 15 subjects as a matrix-verb in indirect reporting.
31. '-Quelle bassesse ä moi! s'etait-elle έ α ϊ έ ε (Stendhal, La chartreuse de Parme vol.2:
126). 'Mais le marchand s'dcria qu'elle avait tort;' (Flaubert, Madame Bovary,
Oeuvres ed. Pleiade vol. 1: 471). - '-Mais . . . balbutia Dussardier.' (Flaubert, Educa-
tion sentimentale, Oeuvres vol. 2: 62). Ί1 balbutia qu'il avait froid' (Mauriac, Le baiser
du lepreux p. 69). - In a similar way the proto-directive verb menacer was accepted in
direct quotations by 17 subject and in indirect reports by 19.
32. For a more detailed analysis see: Fönagy and Magdics 1967: 280-305. I attempted to
verify the hypothesis of complex or stratified melodic patterns by means of mimicking
experiments and semantic tests (Fönagy 1979).
33. The above quoted sample can be compared with other examples taken from recorded
conversations (Fönagy 1983).
300 Ivan Fönagy

34. Sigmund Freud, in his New introductory lectures on psycho-analysis [1933] speaks to
an imaginary public. He formulates and answers potential questions, accepting or
refuting fictitious critical comments. (See Gesammelte Werke (GW) vol. 15, Standard
edition (SE) vol.22.)
35. Frigyes Karinthy's juvenile hero has to offer his condolences to his uncle's widow. At
her question 'Oh, Franci, did you come and see me, didn't you?' he is seized by an
irresistible urge to laugh. His disrespectful and highly embarrassing associations
assume the figure of a marquis:
'And while aunt Stanci leads him through the flat, there is someone in Franci's brain
talking loudly and disrespectfully with his hat pushed to the back of his head. Bowing
deeply he raises his hat with the grace and elegance of a marquis ( . . . ) he says:
Madame, dear aunt Stanci, I have not yet arrived, I am almost here, I am coming up
the street right rushing to get here ( . . . ) ' .
(Tanär ür kerem p. 136-137)
36. Proust refers to this type of defensive oral quotation marks:
' ( . . . ) et je remarquai, comme cela m'avait souvent frappd dans ses conversations avec
les soeurs de ma grand'mere, que quand il ( = Swann) parlait de choses serieuses,
quand il employ ait une expression qui semblait impliquer une opinion sur un sujet
important, il avait soin de l'isoler dans une intention spöciale, machinale et ironique,
comme s'il l'avait mise entre guillemets, semblant ne pas vouloir le prendre ä son
compte.'
(Du cötd de chez Swann, A la recherche du temps perdu vol. 1: 98)
37. Julia Kristeva opposes to citation anonymous references labeled prelivements (samp-
les), and designates the three basic sources of such samples: 'Pris des textes mythiques
( . . . ) , scientifiques ( . . . ) ou politiques (Marx, Lenine, Mao Tse-Toung), les preleve-
ments laissent voir l'engendrement ä travers cette triple orientation qui ramdne sur la
page les trois lieux determinant notre culture' (1969: 332).
38. And then is heard no more; it is a tale
Told by an idiot, full of sound and fury,
Signifying nothing.
(Shakespeare, Macbeth V, v)

Es e besz6det nem veszi [And won't he consider this

Egy örült remülesteli, As a tale full with horror
Zavart 6sz mes6jenek? Born out by the troubled mind of a madman?]
(Petöfi, Szörnyü idö - Horrible time)
39. Thus, the words of the Psalm 124 torrentem pertransivit anima nostra 'When the waters
had overwhelmed us', forsitan pertransisset anima nostra aquam intolerabilem "Then
the proudwaters had gone over our soul' are, irrespectfully, applied to the scene where
Gargantua in eating up some huge leaves of salad, inadvertently consumed some
pilgrims, as well; brought up again by a huge vomitus they had to wade in the lake of
his urine, 'quand nous passasmes la grande Boire ( . . . ) de son urine dont il nous tailla le
chemin (Rabelais, Oeuvres vol. 1 chap. 30: 129-130).
40. In Frigyes Karinthy's sketch, the journalist, very much impressed by a recent reading -
it was actually the Holy Bible - he starts ruminating on his daily professional and
individual problems adopting biblical parallelism, and comes to the conclusion, invaria-
bly formulated in parallel verses, that this literary form would be most convenient for a
journalist who would double in this way his royalties (igy irtok ti p. 279-281).
 Reported speech in French and Hungarian 301

41. 'Berichtende Form' (Behaghel 1899, 1928), represented speech (Jespersen 1924),
anführende Rede (Spitzer 1928a), reported speech (Ullmann 1957), narrated monolo-
gue (Cohn 1966), style indirect partiel (Antoine 1904), style indirect libre (Bally 1912,
1914), indirect libre double (Thibaudet 1922), stellvertretende Rede (Läftman 1929),
pseudo-objective Rede (Spitzer 1928a), subjektiv indirekte Rede (Heinermann 1931),
verkleidete Rede (Kalepky 1899, 1913), (mittelbar) erlebte Rede (Lorck 1921), Rede
als Tatsache (Lerch 1930), semi-direct (Legrand 1922), nachgeahmte Rede, nachge-
äffte Rede (Spitzer 1928 a, Lerch 1930) uneigentlich direkte Rede (Gertraud Lerch
1922), discours direct impropre (Kalik-Teljatnicova 1965-1966).
42. Bally attributes the irregular use of the imperfect in narration to an underlying verb of
saying or thinking: "L'helice se mit en branle. On partait (Alphonse Daudet).
According to Bally on partait "equivaut ä peu-pres ä: 'Evidemment on partrait', il
fallait croire qu'on partait"" (1912: 601-602). Eugen Lerch proposed a different
interpretation: he considered such imperfects as an expression of lively imagination
(1930: 139-234).
43. Gabriel soupira. Encore faire appel ä la violence
( . . . ) Mais enfin il fallaif ce qu'il fallaii. (My italics)
(Queneau, Zazie dans le Metro p. 8)
The underlying speech act is suggested by means of a colloquial utterance in the
Hungarian translation:
Gabriel felsöhajtott. Most megint jön az eröszak.
( . . . ) Dehat jöjjön, aminek jönnie kell ['come what may'] (p. 20).
44. Queneau systematically substitutes ss for ks, bs in indirect speech acts, reproducing the
more or less vulgar phonetic variant of expliquer [esplike], s'exclamer [sesklame],
obstiner [ostine] etc. (products of a phonetic change in progress sk—>s, bs—>s).
45. The 2nd person singular may result from a double shift: substitution of the 3rd person
for the 1st, just as in the cases referred to above, followed by a substitution of the 2nd
person for the 3rd.:
( . . . ) il etait grand temps de remettre sur le metier l'ouvrage un peu negligd, d'eviter au
fils qui etait toute voire raison d'etre de tomber dans les rets d'une aventuriere. (My
italics)
(Andre Wurmser, L'interdiction de sejour p. 120)
46. There are other forms of pseudo-direct discourse, unrelated to free indirect quotation.
Affidavits submitted to the court in the first person, but composed in the language of
the addressee, the Court itself (see: Meir Sternberg 1982: 85-89). - In the alleged
confessions of the victims of faked political trials, transgressions of the pragmatic rules
of direct discourse clearly show that the words spoken by the accused have been
originally written in the third person: they are, in fact, incompatible with the first
person.
G. Bannantine: I declare that ( . . . ) while we compelled the Hungarian refiners by legal
process to pay their debts, in a significant manner, we did not use the same legal
process press the Anglo-Saxons interest to pay theirs. (My italics)
(Communiqud of the Hungarian Ministery of the Interior relative to the sabotage-
affaire of the M A O R T (= Hungarian Oil-Company). Budapest: Szikra 1948, Annex)
The sentence adverb significantly (or: in a significant manner) draws attention to a
relationship which might have passed unnoticed. In the first person singular, it implies
that the speaker calls his own attention to such a phenomenon; he does it moreover in a
very sarcastic manner.
302 Ivan Fönagy

Al. According to Giulio Herczeg, delusional ideas of persecution systematically allude to

the character's verbalized thoughts in Verga's 'Gesualdo' (1963: 45).
48. The possible overlapping of syntactic and semantic features of narration in live speech
was pointed out by Strauch (1974) and also mentioned by Jenny Simonin (1984).
49. This is still more evident in the Hungarian original than in this rough English
translation. For instance Most vigigmenjen, ujra, a padsorok között 'should he walk
back again between the rows of desks' betrays its colloquial character by its broken
rhythm marked by the commas; by the emphatic stress on vέgigmenjen (imperative
form) 'should walk back' - in a simple narrative style the prefix would follow the verb
(menjen vigig); the imperative form of the verb recalling the colloquial pattern (Hat)
most vigigmenjek a padsorok között with a strong stress on vigigmenjek /"ve:
gigmenjek/ 'Shall I walk back again between the rows of desks!'. Senki se mondta
nekem, hogy 'No one told me that' is a colloquial situation-specific sentence.

Text list

A. French texts
Aragon, Louis. Les beaux quartiers. Paris: Denoel s.d.
Aragon, Louis. Aur61ien. Paris: Gallimard 1958.
Aragon, Louis. Les communistes. Mai 1940. Paris: La Bibliotheque Frangaise 1949.
Aragon, Louis. Les communistes, vol.4. Paris: La Bibliothdque Fran?aise 1949.
Arland, Marcel. L'ordre. Paris: NRF 1929.
Balzac, ΗοηοΓέ de. Ursule Mirouet. In: Romans, vol. 3. Paris: NRF, Editions de la
Pldiade.
Bodel, Jean. Le jeu de Saint-Nicolas, Alfred Jeanroy ed. Paris: Champion.
Camus, Albert. La peste. Paris: Gallimard 1947.
Carco, Francis. Nostalgie de Paris: Geneve: Editions du Milieu du Monde 1941.
Carco, Francis. Printemps d'Espagne. Paris: Albin Michel 1929.
C61ine, Louis-Ferdinand. Voyage au bout de la nuit: Denoel 1932.
Chanson de Roland, La Joseph B6dier ed. (on the basis of the Oxford manuscript).
Paris: L'Edition d'Art H. Piazza, s.d.
Chretien de Troyes. Cligds, Alexandre Micha ed. Paris: Champion 1957.
Cocteau, Jean. Les enfants terribles. Paris: Fayard 1931.
Cocteau, Jean. Le grand 6cart. Paris: Fayard 1954.
Colette. Mitsou. Paris: Fayard s.d. 1919.
Daudet, Alphonse. Sapho. Paris: Charpentier 1884.
Diderot, Denis. Jacques le Fataliste et son maitre. Paris. Jacoub 1929.
Fabliaux. Anatole de Montaiglon and Gaston Raynaud eds. Recueil g6n£ral et complet des
fabliaux - des XII - XIV sifccles [1872-1890], vol. 1-2. R66dition. Geneve: Slatkine
1973.
Flaubert, Gustave. Madame Bovary. Oeuvres vol. 1. Paris: NRF Bibliotheque de la
P16iade 1952.
Flaubert, Gustave. L'education sentimentale. Oeuvres vol.2 Paris: NRF Bibliotheque de
la P16iade 1952.
Flaubert, Gustave. Bouvard et P6cuchet. Oeuvres vol.2 Paris: NRF Bibliotheque de la
Pleiade 1952.
 Reported speech in French and Hungarian 303

France, Anatole. Thais. Paris: Clamann L6vy 1897.

France, Anatole. L'ile des Pinguois. Paris: Calmann L6vy s.d. 1907.
Gide, Andr6. Symphonie pastorale. Paris: Gallimard 1925.
Goncourt, Edmond de. Chdrie. Paris: Charpentier 1884.
Laclos, Choderlos de. Les liaisons dangereuses. Paris: Mercure de France 1919.
Lafayette, Madame de. La princesse de Cleves. Paris: Garnier-Flammarion 1966.
Larbaud, Valery. Fermina Marquez. Paris: Gallimard 1926.
Malraux, Andr6. La condition humaine. Paris: Gallimard 1958.
Marceau, Felicien. Bergdre 16g£re. Paris: Gallimard 1965.
Marie de France, Lais. In: Albert Pauphilet ed. Poetes et romanciers du Moyen-Age.
Paris: NRF Editions de la Pleiade 1939.
Martin Du Gard, Roger. Les Thibault. Paris: NRF 1922-1940.
Mauriac, Fransois. Le baiser du lepreux. Paris: Fayard 1926.
Mauriac, Frangois. Le mystfcre Frontenac. Paris: Grasset 1933.
Maupassant, Guy. Notre coeur. Paris: Louis Conard 1909.
Miomandre, Francis de. L'aventure de Beauchamps. Paris: Fayard s.d.
Miomandre, Francis. Les egarements de Blandine. Paris: Le livre moderne illustrd 1936.
Montherlant, Henry de. Les jeunes filles. Paris: Grasset 1936.
Morand, Paul. Magie noire. Paris: Grasset 1928.
Pdlelinage de Charlemagne, Karl Voretsch ed. Halle: Niemeyer 1932.
Proust, Marcel. A la recherche tu temps perdu vol. 1-2. Paris: NRF Editions de la P16iade
1954.
Queneau, Raymond. Zazie dans le metro. Paris: Gallimard 1959.
Rabelais, Fran$ois. Oeuvres vol. 1-2. Londres-Paris: Bastien 1783.
Rolland, Romain. Jean Christophe vol.2. Paris: Albin Michel 1925.
Rolland, Romain. L'äme enchantöe 1, Annette et Sylvie. Paris: Albin Michel 1933.
Romains, Jules. Les hommes de bonne volonte vol. 3, Les amours enfantines. Paris:
Flammarion 1932.
Rousseau, Jean-Jacques. Les confessions. Paris: Firmin Didier 1890.
Rousseau, Jean-Jacques. Emile. In: Oeuvres, Frankfurt am Main: Bechthold 1855.
Sartre, Jean-Paul. Les chemins de la libert6 vol. 1. L'äge de la raison. Paris: Gallimard
1958.
Sartre, Jean-Paul. Les mots. Paris: Gallimard 1975.
Stendhal. La chartreuse de Parme vol. 1-2. Vienne: Mainz s.d. 1839.
Stendhal. Le rouge et le noir. In: Romans et nouvelle. Paris: Gallimard 1947.
Triolet, Elsa. Premier accroc. Paris: Denoel 1945.
Vian, Boris. L'6cume des jours. Paris: Union G6ndrale d'Editions 1966.
Wurmser, Andre. Interdiction de s6jour. Paris: Les öditeurs frangais reunis 1960-1961.
Zola, Emile. L'argent. Paris: Charpentier 1918.
Zola, Emile. Assomoir. Paris: Charpentier s.d. 1877.
Zola, Emile. La bete humaine. Paris: Charpentier 1890.
Zola, Emile. La debacle. Paris: Charpentier s.d. 1892.
Zola, Emile. Nana vol. 1-2. Paris: Charpentier 1924.

B. Hungarian texts
Csokonai Vitez Mihaly. Dorottya. In: Ferenc Schedel ed. Minden munkäi [Collected
works], Pest: Hartleben 1844.
Deri, Tibor. Α kiközösitö. Budapest: Szepirodalmi kiadö 1966.
304 Ivan Fönagy

Heltai, Jenö. Ismeretlen ismerösök. Budapest: U j idök 1964.

Kaffka, Margit. Szinek 6s 6vek. Budapest: Franklin s.d. [1914],
Kaffka, Margit. Välogatott müvei [Selected works], Budapest: Sz6pirodalmi kiadö 1974.
Karinthy, Frigyes. Krisztus es Barabäs. Budapest: Dick Manö 1918.
Karinthy, Frigyes. Tanär ür kerem. Budapest: Atheneum 1916.
Karinthy, Frigyes. U j görbe tükör. Budapest: Atheneum s.d.
Keller, Dezsö. Pest az Pest. Budapest: Sz6pirodalmi kiadö 1967.
Kosztolänyi, Dezsö. Elbeszdlesei [Short stories]. Budapest: Magyar Helikon 1965.
Krudy, Gyula. Puder. Budapest: Singer es Wolfner 1914.
Knidy, Gyula. Szinbäd. Α feltämadäs. Budapest: Singer 6s Wolfner 1916.
Krüdy, Gyula. Szinbäd megtdrese. Budapest: Atheneum s.d. 1925
Mikszäth, Kälmän. Gavallerok. Budapest: Revai 1904.
Mikszäth, Kälmän. A szelistyei asszonyok. Budapest: Legrady 1901.
Mikszäth, Kälmän. Szent P6ter esernyöje. Budapest: Revai 1895.
Molnär, Ferenc. Az aruvimi erdö titka. Budapest: L6grady s.d. 1917.
Möricz, Zsigmond. Erdily vol. 1-3 1935. Budapest: Szäpirodalmi kiadö 1960.
Möricz, Zsigmond. Legy jö mindhalälig. Budapest: Möra Ferenc 1960.
Nemeth, Läszlö. Bun. Budapest: Szipirodalmi kiadö 1966.
Nemeth, Läszlö. Egetö Eszter. Budapest. Magvetö 1956.
Nemeth, Läszlö. Irgalom. vol. 1-2. Budapest: Magvetö 1965.
Szerb, Antal. Utas es holdviläg. Budapest: Magvetö 1964.
Szomory, Dezsö. Harry Russel-Dorsan a francia hadszinterröl. Budapest: Pallas 1918.
Tamäsi, Äron. Abel. Budapest: Szepirodalmi kiadö 1963.
Tömörkeny, Istvän. Hat szäl gyertya. In: Föltetszett a hajnal. Szeged: Tiszatäji Magvetö
1955.

References

Antoine, F. 1904. 'Le style indirect partiel.' Musee Beige 8: 177-193.

Auerbach, Erich 1967. Mimesis. Dargestellte Wirklichkeit in der abendländischen Litera-
tur. Bern, München: Francke.
Authier, Jacqueline. 1978. 'Les formes du discours rapporte. Remarques syntaxiques et
simantiques ä partir des traitements proposes.' Documentation et Recherche en Lin-
guistique Allemande Vincennes (DRLAV) 17, 1-87.
Authier, Jacqueline and Meunier, Andre. 1977. 'Exercices de grammaire et de discours
rapportd.' Langue Frangaise 33, 41-67.
Bakhtine, Mikhail. 1968. 'L'enoncd dans le roman.' Langages 12, 126-132.
Bakhtine, Mikhail. 1978. Esthitique et thiorie du roman. Paris: Gallimard.
Bally, Charles. 1912. 'Le style indirect libre en frangais moderne.' Germanisch-Romani-
sche Monatsschrift 4, 549-556, 597-606.
Bally, Charles. 1914. 'Figures de pensees et formes linguistiques.' Germanisch-Romanische
Monatsschrift 6, 405-422, 456-470.
Bally, Charles. 1941. 'Intonation et syntaxe.' Cahiers de Ferdinand de Saussure 1, 33-42.
Banfield, Ann. 1973. 'Narrative style and the grammar of direct and indirect speech.'
Foundations of Language 10, 1-39.
 Reported speech in French and Hungarian 305

Banfield, Ann. 1979. 'La syntaxe de l'incise narrative et l'attribution de point de vue de
phrase en phrase.' Linguisticae Investigationes 3, 229-243.
Bauche, Henri. 1946. Le langage populaire. Paris: Payot.
Behaghel, Otto. 1877. Über die Enstehung der abhängigen Rede und die Ausbildung der
Zeitfolge im Althochdeutsch. Paderborn: Schöningh.
Behaghel, Otto. 1899. Der Gebrauch der Zeitformen im konjunktivischen Nebensatz des
Deutschen. Paderborn: Schöningh.
Behaghel, Otto. 1928. Deutsche Syntax. Eine geschichtliche Darstellung, vol. 3. Die
Satzgebilde. Heidelberg: Winter.
Bencz£dy, Jözsef, Fäbiän, Päl, and Velcsov, Märtonnd. 1976. A mai magyar nyelv*.
Budapest: Tankönyvkiadö.
Bergheaud, Anne. 1984. 'Dire entre guillemets' (ä paraitre).
Beyerle, Dieter. 1972. 'Ein vernachlässigter Aspekt der erlebten Rede.' Archiv für das
Studium der neueren Sprachen und Literaturen 208, 350-366.
Bodi, Leslie. 1977. Tauwetter in Wien. Zur Prosa der österreichischen Aufklärung
1781-1785. Frankfurt am Main: Fischer.
Calbris, Genevieve. 1982. 'Structure des titres et enseignes.' Le Frangais dans le Monde
166, 26-55.
Catach, Nina. 1968. L'orthographe frangaise ά l'epoque de la Renaissance. Geneve: Droz.
Catach, Nina ed. 1979. La ponctuation. Recherches historiques et actuelles, vol. 1-2.
Paris-Besan$on: C.N.R.S.
Cohn, Dorris. 1966. 'Narrated monologue: definition of a fictional style. Comparative
Literature 18, 97-112.
Compagnon, A. 1979. La seconde main. Paris: Seuil.
Coulmas, Florian. 1979. 'On the sociolinguistic relevance of routine formulae.' Journal of
Pragmatics 3: 239-266.
Cram, Fred D. 1978. 'The syntax of direct quotation.' Cahiers de Lexicologie 33, 41-52.
Damourette, Jacques and Pichon, Adouard. 1911-1930. Des mots ä la pensie. Essai de
grammaire. Paris: Editions d'Artrey, vols. 1-7.
Davidson, Donald. 'On saying that.' Synthese 19, 130-146.
Dojtman, Betty. 1980. 'Desengagement du Je dans le discours indirect.' Poetique 41,
90-107.
Dömötör, Adrienne. 1983. Ά magyar nyelv egyenes idezeteinek 6s ϊάέζδ mondatainak
kapcsolatäröl [Relation between direct quotation and propositional units of reporting in
Hungarian]'. Magyar Nyelvör 107, 472-481.
Driver, S. R. 1914. An introduction to the literature of the Old Testament. Oxford:
University Press.
Ducrot, Ostwald et al. 1980. Les mots du discours. Paris: Editions du Minuit.
Eliot, Thomas S. 1961. On poetry and poets. Three voices of poetry [1953], New York:
Noonday Press.
Fischern, August. 1899. Die indirekte Rede im Altfranzösischen. Doct. thesis, University of
Jena.
Fönagy, Ivan. 1979. 'Artistic vocal communication at the prosodic level.' In: Harry and
Patricia Hollien eds. Current issues in the phonetic sciences. Amsterdam: Benjamins,
245-260.
Fönagy, Ivan. 1980. 'Structure semantique des signes de ponctuation.' Bulletin de la Sociiti
Linguistique de Paris 75: 95-129.
Fönagy, Ivan. 1982 (a). Situation et signification. Amsterdam: Benjamins.
306 Ivan Fönagy

Fönagy, Ivan. 1982 (b). 'He is only joking. Joke, metaphor, and language development.'
In: Ferenc Kiefer ed. Hungarian linguistics. Amsterdam: Benjamins, 33-108.
Fönagy, Ivan. 1983. La vive voix. Essais de psycho-phonetique. Paris: Payot.
Fönagy, Ivan and Judith. 1983. 'L'intonation et l'organisation du discours.' Bulletin de la
Societe Linguistique de Paris 78: 161-209.
Fönagy, Ivan and Magdics, Klära. 1967. A magyar beszid dallama [Hungarian intonation].
Budapest: Akademiai kiadö.
Freud, Sigmund. 1940-1946. Gesammelte Werke. London: Imago, vol. 1-18.
Freud, Sigmund. 1953-1974. The standard edition of the complete psychological works,
vols. 1-24, translated from German under the general editorship of James Strachey.
London: Hogarth Press.
Frei, Henri. 1929. La grammaire des fautes. Paris: Geuthner.
Gaulmyn, M.-M. 1981. 'Citation et manipulation. In: Argumentation. Lyon: Presses
Universitaires de Lyon, 139-151.
Gauvenet, Helene and Moirand, Sophie-Colette. 1974. 'Le discours rapporte.' Le Frangais
dans le Monde 102: 34-40.
Gougenheim, Georges. 1938. 'La presentation du discours direct dans la Princesse de
Cleves et dans Dominique. Le Frangais Moderne 6: 305-320.
Gougenheim, Georges. 1947. 'Du discours solitaire au monologue interieur. Le Frangais
Moderne 15: 242-248.
Gresillon, Almuth. 1984. Polyphonie, proverbe et detournement. Langages 73: 112-125.
Gurk, P. Mc. 1961. Citation marks in early Latin manuscripts. Scriptorium 15: 3-13.
Günther, Werner. 1928. Probleme der Rededarstellung. Untersuchungen zur direkten,
indirekten und "erlebten" Rede. Marburg: Lahn.
Hadrovits, Läszlö. 1969. Α funkcionälis magyar mondattan alapjai [Foundations of a
functional Hungarian syntax]. Budapest: Akadömiai kiadö.
Hall-Partee, Barbara. 1973. "The syntax and semantics of quotation.' In: Stephen R.
Anderson and Paul Kiparsky eds. A Festschrift for Morris Halle. New York, Chicago,
Toronto, Atlanta: Holt Rinehart & Winston.
Heinermann, Theodor. 1931. Die Arten der reproduzierten Rede. Münster: Forschungen
zur Romanischen Philologie.
Herczeg, Giulio. 1963. Lo Stile indiretto libero in italiano. Firenze: Sansoni.
Herczeg, Gyula. 1983. 'Nagy Lajos stilusa.' Magyar Nyelvör 107: 280-295.
Herdin, Nils. 1905. Studien über Bericht und indirekte Rede im modernen Deutsch.
Doctoral Thesis, University of Uppsala.
Hightower, James R. 1959. 'Some characteristics of parallel prose', in: Studia Serica
Bernhard Karlgren. Copenhagen.
Horn, Wilhelm. Sprachkörper und Sprachfunktion. Leipzig, Berlin: Palaestra.
Jespersen, Otto. 1924. The philosophy of grammar. London: Allen & Unwin.
Jones, Charles. 1968. 'Aspects of English indirect strings.' Lingua 20, 250-264.
Jones, Charles. 1968. 'Varieties of speech presentation in Conrad's 'The secret agent'.
Lingua 20: 162-176.
Jakobson, Roman. 1960. Linguistics and poetics. In: Thomas A. Sebeok ed. Style in
language. New York, London: Wiley & Sons, 350-377.
Jakobson, Roman. 1966. 'Grammatical parallelism and its Russian facet.' Language 42,
399-429.
Jakobson, Roman. 1971. Shifters, verbal categories and the Russian verb' [1963]. In:
Selected writings vol.2, The Hague: Mouton,130-147.
 Reported speech in French and Hungarian 307

Kalepky, Theodor. 1899. Mischung indirekter und direkter Rede. Zeitschrift für romani-
sche Philologie 23: 491-513.
Kalepky, Theodor. 1913. 'Zum 'style indirect libre'.' Verschleierte Rede. Germanisch-
romanische Monatsschrift 5, 608-619.
Kalepky, Theodor. 1927. "Verkleidete Rede.' Neophilologus 13, 491-513.
Kalik-Teljatnicova, I. 1965-1966. 'De l'origine du pretendu 'style indirect libre'.' Le
Frangais Moderne 33, 284-294, and 34, 123-136.
Karpf, Fritz. 1928. Die erlebte Rede im älteren Englischen und in volkstümlicher Rede-
weise.' Die Neueren Sprachen 36, 571-581.
Kieckers, Ernst. 1912. 'Die Stellung der Verba des Sagens im Schaltesatz im Griechischen
und in den verwandten Sprachen.' Indogermanische Forschungen 30, 145-185.
Kieckers, Ernst. 1921. 'Der Schaltesatz im späteren Griechisch.' Glotta 11, 179-185.
Kranz, Dieter. 1970. 'Ambivalente Formen des style indirect libre in den Fabeln von
La Fontaine.' Archiv für das Studium der neueren Sprachen und Literaturen 207,
36-42.
Kristeva, Julia. 1969. Semiootiki. Recherches pour une sömananlyse. Paris: Seuil.
Legrand, Emile. 1925. Stylistique frangaise. Livre du maitre. Paris: Gigord.
Lentin, Laurence. 1972. Apprendre ä parier ä l'enfant de moins de six arts. Paris: Edition
ESF.
Lerch, Eugen. 1914. Die stylistische Bedeutung des Imperfekts. Germanisch-Romanische
Monatsschrift 6, 470-489.
Lerch, Eugen. 1930. Hauptprobleme der französischen Sprache, vol. 1-2. Braunschweig:
Westermann.
Lerch, Gertraud. 1922. Die uneigentliche direkte Rede. In: Idealistische Neuphilologie.
Festschrift für Karl Vossler. Heidelberg: Winter.
Libera, Alain de. 1984. De la lecture ä la paraphrase. Langages 73, 17-29.
Lips, Marguerite. 1926. Le style indirect libre. Paris: Payot.
Lorck, Etienne. 1921. Die "Erlebte Rede." Eine sprachliche Untersuchung. Heidelberg:
Winter.
Lowth, Robert. 1779. Isaiah X-XI. Preliminary discourse. London.
McHale, Brian. 1978. Free indirect discourse: a survey of recent accounts. A Journal for
Descriptive Poetics and Theory of Literature (PTL) 3, 249-287.
Martins-Baltar, Michel. 1977. De l'enonce ά l'enonciation. Paris: Didier.
Meyer, Hermann. 1961. Das Zitat in der Erzählungskunst. Zur Geschichte und Poetik des
Europäischen. Stuttgart: Metzlersche Verlagsbuchhandlung.
Miller, George A. and Johnson-Laird, Philip N. 1976. Language and perception. Cam-
bridge, Mass.: The Belknap Press.
Milner, Jean-Claude. 1978. De la syntaxe a /'interpretation. Paris: Seuil.
Molnär, Ilona. 1974. Ά Dicsekszik, hogy . . . mondattipusröl [The sentence pattern He
boasts that . . . ] . ' Magyar Nyelvör 98: 398-404.
Murvai, Olga. 1976. Ά szabad függö besz6d stilushatäsänak kerdesdhez [The stylistic
effects of free indirect discourse].' Magyar Nyelvör 100: 289-296.
Oltean, Stefan. 1981. 'Textual functions of free indirect discourse in the novel 'Mrs
Dalloway' by Virginia Woolf.' Revue Roumaine de Linguistique 26, 533-547.
Ouy, Gilbert. 1979. La ponctuation des premiers humanistes frangais. In: Nina Catach ed.
Ponctuation vol. 2, 56-89.
Paul, Hermann. 1916-1920. Deutsche Grammatik, vols. 1-5. Halle: Niemeyer.
Paul, Hermann. 1920. Prinzipien der Sprachgeschichte [1880].5 Halle: Niemeyer.
308 Ivan Fonagy

Pittenger, Robert Ε., Hockett, Charles F. and Danehey, John J. 1960. The first five
minutes. Ithaca, N.Y.: Martineau.
Plenat, Marc. 1979. 'Sur la grammaire du style indirect libre.' Cahiers de Grammaire
(Universite de Toulouse) 1: 95-140.
Por, Peter. 1984 (in press). Verrätselung. Ein gemeinsames poetisches Verfahren bei
G6rard de Nerval und Theodor Storm. Habilitationsvortrag, Universität Siegen.
Räba, György. 1960. 'Vilägirodalmi hatäsok a fiatal Babits Mihäly költiszetiben [Literary
influences in the early poetic works of Mihäly Babits].' Budapest Akademiai kiadö.
Räcz, Endre. 1973. 'Megjegyzesek Hadrovits Läszlö Ά funkcionälis magyar mondattan
alapjai' cimü müvdröl [Some remarks on 'Foundations of a functional Hungarian
syntax]'. In: Zsigmond Telegdi ed. Ältalänos nyelviszeti tanulmänyok, vol. 9, Budapest:
Akademiai kiadö, 147-163.
Rey-Debove, Alain. 1976. Theorie du signe et du sens, vol. 1-2. Paris: Klincksieck.
Rey-Debove, Jacqueline. 1971. 'Note sur une interprötation autonymique de la litt6rarite.
Le mode du 'comme je dis'.' Littirature 4, 90-95.
Rey-Debove, Jacqueline. 1972. La mitalangue comme systeme de ref6rence au signe.' Le
Frangais Moderne 40, 232-249.
Riffaterre, Michael. 1979. La production du texte. Paris: Seuil.
Sabban, Annette. 1978. 'Verben der Redeeinleitung im Französischen und im Deutschen.'
Zeitschrift für Französisch und Literatur 88: 28-63.
Sanfeld, Karl. 1977. Syntaxe du frangais contemporain, vol. 1. Les propositions subordon-
nees. Geneve: Droz.
Schürr, Friedrich. 1926. Das altfranzösische Epos. Zur Stilgeschichte und inneren Form
der Gotik. München: Hueber.
Simonin-Grumbach, Jenny. 1984. Les plans d'önonciation dans 'Berlin Alexanderplatz' de
Döblin ou de la polyphonie textuelle. Langages 33, 30-56.
Simonyi, Zsigmond. 1881-1883. Α magyar kötöszok, egyuttal az összetett mondat elmilete
[The Hungarian conjunction, with a theory of the complex sentence] vol. 1-2. Budapest.
Akadömiai kiadö.
Sipka, Sändor. 1966. Ά ζ idezes formäi Nemeth Läszlö 'Irgalom' c. regenyeben [The forms
of quotation in the novel 'Irgalom' of Läszlö N6meth].' Magyar Nyelvör 90, 256-268.
Spitzer, Leo. 1926. Stilstudien, vol. 1. Sprachstile. München: Hueber.
Spitzer, Leo. 1928 (a). Stilstudien, vol. 2. Stilsprachen. München: Hueber.
Spitzer, Leo. 1928 (b). 'Zur Entstehung der sog. 'erlebten Rede'.' Germanisch-Romani-
sche Monatsschrift 16, 327-332.
Spitzer, Leo. 1939. Verlebendigende Rede. Vox Romanica 4: 65-85.
Steinitz, Wolfgang. 1934. Der Parallelismus in der finnisch-karelischen Volksdichtung.
Inaugural-Dissertation, Universität Berlin.
Sternberg, Meir. 1982. 'Point of view and the indirections of direct speech.' Language and
Style 15: 67-117.
Strauch, Gerard. 1970. 'Contribution ä l'etude simantique des verbes introducteurs du
discours indirect.' Recherches Anglaises et Amiricaines 3, 226-242.
Strauch, Gerard. 1974. 'De quelques interpolations recentes du style indirect libre.
Recherches anglaises et americaines 7, 40-73.
Szabö, D6nes. 1958. A mai magyar nyelv [Contemporary Hungarian/Budapest: Jegyzetel-
lätö, vol. 1-2.
Szabolcsi, Miklös. 1976. 'Megjegyzes egy magyar stilusjelens6ghez [A comment on a
Hungarian stylistic phenomenon].' Nyelvtudomänyi Közlemenyek 76, 443-452.
 Reported speech in French and Hungarian 309

Tesniere, Lucien 1976. EUments de syntaxe structural [1966], Paris: Klincksieck.

Thibaudet, Albert. 1922. Gustave Flaubert. Paris: Gallimard.
Tobler, Adolf. 1887. 'Vermischte Beiträge zur französischen Grammatik. Neue Reihe 1.'
Zeitschrift für romanische Philologie 11, 433-441.
Tompa, Jozsef ed. 1961-1962. A mai magyar nyelv rendszere [Structure of contemporary
Hungarian] vol. 1-2. Budapest: Akad£miai kiadö.
Ullmann, Stephen. 1964. Style in the French novel [1957].2 Oxford: Blackwell.
Verschoor, Jan Adriaan. 1959. Etude de grammaire historique sur le style direct et les styles
indirects en frangais. These de doctorat, Universite de Paris.
νέζίη, Jean. 1979. Persistance d'usages medievaux dans la ponctuation des premiers
imprimis. In: Nina Catach ed. Ponctuation, vol.2, 90-101.
Vinogradov, Viktor 1926. Problema skaza ν stilistike. Poetika 1. Leningrad:
Vossler, Karl. 1921. Frankreichs Kultur im Spiegel seiner Sprachentwicklung. Heidelberg:
Winter.
Wagner, Burckhardt. 1972. Innenbereich und Äußerung. Flaubertsche Formen indirekter
Darstellung und Grundtypen der Erlebten Rede. München: Fink.
Wagner, Robert-Leon and Pinchon, Jacqueline. 1961. Grammaire du frangais classique et
moderne. Paris: Hachette.
Wunderlich, Dieter. 1969. 'Bemerkungen zu den verba dicendi.' Muttersprache 79,97-107.
Introducing constructed dialogue in Greek and
American conversational and literary narrative*
Deborah Tannen

1. Introduction

The term "reported speech" is a misnomer. Examination of the lines of

dialogue represented in storytelling or conversation, and consideration
of the powers of human memory, indicate that most of those lines were
probably not actually spoken. What is commonly referred to as reported
speech or direct quotation in conversation is constructed dialogue, just
as surely as is the dialogue created by fiction writers and playwrights. A
difference is that in fiction and plays, the characters and actions are also
constructed, whereas in personal narrative, they are based on actual
characters and events. But even this difference is not absolute. Many
works of fiction and drama are also based on real people and events, and
many conversational storytellers - to the consternation of their children
and spouses but the delight of their hearers - embellish and adjust
characters and events.
Many researchers (for example, Labov 1972, Chafe 1982, Ochs 1979,
Tannen 1982, Schiffrin 1981) have observed that narration is more vivid
when speech is presented as first-person dialogue ("direct quotation")
rather than third-person report ("indirect quotation") - and is more
commonly found in conversational narrative (sometimes generally re-
ferred to as spoken discourse) than written expository discourse (but not
of course in written literary discourse, precisely because fiction and
poetry are akin to conversation in workings and effect). But there is
more to it than that. The creation of voices, more than the depiction of
actions, occasions the imagination of alternative and distant worlds that
is the stuff of dreams and art.
Friedrich (1979: 473) suggests that "it is the more poetic levels and
processes of language, however defined, that massively model, con-

* Research for this chapter was done while the author held a Rockefeller Humanities
Fellowship, for which she is grateful.
312 Deborah Tannen

strain, trigger, and otherwise affect the individual imagination." I see

constructing dialogue as one such poetic process. Therefore I have
included it in an extended, ongoing project examining the ways that
poetic linguistic devices are spontaneously used in ordinary conversation
as compared to how they are developed and manipulated in literary
discourse.
In this chapter I argue that constructed dialogue in conversation and
in fiction is a means by which experience surpasses story to become
drama. Moreover, the creation of drama from personal experience and
hearsay is made possible by and simultaneously creates interpersonal
involvement among speaker or writer and audience. After discussing
these theoretical aspects of constructed dialogue, I turn to an analysis of
forms and functions of constructed dialogue in two genres of two
languages: stories told in conversation in American English and Athe-
nian Greek, and excerpts from an American and a Greek novel. Citing
examples from these discourse samples, I first consider the evidence that
dialogue in conversational storytelling is not reported speech; then
examine the ways that dialogue is introduced in the four discourse types
under study; and finally examine more closely the use of dialogue in one
genre - Greek conversational storytelling - and consider its place as one
of a range of features creating involvement in narrative.
My analysis compares the forms and functions of conversational and
literary narrative in American English and modern Greek. The conver-
sational discourse analyzed consists of stories told in conversation either
in dyads or in small groups, recorded by someone who happened to be
there. 1 The literary discourse examined consists of excerpts from novels.
The American novel used is Household Words by Joan Silber (Viking,
1976), an enormously moving and beautifully written novel which won
the Hemingway Award for first novels. The Greek novel used is To Trito
Stefani ("The Third Wedding") by Kostas Taktsis (Ermis, 1970), a work
which has been the focus of analysis by Kazazis (1979). The novel is
taken by modern Greek scholars to accurately represent middle class
Athenian dialect. Hence a secondary gain in this study is to see to what
extent it does. 2
 Introducing constructed dialogue 313

2. Reported speech is constructed dialogue

It cannot be the case that dialogue presented in oral storytelling is being

reported exactly as it was spoken, unless the report is based on the
deliberate memorization of a transcript which was based on a tape-
recording of the talk. 3 Experiments have proven what is intuitively
obvious - that humans cannot keep in their minds the precise words they
have heard, even for a moment. They listen for the meaning and, when
called upon to remember what was said, may reconstruct it into words,
much as Bartlett (1932) discovered that memory for objects and events is
constructive.
When one examines closely instances of what has been called reported
speech in spontaneous spoken narratives, it becomes clear that such lines
of dialogue are not reporting speech actually spoken. Some "quotations"
are of dialogue that was never spoken: 4
You can't say, "Well Daddy I didn't HEAR you."

Others are represented not as something that was said once but as an
illustration of a general phenomenon:
Daisy: The minute the kids get old enough to do these things
themselves, r that's when
Mary: L You do it yourself
Daisy: Yeah that's when I start to say . . . "Well... I don't think I'll go in
the water this time. Why don't you kids go on the ferns wheel. I'll
wave to you."

The reference to both swimming and a ferris wheel in the same "quote"
makes it clear that the quote does not purport to represent a single
instance of speech. Moreover, the introduction is phrased in terms of a
general time span: "when the kids get old enough" ("the minute," of
course, is a graphic exaggeration).
The preceding example also includes a line of dialogue that is con-
structed not by the storyteller but by a listener, who can't possibly be
reporting what she observed because she wasn't there. When Mary says
"You do it yourself," she is casting herself in Daisy's role and role-
playing what Daisy (might have) said to her children, following the story
line that Daisy has established - offering a line of dialogue as a form of
participatory listenership to demonstrate her attentiveness and under-
standing of Daisy's perspective.
Another instance of this device is found in a narrative that will be
314 Deborah Tannen

excerpted at length in a later section. In that excerpt, a medical resident

tells about an event that took place while he was on duty in a hospital
emergency room. At one point a listener offers the line, "We're okay,"
representing the thoughts or dialogue of patients in the emergency room
- dialogue that she is obviously constructing, not reporting, since she
wasn't there.
In other instances, a line of dialogue is presented as the production of
more than one speaker - an impossibility, unless one is talking about a
Greek chorus, which the following speaker is not:
And then all the Americans said "Oh in that case, go ahead."

In a number of the narratives analyzed, speakers "report" the

thoughts of other people - information they could not know - as when a
Greek woman "reports" what a man thought about her:
Sou leei, "Afti den echei kalo skopo."
[He says to himself, "She's up to no good."]

Sometimes an indirect quote fades into a direct one:

It was like he was telling everybody to have your wisdom teeth taken out and I
didn't see any point as long as they weren't bothering me.

In the process of quotation, the addressee metamorphoses from "every-

body" to "you" to "me".
Finally, it is clear that dialogue is not really reported when it includes
vague referents that would have made no sense had they actually been
uttered:
He was sending me out to get tools or whatever (imitates father) "Go get this
and it looks like this and the other."

In these examples, for the various reasons given, the "reported

speech" is clearly constructed dialogue. But even when dialogue could
conceivably have been spoken by the person to whom it is attributed
(and the narrator was in a position to hear) our understanding of the
powers of memory indicate that it probably wasn't.

3. Introducing constructed dialogue

In the discourse samples, I examined all instances of constructed dia-

logue, focusing in particular on how it was introduced. Results are shown
in figure l. 5
 Introducing constructed dialogue 315

Figure 1. Constructed dialogue in four discourse types

Eng. spoken Eng. novel Gr. spoken Gr. novel

Say 36 ( 43%) 72 ( 49%) 85 (71%) 41 ( 69%)

0 22 ( 26%) 24 ( 16%) 26 (22%) 11 ( 19%)

Thought 1 ( i%) 6 ( 4%) 4 ( 7%)

Tell 3 ( 4%) 4 ( 3%) _ * *

Ask 1 ( i%) - - 2 ( 3%)

Go (Gr.: Kano) 11 ( 13%) - 4 ( 3%) -

Like 7 ( 8%) - - -

Other 3 ( 4%) 40 ( 27%) 4 ( 3%) 1 ( 2%)

Total 84 (100%) 146 (100%) 119 (99%) 59 (100%)

* Included in "say" since the expression used (twice) is "sou leei," lit. "s/he says to you"
for "s/he says to him/herself'.

** Included in "say" since Greek does not have two different words, "say" and "tell."

He
Thesaid/she said
most frequent introducers are forms of the verb "say" (Greek leo).
Some typical examples:
English conversation:
Well I went because my-my regular dentist said, "You should have your
wisdom teeth taken out."
American novel:
"Doesn't he look handsome?" Hinda said. (p. 84)
Greek conversation:
Tou leo "En taxei tha'rtho ti Triti to vrady."
(I say to him "Okay I'll come Tuesday evening.")
Greek novel:
De me noiazei, elega.
(It doesn't bother me, I said.) (p. 73)
The extent to which the verb "say" (or "leo") is favored differs for the
various discourse types. It is most overwhelmingly favored in the Greek
samples. In the Greek spoken stories, forms of the verb leo ("say")
account for 71% of instances of dialogue. In the excerpt from the Greek
novel it is a strikingly similar 69%, suggesting that in this sense tfie novel
accurately represents the spoken idiom, as it purports to do.
In the American spoken stories, forms of "say" constitute a majority
316 Deborah Tannen

of the introducers used, but this majority is less than half (43%). In the
sample from the American novel, forms of "say" constitute a majority of
just about half (49%).
A small part of this disparity between the English and Greek samples
is accounted for by the use in English of a variant verb of saying, "tell",
which has no Greek counterpart; however, forms of "tell" constitute
only 4% and 3% respectively of the English conversational and literary
instances of dialogue. Thus, adding these percentages to those for "say"
yields 47% for the American spoken stories and 52% for the American
novel - still about half.
Yet another small influence on the higher percentage of forms of
"leo" ("say") in the Greek samples as compared to "say" in the English
samples is that the count for Greek includes instances of an expression
reporting characters' thoughts: "sou leei" - literally, "she/he/one says to
you," but figuratively, "she/he/one says to her/him/oneself'. 6 This usage
is seen in the following excerpt from a woman's narrative about a
university professor who told her to return on an evening to take an
individual oral exam and then, when they were alone in the deserted
building, made improper advances. The next time she had to appear
before him he again suggested that she return later in the evening to be
examined; this time she insisted on being admitted to the exam immedia-
tely. He complied, and she suggests the thoughts that might have led him
to his compliance:

(1) kai sou leei xero 'go

(2) "borei nachei kai kanena afto
(3) na mou kanei kamia zimia"

[idiomatic gloss]

(1) and he says to himself yknow

(2) "she might have some
(3) make some kind of scandal for me"

The form "sou leei" is found once in the Greek novel as well. The
narrator is telling her friend how her husband chose her as a wife: "Sou
leei: afti edo den einai san tis alles" (p. 69). This may be glossed: "He
thinks, 'This one here isn't like the others'."
Nonetheless, if instances of "sou leei" are subtracted from the count
of instances of "leo" in the Greek data, the percentage of "leo" is still
70% for the stories and 68% for the novel.
 Introducing constructed dialogue 317

Thus, despite the language differences noted, the verb "say" is the
favored dialogue introducer, accounting for roughly half the instances of
dialogue in the American and more than two-thirds of the instances of
dialogue in the Greek discourse samples.

"Go" and "kano" as verbs of saying

Of the introducers that are not forms of "say" (or "leo"), there is a type
found only in the spoken narratives. This variant in English, associated
with a very informal register, is the use of a form of "go" as a verb of
saying. The Greek counterpart, kano (ordinarily translated into English
as "make" or "do"), was not found as frequently as the English "go",
possibly because of the prevalence of this form in a single English
narrative.
13% of the lines of dialogue presented in the English spoken stories
are introduced with forms of "go". But of the 18 speakers whose
narratives make up the English stories examined, only 3 use "go" in this
way. 9 of the 11 instances are found in a single narrative, an excerpt from
which follows. 7
In this story, a hearing woman tells that she left a friend (who knows
some sign language) alone with her mother, who is deaf. When the
mother produced a sign that the friend didn't understand, he panicked
and called for her to come and help. She returned and inspected the sign
that her mother was making:
(1) 'η I look at it
(2) 'η I go
(3) Gee I don't know what the sign is either.
(4) Mom, what's the sign?
—» (5) 'n she goes
(6) Chair.
(7) CHAIR?
Forms of the verb "kano" were used by 2 of 25 Greek speakers, once
each. The continuation of the story about the university professor
illustrates this:
(4) en taxei, en taxei mou leei
—> (5) "Peraste" mou kanei
(4) okay, okay (he) says to me
—» (5) "Come in" he goes
Thus the informal forms "go'V'kano" distinguish the conversational
narratives from the literary ones.
318 Deborah Tannen

Unintroduced dialogue

The American conversational and literary narratives differ widely in how

they introduce dialogue when they don't use "say".
Prior to this study, I thought that a typical page of dialogue in fiction is
one in which lines of dialogue are presented with no introducers, made
possible by the written convention of quotation marks and indentation
(in Greek fiction sometimes represented by dashes and indentation). I
will illustrate this with a short segment from another novel: Judith
Rossner's August (Houghton Mifflin, 1983). In all but one line, speakers
are identified only by their place in the sequence of dialogue. This
segment is from a therapy session between a psychotherapist and her
young patient, Dawn, who speaks the first cited line:
" . . . How come you're so quiet? What are you thinking about?"
"I was wondering whether Tony happened to mention where you were
when the accident took place."
"Accident?"
"Your father's drowning."
"In the house," Dawn said promptly. "With the housekeeper. I mean, she
didn't mention it, but that's where I would've been."
"Mmra."
"Why'd you ask?"
"Because it would be interesting to know the answer."
"I gave you the answer."
"Perhaps."
"I don't like the way you said that."
Silence.
"You're making me nervous."
"Oh?"
"I feel as though you know something, and you think I know it, too, but I'm
keeping it from you." (p. 132)

In speaking, I thought, it is necessary to mark dialogue explicitly with

such expressions as "he said'V'she said". But this study revealed other-
wise. The use of no lexicalized introducer accounted for a significant
percentage of all the discourse types examined, but the percentage was
larger rather than smaller in the conversational narratives.
The conversational stories present dialogue with no introducers in
26% - more than a quarter - of the instances in which they present
dialogue. The Greek conversational stories do so in a comparable 22%.
In contrast, the American novel presents dialogue with no introducers in
only 16% of the instances of dialogue, and the Greek novel a compara-
ble 19%. (The difference between the Greek stories and novel - 19% as
 Introducing constructed dialogue 319

compared to 22% - is not as striking as the difference between the

English stories and novel - 16% as compared to 26%.)
How then do speakers mark dialogue as such, changing their footing
(Goffman 1981) from that narrator to that of character in the drama they
are creating? They do so by changing their voices to take on the
characters' voices.
This device is seen in the segment cited earlier from the story told by a
hearing daughter of a deaf mother, repeated here with arrows indicating
instances of elided introducers:
(1) 'η I look at it
(2) 'η I go
(3) Gee I don't know what the sign is either.
—» (4) Mom, what's the sign?
(5) 'n she goes
(6) Chair.
(7) CHAIR?

In (4) the speaker switches from addressing her friend to addressing her
mother, and in (7) she switches from portraying her mother's words to
portraying her own words in reaction.
The great versatility of the voice in presenting the dialogue of charac-
ters without introducing them is seen in the following excerpt from a
narrative told by a medical resident who had just returned from an all-
night stint at a city hospital about an incident in the emergency room. 8
Three young men appeared, one with a cut on his arm that was bleeding
profusely but was not serious. In telling about the incident, the resident
alternately took on the voices and gestures of himself, other hospital
staff, the wounded young man, the other two young men, other patients
in the emergency room, and a policeman who came to investigate.

(1) They come bustin' through the door -

(2) blood is everywhere
(3) on the walls
(4) on the floor
(5) everywhere
—» (6) [raised pitch] It's okay Billy
—» (7) it's okay
—» (8) we're gonna make it
—> (9) [normal voice] What's the hell wrong with you
(10) We . . . we look at him.
(11) He's covered with blood yknow?
(12) All they had to do was take a wash cloth at home
(13) and go like this [pause]
(14) and there'd be no blood
320 Deborah Tannen

(15) no blood
[listener: You put pressure on it]
(16) These drunk guys come bustin' in
—» (17) all the other patients are like Ugh Ugh
(18) They're bleedin' everywhere yknow
(19) People are passin' out just lookin' at this guy's blood here
(19) [listener: We're okay]
-> (20) I'm like "Get the hell out of here"

The speaker takes on the voices of characters by shifts in pitch, ampli-

tude, voice quality, prosody, and pacing. The number of speakers who
can be thus role-played is not limited to two, as in writing, since the
speaker can affect a range of voices, whereas a writer can not simply
alternate voices in a printed dialogue sequence.
In the novel Household Words, 16% of the lines of dialogue presented
are not lexically introduced. However, there are no extended sequences
of unintroduced dialogue like the one quoted from the novel August.
The closest thing to that in Household Words is the following inter-
change between Rhoda and her husband. Although the identities of the
speakers could easily be understood from the sequence, the author
marks Rhoda's dialogue with "said", but not her husband's:
Rhoda stood at the stove. "Dinner," she said, forking pieces of chicken onto
the plates.
"We have to eat fast. The men are coming in for pinochle at eight. Did you
forget?"
"Sort of," Rhoda said.
"But you wanted them to come. You told me to invite them, it was your
idea."
"I know," she said. "Don't tell me. I made cookies. I know."

The use of the verb "said" for Rhoda, along with her name, though not
for her husband, keeps her in focus as the protagonist.
Household Words has a literary (one might say poetic) sounding prose
style. In contrast, August is more conversational sounding, as observed
by Kendrick (1983) who described its style as "documentary" realism,
commenting that in it, "The give-and-take of real conversation, its
hesitations, repetitions and Freudian slips - all are reproduced with exact
fidelity". To a conversational analyst who has studied transcripts of real
conversation, the attribution of "exact fidelity" seems naive to the point
of absurdity, but presenting dialogue with no lexical introducer is indeed
typical of spoken conversational storytelling; therefore this feature may
contribute to the conversational sound of August, and use of lexical
introducers may contribute to the literary style of Household Words.
 Introducing constructed dialogue 321

Thus, the use of quotation marks in writing, which seem at first to

provide enhanced ability to mark dialogue as such, is actually a very poor
substitute for the great versatility of the human voice to do so.

"Be" + "like" for "say"

A type of introducer that is akin to using none at all is a device found

only in the spoken English stories. It does not appear in the English
novel, and there is no counterpart in Greek in the stories analyzed. This
is the introduction of dialogue with the word "like" following a form of
the verb "to be". Examples are found in the emergency room story
excerpted earlier:
(17) all the other patients are like "Ugh Ugh"
(20) I'm like "Get the hell out of here"
The line of dialogue "Get the hell out of here" does not represent what
the speaker actually said at the time; perhaps it is what he was thinking,
or what he would have liked to say. It is simply the line spoken in the
drama he created based on the night's events, by the character based on
himself.
8% of the lines of dialogue presented in the English spoken stories
were introduced by such forms. Besides the two cited above, the
following occurred:
It was like [spoken with foreign accent] "Would you stay. Why you why you
leaving Santorini?"
He was like [accent] "I give you a week of my life. Will you stay."
I was like - "Nice sister, right?"
'n it's like "Mary, I understand . . . "
And he was like, like "I feel like a failure. Yknow, even, even Brit could do
it."

This usage is less idiosyncratic, in the present study, than "go" as a verb
of saying; the 7 instances were produced by 5 different speakers, almost
a third of the speakers represented in the sample.
"To be" + "like" thus functions as a formulaic introducer, not by its
literal meaning but simply by convention. If the literal meaning functions
at all, it is to suggest that the dialogue is not being quoted but simply
represents the kind of thing that character was saying or thinking.
322 Deborah Tannen

Graphic introducers in literary narrative

The characteristic that sets the novel Household Words off most notice-
ably from the other three discourse types studied is the category "other"
in Figure 1. 4% of the instances of dialogue in the English spoken
stories, 3% in the Greek spoken stories, 2% in the Greek novel, but
27% in the American novel are introduced by verbs other than "say,"
"tell," "think," "ask," "go"/"kano", or "be" + "like." The verbs thus
used are: "explain," "complain," "croon," "coo," "demand," "call (0/
down/out)," "wheeze," "cry out," "mutter," "bellow," "murmur," "go
on," "titter," "grumble," "gasp," "whisper," "hiss," "sob," "scream,"
"suggest," "groan," "intone," "grimace," "yip," "warn," "sniff,"
"want to know," "shout," "wail," "repeat," "supply," "yelp," "snap."
Of these, only five are repeated, once each ("explain," "whisper,"
"scream," "shout," and "suggest").
Following is an excerpt showing how such verbs work in the novel. In
this passage, Rhoda serves lunch to her fifth-grade daughter Suzanne
and Suzanne's classmate Ina Mae. (Verbs other than "say" introducing
dialogue are underlined.)
Suzanne . . . reached out to give Ina Mae a "feeny bird," a rap on the skull
with flicked fingers, as Ina ducked away, screaming, "Get away from me!"
"How about," Rhoda suggested, "clearing off the kitchen table so you can
have some good old peanut butter and jelly sandwiches?"
"Oh boy," Suzanne groaned sarcastically. "Oh boy, oh boy, oh boy."
"THE BOY," Rhoda intoned, beating time with a spoon at the kitchen
sink, "STOOD ON THE BURNING DECK,/ HIS FEET WERE FULL OF
BLISTERS,/ HE TORE HIS PANTS ON A RED-HOT NAIL/ SO NOW HE
WEARS HIS SISTER'S." The girls, unfamiliar with the original poem (a
staple of recitations in Rhoda's childhood) failed to find this wickedly amu-
sing. "Oh, Mother," Suzanne grimaced. "Ina, for Christ's sake, would you
please pass the jelly? I'm starving, you know."
"You poor old thing," Rhoda said. "You're so hungry you could dydee-
dydee-dydee-die." Ina giggled. Rhoda poured a glass of milk for the guest.
"Say when," she suggested.
"I HATE milk," Ina yipped.
"Oh, we never serve milk in this house. This is cow juice. Don't be fooled
by the carton." Rhoda smiled mysteriously.
"She thinks she's funny," Suzanne said. (p. 104)

Graphic introducers are an evaluative device, to use a term coined by

Labov (1972) to describe elements in oral narratives. The author uses
them to hone her description of the characters, their personalities and
states of mind, and their relationships to each other.
 Introducing constructed dialogue 323

Another characteristic of the dialogue in this segment, as in the rest of

the novel and in the Greek novel examined, is the interruption of lines of
dialogue to insert the narrator's commentary:
"How about," Rhoda suggested, "clearing off the kitchen table so you can
have some good old peanut butter and jelly sandwiches?"

This does not happen in the spoken stories, where introducers may be
repeated between but not in the midst of spurts of speech which
Gumperz (1982) calls "tone groups" and Chafe (1980) calls "information
units". Because language in print does not follow the breath units of
speaking, such line breaks can come in places they would not in speech.
In this way, the Greek novel departs from the conventions of spoken
discourse.
Another characteristic of the excerpt from Household Words which is
also characteristic of literary writing is that some verbs introducing
dialogue do not really describe the way the dialogue was spoken but,
under the guise of such service, actually describe something else about
the action or the actors. For example, in: "Oh, Mother,' Suzanne
grimaced," Suzanne could not so much have grimaced "Oh, Mother" -
grimacing is not a behavior that produces speech - but rather she spoke
those words and grimaced. Since this fact is self-evident, using the word
"grimaced" to describe the speech is a more concise and effective way of
describing how Suzanne spoke than would be a more pedestrian sound-
ing, "Suzanne said with a grimace".

A continuum

I suggest that these various ways of introducing dialogue fall along a

continuum. At one pole is no introducer at all, used in informal
conversational narrative because of the great expressive power of the
human voice. At the other pole is the use of graphic verbs as introducers
- a form typical of literary narrative. Perhaps this is a way of compensa-
ting for the loss of expressive voice quality in print by packing more
expressive power into the words chosen - more work is done by the
meanings of the words, less by the way they are spoken. This supports
Chafe's (1982) finding that written discourse is more integrated (packing
more information into fewer words), and Tannen's (1982) hypothesis
that written fiction combines the integration Chafe found in written
discourse with the involvement he found in speech.
324 Deborah Tannen

On the continuum posited, "be" + "like" is next to no introducer,

depending for effect on the way the dialogue is voiced. The form "go" is
similar to "say" in meaning, but it is similar to "be" + "like" in register.
These two forms co-occur in the sign language story (and, it seems, in the
casual speech of middle class American teenagers).

4. Constructed dialogue in action: Greek stories

When hearing some of the Greek stories included in this study, in the
original Greek or in translation, listeners find them very vivid. This
impression seems to reflect a phenomenon frequently observed, and
supported by folk wisdom, that Greeks are good storytellers. Elsewhere
(Tannen 1983) I identify and illustrate the linguistic features that contri-
bute to that impression - features which I suggest contribute to the
creation of involvement: both the involvement of the audience and the
sense of the speaker's own involvement in the storytelling. 9 1 found that
involvement is created by (1) immediacy, portraying action and dialogue
as if it were occurring at telling time and (2) forcing the hearer to
participate in sensemaking.
The features which typify the Greek narratives and which contributed
to involvement are:
1) repetition
2) direct quotation in reported speech
a) dialogue exchanged
b) thoughts of speaker
c) thoughts of man
3) historical present verbs
4) ellipsis
a) deletion of verb of saying
b) deletion of copula
c) deletion of comment or proposition
5) sound-words
6) second person singular
7) minimal external evaluation
Thus, constructed dialogue is one of a range of features that create
involvement, and vividness, in the Greek stories.
15 of the 25 narratives told by the Greek women used constructed
 Introducing constructed dialogue 325

dialogue. The remaining 10 did not report dialogue at all. It would be

mistaken, however, to conclude that the 15 stories which present dia-
logue do so because talk occurred during the incidents they report,
whereas talk did not occur during the other 10.
The Greek oral narratives studied are about being molested by men.
At the same time that I collected those stories, I also collected narratives
told by American women about being molested. The reason I did not use
them for comparison with these Greek stories is that in all of them I
found only one instance of constructed dialogue. I do not believe that
this is because the incidents reported by the American women happened
not to involve talk, whereas those reported by the Greek women did.
Rather, I believe it is characteristic of Greek storytelling conventions to
construct dialogue.
Similarly, it would be misleading to measure the incidence of con-
structed dialogue as a percentage of the instances of dialogue. The
representation of speech in dialogue is a narrative act, not the result of
the occurrence of speech in the episode. By setting up a little play, the
speaker can portray motivations and other subtle evaluations internally -
from within the play - rather than externally - by stepping outside the
frame of the narrative to make evaluation explicit. (The terms and
concepts internal and external evaluation are from Labov 1972).
For example, a speaker I will call Marika "reports" her decision to call
the friend of a friend during a trip to Rhodes in the form of her comment
to her traveling companion:

Tis leo tis xadelfis mou,

"Kaiti, den pame kai ston systimeno ton anthropo
na mi fygoume apo tin Rodo
kai den echoume patisi to podaraki mas?"
"Pame," mou leei.
I say, to my cousin,
"Katie, shouldn't we go see the fellow
we were told to look up,
so as not to leave Rhodes
without having set foot [on his doorstep]?"
"Let's go," she says to me.

By casting the decision in the form of dialogue, Marika creates a

dramatic scene and is thus able to show by her phrasing (i.e. internal
evaluation) that she was motivated by a sense of obligation to behave
properly, not by any desire to spend time with this or any man.
Marika then tells that the man insisted on taking them for a tour of
326 Deborah Tannen

Rhodes, for which excursion he showed up with a friend. She lets us

know what she feared - and builds suspense - by reporting her thoughts
in the form of direct quotation:
Leo, "ti thelei
dyo ekeinoi, dyo emeis,
ti echei skopo na mas kanei?"
I say [to myself], "what does he want,
two [of] them, two [of] us,
what does he intend to do to us?"
In four stories Marika represents her thoughts as direct quotations to
herself, sometimes even addressing herself by name:
"Kala" leo, "Marika edo eimaste tora."
"Okay" I say [to myself], "Marika, here we are now."

Like the Greek speaker previously cited who told about her university
professor, Marika casts the (projected) thoughts of another character as
dialogue. In telling how she chased off a man who had been harrassing
her and her friend, she tells what the man (must have) thought upon
seeing her step toward him brandishing a rock: "sou leei 'afti den echei
kalo skopo'" ("he says to himself, 'she doesn't have a good purpose',"
i.e. 'she's up to no good').
A variation on constructed dialogue - something that is constructed
but not exactly dialogue - that is prominent in the Greek spoken stories
is the use of sound words, or sound non-words, to represent action. 10
There are 13 instances of sound words in the 25 Greek narratives. A few
examples follow.
peftei aftos apano mou
xereif apano mou BAM.
he falls on top of me
yknow on top of me BOM.

In this example, as in the following one, the sound word "BAM" (/bam/)
illustrates the action that has been described.
opou vlepeis ton [name]
opos einai kontochondros
na pesi epano mou paidia
etsi epese PLAF
when you see [name]
as he is short-and-fat
falling on top of me, guys,
like that he fell PLAF
 Introducing constructed dialogue 327

And, finally, an example in which three successive sound words are used
to represent action which is not otherwise described:
vgazo tin petra - DAK!
pali do etsi - DOUK!
ekane ΤΑΚ!
kai exifanisthi aftos.
I take out the rock - DOK!
again here like that - DUK!
he went TOK!
and he disappeared.

I can not reconstruct the actions that the sound words represent, but I
can reconstruct that Marika's "DAK'V'DOUK" (/dak/ /duk/) represent-
ed some form of attack with the rock. "Ekane ΤΑΚ!" ("[It/He] went [lit.
made] /tak/!") would have been disambiguated by a gesture as well.
The sound words that appear in the narratives are: /bam/, /gan/, /ga/,
/dak/, /duk/, /tak/, /mats/-/muts/, /plaf/, /ax/, /a/, and /psit/-/psit/. The last
is somewhat different, I believe; it represents onomatopoetically the
sound with which Greek men get the attention of women and chase away
cats. All the other sound words are composed primarily of the large-
sounding back vowels /a/ and /u/; the abrupt voiceless and voiced stops
/k/ /g/, /t/ /d/, and /p/ /b/; and consonant clusters /ts/ /pi/. The sound words
are phonologically graphic, patterning with similar phenomena in many
other languages (Ohala 1983,1984), and they contribute to involvement
by forcing the hearer to recreate the action represented by the sound.

5. Constructed dialogue as vivid storytelling

Thus, constructed dialogue in the Greek stories is part of a network of

features which create involvement. It seems likely that the use of
constructed dialogue is associated not only with Greek but also with
other ethnic styles that come across as "vivid" - as Kirshenblatt-
Gimblett (1974) and Tannen (1984) have shown for East European Jews,
and Labov (1972) and others have demonstrated for American Blacks.
There is some evidence that Brazilian speech falls into this category as
well, and that constructed dialogue is a dimension of that effectiveness.
In a pilot study comparing how Brazilian and American speakers told the
story of Little Red Riding Hood (Ott 1983), Brazilian speakers used far
328 Deborah Tannen

more constructed dialogue. The American man in the study used 6 such
instances, all formulaic for this fairy tale:
"Grandma, what a big nose you have."
"All the better to smell you my dear."
"Grandma, what big ears you have."
"All the better to hear you my dear."
"Grandma, what a big mouth and big teeth you have."
"All the better to eat you with my dear."
The American woman in the study used 15 instances of dialogue,
including the formulas found in the American man's story, but also
including some improvised variations on them ("What long whiskers you
have"; "The better to wiggle them at you my dear") and the casting of
other parts of the story in dialogue. For example, she has the mother tell
Little Red Riding Hood, "Go to your grandmother's house . . . " The
Brazilian woman who told the same story used 20 instances of dialogue,
and the Brazilian man used 43!
The Brazilian man's version of Little Red Riding Hood represents
almost all action in dialogue which makes the story rich in particularity.
For example, at the beginning (as translated into English by Ott):
One time on a beautiful afternoon, in her city, her mother called her and
said:
"Little Red Riding Hood, come here."
"What is it, mother? I am playing with my dolls, can I continue?"
Long segments are composed only of dialogue. For example, when she is
accosted by the wolf on her way to her grandmother's house:

"Little Red Riding Hood, Little Red Riding Hood".

And Little Red Riding Hood stopped and looked: "Who is there?"
"Ah, who is talking here is the spirit of the forest."
"Spirit? But I don't know you."
"No, but I am invisible, you can't see me."
"But what do you want?" (imitating child's voice)
"Where are you going, Little Red Riding Hood?"
"Ah, I'm going to my granny's house."
"What are you going to do there, Little Red Riding Hood?"
"Ah, I'm going to take some sweets that my mother prepared for her."
"Ah, very good . . . the sweets are delicious, they are, they are, they are, they
are . . . " (licking his lips)
"Do you want one?"
"No, no, no, no. (Accelerated) Spirits don't eat.
Okay, okay. Then, now, yes, yes, you are going to take it to your granny . . .
remember me to her, okay?"
"Okay, bye."
 Introducing constructed dialogue 329

Thus, through constructed dialogue and other features, this Brazilian

speaker created a vivid new story out of a standard fairy tale.

6. Dialogue in a writer's conversation

I had a rare opportunity to observe a naturally-occurring instance of the

sort of linguistic comparison that I have set up experiments to achieve.
Two people independently and spontaneously recounted to me the same
conversation. One was a writer and the other an employee of a publish-
ing company which was publishing the writer's book. It happened that I
knew them both, and in the course of conversation with me, each one,
knowing I knew the other, told me about a long distance telephone
conversation they had had. The author had been dilatory about obtain-
ing permissions to reprint illustrations in his book. Publication was
delayed as a result; the publisher had spent a great deal of time trying to
track down the copyright holders himself; and the author had repeatedly
failed to respond to the publisher's phone messages and letters. The
author finally called the publisher to give him the necessary information.
Both men agreed that these were the circumstances.
The author described the conversation in this way:
I said, "I'm sorry to have been so exasperating." [pause] And there was a long
silence.

The publisher described it this way:

He apologized, but when the time came for me to say "That's all right," I
didn't say it, so there was a long silence.

Though the two accounts do not disagree on facts, I believe the author's
recount of this conversation is more effective, in an aesthetic sense. Both
accounts include constructed dialogue, but for different functions. The
author gave a line of dialogue to represent what he'd said ("'I'm sorry to
have been so exasperating'"). The publisher reported that utterance by
naming the speech act, "He apologized". The publisher used a line of
dialogue to represent what he didn't say ("'That's all right'"). The
author left that line unstated, assuming that I know what is omitted when
silence follows an apology. In other words, the dialogue the author
included was partistet dialogue - what he said. The dialogue that the
publisher included was a general representation of the kind of statement
330 Deborah Tannen

that could have been said but wasn't. (The author's omission of such
dialogue constitutes another poetic process - using ellipsis to force the
hearer to supply part of the meaning).
I think it is not a coincidence that the more effective story (minimal
though it was) was told by the author - a writer of fiction. I don't know
whether or not the words he reported are exactly the words he spoke. I
don't think it matters. It may be that as a writer he has a good memory
for exact wording. But it may also be that he has a good sense of possible
wording, that the words he reported were not exactly the ones he had
spoken, but they had an authentic ring. He seems to have a sense that
retelling his apology in the form of constructed dialogue will be vivid - a
particular apology - and make the sense of what should come next vivid
also.
If I am right about the differentially poetic use of constructed dialogue
in the conversation of different speakers, then the notion of a continuum
from conversational to literary narrative is not a linear one, but rather
various of the devices discussed may turn up in different genres, depen-
ding upon the register employed and the effect desired. In any case, I
hope to have demonstrated that what has been called reported speech or
direct quotation is constructed dialogue, that it makes story into drama,
and that through such drama talk builds on and creates interpersonal
involvement.

Notes

1. The English stories were recorded, selected, and transcribed by students in my

Discourse Analysis class Fall 1983: Gayle Berens, Gina Doggett, Dianne Falvo,
Matthew Glotfelter, Susan Hoyle, Aziz Kamel, Deborah Lange, Kimberly Murphy,
Tulinabo Mushingi, Faith Powell, Fran Smith, and Joseph Wieczorek. Terry Wald-
spurger helped identify constructed dialogue and count words in these narratives. I
recorded the Greek stories in Athens, Greece, in a small group or in dyads. Fileni
Kalou transcribed them; Maria Spanos checked transcriptions and helped with identifi-
cation of constructed dialogue and translation. I gratefully acknowledge all this help.
2. Peter Mackridge, an authority on the modern Greek novel, suggested that I use the
Taktsis novel. I chose the Silber novel not only because I admired it but also because I
was able to tape record its author in casual conversation and thus obtain a discourse
type I needed for another phase of the larger study of which the present analysis is part.
3. Even in transcribing talk from a tape, one must continually replay because one cannot
be sure that one has transcribed correctly. Replaying and carefully listening to the
smallest segments of tape proves that the words "transcribed" were not exactly those
that were heard.
 Introducing constructed dialogue 331

4. Punctuation has different meanings in excerpts from transcripts of talk and printed
sources. In spoken segments, dialogue is bounded by quotation marks to facilitate
reading; commas indicate phrase-final intonation ("more to come"); periods indicate
sentence-final intonation. Three dots ( . . . ) indicate pause. In some cases transcription
is set out in poetic lines rather than run together as prose to facilitate reading. The lines
represent audible chunking, by intonation and prosody, into "tone groups" (Gumperz
1982) or "information units" (Chafe 1980). Italics indicate emphasis. Excerpts from
literary samples are reproduced exactly as they appeared in print. (In the Greek novel
quotations are variously marked by a combination or none of: quotation marks,
dashes, indentation). Here, three dots ( . . . ) indicate ellipsis. Transliteration of Greek
follows guidelines established by Peter Bien and Julia Loomis for the Modern Greek
Studies Association.
5. The discourse samples consist of 18 English conversational stories (about 8000 words),
25 Greek conversational stories; about 9,600 words of the English novel and about
3,000 words of the Greek novel. Because numbers of words in the samples differ, as
well as numbers of instances of dialogue, results are presented in percentages. No
attempt was made to obtain equal numbers of words in samples because this would
imply that there is objective significance in the number of instances of constructed
dialogue - an assumption for which there is no foundation, given the impossibility of
determining the relationship between instances of constructed dialogue and occurrence
of talk in the events on which narratives were based (a non-existent variable for fiction,
in any case). Rather, what is significant is the percentage of each type of introducer in
the instances that occur, and the function of each type, which is not a matter of
numbers at all.
6. There is also a Greek verb skeptomai ("think"), which is found in the sample from the
Greek novel but not in the conversational stories.
7. This story was tape recorded and transcribed by Matthew Glotfelter. Its effectiveness
depends on the knowledge that the person the story is about was present at the time of
its telling. The speaker exaggerates his incompetence and the extremity of the incident
as a way of teasing him and making him the protagonist of a story.
8. This story was taped and transcribed by Kimberly Murphy.
9. This section is based on material included in Tannen 1983. It is improved by comments
by Kostas Kazazis on that paper, for which I am grateful.
10. I did not originally think of these as constructed dialogue; the connection was pointed
out to me by Florian Coulmas.

References

Bartlett, Frederic C. 1932. Remembering. Cambridge: Cambridge University Press.

Chafe, Wallace L. 1980. The deployment of consciousness in the production of a narrative.
In: W. Chafe (ed.), The pear stories: Cognitive, cultural, and linguistic aspects of
narrative production. Norwood, NJ: Ablex, pp. 9-50.
Chafe, Wallace L. 1982. Integration and involvement in speaking, writing, and oral
literature. In: Deborah Tannen (ed.), Spoken and written language: Exploring orality
and literacy. Norwood, NJ: Ablex, pp. 35-53.
Friedrich, Paul. 1979. Poetic language and the imagination: A reformulation of the Sapir
332 Deborah Tannen

Hypothesis. Language, context, and the imagination. Stanford, Ca.: Stanford University
Press, pp. 441-512.
Goffman, Erving. 1981. Footing. Forms of talk. Philadelphia: University of Pennsylvania
Press, pp. 124-159.
Gumperz, John J. 1982. Discourse strategies. Cambridge: Cambridge University Press.
Kazazis, Kostas. 1979. Learnedisms in Costas Taktsis's Third Wedding. Byzantine and
Modern Greek Studies 5. 17-27.
Kendrick, Walter. 1983. The analyst and her analysand. Review of August, by Judith
Rossner. The New York Times Book Review, July 24, 1983, pp. 1, 19.
Kirshenblatt-Gimblett, Barbara. 1974. The concept and varieties of narrative performance
in East European Jewish culture. In: Richard Bauman and Joel Sherzer (eds.), Explora-
tions in the ethnography of speaking. Cambridge: Cambridge University Press,
pp. 283-308.
Labov, William. 1972. Language in the inner city. Philadelphia: University of Pennsylvania
Press.
Ochs, Elinor. 1979. Planned and unplanned discourse. In: Talmy Givon (ed.), Discourse
and syntax. NY: Academic, 51-80.
Ohala, John J. 1983. Cross-language use of pitch: an ethological view. Phonetica 40.
pp. 1-18.
Ohala, John J. 1984. An ethological perspective on common cross-language utilization of
tone of voice. Phonetica 41. 1-16.
Ott, Mary Miglio Bensabat. 1983. Orality and literacy in Brazilian and American storytel-
ling: A comparative study. Ms., Georgetown University.
Schiffrin, Deborah. 1981. Tense variation in narrative. Language 57: 1. 45-62.
Tannen, Deborah. 1982. Oral and literate strategies in spoken and written narratives.
Language 58: 1. 1-21.
Tannen, Deborah. 1983. "I take out the rock - dok!": How Greek women tell about being
molested (and create involvement). Anthropological linguistics. 25: 3. 359-374.
Tannen, Deborah. 1984. Conversational style: Analyzing talk among friends. Norwood,
NJ: Ablex.
Characteristics of direct and reported speech
prosody: Evidence from Spanish
Karen H. Kvavik

1. Introduction

Spanish grammatical and stylistic literature posits that 'international'

differences are major distinguishing features of direct and reported
(indirect) speech. This chapter examines prosodic characteristics of these
two speech types. The first objective of the study is to survey and critique
the literature about their sentence intonations, e.g., 1) that indirect
discourse is characterized by neutral intonation; 2) conversely, that
direct discourse sentences are characterized by greater expressivity, less
uniformity of intonation; 3) that the introducer phrase in reported
speech ('he/she says that . . . ' ) is melodically uniform, as is the rest of the
reported sentence. Second, assumptions in the literature are examined
with new data from an investigation based on a conversational situation
protocol. The third goal develops a 'profile' of these discourse-based
intonations, that is, acoustic and linguistic features which might ultima-
tely contribute to perceptual differences.
Results are given for perceived sentence primary 'stress', intonation
contours or 'tunes', and length and fundamental frequency (sentence
peak and low frequencies, and range). (Here 'contours' and 'tunes' are
used without taking a position on the tone (levels) vs. contours approa-
ches to intonation.) If intonational differences are observed for these
sentences, we should be able to characterize them in an explicit way.
Assuming there are prosodic differences, what linguistic and acoustic
cues might be relevant? In addition, to what degree are traditional
notions about intonation only literary or grammatical conventions? To
what degree can we find empirical data to support them?
The chapter is organized into two main parts: 1) General comments
and overview of Spanish direct and reported speech literature; 2) results
from my investigation of the two sentence types.
334 Karen Η. Kvavik

2. General comments and overview

It is evident that discussing intonation in terms such as 'uniformity' and

'monotony' is imprecise and vague. Even the term 'intonation' is used in
various ways. In a wide sense, it is used to mean 'sentence melody',
comprehending impressions of loudness and length, even stress. In a
narrow sense, it refers to acoustic measurements (fundamental fre-
quency, length, and/or amplitude) or perceptual measurements (pitch,
duration, and/or loudness).
Grammarians, literary critics, and stylisticians use the Sp. terms
discurso 'discourse', or stilo 'style' interchangeably to describe either an
analysis of sentence syntax of spoken or written speech or written
discourse styles.1 Examples of direct vs. reported sentence syntax are 'he
sees the boy' vs. 'he says (that) he sees the boy'. The reported sentence
has an introducer phrase + complementizer 4- embedded complement
phrase; the introducer phrase consists of a performative verb ('he says,
shouts, whispers, exclaims, declares', etc.,). To complicate matters,
literary critics and stylisticians claim to analyze direct and reported
(speech) styles in (written) literary works: Direct speech style is the
author's depiction of a character's speech through supposedly faithful
reproduction of that [written] speech, e.g., 'She said, "I won't go there
again.'" The double quotes (") indicate the direct speech. Reported
speech style describes or reports literary characters' speech acts, either
through the author or through another character with a predominance of
reported sentence syntax, e.g., 'She said . . . She asserted . . . ' .
In grammars it is not always clear whether (written) sentences were
directly observed or created for illustration; literary examples may be
cited. Spoken speech is scarcely acknowledged in the scholarly litera-
ture; it is as if 'direct and reported' were artifacts of the written language.
This is paradoxical because both types are found abundantly in spoken
language. Moreover, all spoken utterances are 'direct' in the sense that
they are all directed to or spoken directly to someone, whether 'direct' or
'reported' syntax. Furthermore, written speech (or written discourse)
styles are described by both grammarians and stylisticians as having
intonation supposedly distinguishing the two [written] styles. In this
roundabout way, we observe that a stylistic analysis of a written text is
characterized in terms of spoken speech, yet this 'spoken' speech is a
given, not analyzed heretofore in any detail. One concludes that the
scholar (or reader) 'hears' or imposes a 'reading' of how the direct or
 Direct and reported speech prosody 335

reported speech supposedly sounds: the imagined or reading style into-

nation (actual? mental?) then is part of the definition of a written or
literary discourse style.
All this is logically flawed. To add to the difficulties, Sp. grammars
and literary analyses echo each other and perpetuate the same notions;
there is very little difference among them, as we shall see. While
impressionistic, scholars nevertheless have indicated how these senten-
ces should sound. This suggests that there are intonative conventions and
stylistic devices which the author/reader/stylistician counts on in text
interpretation. But do these conventions reflect actual speech?
A brief overview of the grammatical and stylistic literature finds direct
and reported speech characterized under the rubric 'noun complement
clauses' (oraciones sustantivas subordinadas 'subordinated noun senten-
ces'), functioning as 'accusative or direct object complement' (acusativo
or complemento directo). There are two types of direct object comple-
ments, estilo directo and estilo indirecto ('direct style' and 'indirect or
reported style'). Thus, the term estilo 'style' here refers to subtypes of
declarative sentence constructions.2
Seco 1975 characterizes 'direct style' sentences as literally reproducing
another's speech. 'Indirect style' refers to 'that which is said' by another
(lo dicho).
En realidad, en el estilo directo no hay subordination, aunque asi lo piensen la
mayoria de nuestros gramäticos. ... No hay relation entre el "verbo de decir" y
"lo dicho"; la relation es exclusivamente logica. ... Es importante notar que,
en el estilo indirecto, el "verbo de decir" y "lo dicho" presentan siempre
entonaciones independientes, prueba bien clara de lo que afirmanos. 'In reality,
in direct style there is no subordination, although a majority of our gramma-
rians may think so. . . . There is no relationship between the verb of "saying"
and "that which is said"; the relationship is exclusively logical. . . . It is
important to note that, in indirect style, the "verb of saying" and "that which
is said" always present independent intonations, a clear proof of what we
maintain' (227, η 1).

As for there being 'no subordination' in direct speech, Seco must mean
that there is no embedded complement with the introducer as the matrix,
because subordinated clauses do occur in direct speech. Also, the
reference to the 'majority of our grammarians' is not illuminated. What
Seco means by 'independent intonations' in indirect style is unclear; one
assumes that he means a tonal rise or fall and perhaps some sort of a
juncture.
Alcina Franch and Blecua (1975) discuss reported 'style' (e.g., sen-
tence types) in interrogatives, but say, interestingly, that estilo directo/
336 Karen Η. Kvavik

indirecto are primarily characteristic of written language, that only

sometimes occur in spoken language (1120). Gili Gaya (1971: 288)
speaks of the two styles as follows:
Se llama directo el estilo, cuando el que habla ο escribe reproduce textualmente
las palabras con que se ha expresado el propio autor de ellas ... En el estilo
indirecto el que habla ο escribe refiere por si mismo lo que otro ha dicho. 'It is
called direct style when he who speaks or writes textually reproduces the
words with which the very author of them has expressed himself . . . In indirect
style he who speaks or writes reports by himself that which another has said.'

The use of 'textually reproduces' in the context of spoken speech is

curious. Tenses and moods alternate depending on the tense in the
introducer phrase (again in a sentence type). He notes that even if the
complement marker que begins an 'apparently dependent' sentence, in
reality, the complementizer is 'mentally subordinate' (288f.): 3

(1) iQue no te has acordato? 'you have not remembered?'

Que no diags estas cosas a nadie 'may you not say these things to
anyone.'
Comments of the Royal Spanish Academy (1975: 516) are identical to
those of Gili Gaya (1971: 288) above. Evidently, Gili must have compi-
led that section of the Academy grammar. Finally, Criado de Val (1976)
maintains direct style has a pause before the quoted sentence, while
reported style shows 'strong syntactic subordination . . . avoiding an
abrupt cut' with the complement (67). Criado's view differs from Seco's,
who says the reported sentence shows independent intonations. Repor-
ted style tends to have an introducer verb, but not necessarily. Finally, in
his intonation manual, Navarro Tomas 1966 has nothing specific to say
about direct vs. reported speech intonations, both considered 'declara-
tive' intonation.
Our knowledge about reported intonations is mainly found in two
literary-stylistic analyses, Verdin Diaz (1970) and Dominguez (1975).
Both authors primarily focus on estilo indirecto libre 'free reported style'.
Although these scholars claim to analyze direct, reported, and free
reported styles in different authors, it is not that simple. Grammatical
analyses meld with literary style analyses through similar terminology.
Verdin takes Criado de Val's definition of 'direct style' which consists
con toda objectividad de las mismas palabras del hablante ('with complete
objectivity of the very words of the speaker'). 4 While Criado says there is
a pause after the introducer, Verdin maintains there is una estrecha
unidad semäntica 'narrow semantic unity' between the introducer and
 Direct and reported speech prosody 337

reproduced speech; when the introducer verb precedes the sentence, la

entonacion de la frase en estilo directo sigue en estos casos una elevacion
tonal, descendiendo gradualmente 'the phrasal intonation in direct style
follows a tonal elevation in these cases' (34). Specifically, how this
intonation is elevated and starts to descend gradually is not made known.
One only presumes it is based on the standard intonation comments of
most Spanish scholars, ultimately based on Navarro Tomas' work on
Castilian. In direct style, Verdin says voice descent depends on where
the introducer phrase is place in the sentence (beg., mid., or end). He
lists introducer verbs (performatives), whose emotional content varies
from 'exclamation to insinuation to doubt' (37-38). For example, the
introducer phrase in direct style permits us to know whether a character
or the author is speaking, although the introducer can be eliminated
from more rapid narration (40); an obvious sign of direct speech in
written language are quotes. All grammatical persons are represented in
direct speech, whereas reported style is marked by a predominance of
3rd person in the complement, e.g., 'he says/said, "I'm leaving soon'"
vs. (ind. style) 'he says/said that he is/was leaving soon'.
Verdin's definition of reported style again is taken from Criado de
Val: 'El estilo indirecto pur ο en realidad comienza por el empleo de una
conjuncion subordinante . . . 'Pure indirect style in reality begins by the
use of the subordinating conjunction . . . ' (51). The conjunction (que)
replaces the direct style pause. Indirect style lacks la vitalidad de los
signos expresivos 'lacks vitality from expressive signs' (52). Compared to
direct style, with its 'variety and richness of expressive signs', indirect
style is distinguished by 'uniformity of syntactic construction'. In a
reported sentence (vs. direct sentence) ningun rasgo caracteristico nos
descubre el tono de voz del hablante 'no characteristic reveals the
speaker's voice tone to us' (52).5
Prosodic features further (negatively) characterize reported style:
Subordination of exclamatory, interrogative, and imperative sentences
cause their respective intonations to be lost; vocatives are also absent,
being a feature of direct speech. As an evaluation of the reported
intonation, Verdin says:
Tanto la frase introductora como la frase reproducida aparecen casi enel mismo
piano respecto a la entonacion, sin olvidar, por supuesto, que son grupos
distintos y que estän caracterizados respetivamente por sus terminaciones
ascendente y descendente. 'The introducer phrase as well as the reproduced
phrase appear almost on the same plane with respect to intonation, without
forgetting, of course, that they are distinct groups and are characterized
respectively by their ascending and descending terminations' (65).
338 Karen Η. Kvavik

Besides absence of pauses and expressive elements, syntax becomes

'monotonous', with a 'special uniform intonation', and with a final
intonative descent (66).
Dominguez (1975), the second of detailed works, only briefly com-
ments about direct and reported 'discourse'.6 In essence, direct discourse
'textually reproduces the morphology, syntax, intonation, and even level
of language characteristics used by the character (12)'. When reproduc-
ed, intonation is 'autonomous', with the same intonation as before its
reproduction. She believes that intonation depends on a syntactic type,
despite her assertion that it is 'autonomous':
La entonacion es autonoma; la misma que tenia antes de su reproduction; es un
signo gramatical, una constante por la que se comprueba el discurso directo ...
La entonacion es el signo revelador del discurso directo. 'Intonation is autono-
mous; the same that it was before its reproduction; it is a grammatical signal, a
constant through which direct discourse is verified . . . Intonation is the
revealing sign of direct discourse' (12).

Since intonation is taken for granted as part of written texts, we must

assume it is supplied somehow - by a stylistic convention or by reader
interaction with the text by overlaying intonation.
According to Dominguez, indirect discourse usually undergoes verbal
transformations regarding tense (this is not an absolute), already discuss-
ed above for Verdin. An introducer verb and conjunction are the
'revealing sign' of indirect discourse. Intonation is 'neutral'. Interrogati-
ves and imperatives lose their 'special' intonation. Indirect discourse
permite ver la actitud del relator, reproduce el discurso en una forma
abstracta y diferente del directo y ademäs tiende a eliminar los elementos
expresivos de la lengua 'permits the narrator's attitude to be seen,
reproduces discourse in an abstract form and tends to eliminate expres-
sive elements' (13).
In summary, for direct speech or style, intonative elements are present
with a characteristic vividness. For reported style, both Verdin and
Dominguez indicate lack of expressivity, and uniformity or neutrality,
where the introducer phrase supposedly terminates in a rise and the
sentence with a fall. The idea that a grammatical or syntactic type (e.g.,
imperative) has a particular intonation is widely held. At least for
English, intonation does not depend on a particular semantic or syntactic
category. Bolinger (1982, 1983) demonstrates how intonation is a sepa-
rate system, akin to gestural and metaphorical systems. Speaker-hearer
inferences are important in determining intonations. By this reasoning
we should not necessarily expect to find a difference between direct and
 Direct and reported speech prosody 339

reported intonations. Since grammarians say both types are declaratives,

we might expect 'declarative' intonation. On the other hand, the litera-
ture says direct and reported utterances have different intonation be-
cause of their syntax. Despite the indication of differences (e.g., 'expres-
sive' vs. 'neutral'), given that both types are declaratives, the differences
might rather turn out to be another manifestation of gradience (Bolinger
1961).

3. An investigation of direct and reported prosody

A conversational-situation protocol was developed to elicit comparable

declarative, imperative, and reported speech sentences; the aim was to
approximate natural speech and to avoid reading or mimicry. The
miniconversations are everyday situations, similar to foreign-language
directed dialogues, but used with the native language, and similar to the
contextual glosses one often finds in linguistic literature.
The protocol is divided into 3 sections: 1) a general explanation and
illustration of situations and responses, given orally by the experimenter;
2) a practice session, with prerecorded situations, and 3) the recorded
session itself, with the prerecorded situations. The subject is presented
with the context and asked to respond with the exact words of the cue to
an imaginary peer, child, the voice on the tape.
What follows illustrates the syntactic structure and contexts employed
in the protocol. Note the syntactic identity of the Sp. fam. imperative
and 3rd sg. declarative in compra una pulsera 'She buys/Buy a bracelet'.
The reported sentence consists of the introducer phrase + que, dice que
'she says that', + embedded sentence, syntactically identical to the direct
declarative or imperative. For example, in a situation describing a trip to
New York. The subject advises a friend to go shopping at Tiffany's. The
friend doesn't know what to do or buy. The subject advises the friend
to . . . buy a bracelet (Imperative answer: COMPRA UNA PULSERA
'Buy a bracelet'). In a subsequent situation, the subject is asked to tell
someone else . . . what the friend buys (Direct answer: COMPRA UNA
PULSERA 'she buys a bracelet'). In the corresponding reported speech
situation she is requested to tell others present what the other person . . .
says she is buying, beginning with dice que: (Reported answer: DICE
Q U E COMPRA UNA PULSERA 'she says she's buying a bracelet')'. 7
340 Karen Η. Kvavik

4. Subjects

The subjects are 4 Cuban-American females, ages 25-28. They are well
educated, with a high degree of Spanish language maintenance. All left
Cuba between 1959 and 1965. Originally, all are from La Habana, except
for Subject 2, who is from Oriente.

5. Procedure

The subjects were seated in a soundproof room with Altec wall speakers
positioned directly behind them, in the Labs, for Recorded Instruction,
U. of Wisconsin. A microphone (Superscope EC-9P) was placed at
approximately 16 cm. on a table in front of them. For the explanation
and practice sessions the experimenter faced the subject, but sat beside
the subject during the experiment, so as to avoid eye contact. Subjects
were asked to repeat disfluent or erroneous responses. The total session
lasted approximately 40 minutes. The answers were recorded on an
Ampex 351 recorder, on which the experimental tape had been similarly
recorded. The laboratory technician operated the tape recorders; the
recording level was maintained at a predetermined setting throughout
the entire session.
A total of 28 sentence pairs (56 sentences) were analyzed for the 4
subjects (14 sentences per subject). The answers were digitized at
10 kHz. at a predetermined volume level. Fundamental frequency (F0)
and amplitude were analyzed by the VOCAL program at the University
of Wisconsin, Waisman Center. The direct and reported items were
paired sequentially at times and separately at other times and then given
in random order. The reported sentences begin with dice que 'he/she says
that', shown only for the first example: 8
1. Abre la ventana vs. Dice que abre la ventana 'She opens the window' vs.
'He/she says that she opens the window'.
2. Empieza a la una 'It begins at 1:00'.
3. Escribe la carta 'He/she writes the letter'.
4. Llama por telefono 'She telephones'.
5. Pasa por la casa del lobo 'She goes by the wolf's house'.
6. Puede salir a las cinco 'She can go out at 5:00'.
7. Termina el trabajo manana 'He finishes his work tomorrow'.
 Direct and reported speech prosody 341

6. Results

In the next sections, data are presented on sentence prominence, intona-

tion patterns, sentence peak high and low frequencies, and on overall
range.

7. Prominence patterns

Sentence primary stress is examined to establish comparative patterns on

the two sentence types, given the conflicting information on basic word
order and intonation in the literature. Essentially, the data show that 1)
'basic' word order and intonation patterns described in the literature
differ from the results, and 2) direct and reported sentence pattern
differences themselves require discussion. 'Sentence stress' here is the
investigator's subjective judgment of the most prominent (lexically
stressed) syllable.
One might expect sentence primary stress to fall on the same syllables
of each answer across subjects, given the protocol situations. That is, one
might expect the strongest accent either on the sentence-initial verb,
final noun, or somewhere in between (in longer sentences). In the
reported sentences the hearer awaits important information in the
embedded clause; it is reasonable that this information receives a
primary stress. In the literature, notions of sentence stress and basic
word order are inextricably linked. Linguists have used sentence stress in
their analyses of word order. However, the literature on word order does
not necessarily agree with intonation literature on what the 'basic' stress
patterns are.
It is generally thought that unmarked word order is SVO, with
primary sentence stress or accent on the rheme and/or new information
(Bolinger 1954, Contreras 1976, Gili Gaya 1961, Suner 1982).9 Intona-
tion literature views the basic sentence intonation pattern for Spanish
with several potential primary stress positions, but not necessarily with
primary stress in the rheme (usually the final position). A widely agreed
upon pattern for declarative intonation is an initial rise to the first
lexically stressed syllable, with a subsequent leveling off, and a descent
on the final stressed syllables, e.g.:
342 Karen Η. Kvavik

2)

This is also schematized as 71211 j /' in level-juncture analysis (Bolinger

1956, Haden and Matluck 1973, Navarro 1966, Stockwell et al. 1956,
Bowen and Stockwell 1960). Tone '2' or the initial stress of the sentence
is the most prominent in the schema. A disagreement with this basic
schema is found in Hockett's Course in Modern Linguistics (trans, with
Sp. analyses by Gregorio and Suärez 1971). These linguists maintain that
/1221 I / is the most 'common and neutral' pattern, with 'pitch promi-
nence' found on the final '2', even though there are two consecutive level
2's. Otherwise, a single level 2 would be the most prominent (45-47).
To summarize, for the basic (level-juncture) intonation pattern, sen-
tence stress is favored on the first stressed syllable, vs. the final stressed
syllable found in the word order literature (Bolinger 1954, Contreras
1976, Hockett 1971 [trans, by Gregorio and Suärez], Sufier 1982, etc.). 10
A third viewpoint is Fant (1980), who finds the first stressed word to
have the highest 'salience' in unmarked sentences, but this 'salience' is
not to be equated with prominence. Only marked sentences have
sentence prominence, according to Fant. It appears that 'basic' (or
favored?) word order, sentence stress, and intonation patterns need
clarification within spoken discourse contexts, so that a basis for compa-
rison is established.
For the prominence results, the subjects were not consistent as to
place of primary focus within direct and reported categories, nor were
they consistent across particular sentences of either type. For direct
sentences, sentence stress falls on the lexically stressed syllable of the
initial verb in 78.6% (22 of 28) of the sentences (e.g., esCRIbe la carta),
with the remaining 21.4% (6 of 28) on the penultimate syllable (e.g.,
escribe la CARta). Subjects 3 and 4 were totally consistent in placement
on the verb-initial position. For SI and S2, stress placement varies; it is
found sentence finally in 5 of 7 sentences for SI, and in 1 of 7 sentences
for S2.
The situation question cues indicate expected stress could vary as to
initial or final position; by taking cue presuppositions into account, there
is still a strong tendency for initial stress, even when not expected. For
example, in 5 sentences, the subject was give a WH-question where new
information would be expected in the verb accusative (noun phrase)
position, e.g., (trans.) 'WHAT does she take?' In two questions, (trans.)
'WHAT is she doing', the expected prominence would be on the initial
 Direct and reported speech prosody 343

verb in the answer. Only 42.9% (or 12 of 28 sentences) showed sentence

stress as expected, whether sentence-initial or final, leaving 57.1% of
sentence stresses in 'unexpected' positions. (8 were expected initially, 7
occurred initially; 20 were expected finally, 71.4% of the total, but only 5
occurred finally, 17.9% of the total). Subject 1 differs in performance
more than the others; she tends towards a final stress position (5 of 7,
including 1 expected initially). These results have several implications: 1)
new information does not receive sentence prominence as much as we
expect; 2) rather, old information (here considered apart from 'topic')
receives the sentence prominence; 3) or, topic receives the stress, rather
than comment (rheme). Thus, linguistic expectations of 'basic' stress
patterns for SVO word order are not borne out by the data.
Reported sentences are less consistent and more problematic regar-
ding expected primary stress placement. Across subjects, 17 of 28
sentences (60,7%) show the introducer phrase dice que receiving the
primary stress. For stress on the embedded clauses (39.3% or 11 of 28), 5
of 11 (or 17.9% of the reported total) have primary stress on the final
word of the sentence, e.g., dice que verb OBJECT. In the other 6
(21.4% of the total), primary stress falls on the initial verb portion of the
embedded sentence, e.g., dice que VERB object.
If expected stress placement is on the embedded complement ('What
is reported'), the data show it there only 39.3%, vs. the initial position
on dice (60.7%). Even if the reported di- of dice que does not always
receive a primary accent, it is still at least the second most prominent
syllable in the longer sentences. Thus, the reported introducer verb
always receives attention with the strongest or at least secondary sen-
tence accent - primary in (17 of 28 examples) and secondary in the
remaining sentences. The longer reported sentences have more options
for primary stress placement, and they are also more unpredictable as to
its placement. The added length means there is more information
content for potential highlighting. As a sidelight, some information
from the imperatives in the study (Kvavik 1984) show that the promi-
nence patterns are more rigid than those of the declaratives: 92.9% (26
of 28) have prominence on the initial verb, compared to 78.6% of the
direct and 60.7% of the reported items. Reported items then are the
most variable of the three types - still with a tendency to accent the
introducer verb. Sentence prominence schemes therefore appear to be
one of the gradient factors differentiating these three types.
The initial prominence placement may be due to discourse-initial
position, which 'brings into focus one who has been out of the picture in
344 Karen Η. Kvavik

the immediate preceding context' (Derbyshire 1977, cited in Sufter 1982:

270). It is possible that the experimental task and cues contribute to the
variability observed here, or that the task itself leads to stereotyped
utterances. However, the overall intonational and initial primary stress
schema (initial high rise and stress) tend to follow that in the Sp.
intonation literature, even though the local initial 'tune' specifics are
different (see below for more discussion). The data results differ greatly
with the word order literature and do not support the new information'
or 'rheme' placements (final position) seen in that literature (e.g.,
Contreras, Suner).

8. Intonation patterns

This section examines initial intonational 'tunes' or contours, based on

F 0 from the computerized analysis. The data show that these initial tunes
are different from the description in the intonational literature. Quanti-
tatively, the direct and reported sentences are different; qualitatively,
the introducer phrase suggests that reported sentences have a very

miter Time (Milliseconds) 220

Figure 1. Direct sentence example for Subject 1:
Termina el trabajo manana 'he finishes the work tomorrow'.
 Direct and reported speech prosody 345

predictable initial tune. The final intonations for direct and reported
declaratives, as expected, have a final fall.11
For the direct sentences, it was predicted that the highest F0 would be
on the first lexically stressed syllable, as shown by the traditional schema
(exs. 1 and 2), e.g., a rise-fall or rise-level tune. While primary sentence
stress tends to occur on the initial lexically stressed syllable, the initial
tunes themselves on the verbs are overall high rising tunes rather than
rise-falls in 85.7% (24 of 28) sentences. The initial rising tune is schema-
tized as a continuous rise from the first stressed syllable to the post-
stressed syllable: ,

3) es cri be vs. an expected es cri be

Figure 1 and Figure 2 show the direct and reported versions of termina el
trabajo manana for SI, and Figures 3 and 4 show the same sentences for
S4 for comparative purposes. Note the initial rising tune on the direct
sentences, Figure 1 and Figure 3 and S4.
There are 4 exceptions to the high rising tune (14.3%) for direct
sentences: Subject 1 (3 of 7) and Subject 3 (1 of 7). Three of the
exceptions begin with falling tunes; the fourth is very nearly level.
Whether the favored overall rise seen here is dialect specific is still an
open question. Evidence against a dialect interpretation is found in Fant

0 200 400 B00 Θ00 1000 1200 1100 1600

A04TERR Time (Milliseconds) 220

Figure 2. Reported sentence example for Subject 1:

Dice que termina el trabajo manana 'he says he'll finish the work tomorrow'
346 Karen Η. Kvavik

ζί ι Ter Time (Milliseconds) 220

Figure 3. Direct sentence example for Subject 4:
Termina el trabajo mafiana 'he finishes the work tomorrow'.

zcmterr Time (Milliseconds) 220

Figure 4. Reported sentence example for Subject 4:
Dice que termina el trabajo mafiana 'he says he'll finish the work tomorrow'.
 Direct and reported speech prosody 347

(1980), who schematizes unmarked Madrilian intonations with a similar

rising melody.12
Surprisingly, all reported sentences (100%) show the initial rise on the
introducer verb dice, instead of a fall beginning on di-to the rest of the
dice que phrase, as expected from the intonation literature. Figure 5
nicely illustrates the initial tune for Subject 2, dice que llama por
telefono.
Figures 2 and 4 show the initial tunes and also can be compared to their
corresponding direct sentences (Figures 1 and 3). The overall introducer
tune falls on the complementizer que (96.4%) before the embedded
clause begins, e.g., Figure 5. (In one exceptional example, there is a high
rising tune, which rises throughout dice que, rather than falling on que,
e.g., SI, dice que empieza a la una.)
The dice que patterns tell the hearer that what follows is 'reportage',
that 'an announcement is forthcoming - reported speech follows'. Note
that high and rising F 0 signals 'disturbance in discourse flow' according to
Menn and Boyce (1982). A n initial rise is strongly hearer-oriented an
cues 'the hearer to the concern, interest, etc. of the speaker' (Bolinger
1982: 525). The literature predicts a final rise in the main clause before
an embedded complement, consistent with the data here, if que were not

C3-1LI ah Time (Milliseconds) 220

Figure 5. Reported sentence example for Subject 2:

Dice que llama por telifono 'she says she telephones'.
348 Karen Η. Kvavik

included (Alcina Franch and Blecua (1975: 463ff.; Navarro Tomas 1966:
78ff.). As noted above, there is no specific information on reported
intonations, other than that there are independent intonations (Seco), or
that the introducer phrase is almost on the same plane as the embedded
clause (Verdin 1970: 65).13

9. Sentence length

Sentence length in milliseconds is given in Table 1.

A Β C D Ε F G Η
S. Sent. Direct Reported D i f f . Mean/9.d. Mean/a.d. Mean/a.d
(R - D) D i r e c t Reported Diff.

1 1. ABR 694 1228 334 1122.14 1553.14 205.89

2. EMP 981 1526 545 195.57 243-68 125.54
3. ESC 1057 1333 276
1. LLA 997 1 401 404
5. PAS 1297 1841 544
6. PUE 1186 1772 586
7. TER 11)13 1771 328

2 1. ABR 959 1317 358 1218.00 1612.57 394.57

2. EMP 1025 1387 362 196.95 253.07 69.68
3. ESC 1152 1545 393
»1. LLA 1171 1 449 278
5. PAS 1 484 1959 475
6. PUE 1311 1737 426
7. TER 1424 1894 470

3 1. ABR 854 1229 375 1037 .57 1410. 86 3 7 3 .. 2 9

2. EMP 876 1291 415 1 6 2 ,. 8 7 201 . 1 7 7 1 .. 9 0
3. ESC 912 1 225 313
4. LLA 1058 1312 254
5. PAS 1245 1717 472
6. PUE 1086 1456 370
7. TER 1232 1646 414

4 1 . ABR 839 1265 426 1 0 6 9 ., 2 9 1 4 9 0 . 71 421 . 4 3

2. EMP 907 1227 320 2 0 0 ..94 271 . 8 7 1 0 6 . 43
3. ESC 912 1260 348
4. LLA 1002 1426 424
5. PAS 1250 1856 606
6. PUE 1234 1552 318
7. TER 1341 1849 508

Total: 1111. 75 1 5 1 6 . 82 4 0 5 - 07
191 . 9 2 2)42.18 93.71
Table 1. Direct and Reported Sentence Length (Msecs.).
Cols. Α - D represent subjects, sentences, and the direct and reported sentence values.
Col. Ε shows the difference between the two types (Reported minus Direct). Cols. F - H
are subject means and (sample) standard deviations for Cols. C-E. Col. I is probability
for differences between the two types within the particular subject (Col. E).
 Direct and reported speech prosody 349

The average difference between the two sentence types (Col. E) ranges
from 373.3 ms. (S3) to 431.0 ms. (SI). As expected, reported sentences
are longer; but this is from their greater segmental length (3 syllables
longer), unless reported sentences were found to have a slower rate.
In order to check whether the reported sentences were spoken at a
different rate, the noun clause complement length of the reported
sentence (identical to the direct sentence) was compared to the syntacti-
cally identical direct speech sentence. There is no statistical difference
between the direct sentences and segmented reported clause lengths.
The timing difference is owed to the introducer phrase - at least in this
data. The question of speech rate differences for reported speech is open
to investigation. It is reasonable to hypothesize that reported speech
might have a different speech rate.14

A B C D E F G Η I
S. S e n t . Direct Reported D l f f . Mean/a.d. Mean/s.d. Mean/s.d. Prob.
(D - R) D i r e c t Reported Dlff.
1 1 . ABR 263 21)1) 19 256 .86 280 .86 -21.00
2. EMP 255 31)2 -87 21 . m 34 .U8 35.18
3- ESC 236 261 -25
1). LLA 227 2 67 -1)0
5. PAS 256 254 2
6. PUE 292 293 -1
7. TER 269 305 -36

2 1 . ABR 300 325 -25 290 .29 31)5, . 0 0 -5D.71

2. EMP 322 382 -60 2 D ,. 8 8 3 1 ·. 0 7 23.88
3. ESC 297 366 -69
I). LLA 246 333 -87
5. PAS 278 320 -D2
6. PUE 309 382 -73
7. TER 280 307 -27

3 1 . ABR 248 269 -21 2 5 7 .,71 295.,57 -37.86

2. EMP 267 328 -61 1 5 . 32 29. 02 21.54
3- ESC 249 316 -67
D. LLA 239 276 -37
5. PAS 248 253 -5
6. PUE 277 318 -D1
7. TER 2 76 309 -33

if 1 . ABR 237 267 -30 21)1). 71 2 8 7 . 29 -1)2.57

2. EMP 248 319 -71 1 6.85 19. 89 26.26
3. ESC 256 263 -7
1|. LLA 215 282 -67
5. PAS 269 281 -12
6. PUE 210 300 -60
7. TER 248 299 -51

Total: 2 6 2 . 39 3 0 2 . 18 -39.79
25.51 37.63 27.99

Table 2. Direct and Reported Sentence Peak Frequency (Hz.).

Cols. Α - D show subjects, sentences, and direct and reported sentences values. Col. Ε is
the difference between the two types (Direct minus Reported). Cols. F-H give subject
means and (sample standard deviations for Cols. C-E. Col. I gives probabilities for
differences between the two types within the particular subject (Col. E).
350 Karen Η. Kvavik

10. Sentence high frequency

The peak F0 was measured at the highest vocalic or sonorant consonant

frequency. First, it is a way of addressing the claim that reported speech
is less 'vivid' with more 'melodic uniformity' through degree of pitch
excursion. Second, the peak F0 can then be used to calculate an overall
sentence range measurement.
Table 2 shows the direct and reported sentence peak frequencies.
Cols. Α - D are subject and individual sentence frequencies; Col. E. is the
difference between the two types. It was assumed on the basis of the
literature that the peak highs of the direct sentences would be higher,
e.g., indicating wider range or more 'vividness' for the listener in direct
speech (cf. Verdin and Dominguez); this would be indicated by positive
values ( + ) in Col. Ε (Direct minus Reported values).
Peak F 0 in direct speech usually occurs on the post-stressed syllable of
the initial verb, e.g., in -be of escribe. There are 5 exceptions for direct
sentence; 4 are for Subject 1 with the stressed vowel of the sentence-final
word, and 1 for Subject 3, which occurs on the stressed syllable of the
initial verb. There are 2 exceptions for the reported items, both items for
SI, again the exception.
The reported sentence results are contrary to expectations. The
negative (—) values in Col. Ε indicate higher peak F c for reported,
rather than direct sentences, except for Si's direct sentences, Abre la
ventana (+19 Hz) and Pasa por la casa del lobo (+2.0 Hz). Graph 1
compares the direct and reported mean peak values for each subject.
The reported values are on the right:
T-tests on the differences within each subject (Col. E) are highly
significant for Subjects 2, 3 (p < .001), ρ < .005, and 4 (p < .01), with no
significant difference for Subject 1. The average differences for each
subject (Col. H) also indicate that the reported have higher peak F c than
direct speech sentences and suggest that the reported sentences are more
'vivid' (see Figure 6). Even the data for SI, with no significant diffe-
rence, do not support a claim of 'reduced intonation' for reported
items.15
It is claimed that longer sentences begin with higher F0 than do shorter
sentences, indicating 'preplanning' (Cooper and Sorensen 1981: 36f.,
Geifer et al. 1983). This might indicate the higher peak F0 for the
reported data is not necessarily attributable to discourse differences or to
any information role F0 might play. On the other hand, Menn and Boyce
(1982) also found higher F0 in longer sentences in their study, but they
 Direct and reported speech prosody 351

PEAK FREQUENCY
Direct v s . R e p o r t e d S e n t e n c e s
350

300

250

Ν
w
>- 200
υ
ζ
LJ
g 150
LJ
tr

100

50

0
Reported
SUBJECT MEANS
[V\l S2 E Z 3 S3 E23 S4

Graph 1. Peak Frequency, Direct vs. Reported Sentences. The graph illustrates subject
means for the direct (left) and reported sentences (right).

maintain that shorter sentences have less information content. Pierre-

humbert (1980: 133, 144) finds that longer sentences only tend to have
higher F 0 , but attributes this to pragmatic or expressive factors, where
the first high peak is a 'free choice'. Liberman and Pierrehumbert (1984)
suggest that stylistic, discourse, and syntactic factors may be involved
with higher sentence initial frequency and do not accept a 'preplanning'
argument. In a study of English and Hebrew, Berkovits (1984) finds
starting values for unfinished sentences no higher than for finished
sentences, where they were predicted from Cooper and Sorensen's
theory. In O'Shaughnessy and Allen's 1983 data, a study of F 0 and
modality, two-clause sentences showed an initial higher peak and lesser
emphasis on ensuing words, due to 'anticipation' or preplanning
(1168f.). Note that their matrix sentence was a reported introducer
phrase 'Bruce said . . . ' , and the embedded clause had the modals they
were testing. Bolstering the pragmatic/information content role of initial
F 0 is evidence from Kvavik (1984) for imperatives vs. declaratives. This
study shows a tendency toward higher initial F 0 for imperatives than
352 Karen Η. Kvavik

their corresponding declaratives, which are segmentally identical and

indicates that pragmatic and expressive factors are at play with the
higher F0; a preplanning argument because of length therefore cannot be
the only factor involved. This discussion of initial F0 should not obscure
the fact that the data results do not support the literature's claim of
'uniformity' or 'less expressivity' in reported sentences.

11. Sentence low frequencies

Intonation literature indicates that the lowest sentence frequency, sen-
tence finally, is a stable point and does not 'vary' significantly (Boyce
and Menn 1979, Liberman and Pierrehumbert 1984). Table 3 gives the
low frequencies for the data:

A Β C D Ε F G Η
S. Sent. Direct Reported D i f f . Mean/s.d. Mean/s.d. Mean/s.d
(D - R) Direct Reported Diff.

1 1. ABR 1 64 1 61 3 143.14 137.57 5.57

2. EKP 137 140 -3 11 . 8 9 11 . 6 5 7.61
3. ESC mi 131 10
4. LLA 131 129 2
5. PAS 151 1 42 9
6. PUE 147 128 19
7. TER 131 132 -1

2 1 . ABR 1 85 1 80 5 183.29 177.43 5.86

2. EMP 188 157 31 8.26 10.75 14.36
3. ESC 183 1 82 1
4. LLA 187 193 -6
5. PAS 180 177 3
6. PUE 167 1 78 -11
7. TER 193 175 18

3 1 . ABR 171 160 11 174.43 161.86 12.57

2. EMP 183 155 28 7.87 6.18 11 . 3 9
3. ESC 1 78 167 1 1
4. LLA 173 1 60 13
5. PAS 165 157 8
6. PUE 185 161 24
7. TER 166 173 -7

1 1. ABR 180 169 1 1 157.86 157.50 -0.67

2. EMP 156 154 2 11.14 8.50 1 0.82
3. ESC 156 1 49 7
4. LLA 149 158 -9
5. PAS 1 48 166 -18
6. PUE 164
7. TER 1 52 1 49 3

Total: 164.68 158.63 6.07

18.29 17.27 11.63
Table 3. Direct and Reported Sentence Low Frequency (Hz.)·
Cols. Α - D are subjects, sentences, and direct and reported sentence values. Col. Ε is the
difference between the two types (Direct minus Reported). Cols. F-H are subject means
and (sample) standard deviations for Cols. C-E. Col. I gives probabilities for differences
between the two types within the particular subject (Col. E).
 Direct and reported speech prosody 353

Cols. A and D of Table 3 show the subject and individual sentence low
frequencies for direct and reported items. These low frequencies are
found in the final sentence vowel. The direct items are based on the 28
sentences, the reported on 27, because S4 has a missing value for
sentence 6, dice que puede salir a las cinco. Only S3 shows a significant
difference between her two sentence types (p < .05). Moreover, the
overall average difference across subjects is slightly higher for the
reported items (6.07 Hz). The data merely hint that reported sentences
may end in a slightly higher frequency than do the direct sentences, that
the lower range may be shifted upwards, along with the higher initial F0.
This may be a subtle cue in pragmatic differences, or an artifact of this
particular investigation from too few subjects and sentences, but should
be kept in mind for future work.

12. Semitone range

Overall sentence (high to low) range as a potential contribution to

discourse signalling is examined in this section. The sentence high to low
Hz. frequencies were converted to semitones; this permits comparison
with range data in Navarro (1966). Table 4 shows semitone data for the
two sentence types:
The subject and individual direct sentence ranges are seen in Cols.
Α-D. (The reader will note a missing data value for S4, sentence 6.) The
direct items (Col. C) range from a low of 4.75 (S2) to a high of 12.46
semitones (SI). Col. F shows direct means for each subject, which range
from 7.59 - 10.12 semitones, with an overall means of 8.15 semitones.
The averages are lower than those of Navarro 1966 for declaratives:
12) Range in Semitones
Navarro: 8
Present Study: 7.6 - 10.1 (Av. 8.15) Direct
10.3 - 12.3 (Av. 11.13) Reported

Reported sentence highest to lowest values across sentences (Col. D)

range from 7.2 (SI) - 1 5 . 5 semitones (also SI). Reported sentence means
for each subject (Col. G) range from 10.26 (S4) to 12.30 (SI), with an
overall mean of 11.13 semitones, wider than Navarro's declaratives (See
ex. 12 above). Differences with Navarro may be due to elicitation
(reading), style, or possibly dialect differences.
354 Karen Η. Kvavik

A B C D E F G H I
S. Sent. Direct Reported Diff. Mean/s.d. Mean/s.d. Mean/s.d. Prob.
(D - R) Direct Reported Dlff.

1 . ABR 8 . 18 7 . 20 0. 98 10.12 12.30 -2.18

2. EMP 1 0 . 76 1 5 . 46 -K. 71 1 .61 2.89 1 .80
3- ESC 8 . 92 1 1 . 93 -3. 02
K. LLA 9 . 52 1 2 . 59 -3. 08
5. PAS 9. 1 Κ 1 0 . 07 -0. 93
6. PUE 1 1 . 88 I t . 31 -2. K6
7. TER 1 2 . 16 1 It. 5 0 -2. OK

1 . ABR 8.37 1 0.23 -1 .86 7.92 11 . 4 8 -3 .56

2 . EMP 9.32 15.39 -6.08 1.93 2.20 1,·Κ3
3 - ESC 8.38 12.09 -3.71
LLA 4. 75 9.ΚΚ -Κ.70
5 . PAS 7.52 1 0.25 -2.73
6 . PUE 10.65 1 3.22 -2.57
7 . TER 6.KU 9.73 -3.29

1 . ABR 6.KK 8.99 -2.56 6.75 1 0 . 36 - 3 .. 6 2

2. EMP 6.5K 1 2.98 -6.ΚΚ 1.06 1 .66 2 .. 0 2
3. ESC 5.81 11 .OK -5.23
K. LLA 5.59 9.ΚΚ -3.8Κ
5. PAS 7.05 8.26 -1 .21
6. PUE 6.99 11 . 7 8 -Κ.80
7. TER 8.80 10.OK -1.2Κ

1 . ABR Κ. 7 6 7.92 -3.15 7.59 10.26 - 2 . 51

2. EMP 8.03 1 2 . 61 -Κ.58 1.83 1 .78 2 . 13
3. ESC 8.57 9 . 8K -1 .26
K. LLA 6.35 10.03 -3.68
5. PAS 10.3K 9.11 1.23
6. PUE 6.59
7. TER 8.Κ7 12.06 -3-58

Total: 8.15 11.13 -2.98

2.02 2.21 1.86
Table 4. Direct and Reported Sentence Range (Semitones).
Cols. Α - D are subjects, sentences, and direct and reported sentence values. Col. Ε is the
difference between the two types. Cols F-H show subject means and (sample) standard
deviations for Cols. C-E. Col. I gives probabilities for differences between the two types
within the particular subject (Col. E).

Finally, average differences between the two sentence types (Col. E,

Direct minus Reported values) range from 2.51 (S4) to 3.62 (S3). A one
sample t-test shows that the individual subject mean differences (Col H)
are significant, (p < .05) or higher.16 Graph 2 illustrates the subject mean
sentence ranges for the two sentence types (Cols. F, G); the reported
values are on the right.
Given comments by Verdin and Dominguez on melodic uniformity in
reported speech, and that the introducer phrase is supposedly on the
same melodic plane as the rest of the sentence (Verdin), it is surprising
that overall range is greater for the reported declaratives. One then is
curious as to why the impressions documented in the literature have not
been substantiated in the present investigation. Reported style is con-
ceived of as a step removed from direct speech, as if direct speech were
 Direct and reported speech prosody 355

SENTENCE RANGE
Direct vs. Reported Sentences

Reported

SUBJECT MEANS
E S S2 E22 S3 S23 S4

Graph 2. Sentence Range, Direct vs. Reported Sentences. The graph illustrates subject
means for the direct (left) and reported sentences (right).

more 'real' and therefore more vivid. Hence, reported style is somehow
more remote, more distant, more muted. These descriptions may well be
literary conventions.

13. Conclusion

Notions about prosodic differences in (written) direct and reported

speech styles and sentence types found in grammatical and stylistic
literature have not been confirmed in the investigation, at least for the
dimensions measured. The data do not support a sentence-final primary
stress position (rheme and/or new information position), but rather a
sentence-initial position in both sentence types. However, direct senten-
ces (and especially imperatives) show a gradiently stronger tendency to
initial sentence stress. Moreover, the initial sentence tune differs from
356 Karen Η. Kvavik

that described in Sp. intonation literature; there is a strong tendency to

have a high rising initial tune, especially on the reported sentence
introducer dice. The reported introducer phrase overall tends to fall on
the que complementizer. Overall sentence range and sentence peak F0
are higher for reported items, rather than lower, contrary to what claims
of 'uniformity' and 'less vividness' might presuppose. It is plausible that
writers and critics rely on intonative stereotypes in text interpretation,
akin to poetic devices.
A number of factors undoubtedly contribute to the differences obser-
ved. For example, the high F0, prominence placement, and risefall tunes
on the introducer phrase + que together contribute to the strong
impression of a forthcoming announcement of 'what will be said'.
Comments by Menn and Boyce 1982 are pertinent to the results of this
study. They examined the role of F0 peak in free conversation. Topic
change has the highest pitch among a variety of discourse types, with
verification questions showing an unpredicted increase. While pitch can
signal 'newness', they conclude it also signals 'amount of disturbance in
the flow of discourse (e.g., 'deflection in discourse flow + degree of
arousal or tension), as well as assertion of authority (370f.). In reported
sentences, the high F0 in the introducer phrase indicates disturbance in
discourse flow by announcing to the hearer that some reportage is
forthcoming. The higher pitch and rise-fall tunes in the introducer are
the speaker's more forceful way of capturing attention, that she is the
information giver. The data in this study confirm Menn and Boyce's with
controlled sentences and also with different sentence types than they
studied.
An interesting outcome are the individual style differences, especially
SI, who is the most anomalous, but nevertheless performs more accor-
ding to linguistic expectations (e.g., in the literature). This implies
stereotypic intonations, already suggested for text interpretation, may be
overlayed or imposed on sentences. In other words, when linguists make
up examples, stereotypic prosodic patterns may be forced on the data. I
am not saying such stereotypic patterns do not exist, rather that they may
not characterize spoken speech as believed. Amplitude differences need
examination; perhaps reputed differences between the types lie here.
The investigation suffers from omissions in research design. More sub-
jects and sentences would be desirable. As a modest investigation, the
results do elucidate factors contributing to gradient differences in direct
and reported sentence prosody, as well as sentence stress and intonations
differing from those in the literature.
 Direct and reported speech prosody 357

Acknowledgments

The acoustical work was facilitated by the Speech Motor Control Labs.,
Waisman Center, University of Wisconsin, supported by NINCDS #NS-
13274. Special thanks to James H. Abbs, Dir., SMCL, Cliff Gillman,
Director, Waisman Computing, and Jean Lentz, and also to Birute
Ciplijauskaite and Zunilda Gertel. The ms. was prepared while a post-
doctoral fellow at the Center for Research in Human Learning, Univer-
sity of Minnesota, supported by NICHHD #T-32-HD07151.

Notes
1. Sp. direct and indirect speech acts have been studied by Haverkate 1979; direct speech
acts are unmistakably impositive, e.g., 'Wash your hands'. Indirect speech acts are
impositive because of subtle factors; besides lexical semantic ones, e.g., contextual
factors such as setting, speaker-hearer social relations, and (it is thought) intonation.
2. Pedagogically oriented texts usually refer to reported speech in the context of the verb
decir 'to say or tell', with two glosses, 'inform, report' or 'command'. The 'inform' gloss
has an indicative verb in the complement, while the 'command' gloss has a subjunctive,
indicating impositivity:
Dice que cantol-a (ind.) 'He/she says that I/he, she sing(s)'.
Dice que cante (subj.) 'He/she tells me/him/her to sing'.
3. Also, que (untranslatable) may be retained in stories or poetry: Que Caperucita volvia
al bosque, y que el lobo ... 'Little Red Ridinghood was returning to the woods and the
wolf . . .* (289). It may be omitted after rogar 'to beg' and temer 'to fear' in 'educated
speech' (289). Keniston 1964 says such omission is infrequent, except after rogar (272).
These notes about social styles have never been followed up in any detail.
4. There is a good historical section and introduction in Verdin (Chs. I and II). Spanish
tradition has little to say about these styles until Todemann 1930, which focusses on
free indirect styl in contrast to direct and indirect styles.
5. Verdin has ample sections on other characteristics of reported sentences: transposition
of verb tense from present to past in the complement (53ff.); changes in adverbs of
place and time, pronouns, possessive and demonstrative adjectives (57-59), e.g.,
Ramön dijo: Estoy aqui 'Ram0n said, "I'm here"' (direct); Ramön dijo que estaba alii
'Ramön said that he was there' (indirect, with transposition of tense and adverbs of
place).
6. An historic panorama of indirect free discourse is given in Chs. II and III.
7. The imperative examples are discussed in Kvavik 1984. Some specific examples from
the protocol are:
1. Hace un buen dia de verano en Wisconsin. De verdad estä haciendo calor. Por
favor, ^quieres mandar a tu compaiiero . . . abrir la ventana?
A B R E LA VENTANA.
21. Hace un buen dia de verano en Madison. En efecto, hay mäs calor cada minuto.
Para que entre aire fresco, una persona hace algo con la ventana. ^Que hace?
A B R E LA VENTANA.
358 Karen Η. Kvavik

23. En el mismo mitin: En este momento la persona de en frente abre la ventana. Un

compafiero te comenta sobre las acciones de esta persona. Por favor, empezando con
'dice que' . . . ^puedes informarme . . . lo que dice sobre la persona de enfrente?
DICE Q U E A B R E LA VENTANA.
8. The contextual situations are about 1) a summer day in a stifling classroom; 2) a
discussion about a class in the timetable; 3) a friend who has been encouraged to write
her parents; 4) and 5) a somewhat perverse story about Little Red Ridinghood, where
her mother describes to a friend what her daughter does; 6) a friend who can go out at
5:00; 7) an absent classmate and the necessity of explaining to the professor when her
homework will be finished. Not all of the contexts are about the same persons. The
reported version dice que termina el trabajo manana 'he says he will finish the work
tomorrow' is not about a delinquent student, rather reporting what the shoe repairman
says.
9. Sutler 1982 reviews a number of points of view with regard to Spanish word order and
the separateness of given/new and theme/rheme with helpful examples. Contreras 1976
equates rheme with new information.
10. To be fair, while Sufier says that sentence prominence ('information focus') falls within
'rheme under conditions of neutral stress and information' (11), it does not have to (16,
022).
11. Imperatives, as expected, also have a final fall, except for one sentence, which ends
with a slight rise.
12. Fant also says that their shape was confirmed acoustically (1980: 13f.).
13. Garro and Parker 1983 find a rise before a restricted relative clause and a fall before a
nonrestricted relative clause in Spanish, both marked by que, different from the que
complement marker.
14. Segmenting dice que from the rest of the reported utterance is not without problems. In
Sp., two identical unstressed vowels coalesce in rapid speech, e.g., que empieza 'that it
begins' usually is [kempiesa]. The word que ([ke or ki] before vowels) was segmented
as carefully as possible; the semivocalization is very brief: [kiabre] 'que abre'.
15. The sentences dice que empieza a la una consistently show the highest differences
across subjects. Since the protocol item involves a task of reiterating something already
reported to friends, e.g., 'what the timetable says', the item could be eliciting
contrastive intonation. Removing that item lowers probability slightly to ρ < .025 for
Subjects 2, 3, and 4, also highly significant.
16. If sentence 2, dice que empieza a la una, were removed, for reasons already given, the
significance results are not different (p < .05) for Subjects 2 and 3; for Subject 4
differences are no longer significant at the .05 level (.10 < ρ < .05). Sentence 2
particularly skewed the results for S4. Data trends for reported sentences still counter
the results predicted by the literature, e.g., that there is reduced overall range.

References

Alcina Franch, Juan, and Jose Manuel Blecua. 1975. Gramätica espanola. Barcelona:
Ariel.
Berkovits, Röchele. 1984. Duration and fundamental frequency in sentence-final intona-
tion. Journal of Phonetics 12, 255-265.
 Direct and reported speech prosody 359

Bolinger, Dwight. 1954. English prosodic stress and Spanish sentence order. Hispania 37,
152-156.
Bolinger, Dwight. 1961. Three analogies. Hispania 44, 133-136.
Bolinger, Dwight. 1961. Generality, gradience, and the all-or-none. The Hague: Mouton.
Bolinger, Dwight. 1982. Nondeclaratives from an intonational standpoint. In: Robinson
Schneider, Kevin Truite, and Robert Chametzky, ed. 1982. Papers from the parasession
on declaratives. Chicago: Chicago Linguistic Society, pp. 1-22.
Bolinger, Dwight. 1983. Intonation and gesture. American Speech 58, 156-174.
Bowen, J. Donald, and Robert P. Stockwell. 1960. Patterns of Spanish pronunciation: A
drill book. Chicago: University of Chicago Press.
Boyce, Suzanne and Lise Menn. 1979. Peaks vary, end points don't. Proceedings of the
1979 Berkeley Linguistics Society. Berkeley, CA: Linguistics Dept, University of Cali-
fornia at Berkeley, pp. 373-384.
Contreras, Heles. 1976. A theory of word order with special reference to Spanish. Amster-
dam: North-Holland Publishing Company.
Cooper, William E., and John M. Sorensen. 1981. Fundamental frequency in sentence
production. New York: Springer-Verlag.
Criado de Val, M. 1976. Gramätica espanola y commentario de textos. 7a. ed. Madrid:
Editorial S.A.E.T.A.
Derbyshire, D. C. 1977. Word order universale and the existence of OVS languages.
Linguistic Inquiry 8, 590-599.
Dominguez de Rodriguez-Pasques, Petrona. 1971. El discurso indirecto libre en la novela
argentina. Rio Grande do Sul: Pontificia Universidade Catölica do Rio Grande do Sul,
Instituto de Letras e Artes, Curso de Pös-graduagäo em Lingüistica e Letras.
Fant, Lars. 1980. Prosody versus intonation. Reports from Uppsala University Department
of Linguistics 6, 1-124.
Garro, Luisa, and Frank Parker. 1983. Relative clauses in Spanish: some suprasegmental
characteristics. Journal of Phonetics 11, 85-89.
Geifer, Carole E., Katherine S. Harris, Ren6 Collier, and Thomas Baer. 1983. Fundamen-
tal frequency declination over changes in length. Paper presented at the 105th meeting,
Acoustical Society of America. Abs. in Journal of the Acoustical Society of America.
Suppl. 1, Vol. 73, S87.
Gili Gaya, Samuel. 1970. Curso superior de sintaxis espanola. 9a. ed. Barcelona: Biblio-
graf, S.A.
Haden, Ernest F., and Matluck, Joseph H. 1973. El habla culta de La Habana: Anälisis
fonolögico preliminar. Anuario de Letras 11, 5-33.
Haverkate, Henk. 1979. Impositive sentences in Spanish. Amsterdam: North-Holland
Publishing Company.
Hockett, Charles F. 1971. Curso de lingüistica moderna. Trans, and adapt, by Emma
Gregorio and Jorge Albert Suärez. Buenos Aires: Editorial Universitaria de Buenos
Aires.
Keniston, Hayward. 1964 [1937], Spanish syntax list. New York: Holt, Rinehart, and
Winston.
Kvavik, Karen H. To appear. Is there a Spanish imperative intonation? In: Robert
Hammond, Melvyn Resnick, and John Jensen, eds. Adas del VIII Simposio de Dialecto-
logia del Caribe Hispänico, 1984.
Ladd, D. Robert, Jr. 1978. Stylized intonation. Language 54, 517-540.
Lenz, Rodolfo. 1944. La oraciön y sus partes. Santiago: Editorial Nascimento.
360 Karen Η. Kvavik

Liberman, Mark and Janet Pierrehumbert. 1984. Intonational invariance under changes in
pitch range and length. In: Mark Aronoff and Richard T. Oehrle, eds. Language and
sound structure: Studies in phonology presented to Morris Halle by his teacher and
students. Cambridge, MA: The M.I.T. Press, pp. 157-233.
Menn, Lise, and Suzanne Boyce. 1982. Fundamental frequency and discourse structure.
Language and Speech 25: 4, 341-383.
Navarro Tomäs, Tomäs. 1966. Manual de entonaciön espanola. 3a. ed. Μέχίοο, D. F.:
Colecciön Mälaga, S.A.
O'Shaughnessy, Douglas, and Jonathan Allen. 1983. Linguistic modality effects on funda-
mental frequency in speech. Journal of the Acoustic Society of America 74, 1155-1171.
Pierrehumbert, Janet Breckenridge. 1980. The phonology and phonetics of English intona-
tion. Unpublished diss., Massachusetts Institute of Technology.
Real Academia Espanola. 1975. Esbozo de una nueva gramätica de la lengua espanola.
Madrid: Espasa-Calpe, S.A.
Seco, Rafael. 1975. Manual de gramätica espanola. 10a, ed. Madrid: Aguilar.
Stockwell, Robert P.; J. Donald Bowen; and Ismael Silva-Fuenzalida. 1956. Spanish
juncture and intonation. Language 32, 641-665. Rpt. in: M. Joos, ed. 1963. Readings in
Linguistics. 3rd ed. New York: American Council of Learned Societies, pp. 406-418.
Suiier, Margarita. 1982. Syntax and semantics of Spanish presentational sentence-types.
Washington, D.C.: Georgetown University Press.
Todemann, F. 1930. Die erlebte Rede im Spanischen. Romanische Forschungen 44.
Verdin, Diaz, Guillermo. 1970. Introducciön al estilo indirecto libre en espanol. Madrid:
Consejo Superior de Investigaciones Cientificas.
Index of subjects

Abkhaz 121, 129, 131 complement sentence 20, 48, 101, 166
Abkhaz, written 130 complementizer 29, 66, 72, 79,85,87,101,
accent 341 123, 125, 130, 148 f., 153, 163 f., 179,
accusative with infinitive 19 184f., 236, 242, 334, 347, 356
African languages 148, 155 complementizer, suppressed 163
agreement 47, 157, 181, 284 constructed dialogue 311, 312, 314, 327,
Altaic languages 166 330
ambiguity 3, 117, 130, 154, 163, 175, 183, content reporting 192
213, 250, 295 context 32, 82, 85, 115, 132, 145, 251
American 311 ff. context, spatio-temporal 221
Amharic 39 context of utterance 167
anaphora 277 contextual information 117
anaphoric 173 control, rules of 65
anaphoric pronouns 57 conversational discourse 312
anaphoric reference 188 f. conversational narrative 323
Andoke 44 coreference 42, 82, 156
aorist 123 Creoles 158
aspect 37, 161 Cuiva 41, 44
Athapaskan 47
authenticity 180 Danish 22, 219f., 248, 251
Avar 134 f., 137, 141 Danish grammar 233
Avar, literary 138 declarative sentences 106, 108, 353
de dicto 1, 3f., 259
back-shifting 14-16, 78, 96, 110f., 115 deictic adjustments 5
Bantu languages 116 deictic center 18
binding 36 deictic changes 16
Bulgarian 22 deictic complexity 238
deictic demonstration 172
Chamling 145, 147, 154, 157, 248 deictic elements 181
Chechen 121, 134, 140 deictic field 291
Chinese 32 ff. deictic field, imaginary 194
citation-suffix 137 f. deictic origins 230
Cofan 40, 44 deictic pronoun shift 239
commitment 223, 225 deictics 34, 145, 156
common places 282 deictic shift 183f., 236, 238f., 243, 249
communication 264 deictic switch 151, 156, 167, 174, 177
communicative functions 30, 42f., 213 deictic terms 25
communicative situation 38 deixis 17, 78, 132, 237
complement 47, 53, 57, 62, 165, 172, 335, deixis, spatial 139
337 demonstrative 201, 207
complementation 164, 168 de re 1, 3 f.
complement clause 16 dialogue 172
complement marker 336 dialogue, constructed 145, 242 ff.
complement pronouns 60 dialogues, fictitious 279
362 Index of Subjects

discourse 179, 197, 283, 334 Guajiro 40, 44

discourse cohesion 194 Guanano 40, 44
discourse, literary 197
discourse shifting 64 hearsay 22, 35, 37, 94, 312
Hindi 153
echo-questions 278 historical present 241, 324
elements of speech 171 historical subjunctive 232
embedding 146, 167, 259 honorific 166
embedding, multiple 82, 95 honorific forms 172
emotives 202 honorifics 172
English 14-16, 19, 21, 32, 34, 36f., 78, 96, honorific speech 174
110, 117, 126, 156, 187, 201, 230, 316, honorific systems 167
338 Hopi 22, 155
English stories 321 Hungarian 14, 201 ff., 211, 255ff., 278f.
erlebte Rede 7, 24, 301 Hungarian texts 271
evidentiality 22, 41 f. hypothetical direct speech 225
expressive elements 246, 338
expressive factors 351 illocutionary force 148, 166, 220, 225
expressive function 287 imperative 79, 86, 111, 337
extended predicate 182, 184 indexical gesture 257
eyewitness reports 78, 93 indirect question 3
infinitive 131
factive presuppositions 205 inner-speech 157, 258, 297
factives 202 inquit-position 272, 274, 298
factivity 202-204, 208, 216 interjections 211, 287
faithfulness 223, 225 interrogative 109, 337
Farsi 153 intonation 39,173,247, 256, 277, 331,334,
formulaic 328 336, 338
formulaic introducer 321 intonation, declarative 339, 341
formulaic sentences 291 intonation, ironic 286
formulas 328 intonation pattern 275, 342, 344
fourth person pronoun 59 intonation, reduced 350
free indirect discourse 292 intonations, reported 337, 348 f.
free indirect speech 42, 95 intonative stereotypes 356
free indirect style 278, 284 involvement 312, 324, 327, 330
French 14, 22, 201, 255 ff., 262 f., 272, 278
Japanese 14, 22, 161 ff., 179ff., 185, 187
genre 163, 279, 330
Georgian 121, 128, 131 Kartvelian 125
German 14f., 19-22, 156-158, 201, 222, Kera 148, 155 f.
235, 240, 251, 286 Kilivila 252
grammatical adaptations 140 Korean 166
grammatical changes 221 Kwakiutl 22
grammatical switches 174
Greek 224, 311 ff. language, spoken 235
Greek, Classical 10, 19f., 125, 141, 163, Latin 15, 19, 21, 141, 248, 257, 262, 298
177, 219, 222, 240, 243, 248, 251 lexical meaning 260
Greek stories 325 lexical switches 177
 Index of Subjects 363

lexicon 51, 65, 298 orality 163

literacy 163, 247, 251 oral narratives 322, 325
literal quote 224 oral reporting styles 247
literary conventions 355 oral storytelling 313
literary discourse 312 oratio obliqua 2, 100, 123
literary form 296, 300 oratio recta 2, 99, 123
literary functions 196 oratio tecta 233
literary language 134, 136, 139, 248
literary narrative 336 Paez 39f., 44
literary representations 255 paraphrase 40, 91 f., 271
literary style 196, 230, 320 performative verb 334
literary texts 262 phatic 171
literary tradition 161, 197 Pidgins 158
live speech 255, 287, 302 point of view 53, 60, 63, 65, 73, 176, 181,
logophoric pronouns 21, 26 183, 196 f.
politeness 168, 274, 292
Mambila 33, 43
presuppositions 205, 207, 216
masdar 123
print 323
meta-language 24
pronominalization 30, 35, 44
metalinguistic taxonomies 25
pronominal reference 79, 125, 129, 138,
metaphor 255, 267
141
metaphorical systems 338
pronominal shift 249
modal particles 211 f., 215
pronominal switches 138, 141, 150, 152 f.
monologues interieurs 258
pronoun 22, 30, 112
mood 86-88, 96, 113, 166, 171, 236, 284
pronoun, anaphoric 89
mood, imperative 111
pronoun choice 81
mood, interrogative 171
pronoun deixis 86, 93, 95
movement 70
pronoun, demonstrative 115
narration 162, 276 pronoun, fourth person 59
narrative 40, 137, 150, 259, 278, 284, 294, pronoun hierarchy 60
296, 311-313, 317 pronoun, interrogative 118
narrative, oral 322, 325 pronoun, personal 163, 170, 221, 236, 238
narrative style 161, 302 pronoun, possessive 115
narrator 137, 157, 192, 197, 225, 229, 255, pronoun, relative 107
319 prosodic features 256, 258, 337
Navajo 22, 39, 73 prosodic patterns 356
Nepali 145, 151 f., 154 prosody 320, 331, 333
Newari 145, 154, 156 pseudo-direct discourse 301
nominalization 85 psycho-ostensive expressions 246

object clauses 260 quasi-direct discourse 8 f., 232 f., 246

object pronoun 56-58 quasi-factivity 203
object-sentence 176 question 105, 107, 109, 119, 131, 146, 156,
Old Armenian 141 204
Old French 10 question, embedded 203
onomatopoetic expression 164 question, indirect 124, 201, 235
opaque contexts 3 question movement 47, 73
optative 219, 251 question movement, unbounded 72-74
364 Index of Subjects

question word 49, 67, 69, 72 style, written 123

question word, adverbial 70 subcategorization 66f., 71
question word, nominal 68 subjunctive 21, 111, 113, 119, 123f., 131,
quotation content 77 158, 177, 241, 251
quotation, foreign language 192 Svan 128, 138
quotation, hidden 280 f. Swahili 99 ff
quotation marker 180, 183, 257, 280, 299 switch-reference 155
quotative particle 164f., 172
taboo reference 83
reference 93, 108 temporal deixis 17
reference identity 155 tense 2, 14f., 22, 37, 78, 109, 118, 126,
referential opacity 92 130, 161, 201, 221, 238, 284, 357
repetition 219, 287 tense marker 114
report verb 20, 22, 84, 86, 88, 90, 146,152 tense shift 240, 284
reproduced speech 337 tense system 15, 230
rhetorical questions 292 Tikar 35 f.
Russian 18f., 141, 248 time references 110, 118, 249
topic 207, 236
Sanskrit 153 topicalization 75
semantic erosion 288 topic-comment structure 207 f.
semantic reduction 262 transcripts of real conversation 320
semantic transfer 267, 269 transitivity 104
semantics 202 truth 37, 203
semi direct 198 Tunica 22
semi-indirect 125 f., 138, 181 type-token identity 11, 13, 25
sentential object 20
sequence of tenses 14-16, 21, 201 uneigentliche direkte Rede 7
situation 169, 251 universal 39, 118, 156, 197, 219, 248
Slave 47 ff. utterance 113
South-Caucasian 121
Spanish 333 ff. vagueness 163
speaker attitudes 210, 212f., 215f verba cogitandi 262
speaker perspective 174 verba dicendi 122, 134f., 165, 262
speech act 77, 89, 93, 151, 246 verba non-dicendi 272 f.
speech act, original 157 verbatim quotation 99, 139, 172, 177
speech act verb 96, 148, 252 verbatim rendition 1, 6
speech-particle 123, 128, 130 verbs of asserting 210
split deixis 19 verbs of communication 186f., 193, 198
spoken discourse 323 verbs of perception 49
spoken language 334 verbs of saying 29, 32, 34f., 37, 49, 164,
spoken stories 326 202, 209, 246, 259, 267, 298, 317 See
spoken stories, Greek 322 also report verb
style 25, 168, 334 verbs of saying, French 269
style indirect Iibre 7, 100 verbs of saying, secondary 268, 270, 275,
style indirect partiel 301 299
style of speech 100 verbs of saying, transitive 261
style, parody of 282 verbs, cognitive 204
style, social 357 verbs, directional 170
 Index of Subjects

verbs, emotive 206, 208, 270 written fiction 323

verschleierte Rede 7 written language 91, 122, 334, 336
vocatives 337 written prose 287
written texts 134, 163, 283
WH-questions 124, 131, 277
word order 243f., 261, 264 Yoruba 18, 77 ff.
writing 10, 13, 24, 251, 257 Yoruba, spoken 80
writing systems 256 Yoruba, written 80, 89
writings, fictional 180
written discourse 180, 311, 323, 334 Zeigfeld 239
Index of names

Alcina Franch, Juan Blecua 335, 348 Celine, Louis-Ferdinand 284

Andersen, H. C. 232 Cervadze, J. 134
Antoine, F. 301 Chafe, Wallace L. 41,311, 323, 331
Anzai, Tetsuo 199 Charachidze, G. 136,139
Ariyoshi, Sawako 180,192f., 198 Chomsky, Noah 48
Aslanov, A. M. 141 Cikobava,A. 134
Aragon, Louis 288, 291-293 Clark, T.W. 151,158
Auerbach, Erich 284 Clements, George N. 26
Austin, John L. 25,220 Cocteau, Jean 298
Authier, Jacqueline 272, 283 Cohen, Marcel 284
Cohn, Dorris 301
Babits, Mihäly 281 Colette, Mitson 263,299
Baden,Jacob 233 Collins, James 22
Bakanidze, G. 141 Comrie, Bernard 16,18,26
Bakhtin, Mikhail 255, 282, 293 Conrad, Joseph 24
Bally, Charles 7 f., 232,272,278, 283 f., Cooper, William E. 350 f.
286,291,295,301 Coulmas, Florian 10,17,167,177,223,
Balzac, Honore de 263,279,297 225, 282, 331
Bamgbose, Ayo 14,18, 26,158 Coutreras, Heles 341 f., 344, 358
Banfield, Ann 5,23,178,249,283 Cram, Fred D. 260
Bartlett, Frederic 313 Criado de Val, M. 336 f.
Baskakov, N. A. 139 Crisp, S . R . 18
Bauche, Henri 259
Baudelaire, Charles 297
Damourette, Jacques 284
Beckett, Samuel 8
Danehey, John J. 255
Behaghel, Otto 268, 283,298,301
Davidson, Donald 259
Bergheaud, Anne 297,280
Derbyshire, D. C. 344
Berkovits, Röchele 351
Deri, Tibor 264,276
Beyerle, Dieter 283, 295
De Silvia, M. W. Sugathapala 24
Bloomfield, Leonard 22
Diderichsen, Paul 240,243 f., 250
Bodi, Leslie 297
Dominguez de Rodriguez-Pasques, Pe-
Bokarev, A. A. 136 ff.
trona 336,338,354
Bolinger, Dwight 338 f., 341., 347
Dömötör, Adrienne 260-262,298
Borman, Μ. B. 40,44
Driver, S. R. 282
Borszem, Janko 356
Ducrot, Ostwald 288
Bowen, J. Donald 342
Diamalxanov, Z. D. 140
Boyce, Suzanne 347, 350, 352, 356
Br0ndum-Nielsen, Johs. 232 f., 250 f.
Bühler, Karl 220, 239, 291 Ebert, Karen H. 158,248
Buscha, Joachim 249 Eideneier, Η. 177
Ehlich, Konrad 11,251
Calbris, Genevieve 282 Endo, Hiroko 177,198
Camus, Albert 257,285 Endo, Shuusaku 182,184,186,191,194,
Catach, Nina 257 198
 Index of Names 367

Erteschik-Shir, Nomi 251 Heringer, James 93

Esxdiiev, Ja. Μ. 141 Hewitt, Β. G. 18,122,124f., 128,130,141
Hightower, James R. 282
Fant, Lars 342, 345 ff. Hildyard, Angela 13,25
Fischern, August 272 Hironaka, Wakako 180,192 f.
Flaubert, Gustave 288,290,299 Hocke«, Charles F. 25, 242,255
Fönagy, Ivan 22,259,280,282,289, 299 Horn, Wilhelm 289
Frake, Charles 25 Hyman, Larry 26
France, Anatole 258,264,278
Frei, Henri 277 ImnaiSvili, J. 123
Fretheim, Thorstein 251 Inoue, Kazuko 181,183,194,198
Freud, Sigmund 300
Friedmann, V. A. 137
Jackson, Ellen 35,43 f.
Friedrich, Paul 311
Jacobsen, J. P. 232
Jakobson, Roman 22,282, 287,291
GadZiev, Μ. M. 139,141
Jakovlev, N. F. 134,140
Gallagher, Michael 198 f.
Jespersen, Karen 249
Garro, Luisa 358
Jespersen, Otto 7f., 15,25,96,301
Gaulmyn, M.-M. 280
Johnson-Laird, Philip N. 270
Gauvenet, Helene 272
Johnston, William 186,191,194
Geifer, Carole E. 350
Jones, Charles 281
Gerdel, Florence L. 39
Josephs, Lewis 199
Gide, Andre 280
Gigineiävili, J. 123
Gili Gaya, Samuel 336, 341 Kaffka, Margit 265
Givön, Talmy 36,43 Kalepky, Theodor 7 f. 301
Goffmann, Erving 319 Kalik-Teljatuicova, J. 255, 283, 292, 301
Goodman, Nelson 12 Kamada, Osamu 181,183
Goodwin, W. W. 100 Karinthy, Frigyes 264,277,283,295,296,
Goody, Jack 11 300
Gragg, Gene B. 158 Karttunen, Lauri 43,216
Gresillon, Almuth 282 Kazazis, Kostas 312, 331
Grierson, G. A. 157 Keenan, Edward L. 158
Gumperz, John J. 323, 331 Keller, Dezsö 277
Günther, Werner 257,272,286 f., 295 Keller, R. E. 24
Gurk, P. Mc. 257 Kempson, Ruth M. 216
Kendrick, Walter 320
Haden, Ernest F. 342 Keniston, Hayward 357
Hadrovits, Läszlo 272,298 f. Kerr, Isabel 41
Hagdge, Claude 21,158 Kieckers, Emst 260,262,298
Haiman, John 36,233 Kiefer, Ferenc 14,216,261,289
Haie, Austin 158 Kirshenblatt-Gimblett, Barbara 327
Hall-Partee, Barbara 3,29f., 34,165,259 K'ldiaSvili, D. 125 f.
Hamzatov, R. 135,139 Kostant, Ann Liller 180,192 f.
Haverkate, Henk 357 Kosztolanyi, Dezsö 257,261, 269
Heinermann, Theodor 301 Krantz, Dieter 294
Heibig, Gerhard 249 Kristeva, Julia 300
Herczeg, Giulio 283, 287,292f., 295,302 Krüdy, Gyula 255
368 Index of Names

Kvavik, Karen H. 343, 351, 357 Mikszäth, Kälmän 266,269,278, 294

Kuno, Susumu 22,43,185,199 Miller, George A. 270
Miller, R. A. 161
Labov, William 311, 322,325,327 Milner, Jean-Claude 280
La Fontaine, Jean de 6 Mishima, Yukio 198 f.
Lang, Ewald 216 Moirand, Sophie-Colette 272
Lanser, Susan 23 M0ller-Kristensen, Sven 232
Lasria, M. 130 Molnär, Ferenc 267
Laudry, F. 295 Montherlant, Henry de 256
Leech, G. 91 f., 298 Möricz, Zsigmond 265
Legrand, Emile 301 Munro, Pamela 34,37
Lerch, Eugen 283,286f., 293,301 Murq'ilinskij, G. B. 139
Lerch, Gertraud 7 f. 301 Murtaza Galiev, Μ. M. 135,138
Leutin, Laurence 276
Lewin, B. 177 Nakau,M. 168
Li, Charles N. 23 Navarro Tomäs, 336, 337,342,348, 353
Libera, Alain de 257 Neubert, Albrecht 24,232
Liberman, Mark 351 f. Neweth, Läszlö 264,266f., 269, 289,
Linsky, Leonard 12 291-293,297
Lips, Marguerite 266,272,275,278, Nitta, Yoshio 181
283ff., 293,295
LolaSvili, J. 125
Ochs, Elinor 311
Longacre, Robert E. 29
Ohala, John J. 327
Lorck, Etienne 7 f., 24,232, 301
Okutsu, K. 177
Lord, Carol 158
Olbrechts-Tyteca, L. 25
Lowth, Robert 282
Olson, David R. 13,25
Lyons, John 13
O'Shaughnessy, Douglas 351
Ott, Mary Mighio Bensabat 327
Maiigov, M. Ju. 140
Ouy, Gilbert 257
Madieva, G. J. 135,138
Malinowski, Bronislaw 251 Oyelaran, Olasope 96
Mansen, Karie 40
Mansen, Richard 40 Parker, Frank 358
Marceau, Felicien 255 Paul, Hermann 26, 268
Martin, Samuel 188 Perelman, Ch. 25
Martin du Gard, Roger 292 Perlmutter, David M. 233,251
Maupassant, Guy de 286 Perrin, Mona 26,43,77
Maurice, Francois 263,285, 299 Petöfi, J. 281,300
Maw, J. 101 f., 107 ff., 118 f. Pichon, Eduard 284
Maynard, S. K. 20,176 f. Pierrehumbert, Janet 351
McCawley, Noriko 199 Pittenger, Robert E. 255
McClain, Υ. M. 161 Platinga, A. 12
McHale, Brian 24,293 Ptenat, Marc 260,272,283 f.
Menn, Lise 347,350, 352, 356 Polanyi, Livia 41,44
Merimee, Prosper 6 Proust, Marcel 300
Meunier, Andr6 272
Meyer, Hermann 279,280,281 Queneau, Raymond 256,259,278,285,
Mikkelsen, Kr. 243 287,292,299, 301
 Index of Names 369

Quine, W. V. 4,12 Szabö,Denas 259

Quirk, Randolph 16, 77,100f., 110,119 Szabolcsi, Mikl<Ss264, 268
Szomory, Dezsö 265
Räba, György 281 Sankoba, B. 129
Rabelais, Francois 300
Räcz, Endre 259, 261 Tamäsi, Äron 265 f.
Rey-Debove, Jacqueline 272, 280,283 Tannen, Deborah 10,145,224,311,323 f.,
Rice, K. 47, 75 327, 331
Riffaterre, Michael 282 Tesniöre, Lucien 270
Rolland, Romain 258,287 f., 291 Theroux, Paul 230 f.
Romains, Jules 297 Thibaudet, Albert 286, 301
Roncador, Mandred von 158 Thompson, Sandra A. 36,43
Rosier, W. 10,24,177 Thurgood, Graham 41
Ross, Claudia 180f., 199 Thurzö, Belväros 262,265
Ross, J. R. 178 Tobler, Adolf 6, 8 f., 24, 287
Rousseau, Jean-Jacques 279, 299 Todemann, F. 357
Tömörkeny, Istvän 289,294
Sabban, Annette 260,268,271 f., 298f. Tompa, Jöszef 259-261, 298
Saidov, Μ. S. 135,138 Toyama, Shigehiko 162 f.
Sanfield, Karl 298 Traugott, Elizabeth Closs 24
Sanidze, A. 123
Sapir, Edward 25 Ullmann, Stephen 287,295
Schauber, E. 47, 73 Updike, John 198
Schiffrin, Deborah 311 Uslar, P. K. 135,137
Schürr, Friedrich 278
Schwartz, Arthur 41, 43 Verdin Diaz, Guillermo 336ff., 348,354,
Searle, John R. 12,25 357
Seco, Rafael 335 f. Verlaine, Paul 256, 280
Seidensticker, Edward 163 Verschoor, Jan Adriaan 258,260,272,
Sgall, Petr 43 276,278
Shakespeare, William 25 Verschueren, Jef 96
Simonin, Jenny 280, 302 Vezin, Jean 257
Simonyi, Zsigmond 272 Vinogradov, Viktor 255
Sipka, Sändor 264 Vogt, H. 122
Slocum, M. C. 39 Voloshinov, V. N. 4 , 7 f . , 10,14,232f.,
Sorensen, John M. 350 f. 243,248
Spitzer, Leo 270, 276,284, 288,293,295, Vossler, Karl 292
301
Steinitz, Wolfgang 282 Wagner, Burckhardt 285,293
Stendhal 287,299 Wagner, Robert-Leon 298
Steinberg, Danny 23 Waltz, Nathan E. 40
Sternberg, Meir 255,259,262, 301 Watt, Ian 11
Stockwell, Robert P. 342 Wierzbicka, Anna 30,38,42 f.
Strauch, Gerard 255,272,283, 302 Wilson, Deirdre 216
Stross, B. 25 Winter, W. 158
Stubbs, Michael 228,250 Witte, Paul 40,44
Sulejmanov, Ja. 134 Wurmser, Andr6 301
Suner, Margarita 341, 342, 344 Wunderlich, Dieter 16,25,223,298
370 Index of Names

Xanmagomedov, Β. G.-K. 139,141 Zirkov, L. I. 136f., 141

XurSilov, Muhammad 135 Zola, Emile 6, 284,286,288, 290,299
Zwicky, Arnold 91 f.
Zabeeh, F. 25 qabdulaev, M. 134