<DOCINFO AUTHOR ""TITLE "Adverbials: The interplay between meaning, context, and syntactic structure"SUBJECT "Linguistik Aktuell/Linguistics Today,

Volume 70"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">

Adverbials
Linguistik Aktuell/Linguistics Today

Linguistik Aktuell/Linguistics Today (LA) provides a platform for original monograph studies
into synchronic and diachronic linguistics. Studies in LA confront empirical and theoretical
problems as these are currently discussed in syntax, semantics, morphology, phonology, and
systematic pragmatics with the aim to establish robust empirical generalizations within a
universalistic perspective.

Series Editors
Werner Abraham Elly van Gelderen
University of Vienna Arizona State University

Advisory Editorial Board

Guglielmo Cinque Ian Roberts
University of Venice Cambridge University
Günther Grewendorf Ken Safir
J.W. Goethe-University, Frankfurt Rutgers University, New Brunswick NJ
Liliane Haegeman Lisa deMena Travis
University of Lille, France McGill University
Hubert Haider Sten Vikner
University of Salzburg University of Aarhus
Christer Platzack C. Jan-Wouter Zwart
University of Lund University of Groningen

Volume 70
Adverbials: The interplay between meaning, context, and syntactic structure
Edited by Jennifer R. Austin, Stefan Engelberg and Gisa Rauh
Adverbials
The interplay between meaning, context, and
syntactic structure

Edited by

Jennifer R. Austin
Stefan Engelberg
Gisa Rauh
University of Wuppertal

John Benjamins Publishing Company

Amsterdam
/
Philadelphia
 TM The paper used in this publication meets the minimum requirements
8

of American National Standard for Information Sciences – Permanence

of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data

Adverbials : the interplay between meaning, context, and syntactic structure /

edited by Jennifer R. Austin, Stefan Engelberg and Gisa Rauh.
p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166–0829 ; v. 70)
Includes bibliographical references and indexes.
1. Grammar, Comparative and general--Adverbials. I. Austin,
Jennifer R. II. Engelberg, Stefan. III. Rauh, Gisa. IV. Linguistik aktuell ; Bd 70.

P284 .A34 2004

15’.76-dc24 2004050187
isbn 90 272 2794 2 (Eur.) / 1 58811 546 1 (US) (Hb; alk. paper)

© 2004 – John Benjamins B.V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or
any other means, without written permission from the publisher.
John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands
John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
 Table of contents

Foreword ix
Current issues in the syntax and semantics of adverbials 1
Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh
Circumstantial adverbs and aspect 45
David Adger and George Tsoulas
Optimizing adverb placement in gap constructions 67
Eva Engels
Domain adverbs and the syntax of adjuncts 103
Thomas Ernst
Depictives and transparent adverbs 131
Wilhelm Geuder
Degree Phrases versus Quantifier Phrases in prenominal
and preverbal positions: A hybrid explanation for some
distributional asymmetries 167
Dagmar Haumann
A feature-based theory of adverb syntax 205
Christopher Laenzlinger
Where syntax and semantics meet: Adverbial positions
in the German middle field 253
Karin Pittner
Left/right contrasts among English temporal adverbials 289
Benjamin Shaer
Author index 333
Subject index 337
 Addresses

David Adger E-mail address:

Affiliation: Queen Mary College, [email protected]
University of London
Postal address: Queen Mary, Thomas Ernst
University of London Affiliation: University of Massachusetts
Mile End Road at Amherst and Marlboro College
London, E1 4NS, UK Postal address: 80 Barrett St. Apt. 13
E-mail address: [email protected] Northampton MA 01060, USA
E-mail address:
Jennifer R. Austin [email protected]
Affiliation: University of Wuppertal
Postal address: Fachbereich A Wilhelm Geuder
Bergische Universität Wuppertal Affiliation: University of Constance
Gauß-Str. 20 Postal address: Universität Konstanz
D-42097 Wuppertal, Germany Fachbereich Sprachwissenschaft / SFB 471
E-mail address: Postfach 5560 / D 174
[email protected] D-78457 Konstanz, Germany
E-mail address:
[email protected]
Stefan Engelberg
Affiliation: University of Wuppertal Dagmar Haumann
Postal address: Fachbereich A Affiliation: University of Erfurt
Bergische Universität Wuppertal Postal address: Universität Erfurt
Gauß-Str. 20 Anglistische Linguistik
D-42097 Wuppertal, Germany Postfach 900221
E-mail address: D-99105 Erfurt, Germany
[email protected] E-mail address:
[email protected]
Eva Engels
Affiliation: University of Potsdam Christopher Laenzlinger
Postal address: Graduiertenkolleg Affiliation: University of Geneva
“Ökonomie und Komplexität Postal address:
in der Sprache” Département de linguistique
Institut für Linguistik / Allgemeine 2 rue de Candolle
Sprachwissenschaft 1211 Genève-4, Switzerland
Universität Potsdam, Postfach 601553 E-mail address:
D-14415 Potsdam, Germany [email protected]
 Addresses

Karin Pittner Benjamin Shaer

Affiliation: University of Bochum Affiliation: Zentrum für Allgemeine
Postal address: Lehrstuhl für Sprachwissenschaft, Berlin
Germanistische Linguistik Postal address: Jägerstraße 10-11
Germanistisches Institut D-10117 Berlin, Germany
Ruhr-Universität Bochum E-mail address: [email protected]
Universitätsstr. 150
44780 Bochum, Germany George Tsoulas
E-mail address: [email protected] Affiliation: University of York
Postal address: George Tsoulas
Gisa Rauh Department of Language and
Affiliation: University of Wuppertal Linguistic Science
Postal address: Fachbereich A University of York
Bergische Universität Wuppertal Heslington, York-YO10 5DD, UK
Gauß-Str. 20 E-mail address: [email protected]
D-42097 Wuppertal, Germany
E-mail address: [email protected]
 Foreword

A number of colloquia were held in Wuppertal in the late 90s and early 00s,
including ones on “Aspect and Adverbs”, “Information Structure”, and “The
Syntax and Semantics of Adverbials”, which aimed at reflecting and promot-
ing the rising interest in interface issues in linguistics. Subsequent to the last
colloquium, the decision was made to gather a number of authors who were
particularly adept at exploring the topic of interface issues in the domain of
adverbials. Some of the authors were participants at one or several of the
colloquia.
We are particularly grateful to those who agreed to review one or more of
the submitted papers: Regine Eckardt, Eva Engels, Gisbert Fanselow, Werner
Frey, Wilhelm Geuder, Hubert Haider, Dagmar Haumann, Joachim Jacobs,
Ingrid Kaufmann, Christopher Laenzlinger, Irene Rapp, Martin Schäfer, Ben-
jamin Shaer, Thilo Tappe, and George Tsoulas.
Helpful in other respects were Ariane Glasner, Heike Baeskow, and Dagmar
Haumann, whom we would hereby like to thank.

Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

Wuppertal
February 2004
 Current issues in the syntax and semantics
of adverbials

Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

University of Wuppertal

The paper presents an overview of topics currently discussed in the domain

of adverbials, focussing on interface issues concerning the interplay between
syntax and semantics. It also serves as an introduction to the volume,
showing in detail how the contributions in this volume relate to the central
topics in the field.

. Introduction

The conclusion of Renate Steinitz’s 1965 dissertation on a transformational

theory of “Adverbialsyntax” (Steinitz 1973) begins with Chomsky’s (1965) now
well-known remark that “adverbials are a rich and as yet relatively unexplored
system, and therefore anything we say about them must be regarded as quite
tentative.” “Gewiß” (Quite) she replies, slightly impatiently, continuing with a
list of what the developing theory should do. It should i) adequately mediate
between semantic and syntactic structures, ii) contain well-motivated means
for the representation of states of affairs, and iii) allow insights into the dis-
course dependence of many syntactic phenomena. Written during a time in
which linguistics underwent a major upheaval, this can undoubtedly be seen
as an insightful evaluation, but it has nevertheless taken another 30 years for
research on the syntax and semantics of adverbials to really take off. Perhaps
this is partly due to the fact that even the mere description of the behavior of
adverbials has continued to produce more and more confusing data than could
probably have been expected back then, but also to the fact that it took quite
a while to develop the theoretical tools for fulfilling Steinitz’s desiderata. Syn-
tactic theories – Principles & Parameters and Minimalism in particular with
respect to adverbial syntax – have made advancements which better allow the
intricate positional behavior of adverbials and its relation to semantic inter-
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

pretation to be captured. Progress in semantic theories, in particular the rise of

event semantics in the succession of Davidson (1967) and insights into the on-
tological basis of natural language semantics (see, among others, Asher 1993),
allow a better understanding of how adverbials relate to events, states, facts, or
propositions. Furthermore, the discourse dependence of syntactic and seman-
tic phenomena can nowadays more easily be captured in linguistic theories.
All these theories have helped to shape our understanding of adverbials, just as
adverbials help to shape these theories.
The present volume has its focus on the interplay between syntax, se-
mantics, and discourse. Although most papers approach this topic from ei-
ther a more syntactic or a more semantic starting point, interface issues arise
throughout the papers. In this introductory article, we want to outline the
phenomena which underlie current discussion of the syntax and semantics of
adverbials, and to point out the questions and major theoretical issues that
arise from these phenomena. Along the way we will show you where and how
the authors of this volume tackle the problems posed by these data.1 Fur-
ther background can be found in a number of helpful introductions, such as
Delfitto (2000), Alexiadou (2002), the introductory chapter in Ernst’s (2002)
monograph, and Lang, Maienborn and Fabricius-Hansen’s (2003) introduc-
tory article to their volume on adjuncts. Furthermore, the papers in the latter
volume (especially Part B on adjunct placement) as well as a thematic issue
in Lingua, introduced by Alexiadou (2004), give a fairly representative sam-
ple of topics currently discussed in the field of research on adverbials. To our
knowledge, there is no state-of-the-art article on the semantics of adverbials to
speak of, but Eckardt’s (2002) recommendable introduction to event semantics
touches on a number of relevant issues.
Before we get started, here is a brief remark on the use of the terms adverb,
adverb phrase, and adverbial. We will generally use the term adverbial, even
though it is not always appropriate as a cover term.2 Often what is said about
adverbials primarily applies to non-extended adverb phrases (AdvPs) that per-
form adverbial functions, and sometimes what is said about non-extended
AdvPs carries over to extended AdvPs (e.g. unfortunately for him, extremely
slowly) and to adverbials that do not project from adverb heads but occur as
PPs, NP/DPs, or CPs (e.g., for two hours, the whole day, when she entered the
room). If we use the term adverb, we do this intentionally to highlight their na-
ture as lexical units or otherwise non-extended expressions. Our usage does not
necessarily correspond to that of the contributors, and no attempt was made
on our part to make their usage of these terms uniform. There is no appropri-
 Current issues in the syntax and semantics of adverbials 

ate cover term because there is still no appropriate cover theory. In Section 5.1
we will return to this issue.
This introductory article proceeds as follows. Section 2 presents the basic
phenomena a theory of the syntax-semantics interface in the domain of adver-
bials is confronted with and outlines the two types of predominant syntactic
theories, namely those which assume that adverbials are specifiers versus those
that conceive of them as adjuncts. We will show how the papers in this volume
contribute to this ongoing debate. Section 3 discusses the issue of syntactic and
semantic adverbial classes and summarizes some of the insights the authors
in this volume have gained on particular classes. In Section 4 some more syn-
tactic phenomena in the domain of adverbials are presented and theoretical
approaches to the phenomena that are dealt with in this volume are sketched.
Section 5 deals with categorial questions. The relation between adverbs, ad-
verb phrases, and adverbials, as well as the one between adverbs and adjectives
is touched upon, and the issue of how adverbials behave in the non-verbal do-
main is raised. Section 6 concludes this introduction and presents an outlook
for further research.

. Fundamental phenomena and debates

. Basic phenomena of adverbial order

Even a brief look at the syntax of adverbials cannot ignore the ensuing im-
pression of having received a very mixed message. While adverbials sometimes
seem to enjoy considerable freedom with respect to syntactic position, at other
times they display extremely tight syntactic restrictions. The following data,
which we take to constitute the core phenomena a theory of adverbials has to
explain, serve to illustrate this behavior.

.. Stable versus variable orders

Adverbials often show a stable order with respect to other adverbials in the
sentence (1, 2). In many other cases, though, the relative adverbial order is
variable (3, 4, 5).
(1) a. Da allora, non accetta mica più sempre i nostri
since then not accept.he at.all any.longer always the our
inviti.
invitations
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

‘Since then, he doesn’t any longer always accept our invitations’

(cf. Cinque 1999: 6)
b. *Da allora, non accetta mica sempre più i nostri inviti.
c. *Da allora, non accetta sempre mica più i nostri inviti.
d. *Da allora, non accetta sempre più mica i nostri inviti.
e. *Da allora, non accetta più mica sempre i nostri inviti.
f. *Da allora, non accetta più sempre mica i nostri inviti.
(2) a. Gina has probably tactfully suggested that we leave.
b. *Gina has tactfully probably suggested that we leave. (Ernst 2002: 131)
(3) a. Fortunatamente ora sei con noi.
fortunately now are.you with us
‘Luckily now you are with us’
b. Ora fortunatamente sei con noi. (cf. Cinque 1999: 13)
(4) a. The speaker never intentionally strays from the topic.
b. The speaker intentionally never strays from the topic. (Ernst 2002: 131)
(5) a. Carol built a treehouse for her brother in the backyard with her new
tools.
b. Carol built a treehouse in the backyard for her brother with her new
tools.
c. Carol built a treehouse with her new tools for her brother in the back-
yard.
d. Carol built a treehouse in the backyard with her new tools for her
brother. (Ernst 2002: 132)

.. Wide versus narrow range of positions

With respect to their position relative to other elements in the clause that occur
in a fixed order, some adverbials allow a wide range of possible positions, while
others display a more restricted behavior, as the English data in (6) and the
French data in (7) show. There are also crosslinguistic differences like those in
(8) which need to be explained:
(6) a. (Stupidly,) they (stupidly) have (stupidly) been (stupidly) buying hog
futures (, stupidly) (Ernst 2002: 2)
b. (Evidently) Horatio (evidently) lost his mind (*evidently) (, evidently)
(Jackendoff 1972: 50)
c. This bridge may (*badly) have (*badly) been (badly) designed (badly)
by Brunel (badly) (Adger and Tsoulas)
d. (*Merely) Albert is (merely) being a fool (*merely) (Jackendoff 1972: 51)
 Current issues in the syntax and semantics of adverbials 

(7) a. Jean a (bien) fait (*bien) son travail (*bien)

Jean has (well) done (well) his work (well)
b. Jean a (parfaitement) fait (parfaitement) son travail (parfaitement)
Jean has (perfectly) done (perfectly) his work (perfectly) (Laenzlinger
1998: 52)
(8) a. Jean (*lentement) mangea (lentement) la pomme
b. John (slowly) ate (*slowly) the apple (Laenzlinger)

.. Effects of positional variance on interpretation

While in some cases the different possible positions do not affect the inter-
pretation, in other cases they do. Invariably, the contribution of the domain
adverbial in (9a) is that the assertion made holds with respect to the domain of
linguistics. On the other hand, (9b) means that it was clever of John to answer
the question if the adverbial occurs sentence-initially or preverbally and that
John answered the question in a clever way if the adverb occurs preverbally or
postverbally, but not when in sentence initial position.
(9) a. (Linguistically,) these examples are (linguistically) interesting (linguis-
tically) (Ernst)
b. (Cleverly) John (cleverly) answered the question (cleverly)

.. Order variability of other sentence elements

When there is a stable order of adverbials, certain other elements seem to be
rather free to occur in between adverbials in some languages, such as partici-
ples in Italian (10a) or objects in German in (10b), but not in others, such as
participles in German in (11).
(10) a. Da allora, non hanno (rimesso) di-solito (rimesso) mica
Since then not have.they (put) usually (put) at.all
(rimesso) più (rimesso) sempre (rimesso) completamente
(put) any.longer (put) always (put) completely
(rimesso) tutto bene in ordine
(put) everything well in order
‘Since then, they haven’t usually not any longer always put everything
well in order’ (cf. Cinque 1999: 45)
b. weil Hans (die Sonate) offen-gestanden (die Sonate)
because Hans (the sonata) frankly (the sonata)
wahrscheinlich (die Sonate) oft (die Sonate) nicht gespielt hat
probably (the sonata) often (the sonata) not played has
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

‘because frankly Hans probably has not often played the sonata’
(cf. Laenzlinger)
(11) Hans hat (*gefrühstückt) heute morgen (*gefrühstückt)
Hans has (*had.breakfast) this morning (*had.breakfast)
wahrscheinlich (*gefrühstückt) in-Ruhe (gefrühstückt)
probably (*had.breakfast) quietly (had.breakfast)
‘Hans probably had breakfast this morning quietly.’

These data raise a number of questions a theory of the interplay of syntax and
semantics in the domain of adverbials will have to answer:
– Are the ordering restrictions which can be observed in the examples above
to be explained by syntactic principles or by semantic ones?
– In which type of positions do adverbs and adverbials occur within local
syntactic structures and how are they licensed there?
– What kind of movement operations for adverbials or other parts of the
clause are involved in generating their appropriate position?
– To what degree can the semantic contribution of adverbials to the rest of
the clause be derived compositionally (in particular with respect to the
phenomena in (9), which show that adverbials sometimes seem to have
the same interpretation in different positions and sometimes different in-
terpretations in the same position)?
– Which classes of adverbials need to be distinguished for syntactic and for
semantic reasons?

. Heads, complements, specifiers, and adjuncts

Browsing through the papers of this volume, one will notice that as concerns
the syntactic positions for adverbs and adverbials, a wide variety of options
is being explored. If we assume that syntactic structure corresponds to an X-
bar schema,3 three conceptual distinctions have to be taken into consideration:
(i) whether adverbs are conceived of as projecting or non-projecting elements;
(ii) if they are projecting elements whether they project onto phrases that can
function as adverbials or onto other types of phrases; (iii) if they project onto
adverbial-like phrases, whether these phrases occur as adjuncts, specifiers, or
complements.
Some theories consider adverbs to be heads which don’t project maximal
phrases at all (e.g. Travis 1988). This is partly motivated by the assumption that
adverbs don’t allow complements, and indeed, many do not, as the difference
between the adverb in (12a) and the adjective it is related to (12b) shows:
 Current issues in the syntax and semantics of adverbials 

(12) a. *proudly of his achievements

b. proud of his achievements

There are adverbs, though, that do take complements (e.g. independently of

him).4 The adverbs-as-heads approach has also been criticized for failing to
account for the fact that adverbs can undergo topicalization and focus move-
ment, which are considered to be processes that are not open to word-level
categories. Furthermore, it has been observed that adverbs do not block head
movement of verbs, as would be expected if they were non-projecting heads
(cf. Alexiadou 1997: 36, 2002: 4ff.).
Thus, the prevailing view is that adverbs do project maximal phrases.
Two basic variants of this idea can be distinguished: the majority view is that
maximal projections of adverbs are phrases that function as adverbials which
usually c-command the entities they modify. Less wide-spread is the idea that
maximal projections of adverbs dominate the entities the adverbs are seman-
tically related to. The latter idea is briefly outlined in Sportiche (1995/1998),
who suggests a certain kind of head-complement or head-specifier relationship
between an adverb and its modifiee:
Informally speaking, I will assume that adverbs and adjectives are all domi-
nated by a projection whose head takes the modifiee of the adverb or the ad-
jective as an argument, that is, as complement (or sometimes a specifier). The
general intuition is that adjectives and adverbs bear the same kind of relation
to their modifiee that determiners bear to their noun phrases or predicates to
their arguments. (Sportiche 1995/1998: 399f.)

The phrase structure options in (13) are the ones proposed by Sportiche
(1995/1998: 400), in which the “modifiee” of the adverb is the VP:
(13) a. John will stupidly answer
...[AdvP [ Adv’ [Adv stupidly] [VP answer]]]

b. John will answer stupidly

...[AdvP [VP answer][Adv’ [Adv stupidly]]]

Although none of the authors of this volume assumes that adverbs are heads
of this kind with respect to adverbials on the clausal level, Haumann’s contri-
bution to the distribution of APs and AdvPs supplemented by degree adverbs
(e.g. so wonderful, more carefully) is built on structures that are comparable to
the ones in (13). Following Corver (1997), Haumann assumes two classes of
degree adverbs to be heads of their own functional projections: adverbs like as,
so, too head a degree phrase (DegP) which dominates a quantifier phrase (QP)
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

headed by adverbs like enough, less, more (see Section 5.3 for further details).
Thus, the AP or AdvP that the degree adverbials are semantically related to is
dominated by the maximal projection of the degree adverbs:
(14) a. less softly
[QP [Q less] [AdvP softly]]
b. so vividly
[DegP [Deg so [QP [Q e][AdvPvividly]]]

In contrast to approaches like Sportiche’s, the vast majority of theories assume

that adverbs project phrases which function as adverbials. Given this, there are
three positional options in local trees for AdvPs (and partly also for adver-
bials of other phrasal categories): they can occur as complements, specifiers,
or adjuncts.
As regards the topic of whether any adverbials are complements, various
approaches differ as to which adverbials are considered to be complements,
and as regards the syntactic (and semantic) argumentation that supports these
assumptions. A traditional view holds that the obligatoriness of adverbials with
certain verbs would justify considering them complements. However, it seems
that exactly for this class, semantic requirements placed on the complement by
the verb are not particularly specific:
(15) a. Sie wohnt in Wuppertal / luxuriös / zur Miete
she resides in Wuppertal / luxuriously / to.the rent
‘She lives in Wuppertal / luxuriously / in a rented place’
b. Er befindet sich in München / wohl
he finds himself in Munich / well
‘He is in Munich / well’ (cf. Jacobs 1994: 19)

Here we also mention approaches which consider various classes of adverbials

to have base positions which are sometimes regarded as being reserved (in
some sense) for DP complements of the verb, for example in verb-sister po-
sition. The role of semantic considerations in justifying this decision varies.
McConnell-Ginet (1982) argues that manner adverbials are the innermost
complements of verbs, the semantic justification being that the meaning of the
verb changes when it combines with a manner adverbial. In contrast, Larson
(1988: 345f., Note 11) claims to follow McConnell-Ginet (1982) in regards to
his assumption that adverbs are the innermost complements of verbs, but the
role of semantics is difficult to make out. He tacitly suggests that this holds for
manner, temporal, and locative adverbials (Larson 1988: 382f.). These adver-
bials are assumed by him to be ones that receive oblique thematic roles that
 Current issues in the syntax and semantics of adverbials 

are low in the thematic hierarchy he adheres to, in which an agent role is high-
est, followed by the theme role and the goal role, oblique roles coming last.
The principle linking thematic roles to syntactic positions that he assumes as-
signs the verb-closest position to the expression satisfying the hierarchy-lowest
role. The generation of shell structures, in which a verbal predicate corre-
sponds to more than one head, results from the conflict between a principle
that demands that a position be made available for every thematic role that
a predicate determines, even if the role is not realized, and the assumption
of a particularly restrictive version of X-bar theory in which there is but one
complement position and one specifier position potentially available for ex-
pressions thematically related to a head. However, Larson (1988) continues to
allow adjunction; in fact he assumes that the expression in a double object con-
struction (send Mary a letter) corresponding to the “dative object” in a dative
object construction (send the letter to Mary), namely Mary, has undergone a
process of demotion and is thus adjoined to a V’ projection. The issue of when
and whether members of such classes of adverbials as manner, locative, and
temporal ones could be demoted and thus appear in adjoined positions in-
stead of as sisters to the verb or specifiers of the verb, is left uncommented on
by Larson (1988).
Stroik (1996), whose analysis follows in Larson’s wake, proposes that only
temporal and locative adverbials are structural sisters of the verb. He also does
not provide any kind of substantial claim about the semantic relation between
verb and adverbial, other than the vague suggestion that the latter is a modifier
of the verb and not an argument.5
In light of event semantics, lower adverbials (e.g. in the park) generally are
more similar to arguments in a way than to some of the higher adverbials (e.g.
probably). Many higher adverbials are traditionally represented as operators,
while lower adverbials often are predicates of events and are integrated into the
semantic representation similarly to arguments. Thus, in a neo-Davidsonian
representation, the sentence Paul probably jogged in the park roughly comes
out as PROBABLY (JOG (e) & AGENT (Paul, e) & IN (park, e)), where the ar-
gument Paul and the locative in the park both end up as predicates over events,
while probably is an operator on the whole proposition.
Many approaches assume that adverbials occur as specifiers of projections
which dominate the verbal projection. This approach became popular in the
late 90s and has developed a number of different varieties (see for example
Cinque 1999, 2004; Alexiadou 1997; Laenzlinger 1998). Among the authors of
this volume, Laenzlinger and Adger and Tsoulas present specifier analyses of
adverbials (see Section 2.3).
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

As equally wide-spread as the specifier approach is the adjunct approach

to adverbials. The adjunct approach is reminiscent of traditional assumptions
about adverbs within X-bar theory and has experienced a renaissance within
the last years (see for example Ernst 2002; Frey & Pittner 1998; Haider 2000).
The core assumption is that adverbials are adjoined at different levels of the
clause. Within this volume, it is particularly Ernst, Engels, and Pittner who
present further developments of this idea (see Section 2.4).
In the last few years, a lively debate has been going on between propo-
nents of the specifier theory of adverbials and those of the adjunct approach.
In general, the specifier appoach is more friendly to the idea that syntax deter-
mines semantics, while the adjunct approach is more friendly to the idea that
semantics determines syntax. The prototypical proponents for the two kinds of
theories are Ernst (2002), who assumes that adverbials occur in adjunct posi-
tions and that positional restrictions are by and large predictable from semantic
principles, and Cinque (1999), who assumes a cascade of functional projec-
tions, each of which can host an adverbial of the respective class in its specifier
position. Both approaches have yielded a number of varieties, each of which
provides a different answer as to how syntax and semantics interact in the do-
main of adverbials. Since a number of papers in this volume contribute to this
debate, we will devote the next two sections to specifier and adjunct approaches
to adverbials.
However, we cannot review in detail here how specifier approaches and
adjunct approaches deal with the vast variety of challenging data they face. In
general, one might expect specifier approaches to fare rather well with data
that reveal a rigid order of adverbials and multiple positions for other elements
of the sentence in between adverbials, while changes in the relative order of
adverbials and multiple positions of adverbials are more of a challenge. Ad-
junct approaches, on the other hand, in the ideal case should struggle less in
explaining varying orders of adverbials with respect to each other and with re-
spect to other elements in the sentences, while increasing rigidity of adverbial
order forces more elaborate explanations. That this is not more than a rough
generalization will become clearer when dealing with particular phenomena,
especially in Sections 3 and 4.

. Adverbials as specifiers

A prototypical specifier theory of adverbial syntax can be conceived of as ad-

hering to the following basic ideas:
 Current issues in the syntax and semantics of adverbials 

– Adverbials occur as specifying phrasal categories (that is, as AdvPs) in a

Spec-head configuration.
– Specifier approaches to adverbials are usually in the spirit of Kayne’s (1994)
antisymmetric syntax, i.e. they assume [ZP [X YP] XP ]XP as the only admis-
sible local structure; there is no right-branching and there are no adjuncts
as in the traditional conception of X-bar theory.
– Different base positions for adverbials are assumed whenever ordering re-
strictions between adverbials are observed from which deviance is only
possible if it can be explained by movements of projections containing the
adverbial, such as topicalization or wh-movement.
– The cascade of functional projections is a product of Universal Grammar
and projected in every sentence.

This characterization is probably closest to Cinque’s (1999) approach to adver-

bials. Cinque (2004) elaborates on arguments in favor of a specifier approach.
Many of these arguments go beyond the question of whether particular posi-
tions of particular adverbials are correctly predicted, and include the following:
(i) Functional particles and auxiliaries are overt manifestions of heads of the
functional projections adverbials merge into as specifiers. Crosslinguistically
the number, type, and order of those head morphemes match the number,
type, and order of the corresponding adverbs. (ii) The acquisition of different
parts of verb morphology mirrors the acqusition of the adverbs corresponding
to these morphological categories. (iii) The specifier approach reflects that we
find the same functional projections and corresponding adverbials in the lan-
guages of the world, while certain semantic distinctions that could conceivably
be expressed by adverbials don’t show up in any language, e.g. particles and ad-
verbials expressing the speaker’s sentimental attitude toward the assertion. This
kind of observation would have to be expressed in Universal Grammar inde-
pendently if ordering restrictions among adverbials also fall out from scope
relations. (iv) Each of the functional projections yields a default interpretation
even if not filled with lexical material.
While the characteristics mentioned at the beginning of this section are
typical for specifier theories, actual approaches can differ from this characteri-
zation in several respects:
– Cinque (2004) even assumes obligatory adverbials to be in specifier posi-
tions, and maintains that this is a rather unremarkable assumption given
a Larsonian VP-analysis, in which most verb complements are assumed to
enter structures in specifier positions.
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

– In contrast, Alexiadou (1997) assumes that some adverbs are not merged
into specifier positions, but rather move into the specifier positions of cer-
tain functional projections from complement positions of the verb. This
holds particularly for manner adverbials and other lower adverbials.
– With respect to right-branching it should be noticed that Laenzlinger
(1998: 80ff.) allows sentence-final adverbials to occur as right-branching
specifiers.
– Laenzlinger (1998) also deviates from Kayne (1994) in assuming that spec-
ifiers are not unique; heads project two specifiers, an A-position and an A’-
position, the latter of which can license adverbials. This is consonant with
Chomsky’s (1995) Minimalist Program, in which more than one specifier
is conceivable.
– In contrast to Cinque (1999: 106), who suggests more than 30 different
specifier projections that are present in the structure of any sentence, Adger
and Tsoulas assume that if the functional projections are not filled with
lexical material, the whole range of projections is not necessarily present in
every sentence.

Two of the papers in the volume at hand contribute to our understanding of

specifier theories, both of them formulated in slightly differing Minimalistic
frameworks. Adger and Tsoulas work at tightening up the specifier approach
to adverbials by showing that the licensing of manner and locative adverbials
is not independent of the licensing of objects (see Section 3.2, 3.3). Both the
head of AspP, related to telicity, and the head of vP, associated with agentiv-
ity, account for object placement. At the same time, these projections are the
ones that provide the architecture in which the features of locative and man-
ner adverbials are checked. Thus, the motivation for assuming the functional
projections in question is not merely to provide a host for these classes of
adverbials. On their approach, this is a side effect.
From Laenzlinger’s paper it becomes clear which kind of movement oper-
ations within a specifier approach can serve to account for adverb positioning.
The paper explores how much empirical mileage can be bought with a handful
of general principles in an account of the crosslinguistic variance found in the
Vorfeld (the complementizer layer), the Mittelfeld (the inflectional layer), and
the Nachfeld (the lexical layer as a Larsonian VP-shell). The languages given the
most attention are French and German. His basic assumption is that adverbials
occur in a specifier-head feature checking configuration in which interpretable
features are checked by pure Merge. He also identifies an interesting distinction
between adverbials that is of an apparently semantic nature, which he accounts
 Current issues in the syntax and semantics of adverbials 

for in terms of a difference in licensing conditions. The adverbials he identi-

fies as qualificational ones (e.g. quickly, seriously, courageously, really), in being
subject only to Checking theory, are more mobile with respect to their checking
configuration, while quantificational adverbials (e.g. often, almost, a lot, hardly)
are subject to additional licensing conditions, and are less mobile with respect
to the head that licenses them.
The following assumptions are central to Laenzlinger’s analysis: (i) Lexical
verbs project onto Larsonian VP-shells. This constitutes the thematic domain
and, as such, excludes lexical material which is not “argumental” in the sense
of being thematically related to the verb. Head-movement of verbs is restricted
to taking place within VP shells. (ii) The subject originates within the thematic
domain of the verb in a specifier position, as does a direct object, while an in-
direct object originates in complement position. They vacate their respective
base positions and raise overtly to specifier positions of the functional projec-
tions SubjP or ObjP, which potentially occur more than once, to have their
EPP features checked.6 (iii) Verb movement beyond the outer VP shell is con-
ceived of as movement of the remnant VP,7 which targets the specifier position
of IP in simple tenses, or, in compound tenses, the specifier position of AuxP.
VP-remnant movement, which is not available in all languages, is seen as trig-
gered by the requirement that selectional features on I or Aux be checked under
specifier-head agreement with the VP-remnant.
VP-remnant movement is involved in the different positions adverbials
have with respect to the finite verb in French vs. English (16). While both lan-
guages show verb projection movement to Spec-Infl and – with compound
tenses – to Spec-Aux, only French not only moves auxiliaries but also finite
verbs up from Aux and Infl to Subj.
(16) a. Jean donne probablement déjà souvent une pomme à Marie
Jean gave probably already often an apple to Mary
‘John probably already often gave an apple to Mary’
b. John probably already often gives an apple to Mary.

Different scrambling options account for greater freedom of object positions in

German with respect to adverbials (cf. (10a) above) in contrast to their more
restricted behavior in French. Laenzlinger assumes that there are two ObjPs
above VP in French while two unordered ObjPs occur recursively between any
two adverbial-related functional projections in German.
AdvP movement is assumed to be involved where adverbs in the German
Vorfeld occur as topic (17a) or focus (17b):
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

(17) a. Oft hat Tom dieses Buch gelesen

often has Tom this book read
‘Tom often read this book.’
b. Gestern hat Tom dieses Buch gelesen
yesterday has Tom this book read
‘(It was) yesterday (that) Tom read this book.’

For adverbials like souvent ‘often’, which occur in different positions with dif-
ferent interpretations (18), and similarly for repetitive adverbials like de nou-
veau ‘again’, and adverbials with factive vs. manner readings like intelligemment,
it is assumed that they target different positions when they merge. Souvent can
merge in a Spec position above and below NegP.
(18) a. Jean n’a pas souvent pleuré aux enterrements.
Jean has not often cried at.the burials
‘Jean did not often cry at the burials’
b. Jean n’a souvent pas pleuré aux enterrements.

Furthermore, Laenzlinger assumes movement of projections containing adver-

bials into topic positions (see Section 4.4, 46c) and so-called snowball move-
ment to account for postverbal adverbials that occur in reverse order (see
Section 4.3).

. Adverbials as adjuncts

Although it was standard to analyze adverbials as adjuncts in earlier versions of

generative grammar, this analysis was also frequently rejected as not being re-
strictive enough. The more sophisticated adjunct analyses that have been devel-
oped during the last years are presented as alternatives to specifier approaches.8

The basic assumptions of a prototypical adjunct theory can be summarized

as follows:
– Adverbials occur as AdvPs or other phrasal categories adjoined to any
available projection in which they are interpretable.
– Adverbials can be adjoined on the left or on the right.
– Only very few, if any, syntactic principles play a role in determining the
position of adverbials.
– Adverbials don’t move except for undisputed cases like wh-movement and
topicalization, and maybe weight-induced rightward movement.
 Current issues in the syntax and semantics of adverbials 

Ernst’s (2002) adjunct theory, which is summarized in his article in this vol-
ume, adheres to the basic ideas above. But as with the specifier approach, there
are several other adjunct-based theories on adverbials which deviate in one
respect or another from these ideas:
– Some approaches, such as Haider’s (2000) and Costa’s (2004), do not al-
low right-adjunction. Building on proposals by Larson (1988) and Stroik
(1990, 1996), Haider (2000: 95, 108) proposes that only preverbal adver-
bials are adjoined while (non-selected) postverbal ones are embedded.
Embedding is to be understood as contrasting with adjoined. Haider’s pro-
posal departs from Larson’s and Stroik’s in that he assumes that postverbal
adverbials are embedded in an empty VP below the VP-shell structure of
the lexical verb.9
– Some approaches assume that additional syntactic regulation is necessary;
see the summaries of Engels and Pittner below.
– Some approaches restrict the types or levels of projections to which an ad-
verbial may adjoin. For example, Costa (1998, 2004) assumes a ban on
adjunction to certain maximal projections, such as AgrP.
– Some approaches, for example that of Pittner, assume additional adverbial
movement, e.g. in scrambling operations.

Among the authors in this volume Engels, Ernst, Geuder, and Pittner explicitly
assume adverbials to be adjuncts. In particular, Engels and Pittner contribute
answers to the question of which additional syntactic means are required in
an otherwise semantics-driven theory of adverbials. Engels assumes additional
optimality-ranked syntactic constraints to account for the occurrence of adver-
bials in front of gaps, and Pittner defines base positions for adjoined adverbials
in terms of c-command relations.
Engels investigates the behavior of a wide range of adverbials in front of
several types of VP gaps (e.g. (19a) vs. (19b)). She develops one of the first ap-
proaches to adverbial syntax as well as to gap constructions within Optimality
Theory (for more details, see Section 4.2):
(19) a. They used to be Socialists, and Communists they clearly were___, (too).
b. *They used to be Socialists, and Communists they were clearly___, (too).

Engels follows Ernst (2002) in assuming that adverbials can be adjoined to

whatever sister constituent has the required semantic type. Her theory accounts
for the data on the placement of adverbials in front of gaps in terms of sev-
eral ranked constraints on semantic compatibility, movement, empty heads,
and phonetic realization. Her comparison of English and French data shows
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

how different rankings of some of the constraints cover the language-specific

differences in pre-gap adverbial placement. Among other things, the investi-
gation yields a result that disfavors a purely semantic explanation of adverbial
distribution. With respect to particular pre-gap constructions, evidential and
subject-oriented adverbials pattern together against epistemic, temporal, and
frequency adverbials. From a semantic point of view, one might have expected
evidential and epistemic adverbials to pattern together, since as fact-selecting
adverbials or proposition-selecting ones (in the sense of Ernst 2002), they are
the semantically most closely related adverbials in the group mentioned; the
other three types being event-related adverbials. Thus, the partitioning of ad-
verbial classes with respect to their pre-gap occurrences points to an arbitrarily
determined distribution rather than a semantically determined one, and thus
indicates a need for additional machinery.
Pittner’s paper on the positions of adverbials in the German Mittelfeld
argues for a strong influence of semantics on the position of adverbials, but
still assumes syntactic means in defining base positions for adverbials. Her
article is built on earlier work partly carried out together with Werner Frey
(Frey & Pittner 1998; Pittner 1999). According to her there are five (adjoined)
base positions which are determined by the semantic type of their sister con-
stituent (proposition, event, process) and defined by c-command relations
(20). For example, the base position of event-related adverbials is defined as
c-commanding the base position of the highest argument and those of event-
internal adverbials.
(20) sentence adverbials (probably, apparently...)
< frame adverbials (from a legal point of view...)
< event-related adverbials (yesterday, usually...)
< event-internal adverbials (with a screwdriver, with a friend,
in the garden, voluntarily...)
< process-related adverbials (carefully, shyly...)

There can be several semantic categories of adverbials associated with base po-
sitions that meet the same definition. For example, as event-related adverbials,
temporal and causal adverbials have the same base position. Adverbials associ-
ated with different base positions can scramble and thereby produce quantifier
scope ambiguities, such as can be found for varying orders of temporal ad-
verbials and locative ones. Within a base position, though, adverbials of the
different semantic categories associated with this position can be freely ad-
joined in any order without giving rise to scope ambiguities. For example,
 Current issues in the syntax and semantics of adverbials 

scope ambiguities are not found for varying orders of temporal adverbials and
causal ones.
In determining the five base positions, a number of tests are called into
play, which were developed to detect argument base positions in the Mit-
telfeld from which scrambling takes place. These involve data from (i) infor-
mation structure (assuming that focus projection from a verb-adjacent con-
stituent indicates the base position of this constituent), (ii) theme-rheme order,
(iii) complex fronting of participle plus adjacent adverbials, (iv) quantifier
scope (assuming that scope ambiguities arise in scrambled but not in base
order), (v) wh-pronouns (assuming that existentially quantified wh-pronouns
are always in base position), (vi) principle C effects.

. Syntactic and semantic classes of adverbials

. Classifying adverbials

As we have seen, adverbials (or adverb phrases or adverbs) do not form a co-
herent syntactic class in that they all share the same syntactic distribution. Yet it
is not the varied positional behavior alone that has made research on adverbials
a hot spot in linguistic research, but rather the intricate interplay between syn-
tactic and semantic structures. Adverbials that behave syntactically alike usually
also constitute a semantic class. For example, the positional behavior of adver-
bials like linguistically in (9a) is typicial for the whole range of so-called domain
adverbials, which serve to characterize the domain with respect to which the
proposition is said to hold (cf. Ernst). Nonetheless, there is neither consen-
sus about the extent to which the syntactic and semantic classes coincide, nor
about the number and kinds of classes to be distinguished. Note that some
approaches do not make use of any syntactic classification (in the sense of re-
strictions on the occurrence of single adverbials in different base positions).
Any kind of syntactic classification in adjunct approaches like Ernst’s has an
epiphenomenal status. In theories which distinguish base positions, the rela-
tion between base positions and semantic classes can be such that (i) each base
position is related to several semantic classes of adverbials (e.g. Pittner), (ii)
base positions and semantic classes coincide as in a prototypical specifier ap-
proach, or (iii) more base positions than semantic classes are distinguished,
which is the case if one and the same semantic class is assigned a higher and a
lower base position in the clause (cf. Cinque 1999: 106).
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

In most cases when we talk about adverbial classes as classes exhibiting

particular syntactic characteristics, the classes get a label that suggest a se-
mantic basis of the classification. Picking randomly from different classifica-
tions and ordering them roughly from syntactically higher to lower adverbials,
there are speaker-oriented speech act adverbials (frankly) and speaker-oriented
evaluative ones (fortunately), evidential adverbials (evidently), epistemic ad-
verbials (probably), domain adverbials (linguistically), subject-oriented or
agent-oriented adverbials (deliberately), temporal adverbials (now), locative
adverbials (here), quantificational adverbials (frequently), manner adverbials
(slowly), degree adverbs (very), etc. In the course of research on adverbials,
a wide variety of different systems of adverbial classes has been suggested,
ranging from Thomason and Stalnakers (1973) coarse distinction of predi-
cate vs. sentential operators (syntactically VP vs. sentence adverbials) to the
much more elaborate collections of classes found in the works of Laen-
zlinger (1998: 42), Cinque (1999: 106), Ernst (2002: 9), Frey and Pittner (1998)
and others.
The way labels for adverbial classes are used in the literature is sometimes
mysterious, and not only raises questions as to how a particular class is defined,
but also as to what is actually classified. It might seem that the classification
pertains to lexical items (or to phrases in the case of complex adverbials), but
this is not necessarily so. We will have a look at three phenomena where two
different positions of an adverbial are accompanied by two different interpre-
tations, and will see how this affects the classification of the adverbial. Firstly,
as is discussed in Adger and Tsoulas, the change in the position of the adverbial
in (21a) vs. (21b) is accompanied by a change in meaning. In (21b) the testing
of each bulb was slow, while in (21a) the testing of all the bulbs is said to have
been slow.
(21) a. He has been slowly testing some bulbs.
b. He has been testing some bulbs slowly. (Thomason & Stalnaker 1973)

Here, both occurrences of slowly are usually referred to as manner adverbials.

The difference in interpretation would probably be linked to a scopal differ-
ence, while lexically, being event modifiers, both occurrences of the adverbial
would be assumed to have the same semantic representation. It looks like the
semantic label indeed pertains to the lexical item slowly.
The situation is different in (22). The adverbial cleverly in (22a) is usu-
ally referred to as an agent-oriented or subject-oriented adverbial meaning ‘it
was clever of John to hide the present’ while in (22b) it is a manner adverbial
meaning ‘John hid the present in a clever way’.
 Current issues in the syntax and semantics of adverbials 

(22) a. Cleverly, John hid the present.

b. John hid the present cleverly.

Another long-standing debate in adverbial semantics is whether the interpre-

tational difference between (22a) and (22b) is due to different lexical items, in
which two variants of cleverly are distinguished by representing them as hav-
ing a different logical type, or whether the difference can be explained without
polysemy by assuming only scopal differences to be at work (cf. the discus-
sion in Delfitto 2000). In the first case, the different semantic class labels would
be attached to different variants of the adverbial, in the second case the same
adverbial would be assigned to two different semantic classes.
Finally, in cases like (23), the first adverbial is often labelled a frame adver-
bial, with the approximate meaning ‘as far as the time when he was in West-
phalia is concerned, he was a happy man’ while the second one is a standard
locative conveying that the hiking event took place in Westphalia.
(23) a. In Westphalia, he was a happy man.
b. He went hiking in Westphalia.

If we assume that the locative adverbial is a first-order predicate that is under-

specified with respect to selectional restrictions, in that it can either predicate
over an event variable or a topic variable,10 then the adverbial itself is not am-
biguous and the difference in meaning is due to the scope differences induced
by the different position of the adverbials. Again, the classification would not
pertain to the expression itself, but rather to the way it functions in the clause.
Thus, when we talk about semantic classes of adverbials we don’t seem to
talk about classes of lexical items (or phrasal expressions), but rather about
particular processes or templates of semantic composition tied to particu-
lar positions in the syntactic structure. Thus, locative adverb might refer to
an expression triggering a process of intersective event-modification in which
the location of an event is narrowed down, while frame adverb refers to any
expression triggering some kind of intersective modification restricting the
sentence topic. The semantic representation of adverbials as lexical elements
might prove flexible enough to let the adverbial enter into one or more of these
compositional processes.
The often fairly sloppy use of class labels for adverbials reveals that the se-
mantic counterpart of the syntax-semantics interface of adverbial structures
still needs a lot of work. It has long become obvious that a simple distinc-
tion between S-related sentence operators and VP-related predicate operators
is much too coarse. Furthermore, treating adverbials like cleverly as predicate
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

operators has been proven simply inadequate, since this would not yield the
right inferences, does not allow for an extensional treatment, and fails to cap-
ture the distinction between the manner and subject-oriented reading of the
adverbial (cf. the discussion in Delfitto 2000; Eckardt 2002). The rise of event
semantics has considerably furthered our knowledge of the meaning of ad-
verbials, in particular with respect to those adverbials that usually appear in
the lower part of the clause and that can in some way or another be related
to the event argument of the verb. Progress has been less breathtaking for
the semantics of higher adverbials ranging from speaker-oriented ones down
to subject-oriented adverbials, which seem to be not as accessible for event
semantics.11
There are a number of general questions that a semantic classification of
adverbials raises:
– Which ontological sorts (events, states, facts, propositions, etc.) do we
have to distinguish and which sorts of entities do adverbials of the various
classes select?12
– How much underspecification and how much polysemy is involved in
capturing the multifunctionality of many adverbials and what kinds of
type-shifting operations have to be assumed to account for this polysemy?
– What is syntactically determined in the semantic contribution of ad-
verbials, what is determined by pragmatics, discourse connection, world
knowledge, etc.?
– What defines semantic classes of adverbials?

As far as considerations about the semantic and syntactic properties of adver-

bial classes go, the papers of the volume at hand contain new contributions to
the whole domain of adverbial classes. The papers of Pittner, Laenzlinger, and
Engels are not devoted to single classes, but rather aim at explanations of posi-
tional restrictions of a larger range of them. In this section, we will summarize
some of the contributions by the authors of this volume as they concern single
adverbial classes, in particular manner (Geuder, Pittner, Adger and Tsoulas),
locative (Pittner, Adger and Tsoulas), and domain adverbials (Ernst). The pa-
pers by Haumann on degree adverbs and Shaer on temporal adverbials are
dealt with in later sections (5.3, 4.4).

. Manner adverbials

Geuder outlines the semantic differences between depictives, manner adver-

bials, and a class of adverbials he calls transparent adverbs. Depictives are stage-
 Current issues in the syntax and semantics of adverbials 

level predicates of states of individuals in subject or object position which are

linked to the time of the event denoted by the verb (24a). For manner ad-
verbials he claims – in contrast to what has been commonly assumed – that
they predicate over an event without requiring that the respective state holds
for the individual: “manner readings are ‘opaque’ with respect to the property
of individuals denoted by the underlying adjective.” In (24b) the manner ad-
verbial does not imply that the person pretending to be an officer was really
angry when he was shouting. Finally, transparent adverbials are semantically
in-between depictives and manner adverbials. In contrast to manner adver-
bials, but in accordance with depictives, they strictly entail the property of
individuals denoted by its adjectival base during the event time (24c). In con-
trast to depictives, though, there is an implication of a dependency relation
between the state and the event, and not merely an implication that they over-
lap. In (24c) the leaving is most likely understood as depending on the emotion
of anger. The direction of this dependency relation is variable. In (24d) it is the
event that causes the emotional state.
(24) a. John left the meeting angry.
b. – How did you manage to make them believe you were a real officer?
– Well, I kept shouting at them all the time real angrily.
c. John angrily left the meeting.
d. He discovered sadly that the solution was incorrect.

Geuder shows in his paper that although the semantic difference between
manner and transparent adverbials corresponds to a syntactic difference, this
difference is just a tendency in that transparent adverbials tend to occur in pre-
verbal position (25) and manner adverbials in postverbal position (26). But
manner adverbials can occur in preverbal position, too, especially with heavy
material behind the verb (27). Transparent interpretations are sometimes also
possible postverbally.
(25) a. He sadly discovered a mistake.
b. He angrily forwarded me a letter.
(26) a. She walked carefully.
b. She walked carefully on the ice.
(27) a. ?She carefully walked.
b. She carefully walked on the ice.

Manner adverbs have always attracted a lot of attention. Semantically, their

contribution seems to be that of the prototype of an adverb in that it corre-
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

sponds to our schoolbook grammar knowledge that adverbs describe how the
action denoted by the verb is performed. Descriptions of the syntax of man-
ner adverbials are characterized by a wide variance of assumptions. They have
been treated as adjuncts, specifiers, or complements, and their position with
respect to the arguments of the verb is an ongoing matter of dispute. Most de-
bates on the position of manner adverbials are based on the syntactic relation
between manner adverbial and direct object. This also holds for the approaches
of Pittner and Adger and Tsoulas in this volume.
Pittner discusses Eckardt (2003), who points out differences between the
behavior of manner adverbials with indefinite versus definite NP objects. These
differences, among other things, lead Eckardt to assume that manner adverbs
can adjoin above the object, and that in case of object-adverb-verb order, the
pre-adverb position of an indefinite object is licensed by topicality.
Pittner, on the other hand, argues that manner adverbials minimally c-
command the verbal complex, i.e. their base position is below the object, since
(i) an indefinite object wh-pronoun occurs to the left of the adverbial, (ii)
quantifier scope ambiguity can be observed when the adverbial occurs to the
left of the object, (iii) complex fronting of adverbial plus participle is possible,
and (iv) ambiguous adverbials occur in postobject position when a manner
reading is intended (28).
(28) a. Du musst langsam das Fenster zumachen.
you must slowly the window close
‘You must close the window soon’
b. Du musst das Fenster langsam zumachen.
you must the window slowly close
‘You must close the window slowly’ or ‘You must close the window
soon’

The crucial assumption made by Pittner is that indefinite proto-patient ob-

jects, and only those, are integrated into the verbal complex (in the sense of
Jacobs 1993), and are thus capable of projecting focus. This is why the manner
adverbial cannot intervene between verb and object in (29).
(29) weil sie schüchtern einen Prinzen geküsst hat
because she shyly a prince kisses has
‘because she kissed a prince shyly’

Adger and Tsoulas cash in on observations going back to Harris (1968) that in
general only those transitive verbs that select affected objects also allow manner
adverbials. These verbs assign structural case and are therefore true agentive
 Current issues in the syntax and semantics of adverbials 

verbs, for example (30a). In (30b) to (30d) the verbs assign inherent case and
are considered unaccusative:
(30) a. He dried the dishes carefully.
b. *John resembled Sue slowly.
c. *He aggravated me revoltingly.
d. *She desired a raise enthusiastically.
e. John walked jerkily to the cliff edge.

These facts prompt Adger and Tsoulas to assume that the licensing of manner
adverbials is connected to the licensing of agents. Their analysis is worked out
within a Minimalist framework in which a lower functional structure [Asp1P [vP
[Asp2P [VP ]]]] is proposed, with Asp1P related to sentential aspect and Asp2P
to lexical aspect. “Little” v probes for φ–features, which are satisfied by the
object, and at the same time v licenses manner adverbials by an EPP-feature.13
While the object, as the probe’s goal, deletes the φ–features when moved into
the appropriate Spec position, the merging manner adverbials don’t delete the
EPP-features. This allows the iteration of manner adverbials as inner specifiers
of vP. Verbs like (30b) to (30d) do not project a vP and thus do not license
manner adverbials, while seemingly intransitive verbs like (30e) are considered
to be transitive verbs with incorporated objects. Thus, they project vP and,
correspondingly, allow manner adverbials.

. Locative adverbials

The variable nature of locative adverbials is precisely accounted for by Maien-

born (1996, 2001). She proposes three structural options: (i) locative adverbials
may occur occur as V-adjuncts, e.g. an den Ohren in (31), which are interpreted
as event-internal modifiers (ii) locative adverbials may occur as VP adjuncts,
e.g. auf dem Marktplatz in (31), in which case they locate the event as a whole,
or (iii) locative adverbials may appear as adjuncts within the C-domain, e.g. in
den Anden in (31), acting as frame adverbials:
(31) In den Anden werden Schafe vom Pfarrer auf dem Marktplatz
in the Andes are sheep from.the priest on the marketplace
an den Ohren gebrandmarkt.
at the ears branded
‘In the Andes sheep are branded on their ears in the market square by the
priest.’ (Maienborn 2001: 199)
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

Both Adger and Tsoulas and Pittner treat the syntax of locatives. More specif-
ically, they deal with the syntax of those canonical locatives that Maienborn
treats as VP-adjuncts. Pittner is particularly concerned with the relation be-
tween locative and temporal adverbials. In contrast to Laenzlinger (1998) and
Haider (2000), she assumes that locatives and temporals differ in their base po-
sition, canonical locatives being lower than temporal adverbials. Data support
comes from focus projection, quantifier scope, theme-rheme ordering, and the
position of indefinite wh-pronouns.
Adger and Tsoulas provide an explanation for locative adverbials that re-
lates to the one offered for manner adverbials. While the licensing of manner
adverbials is connected to the licensing of agents, here it is the functional struc-
ture for lexical aspect that is involved in licensing the locative adverbials. The
connection between lexical aspect and locatives is illustrated in (32) where a
prepositional locative (32a) alternates with a DP (32b). It is only the latter that
leads to a telic interpretation:
(32) a. They climbed up the mountain.
b. They climbed the mountain.

Quantized object DP/NPs (e.g. that house in contrast to houses) have in com-
mon with locatives that they are related to a quantization feature in Asp2P. This
commonality is justified by Adger and Tsoulas’s assumption that both quan-
tized objects and locatives measure the physical extent of the event. The telic
interpretation that arises with quantized objects, but not with locatives, is pro-
posed to be a result of the feature mechanism which guarantees that only a
quantized object that moves to SpecAsp2P leads to the deletion of the inter-
pretable quantization feature. The merging locative does not have this ability.

. Domain adverbials

The domain adverbials which are treated in Ernst’s paper are particularly in-
teresting because they show a wide range of possible positions without any
meaning differences, as in (33a). There are two semantically motivated restric-
tions for the use of these adverbials, though. Firstly, it is not their only function
to tie the eventuality to a particular domain as in (33a); they can also modify
another adverb as philosophically does with clearly in (33b), which could have
been uttered in a conversation about immediate causation. The final position
of the adverbial is excluded because it has to take scope over the element it
modifies. Note that the sentence-final position of the adverbials in (33), as a
 Current issues in the syntax and semantics of adverbials 

right-adjoined position, has scope only over the lower VP that it modifies in
the framework Ernst assumes:
(33) a. (Physically,) the climb (physically) had (physically) become (physically)
more difficult (physically.)
b. (Philosophically,) Bill did not clearly start the car himself (*philosophi-
cally.)

To account for the fact that a domain adverbial can take the scope indi-
cated by its surface position in the sentence, the proposed analysis involves
domain adverbials as overt realizations of complex operators of the format
‘IN-POLITICAL, ETC.-DOMAIN’ which can home in on the eventuality, some
part of it, or some other element higher than down in the basic event. ‘IN-X-
DOM [...]’ is to be interpreted as: “some contextually salient entity within [...]
is in the X domain”. Secondly, Ernst argues that domain adverbials in sentence
initial position may not only have a regular reading in which our understanding
of some predicate is restricted in its scope to a particular real-world domain,
but also a topic reading, in which they specify a previously given topic as in
(34), where economically and politically do not restrict the predicate or any
other element in the sentence, but rather the topic given in the question:
(34) – What have they done in their last two years in office?
– Well, economically, they have passed new tax legislation; politically, they
have raised far more money for the party than was expected.

This behavior is attributed to the existence of an optional topic rule which is

tied to the sentence-inital position, This rule affects not only domain adver-
bials, but also temporal, locative, and other ones.

. Structural intricacies

. More phenomena

Apart from the phenomena we have already encountered in Sections 2 and 3,

there are a number of other intricate phenomena which pose a challenge to the-
ories of the positioning of adverbials and those of the syntax-semantics inter-
face. Some of the more frequently discussed ones are briefly summarized here.
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

.. Stacking
Series of adverbials of the same adverbial class (35a) or sometimes even differ-
ent ones (35b) can seemingly form a single constituent. This is revealed by their
position in the German Vorfeld which is assumed to allow only one constituent.
Not all combinations of adverbials are admissible in a stacking configuration,
though (35c):
(35) a. Unter einer Laterne in der Schönhauser Allee hielt das
under a streetlight in the Schönhauser Avenue stopped the
Auto.
car
‘The car stopped under a streetlight on Schönhauser Avenue.’
(cf. Steinitz 1973: 48)
b. In zwei Tagen am Strand wenn die Sonne untergeht wirst
in two days at.the beach when the sun under.goes will
du es sehen.
you it see
‘When the sun goes down at the beach in two days you will see it’
(cf. Haider 2000: 101)
c. *Mit großer Wahrscheinlichkeit in Wuppertal werden wir
with high probability in Wuppertal will we
uns treffen.
each.other meet

.. Mirror order

If adverbials occur to the right of the verb (36b) they often appear in reverse
order compared to preverbal position (36a) (cf. Section 3.3).
(36) a. Er hat letztes Jahr in Wien mit einer Büroklammer ein
he has last year in Vienna with a paper.clip a
Schloßi geknackt, dasi . . .
lock picked which ...
‘Last year in Vienna he picked a lock with a [paper clip]i , whichi ...’
b. Er hat ein Schloßi geknackt, mit einer Büroklammer in
he has a lock picked with a paper.clip in
Wien letztes Jahr, dasi . . .
Vienna last year which . . .
‘Last year in Vienna he picked a lock with a [paper clip]i , whichi ...’
(cf. Haider 2004)
 Current issues in the syntax and semantics of adverbials 

.. Edge effects

Within head-initial phrases, modifiers before the head cannot be extended to
the right; cf. the adverbial with respect to the left-headed English VP (37b)
versus the same adverbial in the corresponding right-headed German VP (37a).
(37) a. Er hat es [sehr viel sorfältiger als jeder andere] analysiert.
he has it very much carefullier than any other analyzed
b. He has [(much more) carefully (*than anyone else)] analyzed it. (Haider
2004)

.. Obligatory adverbials

With some verbs, accompanying adverbials are obligatory; this raises the ques-
tion whether such obligatory adverbials should be treated as complements, and
whether this also holds of non-obligatory adverbials of the same class as the
obligatory ones (cf. Section 2.2).
(38) a. John worded the letter carefully.
b. *John worded the letter. (Jackendoff 1972: 64)

.. Adverbials before gaps

Adverbials show particular restrictions with respect to the possibility of their
occurrence in front of gaps. It seems that adverbial class, the type of gap, and
other conditions jointly determine the grammaticality of adverb-gap construc-
tions, as in the following examples discussed in Engels (see also Section 3.2).
(39) a. John hasn’t gotten along well with Grandpa lately.
– Has he ever __? (Baker 1981: 313)
b. John intended to finish all he set out to do.
– *Has he clearly __? (Ernst 1983: 548)

.. Left periphery

Adverbials which usually occur lower in the clause can show up in the left
peripheral position of the clause, where they seem to perform particular
discourse-related functions (Section 3.4, 4.4).

.. “Lite” adverbs

Some adverbs (which are often but not always short) in VO languages are
confined to a preverbal position close to the verb; they cannot also occur
postverbally, as e.g. manner adverbials can:
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

(40) a. The government has (hardly) proven his case (*hardly.) (Ernst 2002: 172)
b. Jean a (bien) fait (*bien) son travail (*bien)
Jean has (well) done (well) his work (well) (Laenzlinger 1998: 52)

Three of these phenomena are treated in detail in papers of this volume and are
briefly reviewed here: adverbials in front of gaps, mirror order of adverbials,
and adverbials in the left periphery.

. Adverbials and gaps

A phenomenon that has long been observed but only sporadically treated in
research on adverbial syntax is the occurrence of adverbials in front of gaps.
This is the topic of Engels’s paper. She investigates the possibility of adver-
bials appearing in front of VP gaps that are due to VP fronting, topicalization,
wh-movement, or VP ellipsis (which is understood as VP fronting with subse-
quent deletion). She shows that three factors influence the grammaticality of
adverbials before a gap: (i) the type of adverbial (her study covers epistemic
adverbials like probably, temporal adverbials like always, frequency adverbials
like often, evidential adverbials like clearly, and subject-oriented adverbials like
wisely), (ii) wide-scope versus narrow-scope reading of the adverbial,14 and (iii)
the syntactic construction the adverbial appears in, in particular the distinc-
tion between subject-auxiliary inverted constructions versus non-inverted con-
structions, and constructions with wh-movement versus constructions with
VP fronting, topicalization or VP ellipsis (the latter three involving movement
in clause-initial position). The interaction of these parameters leads to a num-
ber of generalizations, some of which will be mentioned here. One of them
says that in non-inverted constructions, wide scope adverbials cannot occur
postverbally immediately before a gap (41), where (42) shows that it is not
the post-auxiliary position of the adverbial, but rather the occurrence of overt
material between the adverbial and the gap that is crucial.
(41) a. Fred has never been rude to grandfather, but (been rude to grandfather)
John probably / always / often / clearly / wisely has __ (Baker 1981)
b. *Fred has never been rude to grandfather, but (been rude to grandfather)
John has probably / always / often / clearly / wisely __
(42) Bill is living in France, but John is probably not __ (Baker 1971)

With narrow scope adverbials in non-inverted constructions, it is the type of

gap and the type of adverbial that determine the grammaticality of a pre-gap
occurrence of the adverbial. This holds similarly for inverted constructions,
 Current issues in the syntax and semantics of adverbials 

in which all adverbials can immediately precede a gap resulting from wh-
movement (43), while only epistemic (44a), temporal (44b) and frequency
(44c) adverbials, but not subject-oriented (44d) and evidential (44e) ones, can
occur immediately before a gap resulting from VP ellipsis or fronting (see also
Section 2.4).
(43) a. Where is she often __?
b. In which pub was he wisely __?
(44) a. Terry knows how to build an h-bomb. Does he really __? (Ernst 1983)
b. John has never gotten along well with Fred, and Mary hasn’t always __
c. John has to take his medicine, and (take his medicine,) he must regularly
__
d. John intended to finish all he set out to do. *Has he wisely __? (Ernst
1983)
e. John is a burglar, but a murderer, *he isn’t clearly __

. Right periphery and mirror order

Given that there is a typical or standard order of preverbal adverbials crosslin-

guistically, it can be observed that in languages in which these adverbials op-
tionally or obligatorily appear in postverbal position, they can either appear
there in the standard order or in an order that is a mirror image of the stan-
dard order. This phenomenon is equally puzzling for both adjunct and specifier
approaches to adverbial syntax. The mirror order is illustrated in postverbal
position in (45b), in which a reversal of the standard preverbal order (45a) can
be observed:
(45) a. Jean l’a maintenant souvent rapidement refait.
Jean it.has now often quickly redone.
b. Jean l’a refait rapidement souvent maintenant.
Jean it.has redone quickly often now (Laenzlinger)
Laenzlinger, working under the specifier hypothesis, assumes that mirror or-
der arises as a consequence of snowball movement, in which the VP-remnant
moves above the rightmost adverbial and then the projection containing the
verb and the adverbial moves further up, recursively collecting further adver-
bials. French also allows the standard adverbial order in postverbal position.
This is achieved by VP-remnant movement to AuxP and object movement to a
higher ObjP above the relevant adverb projections.15
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

Pittner tackles mirror order phenomena as an adherent of an adjunct ap-

proach to adverbial positioning. Her data come from English and German. In
English, adverbial order is reversed in postverbal position, while in German,
mirror order phenomena can be observed in the right extraposition field. Mir-
ror order, she argues, is to be expected under the assumption of right adjunc-
tion. But according to Larson (1988), binding data and negative polarity data,
which are taken to indicate c-command, show that the object c-commands
these adverbials, which excludes right adjunction. On the other hand, Lar-
son cannot explain the scope data that show up in right position. Rejecting
Cinque’s movement approach to the dilemma, Pittner assumes that adverbials
cannot be interpreted in final position. Instead, she assumes a “modification
marker” which links adverbials to their preverbal place. This is restricted by
the condition that lines of modification markers may not cross. Order and sco-
pal relations are now explained by modification markers, and binding data by
c-command.

. Left periphery and discourse connection

There is something special about adverbials in sentence-initial position in that

adverbials which are usually found in other positions in the standard sequence
of adverbials can appear in this position in the service of performing partic-
ular discourse functions. The most obvious function is that of determining
the topic. An optional topic rule linked to the sentence-initial position is pro-
posed by Ernst that accounts for the topic interpretation of many adverbials
base-generated in this position. This is in the general spirit of Ernst’s (2002)
adjunct approach. Laenzlinger, assuming adverbials to be specifiers, instead
argues for certain movement operations which put constituents containing ad-
verbials in the specifier of a TopicPhrase. Movement of an AdvP is involved
when adverbials are topicalized (SpecTopP = SpecCP) in main clauses (46a).
Topicalization can also be observed in subordinate clauses, where adverbial
and object scrambling interact, resulting in both phrases moving into recursive
TopPs above SubjP (46b). Finally, German also allows adverbials to be preposed
together with an adjacent participle (and possibly the object) (46c). This is ac-
counted for by VP-remnant movement which can only affect projections below
ModP (46d).
(46) a. Oft hat Hans dieses Buch gelesen
often has Hans this book read
‘Hans often read this book’
 Current issues in the syntax and semantics of adverbials 

b. ...,weil wahrscheinlich diesen Brief der Mann meinem Onkel

because probably this letter the man my-dat uncle
gesendet hat
sent has
‘because the man has probably sent this letter to my uncle’
c. Die Sonate gut gespielt hat Hans.
the sonata well played has Hans
‘(It was) the sonata (that) Hans played well.’
d. *Leider die Sonate gespielt hat Hans.
unfortunately the sonata played has Hans
‘Hans unfortunately played the sonata.’

The most thorough investigation of left-peripheral adverbials, though, is pre-

sented in Shaer’s contribution. His starting point is the widely held assumption
that a temporal adverbial in the right periphery (RP) yields ambiguity while the
same adverbial in a left peripheral (LP) position doesn’t. Shaer shows that this
assumption is wrong. Inter alia, it has been claimed that (i) quantifier scope
ambiguities arise only for temporal adverbials in RP position (47a), but not
LP position (47b) (de Swart 1999), (ii) the present perfect yields an up-to-now
and an existential reading in RP position (47c), but only the up-to-now read-
ing in LP position (47d) (Dowty 1979), and (iii) in non-generic sentences, RP
adverbials can be interpreted as denoting a position on the time line that is
definite or indefinite (47e), while LP adverbials are confined to the reading in
which the position on the time line is definite (47f) (Hitzeman 1997):
(47) a. Every student on campus went hiking on a beautiful Sunday in spring.
(de Swart 1999)
b. On a beautiful Sunday in spring, every student on campus went hiking.
c. John has lived in Boston for four years. (Dowty 1979)
d. For four years, John has lived in Boston.
e. John will live in Brooklyn in the summer. (Hitzeman 1997)
f. In the summer, John will live in Brooklyn.

Shaer convincingly shows that LP adverbials can have both interpretations,

too, if the right context is provided. In (48a) the LP adverbial has narrow
scope, in (48b) it induces an existential reading of the perfect, and in (48c)
it has a reading in which its position on the time line is indefinite. Thus, the
“high” interpretations (wide scope, up-to-now perfect, position-definite) of-
ten suggested for LP temporal adverbials are no more than interpretational
tendencies:
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

(48) a. Every Sunday, a certain group of students goes hiking in the foothills.
They always wear Norwegian sweaters, and so are known by their class-
mates as “the Norwegians.”
b. For two weeks, John has been a milkman. For three weeks, he has been a
paperboy. Now he wants a desk job.
c. For two hours, John has been unconscious, but not even John knows
exactly which two hours during the past twenty-four these were.

Shaer argues that LP adverbials are not moved into a position that is struc-
turally higher than the RP position. He rejects this analysis, which might ac-
count for the similarities between the interpretation in LP and RP positions, on
the basis of data from binding, island constraints, and VP-ellipsis. Instead, he
assumes that LP temporal adverbials are base-generated in an extra-sentential
position that allows a particular link to the discourse. The “orphan” analysis
puts adverbials in a position which is independent from the host sentence, in-
volves no hierarchical relations to other elements in the clause, and therefore
also no c-command relations. The “orphan” analysis is shown to be consistent
with the binding, island, and VP ellipsis data that were shown to be problematic
for a movement account.16
The orphan analysis alone would lead us to expect a greater freedom of
interpretation for LP adverbials than is actually found. A number of interpre-
tational constraints can be explained by lexical properties of the adverbials and
how they can interact with the aspectual meaning of the VP. Furthermore, in
order to account for the fact that “high” interpretations (wide scope, up-to-
now perfect, position definite) of adverbials are more salient with LP than with
RP adverbials, Shaer proposes a solution which draws on the similarity with
other preposing constructions such as topicalization, inversion, and left dislo-
cation. Based on Birner and Ward (1998), he develops a pragmatic approach
that allows LP adverbials to engage in a relationship with elements that have
been evoked in prior discourse. Discussing a number of examples, he shows
how this connection makes a “high” interpretation of LP adverbials more likely
while still allowing a “low” interpretation when particular elements are present
in the context.
 Current issues in the syntax and semantics of adverbials 

. Categorial issues

. Adverbs, AdvPs, and adverbials

Adverbs have always given rise to categorial considerations of different sorts,

three of which we would like to touch upon, namely the questions (i) what
the relation is between adverbs, adverb phrases, and adverbials, (ii) what cat-
egory adverbs belong to and how they are related to other categories, in
particular to adjectives, and (iii) how adverbs and adverbial-like phrasal ex-
pressions behave within projections dominated by non-verbal categories like
adjectives or nouns.
As has been already addressed in Section 1, the relation between adverbs,
adverb phrases, and adverbials is a difficult matter. The urgency of distinguish-
ing these concepts, and the degree of explicitness involved in doing so depends
on the syntactic approach pursued and the phenomena to be tackled. The gen-
eral concensus in both specifier and adjunct approaches to adverbs is that single
adverbs are assumed to be heads of adverb phrases. Usually, non-extended ad-
verb phrase is what is meant when the term adverb is used, and what is said
about adverbs carries over to a certain degree to extended AdvPs and adverbials
of other categories. However, this is not always the case. Firstly, there might
be exceptions like degree adverbs, which have been treated as heads of certain
functional projections (see, for example, Haumann’s contribution). Secondly,
as soon as adverbs are extended by complements (e.g. unfortunately for him),
they are banned from certain positions in some languages, apparently those
with head-initial verb projections. This phenomenon is discussed under the
label edge effects in Section 4.1. Things don’t get much clearer when the term
adverbial(s) comes into play. There are indeed items of other categories that can
fulfill the semantic functions associated with single adverbs (49a) or extended
ones (49b). Instead of temporal adverb phrases, we can as well use tempo-
ral PPs (49c), temporal DPs (49d), finite temporal clauses (49e), or non-finite
temporal clauses (49f). A common function ties them together as temporal
adverbials.
(49) a. She left immediately.
b. She left right then.
c. She left after her brother.
d. She left that night.
e. She left when the curtain fell.
f. She left before having finished her lunch.
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

From a syntactic point of view, it is important to clearly distinguish among

these terms, since distributional differences between adverbials, non-extended
adverbs, and AdvPs are involved. There are also complicated effects of the com-
bined factors of relative phonological weight and phrasal complexity on the
mobility of constituents. In addition, whether lighter items have a freer dis-
tribution than heavier ones or the other way round depends on the zone of
the clause in which they occur. Furthermore, differences in the positional po-
tential of adverbials depend on whether the adverb is extended to the right or
to the left.
Although the difference between adverbials as phrases headed by adverbs,
rather than an item of some other category, such as a preposition, nominal,
or complementizer, is not a central topic in this book, there is an assumption
made by Adger and Tsoulas relevant to this topic that should be mentioned
here. According to Costa (1996), the base position of postverbal manner ad-
verbs (50a), but not that of temporal ones (50b), is closer to the verb than a
PP argument, here at a picture of t, as the extraposition data show. Adger and
Tsoulas show that PP locative adverbials (50c) and PP manner adverbials (50d)
behave like simple temporal adverbs. This leads them to the conclusion that
the whole class of PP adverbials assumes a different postverbal position than
simple manner adverbs.
(50) a. Which woman did he glance quickly at a picture of t?
b. *Which woman did he glance yesterday at a picture of t?
c. *Which woman did he glance in the Louvre at a picture of t?
d. *Which woman did he glance in a sultry way at a picture of t?

With respect to adverbial clauses, it is interesting to note in this context the view
put forth by Pittner that their base position is the same as other adverbials of
the same semantic class.

. Adverbs and adjectives

One of the longstanding debates in adverb research is the attempt to deter-

mine the categorial status of adverbs.17 On the one hand, adverbs seem to be
related to adverbials which are phrases, like PP adverbials or clausal adver-
bials, in that they perform similar syntactic and semantic functions, on the
other hand, they seem to be very close to adjectives to which they are often re-
lated from a morphological and lexical semantic view. That is, many languages
have morphological means at their disposal to derive adverbs from adjectives
(cf. English quick-ly, Spanish elegante-mente ‘elegantly’ , German interessant-
 Current issues in the syntax and semantics of adverbials 

erweise ‘interestingly’), and some at least sometimes don’t distinguish adverbs

from adjectives morphologically.18 In German, for example, adverbs and pred-
icative adjectives like schnell ‘fast’ are morphologically unaltered stems. Thus,
the relationship between adjectives and adverbs raises a number of questions:
– Does the sequence of adjectives within DPs (in particular those with dever-
bal nominalizations) show the same ordering restrictions as the sequence
of adverbs in the verbal domain?
– Since adjectives often agree with the noun they modify, should it be as-
sumed that adverbs also are in an agreement relation with the verbal pro-
jection they attach to? (This is very suggestive if it is assumed that adverbs
occur in a Spec-head configuration and -ly is an agreement marker.)
– How are particular morphological means for the derivation of adverbs
related to particular functions to be performed by these adverbs? (For ex-
ample, adverbs ending in -weise in German only function as sentential
adverbs of certain kinds.)
– What is the crucial semantic difference between a stem functioning as an
adverb versus one functioning as an adjective?

The latter question is addressed in Geuder’s paper and embedded in an analysis

of manner adverbials, transparent adverbials, and depictives. As we have seen
above in Section 3.2, Geuder shows that manner adverbials are distinguished
from transparent adverbials and depictives in that the former predicate over
events while the latter modify states. The difference between transparent adver-
bials and depictives lies in the fact that while both require a temporal overlap
between the event denoted by the verb and the state they pertain to, only trans-
parent adverbials imply a dependency relation between state and event. Thus,
the morphological distinction between adverbs and adjectives in English is not
one between event and state predicates, as has sometimes been assumed, but
between predicates that require a dependency relation between event and state
and those that do not.

. Adverbs and adverbials outside the verbal domain

Adverbs and adverbials not only occur in the verbal domain, but also in nomi-
nal, adjectival, and adverbial projections:
(51) a. His submitting/writing the paper quickly surprised me.
b. die angeblich in Italien seit langem glückliche Diva
the allegedly in Italy since long happy diva
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

‘the diva allegedly happy in Italy for a long time now’

c. extremely quickly

Since many adverbials do not perform functions in the non-verbal domain

which are essentially different from those in the verbal domain,19 any assump-
tions about lexical, syntactic, and compositional matters will have to be general
enough to account for that similarity. In particular, the following questions will
have to be answered:
– How do adverbs and adverbials semantically relate to the adjectival, nomi-
nal, adverbial elements they modify?
– What kind of hierarchical order of adverbial elements can be observed
within projections outside the verbal domain? Does it correspond to the
hierarchy observed in the verbal domain?
– Which semantic classes of adverbials can occur within the nominal domain
(in particular with deverbal nominals) and within the adjectival domain?
– What is the categorial status of adverbial elements in non-verbal projec-
tions and in which parts of local structures (head, specifier, complement,
adjunct) do they occur (see also Section 2.2)?
– Which functional projections have to be assumed within the nominal and
adjectival domains?
– How can the distributional properties of projections outside the verbal
domain that contain adverbs or adverbials be explained?

The last three questions, in particular the very last one, are addressed in Hau-
mann’s paper, who investigates the behavior of APs and AdvPs with degree
adverbs. Earlier work by Bresnan (1973) and others has shown that three classes
of degree adverbs can be distinguished, we’ll call them Class 1, Class 2, and
Class 3 adverbs here. Class 1 adverbs (e.g. abundantly, fairly, very, etc.) differ
from Class 2 adverbs (as, so, too) and Class 3 adverbs (enough, less, more) in
that they cannot precede the indefinite article together with the adjective:
(52) a. as beneficial / happy enough / too attractive / less complicated / more
economic / so wonderful an arrangement
b. *abundantly clear / *fairly clear / *very clear an argument

Class 2 and Class 3 adverbs also differ from Class 1 adverbs in that they resist
though attraction (53) and license result clauses (54):
(53) a. *as beneficial though the arrangement was
b. very clear though the argument was
(54) a. At the end of the day, a system is only as good as the user wants it to be.
 Current issues in the syntax and semantics of adverbials 

b. *At the end of the day, a system is only fairly good as the user wants it to
be.

Finally, Class 3 adverbs have a wider distribution than Class 2 adverbs in that
they occur within extended nominal, prepositional, and verbal projections:
(55) a. There had not been enough evidence / *as evidence.
b. These aspects of language performance are more under conscious con-
trol / *too under conscious control.
c. This meant they could drink more / *too.

Based on Corver’s (1997) Split-DegP Hypothesis, Haumann assumes Class 2

adverbs to be heads of degree phrases and Class 3 adverbs to be heads of quan-
tifier phrases. Her contribution is based on two important observations. Firstly,
in the left-peripheral position within the extended nominal projection (the
so-called Big Mess Construction), only headed DegPs and QPs are admissi-
ble (56), while in the right-peripheral position (postcopular, postnominal, or
postverbal), headed DegPs and QPs as well as non-headed ones are equally
acceptable (57):
(56) a. I found the bright lights of the big city *attractive a prospect to refuse /
too attractive a prospect to refuse.
b. It turned out to be *ridiculous an idea than we thought / less ridiculous
an idea than we thought.
(57) a. In fact, his decision was revolutionary / less revolutionary than it ap-
peared.
b. I’ve never met a pathologist cheerful about her work / so cheerful about
her work.
c. The violins had been taking the tempo fast / slightly too fast.

Secondly, headed QPs are admissible in prenominal and preverbal positions,

but headed DegPs are not:
(58) a. Many teachers do not have a *so strong grasp of the language to teach it
effectively.
b. Many teachers do not have a strong enough grasp of the language to teach
it effectively.
(59) a. They will *as carefully test the completeness of the text.
b. They will more carefully test the completeness of the text.

The two observations require different explanations. The data in (56) and
(57) are explained by the assumption that in preverbal position, the differ-
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

ence between DegPs and QPs arises because DegPs require a result clause for
interpretation, which provides the relative standard for the degree expressed,
but fail to license one because the QP as the antecedent of the result clause
would be too deeply embedded. With QPs on the other hand, which share
the embedding problem, the standard can be inferred. With respect to (58)
Haumann argues that in prenominal position, the difference between DegPs
and QPs arises from the fact that DegPs are parasitic on predicative adjectives,
and are thus unacceptable in prenominal position, while QPs co-occur with
attributive and predicative adjectives. Thus, the presence of DegPs and QPs
within the extended adjectival projection is not only licensed by the referential
argument of the adjective, the degree argument d, but also by the presence of an
external argument in the adjective’s argument structure, which is only present
with predicative adjectives and is realized as DP or PRO.

. Conclusion and outlook

What has been achieved by this volume? For one, it has provided an insight
into the vivid debate about whether adverbials are adjuncts (Engels, Ernst,
Geuder, Pittner), specifiers (Adger and Tsoulas, Laenzlinger), or – in partic-
ular domains – even heads whose maximal projection dominates the modified
phrase (Haumann), or syntactic “orphans” (Shaer). Also, besides having im-
proved our understanding of many single phenomena and adverbial classes,
it has shown how crosslinguistic variation can be handled in a feature-driven
Minimalistic framework (Laenzlinger), or – as a new approach within research
on adverbials – in Optimality Theory (Engels). Interesting crosslinguistic issues
have also been addressed by Geuder and Pittner.
Assumptions about adverbials have stimulated the development of syntac-
tic and semantic theories to a great extent, yet in our opinion there remains a
lack of solid in-depth descriptions of many of the phenomena that are crucial
for these theories. Thus, we would like to emphasize that the papers by Engels,
Geuder, Haumann, and Shaer in particular have confronted us with new data
or have corrected older descriptions of data with the consequence that former
theoretical approaches had to be revised. Strikingly, the two papers that have
approached the interface issue from a more semantic point of view (Shaer,
Geuder) have both come to the conclusion that with respect to the phenom-
ena they have treated (temporal adverbials in the left periphery, distinction
between transparent and manner adverbs), syntactic position is merely sug-
gestive of a particular interpretation rather than responsible for determining it
 Current issues in the syntax and semantics of adverbials 

in a strictly compositional way. If more detailled research on the semantics of

particular syntactic positions for adverbials reveals that these are often linked
just to tendencies of semantic interpretation, then this will certainly have an
impact on how we formally handle interface issues.
Looking ahead, the overview provided in this introductory article suggests
that future research might be confronted with the following issues:
– A lot more semantic work on many adverb classes, in particular the higher
adverbs, is required to answer the question in how far adverbial syntax is
detemined by adverbial semantics.
– The exploration of the parallels between adverbs and adjectives will cer-
tainly be a future topic, and it will be one that in particular elicits more
work on the adjectival side of things.
– Although we are certainly not the first to notice this, it bears repeating that
more work on adverbials in domains other than the verbal one is still a
desideratum.
– Although the intensity with which research on information structure and
research on adverbials has been carried out in the last decade is roughly
comparable, both lines of research seem not as tightly integrated as they
need to be to further our knowledge of how position and interpretation
are related in the domain of adverbials. The semantics of what is dealt with
syntactically under the heading topicalization of adverbs or left-periphery
(notwithstanding Shaer’s progress in this area) still remains rather obscure,
and phenomena relating to focus (e.g. the non-focussability of certain ad-
verbials, or the interaction of focus and adverbial position as e.g. addressed
by Pittner) are often rather too sparsely discussed in syntactic work on
adverbials.
– Closely connected with the last issue is the expectation that future research
will reveal more about parenthetical and afterthought constructions and
their particular discourse relations.20
– A domain that is hardly touched on in the interface literature on adverbials
is language change. One might expect interesting insights from questions
like how basic changes in word order have affected the position and in-
terpretation of particular adverbials or – considering the diversity of ad-
verbials with respect to their inner morphological or syntactic structure –
how particular expressions came to work as adverbials.
– Although Cinque (1999) in his seminal work has drawn on a considerable
amount of crosslinguistic investigation, most research still centers on the
better-known Indo-European languages. Significant modifications to the-
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

ories of adverbial positioning are likely to be in store when data from a

broader range of languages is treated less peripherally.
– It is usually considered a matter of taste or heuristics whether highly intri-
cate theories should be developed first and then be gradually confronted
with the wealth of data, or if thorough empirical work should precede the
development of theories. Either way, it seems high time to supplement
the highly developed state of theory construction with more descriptive
corpus-oriented work.

And now enjoy yourself with the following papers. May they enhance your
knowledge about the syntax and semantics of adverbials and stimulate your
own thoughts on the topic.

Notes

. References in bold print pertain to contributions in this volume.

. The term adjunct is also not entirely satisfactory as a cover term, since it excludes se-
lected adverbials. The identification of the criteria by which it is decided that an adverbial
is selected (i.e. by the verb) is, of course, a very interesting matter for investigations into
the nature of how syntax, semantics, and pragmatics interact. In addition, the term adjunct
already has a special status in generative grammar.
. In the Minimalist Program, the X-bar schema as such is done away with as an indepen-
dent primitive of grammar. On this Chomsky (2000: 26) intones: “... no indices, no phrase
structure per X-bar theory — all that has to go.” It is the definition of the operation Merge
that deals X-bar theory the death blow as a theoretical primitive. A clear presentation of this
can be found in Adger et al. (1999: 6): Given that Merge defines constituents of the struc-
ture as sets of sets within the structure, they comment that : “Using sets in this way means
that no appeal can be made to the notions of X-bar or XP as generalizations over structure:
they can only be thought of as derived concepts stemming from relationships in particular
structures. The more commonly used tree structures may still be used as a matter of ease
and familiarity of notation, but they contain by their very nature extra information which is
actually inaccessible to the computational system.”
. As can be expected with selectional phenomena, there are idiosyncratic differences be-
tween languages as to which adverbs can take complements. There also appear to be sys-
tematic differences with respect to the positions that extended AdvPs can appear in (see
Section 4.1 on edge effects.)
. A number of semantic and syntactic arguments for treating adverbials of certain classes
(e.g. temporal and manner adverbials) as complements can be found in Alexiadou (1997).
The data drawn to support this view come from semantic selection, referentiality, extraction,
and adverb incorporation.
 Current issues in the syntax and semantics of adverbials 

. An EPP-feature is a non-semantic selectional feature that determines that something

occupies the specifier of certain functional categories. The name comes from the pre-
Minimalism Extended Projection Principle, which held that every sentence has a subject.
. A VP/vP-remnant is a VP-shell structure which has been emptied of its DPs, leaving only
the verb.
. Two prominent critics of the specifier approach are Thomas Ernst (Ernst 2002) and
Hubert Haider (Haider 2000, 2004).
. See Rosengren (2003: Sections 3.5 and 4) for refinements of Haider’s (2000) analysis, in
particular her suggestion that postverbal modifiers in English and Swedish are adjoined to
an empty VP following the lexical VP-shell, which is “MAC (Minimal Argument-Projection
Complex)” in her terminology. Furthermore, a discussion of constraints on high right
adjunction can be found in Ernst (2003).
. Cf. also Maienborn (2001) for the assumption that frame adverbials predicate over
sentence topics.
. Intriguingly, Cinque (2004) essentially claims a degree of immunity to criticism of his
specifier approach, coming from adjunct-approach advocates in particular, until more elab-
orate work on semantic classes of adverbials is carried out.
. This consideration is given particular prominence in Ernst’s (2002) approach.
. As mentioned above in Section 2.3, v is the head in a Larsonian VP-shell that is associated
with agentivity. φ – features is a cover term for person, number, and gender features, which
are standardly considered to be interpretable features when part of a nominal category, but
not when part of a functional category. See Note 6 on EPP-features.
. Wide-scope interpretations are available for both pre-auxiliary and post-auxiliary posi-
tions of the adverbial. Both (i) and (ii) can mean that John is often able to lift 200 pounds,
while the interpretation that John is able to lift 200 pounds several times in a row is confined
to (i).

(i) John can frequently lift 200 pounds (only narrow scope)

(ii) John frequently can lift 200 pounds (wide or narrow scope)

. Thus, there is at least one environment in which both “standard” and “mirror” orders
can be found, in addition to even more complicated factors. Syntacticians sometimes can-
not resist the temptation to play down the optionality found in one of the environments
(here roughly preverbal versus postverbal environments) with respect to some word order
property. However, the general pattern Laenzlinger describes is the most expected one from
the perspective of the Greenbergian tradition of word order universals research. If a large
sample of languages is investigated with respect to the options they allow given two bipolar
parameters, the most common result is that only one combination is restrictive enough to
form an implicational statement, here for example “If adverbials A and B occur preverbally
(in language L), then their order preverbally is AB”, which contains no prediction about the
occurrence of postverbal adverbials or serializations of adverbials found there. It only rarely
happens that the syntactician’s dream comes true that a vice versa component can be added
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

to such a statement, for example that the postverbal occurrence of adverbials A and B in
language L has the order BA.
. Cf. Shaer (2003) for similar considerations about fronted manner adverbials.
. Cf. the overviews in Alexiadou (2002) and Delfitto (2000) and the discussion in Alexi-
adou (1997).
. Ortmann (2003), citing data from Punjabi, Lakota, and German, points out that there
is reason to suspect that the derivational direction of adverb derived from adjective is not a
privileged one crosslinguistically. In Punjabi, for example, in addition to the fact that adjec-
tives can also be derived from adverbs, there are several other word classes, such as pronouns
and demonstratives, that form derivational bases for adverbs.
. This justifies the use of the term adverbial in domains other than the verbal one.
. Cf. the similar conlusions in Lang, Maienborn and Fabricius-Hansen (2003: 23f.) on
these last two issues.

References

Adger, D., Plunkett, B., Tsoulas, G., & Pintzuk, S. (1999). “Specifiers in generative grammar.”
In D. Adger, S. Pintzuk, B. Plunkett, & G. Tsoulas (Eds.), Specifiers: Minimalist
Approaches (pp. 1–18). Oxford: Oxford University Press.
Alexiadou, A. (1997). Adverb Placement. A Case Study in Antisymmetric Syntax. Amsterdam
and Philadelphia: Benjamins.
Alexiadou, A. (2002). “The syntax of adverbs: Puzzles and results.” GLOT international, 6,
33–54.
Alexiadou, A. (2004), “Adverbs across Frameworks.” In A. Alexiadou (Ed.), Taking up the
Gaunlet – Adverbs across Frameworks [Lingua, Volume 114, Issue 6] (pp. 677–682).
Asher, N. (1993). Reference to Abstract Objects in Discourse. Dordrecht, Boston and New
York: Kluwer.
Baker, C. L. (1971). “Stress level and auxiliary behavior in English.” Linguistic Inquiry, 2,
167–181.
Baker, C. L. (1981). “Auxiliary-adverb word order.” Linguistic Inquiry, 12, 309–315.
Birner, B. J., & Ward, G. (1998). Information Status and Noncanonical Word Order in English.
Amsterdam and Philadelphia: Benjamins.
Bresnan, J. (1973). “Syntax of the comparative clause constructions in English.” Linguistic
Inquiry, 4, 275–343.
Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.
Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press.
Chomsky, N. (2000). “Language and its design: The Delhi Lecture, January 1996.” In
N. Mukherji, B. N. Patnaik, & R. K. Agnihotri (Eds.), The Architecture of Language
(pp. 1–38). New Delhi: Oxford University Press.
Cinque, G. (1999). Adverbs and Functional Heads. A Cross-Linguistic Perspective. New York
and Oxford: Oxford University Press.
 Current issues in the syntax and semantics of adverbials 

Cinque, G. (2004). “Issues in adverbial syntax.” In A. Alexiadou (Ed.), Taking up the

Gaunlet – Adverbs across Frameworks [Lingua, Volume 114, Issue 6] (pp. 683–710).
Corver, N. (1997). “Much-support as a Last Resort.” Linguistic Inquiry, 28, 119–164.
Costa, J. (1996). “Adverb positioning and V-movement in English: some more evidence.”
Studia Linguistica, 50, 22–34.
Costa, J. (1998). Word Order Variation: A Constraint-Based Approach. PhD, Leiden.
Costa, J. (2004). “A multifactorial approach to adverb placement. Assumptions, facts, and
problems.” In A. Alexiadou (Ed.), Taking up the Gaunlet – Adverbs across Frameworks
[Lingua, Volume 114, Issue 6] (pp. 711–753).
Davidson, D. (1967). “The logical form of action sentences.” In N. Rescher (Ed.), The Logic
of Decision and Action (pp. 81–95). Pittsburgh: University of Pittsburgh.
Delfitto, D. (2000). “Adverbs and the syntax/semantics interface.” Rivista di Linguistica, 12,
13–52.
Dowty, D. R. (1979). Word Meaning and Montague Grammar. The Semantics of Verbs and
Times in Generative Semantics and in Montague’s PTQ. Dordrecht and London: Reidel.
Eckardt, R. (2002). “Event semantics.” In F. Hamm & T. E. Zimmermann (Eds.), Semantics
[Linguistische Berichte, Special Issue 10] (pp. 91–127).
Eckardt, R. (2003). “Manner adverbs and information structure: Evidence from the
adverbial modification of verbs of creation.” In E. Lang, C. Maienborn, & C. Fabricius-
Hansen (Eds.), Modifying Adjuncts (pp. 261–305). Berlin and New York: Mouton de
Gruyter.
Ernst, T. (1983). “More on Adverbs and Stressed Auxiliaries.” Linguistic Inquiry, 14, 542–549.
Ernst, T. (2002). The Syntax of Adjuncts. Cambridge: Cambridge University Press.
Ernst, T. (2003). “Semantic features and the distribution of adverbs.” In E. Lang,
C. Maienborn, & C. Fabricius-Hansen (Eds.), Modifying Adjuncts (pp. 307–334). Berlin
and New York: Mouton de Gruyter.
Frey, W., & Pittner, K. (1998). “Zur Positionierung von Adverbialen im deutschen
Mittelfeld.” Linguistische Berichte, 176, 489–534.
Haider, H. (2000). “Adverb placement – Convergence of structure and licensing.” Theoretical
Linguistics, 26, 95–134.
Haider, H. (2004). “Pre- and postverbal adverbials in OV and VO.” In A. Alexiadou (Ed.),
Taking up the Gaunlet – Adverbs across Frameworks [Lingua, Volume 114, Issue 6]
(pp. 779–807).
Harris, Z. (1968). The Mathematical Structure of Language. Chicago: University of Chicago
Press.
Hitzeman, J. (1997). “Semantic partition and the ambiguity of sentences containing tempo-
ral adverbials.” Natural Language Semantics, 5, 87–100.
Jackendoff, R. S. (1972). Semantic Interpretation in Generative Grammar. Cambridge, MA
and London: MIT Press.
Jacobs, J. (1993). “Integration.” In M. Reis (Ed.), Wortstellung und Informationsstruktur
(pp. 63–116). Tübingen: Niemeyer.
Jacobs, J. (1994). Kontra Valenz. Trier: Wissenschaftlicher Verlag.
Kayne, R. S. (1994). The Antisymmetry of Syntax. Cambridge, MA: MIT Press.
Laenzlinger, C. (1998). Comparative Studies in Word Order. Adverbs, Pronouns, and Clause
Structure in Romance and Germanic. Amsterdam and Philadelphia: Benjamins.
 Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh

Lang, E., Maienborn, C., & Fabricius-Hansen, C. (2003). “Modifying (the grammar of)
adjuncts: An introduction.” In E. Lang, C. Maienborn, & C. Fabricius-Hansen (Eds.),
Modifying Adjuncts (pp. 1–29). Berlin and New York: Mouton de Gruyter.
Larson, R. K. (1988). “On the double object construction.” Linguistic Inquiry, 19, 335–391.
Maienborn, C. (1996). Situation und Lokation. Die Bedeutung lokaler Adjunkte von
Verbalprojektionen. Tübingen: Stauffenburg.
Maienborn, C. (2001). “On the position and interpretation of locative modifiers.” Natural
Language Semantics, 9, 191–240.
McConnell-Ginet, S. (1982). “Adverbs and logical form.” Language, 58, 144–184.
Ortmann, A. (2003). “Zur semantischen und morphologisch-typologischen Abgrenzung
des Adverbs vom Adjektiv.” Ms., Universität Tübingen.
Pittner, K. (1999). Adverbiale im Deutschen. Untersuchungen zu ihrer Stellung und
Interpretation. [Studien zur deutschen Grammatik 60]. Tübingen: Stauffenburg.
Rosengren, I. (2003). “Clause-final left-adjunction.” In E. Lang, C. Maienborn, &
C. Fabricius-Hansen (Eds.), Modifying Adjuncts (pp. 335–362). Berlin and New York:
Mouton de Gruyter.
Shaer, B. (2003). “‘Manner’ adverbs and the association theory: Some problems and
solutions.” In E. Lang, C. Maienborn, & C. Fabricius-Hansen (Eds.), Modifying Adjuncts
(pp. 211–259). Berlin and New York: Mouton de Gruyter.
Sportiche, D. (1995/1998). “Sketch of a reductionist approach to syntactic variation and
dependencies.” In D. Sportiche (Ed.), Partitions and Atoms of Clause Structure: Subjects,
Agreement, Case, and Clitics (pp. 379–419). London: Routledge, [first appeared 1995
in H. Campos & P. Kempchinsky (Eds.), Evolution and Revolution in Linguistic Theory:
Essays in Honor of Carlos Otero, Washington D.C.: Georgetown University Press].
Steinitz, R. (in collaboration with E. Lang) (1973). Adverbialsyntax. 3rd ed. Berlin:
Akademie-Verlag.
Stroik, T. (1990). “Adverbs as V-sisters.” Linguistic Inquiry, 21, 654–661.
Stroik, T. (1996). Minimalism, Scope, and VP Structure. Thousand Oaks, London and New
Delhi: Sage Publications.
de Swart, H. (1999). “Position and meaning: Time adverbials and context.” In P. Bosch, P.,
& R. van der Sandt (Eds.), Focus: Linguistics, Cognitive and Computational Perspectives
(pp. 336–361). Cambridge: Cambridge University Press.
Thomason, R., & Stalnaker, R. (1973). “A semantic analysis of adverbs.” Linguistic Inquiry,
4, 195–220.
Travis, L. (1988). “The syntax of adverbs.” McGill Working Papers in Linguistics. Special Issue
on Comparative Germanic Syntax, 280–310. Montreal: McGill University.
 Circumstantial adverbs and aspect*

David Adger and George Tsoulas

Queen Mary College, University of London / University of York

This paper argues that locatives and certain manner adverbs are not, as is
classically assumed, adjoined to VP nor are they the sole specifiers of
dedicated functional heads within an articulated clausal structure as in
Cinque’s (1999) proposal. Rather we argue that they are licensed inside the
verb phrase, as inner specifiers of functional heads whose main semantic
function is to introduce and contribute towards the interpretation of
arguments. More specifically, we propose that locatives are licensed in the
specifier of an aspectual head whose semantic function is to mark telicity and
license quantized arguments. Similarly, manner adverbs in -ly are licensed as
inner specifiers of little v, whose usual function is to introduce agents. This
proposal allows us to capture a range of structural and interpretational facts
about the semantic interaction between locatives and aspect on the one hand,
and agents and manner modification on the other.

. The syntax and semantics of manners and locatives

In this paper we focus on the syntactic licensing and structural position of

manner adverbials (specifically those in -ly in English) and locatives. We show
that there is a correlation between the licensing of arguments and the licensing
of these elements. We argue that the observed correlations can be captured by
developing a theory of the licensing abilities of a series of functional heads in
the verb phrase.
It has been known since Harris (1968) that there is a correlation between
the ability of certain verbs to take a manner adverbial and their ability to sub-
categorise for an affected object. Thus, verbs like resemble, have, cost etc. are ill
formed with manner adverbs:
(1) *John resembled Sue slowly.
(2) *John had flu worriedly.1
 David Adger and George Tsoulas

(3) *The slave cost 600 denarii wholeheartedly.

These verbs, although they take an object which is apparently marked with
accusative case, do not allow of -insertion in their nominalisations:2
(4) a. *John’s resembling of Sue.
b. *John’s having of flu.
c. *The book’s costing of £30.

We can take this as showing that these verbs do not mark their objects with
true structural accusative, but rather with inherent accusative case (see Torrego
1998 for evidence that accusative can be inherent).
A fact which is linked to this is Chomsky’s (1965) observation that certain
verbs which do not passivise also cannot occur with manner adverbials, a fact
which he explained by allowing a passive-triggering morpheme to be generated
in the subcategorised manner position:
(5) a. *Sue was resembled by John.
b. *Flu was had by Maria.
c. *£50 were cost by Rhyme and Reason.

Further facts which confirm this type of proposal come from the incompatibil-
ity of manner adverbs with psychological predicates:
(6) a. *He aggravated me revoltingly.
b. *They enthused the audience excitedly.
c. *We loathed the pig-farmer fiercely.
(7) a. *She desired a raise enthusiastically.
b. *You loved him deeply. (ok on an extent reading)

Again, we find a link here between case and manner modification: psych-verbs
and verbs of desire are usually assumed to be unaccusative (Belletti & Rizzi
1988), and therefore to be unable to assign structural accusative case.
One of the main ideas of this paper is that the functional structure that is
involved in the licensing of manner adverbials is correlated with the functional
structure that is involved in licensing Agents. In turn, this same functional
structure is partially correlated with the assignment of accusative case to an
affected object, leading to the observations above. Note that we do not di-
rectly correlate the ability to license manner adverbials with the ability to select
an affected object, since manners may perfectly well appear in the absence of
such objects:
(8) John walked jerkily to the cliff edge.
 Circumstantial adverbs and aspect 

(9) We responded politely to his request.

Rather we treat this correlation as an indirect one, mediated by functional

material which licenses Agents (see Section 4).
Turning to locatives, we find a clear link between the aspectual specifica-
tion of a predicate and its interaction with these adverbial phrases. As is well
known (see Levin 1993 for discussion), certain verbs allow locative preposi-
tions to drop, with a corresponding change in the lexical aspect of the predicate:
the predicate without preposition denotes an action with a final result.
(10) a. They climbed up the mountain.
b. They climbed the mountain.
(11) a. They shot at the tree.
b. They shot the tree.

In the (b) examples, the event necessarily culminates in a resultant state (in
fact, it is telic), while this is not the case in the (a) examples, where there is no
specification for telicity (that is, the predicate is compatible with a non-telic
interpretation).
A striking correlation between locatives and telicity is noted by Borer
(1998b). She shows that, in Hebrew, certain inversion processes are sensitive
to the telicity of the predicate, so that inversion is impossible with atelic verbs,
giving the contrast between the (a) and (b) examples.
(12) a. Parcu mehuma. (telic)
erupted riot
‘A riot erupted’
b. *‘avad ganan. (atelic)
worked gardener
‘A gardener worked’

Note that with the atelic predicate in (12b) inversion is impossible. Interest-
ingly, when a locative element is inserted, inversion becomes possible:
c. ‘avad sam/kan/ecli ganan.
worked here/there/at.my.house gardener
‘A gardener worked here/there/at my house’

Again, what we seem to have here is a close link between the expression of
a locative and the aktionsart of the predicate, which may then have further
syntactic effects. Of course, all of these constructions involve other interfering
factors (sentence aspect, focus/presupposition structure) and it is not our in-
 David Adger and George Tsoulas

tention to provide an analysis of all correlations between locatives and telicity.

What we will try to do is outline a framework which is capable of making the
necessary links, and show in a couple of core cases how the analytic potential
of the system plays out.
In this paper, then, we seek to give a unified account of the way that manner
and locative adverbials interact with case and aspect specifications. In doing
so, we are led to a particular view of these low-VP adverbials which essentially
places them within the low functional structure of VP, and where their licensing
takes place through featural mechanisms similar to the mechanisms which are
thought to be involved in the licensing of true DP arguments.

. Some further data

Many accounts of manners and locatives place them low within the VP, with
locative adverbials hierarchically superior (Andrews 1982; Bowers 1993; Ernst
2002). This approach can make sense of the fact that manner adverbials appear
closer to the verb than locatives:
(13) We tortured the general slowly in the garden.
(14) ?We tortured the general in the garden slowly.

Example (14) is only well formed with a clear prosodic break between the
two adverbial phrases, suggesting that the manner adverbial has been “moved”
rightwards (possibly in the prosodic component of the grammar – Zubizarreta
1998), or that, at least, it is not in its canonical position.
One traditional approach would be to assume that the manner adverbial is
generated lower down than the locative, and both are right adjoined to some
projection of the verb (15). This data then follows naturally.
(15) . . . [[[. . . V . . .] Manner] Locative]

This type of approach also captures the fact that manner adverbials are far more
restricted in their syntactic distribution than, say, temporals:
(16) This bridge may (*badly) have (*badly) been (badly) designed (badly) by
Brunel (badly).
(17) *We slowly often tortured the general.
(18) We often slowly tortured the general.
 Circumstantial adverbs and aspect 

Note that, in (16), the manner adverbial is restricted to positions very close
to the verbal predicate itself, or to rightward positions which again may be
assumed to arise because of prosodic factors. Similarly, (17) and (18) show
that, while a manner adverb may appear to the left of the verb, it may only do
so if it is adjacent to it.
It is also well known that preverbal manner adverbials have a different
interpretation in English from postverbal ones (Thomason & Stalnaker 1973):
(19) a. He has been slowly testing some bulbs.
b. He has been testing some bulbs slowly.

In (19b), each particular test must be slow, whereas this is not the case in (19a).
One approach to this might be to assume that the preverbal adverb is ac-
tually adjoined to the lexical verb itself, rather than to some higher projection,
giving the following (rough) structures:
(20) . . . been [slowly testing] some bulbs
(21) . . . been [testing some bulbs] slowly

In (20), the adverb modifies the event denoted by the verb directly, with the
semantic consequence that the testing event is slow. In (21), the adverb mod-
ifies the VP, including the object so that each event of bulb-testing is slow. We
are not wedded to this analysis, but we do assume that there is something syn-
tactically special about pre-verbal manner adverbs, and we concentrate in our
discussion on post-verbal ones.
One further argument for the position of post-verbal manner adverbs can
be constructed on the basis of Costa’s (1996) discussion of adverb placement.
Costa argues that PP arguments which occur after manner adverbs are in their
base position, since they do not show any of the “freezing” effects one would
expect of PPs in extraposed positions:
(22) Which woman did he glance quickly at a picture of t?
(23) *Which woman did he glance yesterday at a picture of t?

Compare the sentence with a locative:

(24) *Which woman did he glance in the Louvre at a picture of t?

This paradigm suggests that locatives are in a zone of the sentence after which
extraposed elements appear, whereas manners are not (or, at least, do not have
to be). Note, again, however, that this does not push us into an analysis where
 David Adger and George Tsoulas

locatives are generated higher than manners, since the extraposition zone may
actually be rather low down (see, for example, Haider 1997, or Kayne 1994).
Two final points to note in this discussion of the relationship between man-
ners and locatives are, first, that PP manner adverbials induce the same effect
as locatives and temporals (25), and, second, that these manner phrases may
not occur preverbally without comma intonation (26):
(25) *Which woman did he glance in a sultry way at a picture of t?
(26) *I have in a sultry way kissed him.

It is possible, then, that -ly manner adverbs and the whole class of PP adverbials
are differentiated in their syntactic position. This discussion suggests that the
manner adverbs under consideration are syntactically positioned close to the
surface position of the verb, while locatives are more distant.
There is evidence, though, that locatives at least are actually lower down
than the surface position of the object. Note that in (27), the quantifier in object
position can bind the pronoun in the locative, suggesting it c-commands it (see
also Pesetsky 1995: 161):
(27) Maire tortured every rabbit i in itsi hutch.

A theory which assumes that locatives are right adjoined will have to deal with
this data by assuming that the object raises to a position higher up than VP
(perhaps [Spec, AgrO]), and that the verb raises higher still.
This set of data forms the basic desiderata of a theory of the positions
of these adverbials. The conservative position is that manner and locative are
both adjoined low to VP, and that the object and subject raise to positions
outside VP:
 Circumstantial adverbs and aspect 

(28) TP

DP T'

T FP

F AgrOP

Vk DPi AgrO'

AgrO VP

VP LOC

VP MAN

tk ti

However, this approach does not explain why we find the extraction contrast
discussed above, since both locatives and manners are generated adjoined to
VP. In addition, this approach cannot tie the putative differential positions of
manners and locatives down to their syntactic licensing, since nothing is said
here about licensing, and indeed the most recent proponent of this kind of
structure puts the fact that manners are closer to the surface position of the
verb than locatives down to purely semantic factors (see e.g. Ernst 2002).

. Recent accounts

Less conservative positions have also been adopted recently. The structure in
(28) of course involves right adjunction, an option which UG has been argued
to lack (see, notably, Kayne 1994). Right adjunction is barred under Kayne’s
assumptions because adjunction is barred, and because rightward merge to a
head is barred unless the merged element is a complement. These constraints
follow from a particular view of the way that PF linearisation is read off syn-
tactic structures.
Cinque (1999) develops some of Kayne’s ideas within the context of a the-
ory of adverbs, and essentially argues that adverbials are to be found in left
specifier positions of various functional heads. He points out two ways of deal-
ing with manner and locative adverbials: either they are generated as specifiers
 David Adger and George Tsoulas

of light verb shells within the VP, or they are generated as complements of these
heads, with the VP in their specifier (an idea he attributes to Nilsen 1998).
Nilsen’s proposal is similar to ideas first proposed within the Generative Se-
mantics tradition by Geis (1970), and it is an idea he extends to all circumstan-
tial adverbials. The core notion is that light v heads are involved which mediate
a predication relation between the VP, which has already been constructed,
and the adverbial. So for a sentence like (29) the VP smoked banana-skins is
in the specifier of a light verb whose complement is in the park and whose se-
mantic function is to locate the event denoted by the VP at the appropriate
spatial point:
(29) Johnny smoked banana-skins in the park.
(30) vP

VP v'

smoked banana-skins v PP

in the park

This is, in many ways, an attractive idea, and leads to a fairly clean semantics for
these adverbials. However, note that it still says nothing about the relative hi-
erarchical ordering of manners and locatives. In addition, there is no evidence
for the projection of these extra v heads: they have no phonological content,
and their semantics is either vacuous, or uniformly predicational. If the latter,
then it is impossible even to state the ordering of manners and locatives, since
to do so would require the v heads to have different semantics from each other,
or different syntactic properties.
A further problem here is the binding data noted above. Nilsen’s approach
would assign a sentence like (27) the structure indicated:
(31) Maire [tortured every rabbit] [in its hutch.]

In such a structure, no c-command relation holds between the quantifier

phrase and the pronoun it binds.
Cinque’s own proposal suffers from the same problem. He suggests that
this type of adjunct is generated in the specifier of a light verb shell above VP:
 Circumstantial adverbs and aspect 

(32) vP

PP v'

in the park v vP

AdvP v'

v VP
quickly

smoked banana-skins

Note that in order to achieve the correct word order, we need to posit fur-
ther functional structure. The idea is that VP will raise into the specifier of a
functional head (not shown above) immediately above the head that licenses
the manner adverb, and then the projection of this head will raise into the
specifier of a further head above the locative, giving rise to a type of “leapfrog-
ging” movement. Again, we reject this proposal on the grounds that it requires
more functional structure than is motivated by the phonology or semantics of
the constructions, and moreover it suffers from the same empirical problem
we saw with Nilsen’s account, since the object will not c-command a bound
pronoun in the locative.

. An alternative

The three systems briefly discussed above, of course, were not primarily con-
cerned with dealing with the data we outlined in Section 1. However, we think
that proper attention to this data actually allows us to motivate the correct
functional structure required to give an answer to questions of how these
adverbials are licensed.
Following much recent work, we adopt an articulated structure for the verb
phrase, consisting of the lexical VP, surmounted by a number of functional
heads which encode particular semantic relations. In particular, we adopt the
idea defended by Travis (1991, 2000) and others (Tenny 1994; Borer forthcom-
ing; Ritter & Rosen 1998) that aktionsart (lexical aspect) is marked syntactically
by an aspectual functional head which takes the verb itself as its sister. This head
is itself the complement of the Agent introducing head “little” v (Kratzer 1995;
Hale & Keyser 1993; Chomsky 1995). This gives us the following structure:
 David Adger and George Tsoulas

(33) vP

v AspP

Asp VP

Johnson (1991), Lasnik (1995), and others, argue, on the basis of diverse phe-
nomena such as gapping and the interaction between particles and double
object constructions, that the main verb in English raises out of the verb phrase
proper to some higher head. The precise semantic nature of this head is not
relevant for the discussion here, but we take it to encode sentence aspect, as
opposed to the lexical aspect mentioned above. For convenience we will la-
bel the higher head Asp1 and the lower (the vP-internal one) Asp2. The full
structure we adopt for the English verb phrase, then, is:
(34) Asp1P

Asp1 vP

V DP v'

v Asp2P

tv Asp2
VP

tv V DP

tv

The DP in [Spec, vP] is interpreted (at least for some varieties of v, see Kratzer
1995) as the agent, and the verb raises through Asp2, and v to Asp1 in English.
Our analysis of the adverbial facts will be couched in terms of the frame-
work outlined in Chomsky (2000). In this framework, functional heads are
assumed to bear features which set up dependencies with formatives that the
head c-commands. These dependencies are formed when the functional head
concerned is specified with uninterpretable features. These features are termed
the probe. A probe essentially seeks matching features within its c-command
domain (these matching features are the goal). The relationship between probe
 Circumstantial adverbs and aspect 

and goal is constrained by locality. Essentially, the relevant kinds of structures

are like those in (35):
(35) [ H{probe} [ ..... XP{goal} ... ]] (probe=goal)

The relation between the head H specified with the probe, and the formative
specified with the goal, we will call the H-associate relation, extending Chom-
sky’s terminology. The formation of an H-associate relationship results in the
deletion of the uninterpretable features involved in the relationship. Since it is
the probe that is uninterpretable, the probe deletes.
In addition to probes, heads may also be specified with EPP features. These
features are selectional (i.e. involve category information) and are also un-
interpretable. An EPP feature is satisfied when a category of the appropriate
featural specification is merged with the head bearing the feature. The XP that
is merged can be either the goal itself, or some other phrase (for example, an
expletive):
(36) [ XP H{probe, EPP} [ ..... XP{goal} ... ]] (probe=goal)

The deletion of the EPP feature is, in general, parameterised, so that some lan-
guages allow multiple subject constructions, as discussed in Section 4.10.3 of
Chomsky (1995). The system outlined there can be thought of as a set of con-
ditions on the deletion of the EPP feature: a language does not have EPP (VSO
languages); has EPP but merge into [Spec, HP] causes EPP to delete (SVO with-
out Multiple Subject Constructions (MSCs)); has EPP but allows one element
to merge without deleting EPP (SVO with MSCs); or allows arbitrarily many
merges without deleting EPP (polysynthetic languages). We will adopt Chom-
sky’s idea that the EPP feature must be satisfied, and that when an H-associate
relation is set up, the XP (usually determined by the goal of H’s probes) will
be forced to raise to merge with the projection of H, satisfying EPP. In this
situation, EPP deletes.
We will, however, extend this picture, adapting Chomsky’s idea that dele-
tion of EPP is an option that UG allows variation for. The core extension is
the assumption that, when EPP is satisfied by an element which hasn’t induced
the H-associate relation, the EPP feature does not have to delete immediately.3
Within a phase–based model (Chomsky 2001), we may suppose with Pesetsky
and Torrego (2001) that deletion of features takes place at the phase level. Thus,
for what concerns us here, an EPP feature that has not deleted immediately re-
mains active until the current phase is completed at which point the feature is
deleted, perhaps as part of the TRANSFER operation of Chomsky (2002).
 David Adger and George Tsoulas

In essence, then, a single EPP feature may trigger Merge of an arbitrary

number of non-Agreeing XPs. If no Agreeing XP is merged at all, the EPP fea-
ture is deleted at phase-level. If an Agreeing XP is Merged, then EPP is deleted
immediately.
It follows that we allow an arbitrary number of non-agreeing XPs to merge
with H, in the general case:
(37) [ XP YP ... ZP H{probe, EPP} [ ..... XP{goal} ... ]] (probe=goal)

In (37), EPP will only be forced to delete when XP is merged with the projection
of H, since only XP is determined by the goal of H’s probe.
This system gives rise to a potentially infinite number of adverbials in inner
specifier positions, constrained by only processing considerations, and the par-
ticular specification of the EPP feature (i.e. what it is a selectional feature for).
Before seeing how this picture pans out in detail, we should state how the
semantic interpretation of these structures is governed. Adapting proposals by
Borer (1998a, b) among others (McClure 1995; van Hout 1996), let us assume
that a telic interpretation of Asp arises when Asp’s probe finds a matching goal.
Put another way, only telic specifications of Asp have an uninterpretable probe
which will match features of the object. The features that are relevant, in this
case, are features governing the quantization of the object (following Verkuyl
1993). We will call this feature [Quant] and assume that it is interpretable on
DP, but not on telic Asp. The specification of telic asp is then Asp[uQuant,
EPP], where the u prefix signals that the feature is uninterpretable (follow-
ing the notation of Pesetsky and Torrego 2001). This allows us to capture the
well-known fact that quantized objects give rise to telic readings of certain
predicates:
(38) We built that house. (telic)
(39) We built houses. (atelic)

Having outlined the basic system of assumptions we adopt, the particular cases
to be considered are as follows: when Asp’s probe finds a matching goal in the
VP but no locative phrase has been constructed from the numeration, the Asp-
associate relation is established, a telic interpretation results, and the object
raises to [Spec, Asp] to satisfy the EPP feature of Asp as in (40) (irrelevant
details omitted):
 Circumstantial adverbs and aspect 

(40) vP

DP v'

v Asp2P

DP Asp2'

Asp2 VP

V DP

Since an Asp-associate relation has been established, the EPP feature of Asp
deletes and no further Merge into [Spec, AspP] is possible. However, if a loca-
tive phrase is accessible at a point in the derivation before DP raises, it may
satisfy the EPP feature of Asp. By ‘accessible at this point of the derivation’, we
simply mean previously constructed in the derivation, or accessible as a simple
lexical item from the numeration. It may be the case that these locatives have
been adjoined to VP, and move to an inner [Spec, AspP], but this is not neces-
sary, and in fact might be ruled out, since nothing will force them to Merge with
VP. The minimal solution is that they are simply present in the “workspace” of
the derivation.
Note that the EPP feature of Asp must be of the correct selectional type
to allow a locative to Merge, and we will assume that locative PPs and quan-
tized DP objects both are specified with interpretable (and therefore selectable)
features which relate an event and an individual in terms of spatial measure:
a locative measures out the physical extent over which an event is delimited,
while a quantized DP measures out the physical extent of the result of the
event. Locatives, then, are specified as [Quant]. Of course, the idea that cer-
tain predicates can select for particular semantic properties like this, and that
such selection results in particular prepositions, or particular case forms, is a
traditional one. The preposition within the locative establishes a probe-goal
relationship with the DP in terms of its case requirements. Asp’s [uQuant] fea-
ture, on the other hand, cannot match with [Quant] on DP because [Quant]
on the locative P intervenes.
Our proposal leads to the possibility that a number of locatives may be
merged with Asp. However, once the object has been merged, then the EPP
 David Adger and George Tsoulas

feature deletes, and no further merge of locatives is possible, since merge of

locatives is only made possible by the presence of an EPP feature. This gives the
basic structure:
(41) [Object Locative Locative . . . Asp t]

This predicts that the object will c-command the locative, as we noted in
Section 2 with respect to example (27) repeated here as (42):
(42) Maire tortured every rabbit in its hutch.

Consider now a case where there is no object within the verb phrase, but where
a locative has been constructed. The locative will merge with Asp, satisfying
its EPP feature, but no telic interpretation will result, because no probe-goal
relation has been established. Asp, in this case, does not bear [uQuant] and is
therefore not (obligatorily) telic. This is precisely what is behind the locative
preposition drop phenomenon repeated below:
(43) a. They climbed up the mountain.
b. They climbed the mountain.

(43a) converges because the EPP feature of Asp is satisfied by the locative. The
locative preposition enters into a P-associate relation with the φ-features of
the DP the mountain resulting in it being essentially “Case-licensed” (more
on which, see below). Asp itself cannot enter into an Asp-associate relation
with this DP, since this would give rise to a locality violation (the preposition
bearing [Quant] is closer). This, in turn, means that Asp cannot be specified as
telic since only telic Asp has a quantization probe, giving rise to the appropriate
interpretation. If there is no object, we have a violation of EPP:
(44) They climbed.

(44) can only be construed as unspecified object drop, the unspecified object
satisfying EPP of Asp, but not giving rise to quantization effects; hence (44)
may be interpreted as atelic.
(43b), on the other hand, does involve an Asp-associate relation with the
object DP, EPP is satisfied and the derivation converges.
Let us now turn to the alternation in Hebrew, discussed by Borer. Recall
that a lexically telic verb allowed inversion, but that an atelic one did not, unless
it occurred with a locative clitic:
(45) a. Parcu mehuma. (telic)
erupted riot
‘A riot erupted’
 Circumstantial adverbs and aspect 

b. *’avad ganan. (atelic)

worked gardener
‘A gardener worked’
c. ’avad sam/kan/ecli ganan.
worked here/there/at.my.house gardener
‘A gardener worked here/there/at my house’

On our account, a verb like parcu ‘erupt’ lexically selects an Asp with active
probes (that is, it bears an uninterpretable quantization feature) while a verb
like ’avad ‘work’ does not (this distinction is in addition to the different Merge
positions of the single argument in each case). In (45a), the probe of Asp estab-
lishes the Asp-associate relation with the object (which then raises to satisfy the
EPP feature of Asp). The object bears a case feature which needs to be checked.
Following Chomsky (2001), we assume that case checking (which we can im-
plement as deletion of a uCase feature) is parasitic on H-associate relationships
established in terms of φ-features only (see the discussion of the case licensing
properties of “little” v below).
We then have the following schematic derivation:
(46) Asp[uQuant, EPP] . . . DP[φ, Quant, uCase] ⇒
(47) Asp[uQuant, EPP] . . . DP[φ, Quant, uCase] ⇒
(48) DP[φ, Quant, uCase] Asp[uQuant, EPP] . . . t

In (45b), no Asp-associate relation is established because the subject is merged

in the specifier of little v, and is therefore not c-commanded by Asp, and the
single argument ganan must instead be checked by T’s φ probes, which also
case license it (note, since this verb is agentive, the DP argument is merged in
[Spec, vP] and is therefore too high to be a goal for Asp’s probes):
(49) T[uφ, EPP] . . . DP[φ, Quant, uCase] ⇒
(50) T[uφ, EPP] . . . DP[φ, Quant, uCase] ⇒
(51) DP[φ, Quant, uCase] T[uφ, EPP] . . . t

If this relationship is established, ganan must raise to satisfy the EPP features
of T.
The next question is how the EPP features of T are satisfied in (45a). In this
case some null element must fill the specifier of TP, and we follow Pinto (1997),
who convincingly argues on the basis of Italian, that verbs of this aspectual class
can always select a covert locative. For us, this locative is first merged to satisfy
the EPP feature of telic Asp. From this position it raises to the specifier of T. The
 David Adger and George Tsoulas

appearance of inversion here, then, arises because of a null locative in subject

position, rather than the subject itself.
Turning to (45c), this example is parallel to (45a), except that, again follow-
ing Pinto’s work, the null locative is not available for atelic predicates. However,
it is of course possible to select an overt locative in the numeration, and in this
case this can satisfy the EPP feature of T, so that the subject remains in situ,
giving rise to the apparent inversion.
This system captures the core of the semantic and syntactic dependencies
between locatives, objects and aspectual specification, although clearly there is
more to be said. The system also automatically derives the apparent iterability
of locatives, as well as the fact that objects must c-command locatives. In ad-
dition, the structures proposed involve only heads that are interpretable at the
interfaces, rather than appealing to semantically empty functional structure.
If we assume that T behaves in a similar fashion to Asp, then the system we
have outlined leaves open the possibility that locatives may satisfy an EPP fea-
ture of T, after establishment of a T-associate relation with the single argument
of an intransitive, as in the Hebrew example discussed above. In such cases, as
in the cases above, the locative and subject are in the same minimal domain
and therefore equidistant to the probes of T. Under this scenario, T establishes
an associate relation with both, but only one raises to satisfy (and delete) the
EPP feature of T. The familiar case of Locative Inversion (Bresnan 1994) may
exemplify exactly this situation:
(52) a. In the garden sat a gnome.
b. A gnome sat in the garden.

Note that the probe of T here would have to be person features only, since
agreement is not triggered by conjoined inverted locatives. Of course, factors
such as focus play an important role in Locative Inversion constructions, as
does the thematic structure of the predicate (something we take as being at
least partially reducible to aspectual considerations). However, the suggestion
seems promising.
Let us turn now to manner adverbials. As noted in Section 2, -ly manner
adverbs appear closer to the verb than locatives do, although there is no clear
evidence that -ly manners c-command locatives or vice versa, and it appears
that prepositional manners and locatives appear in the same position.
Recall that Asp induces an Asp-associate relationship with the object. The
goal in the object, we took to be features associated with semantic quantiza-
tion. We did not assume that the goal of Asp’s probes was related to Case, or
to φ-features. In our system, it is v that establishes a v-associate relationship
 Circumstantial adverbs and aspect 

with these features of the object. The licensing of the object takes place, then,
via at least two different featural relationships between functional heads and
different goals within the object. Asp probes for quantization, while v probes
for φ-features.
To maintain maximal parallelism between the components of the extended
verbal projection, v also has an EPP feature, which is satisfied and deleted by
an XP determined by the goal of v’s probes – in this case the object. This is
the standard assumption within Minimalist approaches to clause structure (see
the extended discussion in Chomsky 1995, 2000; Lasnik’s 1995 argument that
movement to [Spec, AgrP] is always driven by EPP considerations). The object
will then raise to [Spec, vP], which we assume is its surface position in English
(Johnson 1991, Koizumi 1993, and the papers collected in Lasnik 1999), with
the verb raising outside the verb phrase, as discussed in the introduction to this
section. This gives us the following structure:
(53) Asp1P

Asp1 vP

SUBJ vP
V

OBJ v'

v Asp2P

tv Asp'
LOC

Asp2 VP

V tOBJ
tv

tv

In the same way that locatives satisfy the EPP feature of Asp without causing
it to delete, -ly manner adverbs will satisfy the EPP feature of v, again with-
out inducing deletion. This means that manner adverbials are inserted into the
structure as inner specifiers of v.
What is the actual structure of manner adverbials themselves, and why are
they licensed by the EPP feature of v? We tentatively suggest that manners are
 David Adger and George Tsoulas

actually nominalisations of copies of the verb in the numeration, which are

modified by adjectival modifiers. So a manner like slowly in the phrase slowly
ran is essentially derived by copying the root ran in the derivation, modifying
it with an adjectival predicate slow and then projecting a nominal functional
structure above the new composite, so that the whole phrase is morphologi-
cally interpreted as ‘slowly’. This process can be seen overtly at work in many
languages (see, for example, the discussion of manner modification in Modern
Hebrew, in Glinert 1995), and makes sense of the fact that manners relativise
to their predicate, so that slowly running might involve speedier movement
than quickly walking. Since the resulting phrase is nominal, it may satisfy the
nominal EPP feature of v.
(54) Asp1P

Asp1 vP

SUBJ vP
V
OBJ vP

MAN v'

v Asp2P

tv Asp'
LOC

Asp2 VP

V tOBJ
tv

tv

Inspecting the structure above, we predict a word order for English which is V
Obj MAN LOC. This is of course the correct prediction. Note that in this struc-
ture the manner adverb c-commands the locative, something that we argued
was at least a possibility in Section 2.
We also immediately predict that it is possible to iterate manner adverbials,
as can be seen in the following examples:
(55) They played loudly badly. (from Ernst 2002)
 Circumstantial adverbs and aspect 

This is, of course, because of the way that EPP features of functional heads
can be satisfied, but not deleted, by elements that do not establish H-associate
relations with the functional head’s probe.
How do we account for the correlations noted in Section 1, between the
case licensing potential of the verb and its ability to occur with manner ad-
verbs? Recall that the probe-goal relationship set up by v is similar to the tradi-
tional case relationship (although technically it will involve uninterpretable φ-
features, rather than case features). Clearly, a version of v which is not endowed
with the appropriate probe will not license a case marked object (and such
verbs will not have an Agent in their specifiers either, following the usual im-
plementation of Burzio’s generalization in this framework – Chomsky 1995).
It follows that verbs like resemble etc. will not check structural accusative case
of their object, thus accounting for their anomalous behaviour in of -insertion
environments. There are a number of ways to implement this, but the simplest
is just to assume that these verbs do not have a v embedded in their structure
at all. Given that these predicates have an impoverished structure, specifically
lacking in an EPP feature, they will not be able to license manner adverbials,
explaining the old observation that predicates which do not assign structural
accusative, do not take manner adverbials.
(56) *John resembled Sue slowly.
(57) *John had flu worriedly.
(58) *The slave cost 600 denarii wholeheartedly.

If passivisation is simply an operation on v, say deleting the probe of v, then we

also explain why these predicates do not passivise:
(59) a. *Sue was resembled by John.
b. *Flu was had by Maria.
c. *£50 were cost by Rhyme and Reason.

One final point to note is that, as mentioned in Section 1, the kind of system
we adopt assumes that agentive unergatives do actually contain a little v (cf.
Hale & Keyser 1993, who argue that such verbs are actually transitive where
the object has incorporated into the verb). It follows, then, that it is possible to
have a manner adverb in the absence of an affected object:
(60) John walked jerkily to the cliff edge.

Following Hale and Keyser (1993), we assume that such unergatives do contain
a little v, which is the licenser of the manner adverb.
 David Adger and George Tsoulas

. Conclusion

We have argued that manner and locative adverbials are licensed by functional
structure which is generated low down in the verb phrase, and that has in-
dependent semantic motivation (aspectual specification, and specification for
agentivity). The heads that we proposed were involved in licensing locatives
and manners do so secondarily; their prime function is to build up the licensing
and interpretations of arguments, and the mechanisms whereby low adverbials
are licensed are simply a subset of those mechanisms which license arguments.
The distinction between true arguments and low adverbials arises because of
the way that the syntactic licensing of both classes interacts with interpretative
mechanisms.

Notes

* A much earlier version of this paper appears as Adger and Tsoulas (2000). Many thanks
to two anonymous reviewers and to the volume editors for helpful comments.
. Note that in cases like (i), only an extent interpretation is available for the adverb:

(i) John had flu badly.

. Note that there are verbs which disallow of -insertion but which are fine with adverbs.
These seem to be copular verbs:

(i) *John’s becoming of a fool/John’s being of a fool.

(ii) John became a Buddhist willingly/John was a doctor wholeheartedly.

Presumably the lack of of -insertion arises because these verbs take predicates as their com-
plements, rather than true objects.
. This is also reminiscent of the theory of adjunction developed in Saito and Fukui (1998)
where they argue that multiple adjunction at the X’ level is possible until a specifier which
“agrees” with the head is merged. In their terms, an agreeing specifier closes off the projec-
tion.

References

Adger, D., & Tsoulas, G. (2000). “Aspect and lower VP adverbials.” In A. Alexiadou &
P. Svenonius (Eds.), Adverbs and Adjunction [Linguistics in Potsdam 6] (pp. 1–18).
Potsdam: University of Potsdam.
 Circumstantial adverbs and aspect 

Andrews, A. (1982). “A note on the constituent structure of adverbials and auxiliaries.”

Linguistic Inquiry, 13 (3), 313–317.
Belletti, A., & Rizzi, L. (1988). “Psych-verbs and Theta Theory.” Natural Language and
Linguistic Theory, 6 (2), 291–352.
Borer, H. (1998a). “Deriving passive without theta roles.” In S. Lapointe et al. (Eds.),
Morphology and its Relation to Phonology and Syntax, (pp. 60–96). Stanford: CSLI.
Borer, H. (1998b). “Licensing aspectual nodes.” Paper presented at the Afro-Asiatic
Linguistics Conference, SOAS, London.
Borer, H. (Forthcoming). The ExoSkeletal Trilogy. Oxford: Oxford University Press.
Bowers, J. (1993). “The syntax of predication.” Linguistic Inquiry, 24 (4), 591–656.
Bresnan, J. (1994). “Locative inversion and the architecture of Universal Grammar.”
Language, 70 (1), 72–131.
Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.
Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press.
Chomsky, N. (2000). “Minimalist inquiries.” In R. Martin et al. (Eds.), Step by Step: Essays
on Minimalist Syntax in Honor of Howard Lasnik (pp. 89–156). Cambridge, MA: MIT
Press.
Chomsky, N. (2001). “Derivation by Phase.” In M. Kenstowitz (Ed.), Ken Hale: A Life in
Language (pp. 1–52). Cambridge, MA: MIT Press.
Chomsky, N. (2002). “Beyond explanatory adequacy.” Ms., MIT.
Cinque, G. (1999). Adverbs and Functional Heads. Oxford: Oxford University Press.
Costa, J. (1996). “Adverb positioning and V-movement in English: some more evidence.”
Studia Linguistica, 50 (1), 22–34.
Ernst, T. (1984). Towards an Integrated Theory of Adverb Position in English. Bloomington
IN: Indiana University Linguistics Club.
Ernst, T. (2002). The Syntax of Adjuncts. Cambridge: Cambridge University Press.
Geis, J. (1970). “Lexical insertion of locative and time prepositions.” In Papers from the Sixth
Regional Meeting of the Chicago Linguistics Society (pp. 226–234). Chicago: Chicago
Linguistics Society.
Glinert, L. (1995). The Grammar of Modern Hebrew. Cambridge: Cambridge University
Press.
Haider, H. (1997). “Extraposition.” In D. Beerman, D. LeBlanc, & H. van Riemsdijk (Eds.),
Rightward Movement (pp. 115–151). Amsterdam and Philadelphia: Benjamins.
Hale, K., & Keyser, S. J. (1993). “On argument structure and the lexical expression of
syntactic relations.” In K. Hale & S. J. Keyser (Eds.), The View from Building 20 (pp. 53–
109). Cambridge, MA: MIT Press.
Harris, Z. (1968). The Mathematical Structure of Language. Chicago: University of Chicago
Press.
van Hout, A. (1996). “Event semantics of verb-frame alternations.” PhD, Tilburg.
Johnson, K. (1991). “Object positions.” Natural Language and Linguistic Theory, 9 (4), 577–
636.
Kayne, R. (1994). The Antisymmetry of Syntax. Cambridge, MA: MIT Press.
Koizumi, M. (1993). “Object agreement phrases and the split VP hypothesis.” In MIT
Working Papers in Linguistics 18: Papers on Case and agreement, I (pp. 99–148).
 David Adger and George Tsoulas

Kratzer, A. (1995). “The event argument and the semantics of verbs.” Ms., University of
Massachusetts at Amherst.
Lasnik, H. (1995). “Case and expletives revisited: on Greed and other human failings.”
Linguistic Inquiry, 26 (4), 615–633.
Lasnik, H. (1999). Minimalist Analysis. Oxford: Blackwell.
Levin, B. (1993). English Verb Classes and Alternations. Chicago: University of Chicago Press.
McClure, W. (1995). “Syntactic projections of the semantics of aspect.” PhD, Cornell.
Nilsen, Ø. (1998). “The syntax of circumstantial adverbials.” Ms., University of Tromsø,
[published 2000 Oslo: Novus Press].
Pesetsky, D. (1995). Zero Syntax. Cambridge, MA: MIT Press.
Pesetsky, D., & Torrego, E. (2001). “T-to-C movement: causes and consequences.” In
M. Kenstowitz (Ed.), Ken Hale: A Life in Language (pp. 355–426). Cambridge, MA:
MIT Press.
Pinto, M. (1997). “Licensing and interpretation of inverted subjects in Italian.” PhD,
Utrecht.
Ritter, E., & Rosen, S. (1998). “Delimiting events in syntax.” In M. Butt & W. Geuder (Eds.),
The Projection of Arguments: Lexical and Syntactic Constraints (pp. 135–164). Stanford:
CSLI.
Saito, M., & Fukui, N. (1998). “Order in phrase structure and movement.” Linguistic Inquiry,
29 (3), 439–474.
Tenny, C. (1994). Aspectual Roles and the Syntax/Semantics Interface. Kluwer: Dordrecht.
Thomason, R. H., & Stalnaker, R. C. (1973). “A semantic theory of adverbs.” Linguistic
Inquiry, 4 (2), 195–220.
Torrego, E. (1998). The Dependencies of Objects. Cambridge, MA: MIT Press.
Travis, L. (1991). “Inner aspect and the structure of VP.” Ms., McGill, Montreal.
Travis, L. (2000). “Event structure in syntax.” In C. Tenny & J. Pustejovsky (Eds.), Events
as Grammatical Objects: the Converging Perspectives of Lexical Semantics and Syntax
(pp. 145–185). Stanford: CSLI.
Verkuyl, H. (1993). A Theory of Aspectuality: The Interaction of Temporal and Atemporal
Structure. Cambridge: Cambridge University Press.
Zubizarreta, M.-L. (1998). Prosody, Focus and Word Order. Cambridge, MA: MIT Press.
 Optimizing adverb placement
in gap constructions*

Eva Engels
University of Potsdam

The paper investigates the distribution of adverbs in English gap

constructions, i.e. constructions in which the V(erb)P(hrase) in its base
position is deprived of its phonetic material by VP Ellipsis, VP Fronting,
Topicalization, or wh-Movement. The acceptability of adverb adjunction to a
phonetically empty constituent is shown to be dependent on the type of
adverb, its scope, and the syntactic construction. The interaction of these
factors is sketched in an Optimality Theoretic framework: the hierarchical
ranking of several violable constraints determines whether or not a certain
type of adverb may immediately precede a gap in a particular syntactic
construction. A brief look at French c’est...que Clefts and VP Fronting
constructions reveals that cross-linguistic variation in the grammaticality of
pre-gap adverb placement can be accounted for by differences in the
language-specific constraint rankings.

Overview

This paper examines the distribution of adverbs in English constructions in

which the VP in its base position does not contain any phonetic material
because of VP Ellipsis, VP Fronting, or argument extraction out of a cop-
ula construction by wh-Movement or Topicalization, henceforward gap con-
structions. The grammaticality of adverb placement in front of a gap will be
shown to be influenced by three factors: the type of adverb, its scope, and
the syntactic construction (subject-auxiliary inversion vs. non-inversion, wh-
Movement vs. Topicalization / VP Fronting / VP Ellipsis). In constructions in
which subject-auxiliary inversion does not arise, e.g. declarative clauses with
a fronted constituent and embedded questions (henceforward non-inversion
constructions), the order of the finite auxiliary and an adverb with a wide scope
reading is optional as long as some overt element follows the adverb; i.e. the
 Eva Engels

order finite auxiliary – adverb is ungrammatical if no element intervenes be-

tween the adverb and the gap. By contrast, if the adverb takes narrow scope
or subject-auxiliary inversion takes place (e.g. matrix questions), the accept-
ability of adverb occurrence in front of a gap depends on the type of adverb
and the type of gap: all types of adverbs may precede a gap induced by wh-
Movement, whereas only epistemic, frequency, and temporal adverbs, but not
subject-oriented and evidential ones may appear in front of a gap caused by
Topicalization, VP Fronting, or VP Ellipsis.
Section 1 presents the relevant data. Previous approaches to adverb place-
ment in gap constructions are reviewed in Section 2. Section 3 outlines an
Optimality Theoretic account. Considering syntactic structures to be subject
to violable constraints, the dependence of adverb occurrence in front of a gap
on adverb type, scope, and syntactic construction follows from the hierarchical
ranking of the constraints: the structure that optimally satisfies the constraint
hierarchy arises as output. Further support for the Optimality Theoretic ap-
proach of differently ranked universal constraints to explain cross-linguistic
variation is provided by a preliminary investigation of French data. Section 4
summarizes the results.

. Data

Generally speaking, adverbs may precede or follow a finite aspectual auxiliary

in English:
(1) a. George and Martha probably never have seen a real politician.
b. George and Martha have probably never seen a real politician.
(Baker 1981: 309)

However, this order variability of adverb and finite auxiliary does not arise in
certain constructions. First, while both pre- and post-auxiliary adverb place-
ment permit wide scope interpretation of the adverb, a narrow scope reading
of an adverb is restricted to post-auxiliary position:
(2) a. John can frequently lift 200 pounds.
‘John is often able to lift 200 pounds.’
‘John is able to lift 200 pounds several times (in a row).’
b. John frequently can lift 200 pounds.
‘John is often able to lift 200 pounds.’
#‘John is able to lift 200 pounds several times (in a row).’
 Optimizing adverb placement in gap constructions 

Hence, the order variability of adverb and finite auxiliary only arises under the
wide scope reading of the adverb; in order to be interpreted as taking narrow
scope, an adverb must follow the element that outscopes it.
Secondly, some overt element has to intervene between an adverb with a
wide scope reading and a gap in non-inversion constructions. Consequently,
independent of whether the adverb or the finite auxiliary is stressed (Baker
1971: 170), the order wide scope adverb – finite auxiliary is obligatory if no
other overt element precedes the gap. This is illustrated for epistemic, tempo-
ral, frequency, evidential, and subject-oriented adverbs in VP Ellipsis and VP
Fronting constructions in (3) as well as in copula constructions from which
the complement is extracted by Topicalization (4, 5, 6) or wh-Movement (7, 8,
9).1 Note that since both VP Fronting and Topicalization of the complement of
the copula be involve movement of some constituent to the clause-initial posi-
tion and have the same effect on adverb placement, they will be summarized as
Fronting.
(3) Fred claimed that he has never been rude to Grandfather, but (been rude to
Grandfather,)
a. he {probably / always / often / clearly / wisely} has __.
b. *he has {probably / always / often / clearly / wisely} __.
(4) They used to be Socialists, but Communists
a. they never were __.
b. *they were never __. (Sag 1978: 148)
(5) They used to be Socialists, and Communists
a. they {probably / clearly} were __ (, too).
b. *they were {probably / clearly} __ (, too).
(6) John isn’t a heartthrob, but a flatterer
a. he {wisely / usually} is __.
b. *he is {wisely / usually} __.
(7) I don’t know how happy
a. they {ever / really} were __.
b. *they were {ever / really} __. (Sag 1978: 148)
(8) [Everywhere Mary goes there is a disaster so that nobody wants to be near
her.] I wonder where
a. they {usually / wisely} were __ (when Mary was in London).
b. *they were {usually / wisely} __ (when Mary was in London).
 Eva Engels

(9) [They always say that a cardiac transplantation is a routine operation, but
they never mention its long-term consequences.] I wonder how harmless
a. such an operation provably is __.
b. *such an operation is provably __.

The sentences in (10) and (11) show that the occurrence of an overt element
preceding the gap allows for post-auxiliary adverb positioning, while overt ma-
terial following the gap does not have this effect. These facts support the view
that it is the placement of the adverb in front of the gap that rules out the order
finite auxiliary – wide scope adverb in (3) to (9).
(10) a. John’s often been arrested, although Mary’s never been __.
(Wilder 1997: 348)
b. Bill is living in France, but John is probably not __.
(11) I wonder (Baker 1971: 170)
a. where Gerard usually is __ at this time of day.
b. *where Gerard is usually __ at this time of day.

In avoiding the position in front of a gap, adverbs pattern like contracted

auxiliaries. The acceptability of auxiliary contraction seems to depend on the
phonological context: a contracted form requires the presence of a stressed syl-
lable to its right preceding the gap; as illustrated in (14), a weak pronoun does
not suffice (see Baker 1971, 1981; Wilder 1997).
(12) I wonder
a. where the party is __ tonight.
b. *where the party’s __ tonight. (Zagona 1988: 106)
(13) a. I don’t know where John’s been __.
b. They said that Paul would be polite, but (polite,) he’s probably not __.
(Wilder 1997: 348)
(14) a. What’s that __?
b. *What’s it __? (Wilder 1997: 351)

Hence, neither auxiliary contraction nor the order finite auxiliary – adverb is
ruled out in a gap construction per se: rather, adverbs and contracted auxil-
iaries should not occur immediately in front of a gap. Pre-auxiliary position-
ing may prevent a wide scope adverb from preceding a gap. Consequently,
no adverb that takes wide scope may appear in front of any type of gap in
non-inversion constructions.
 Optimizing adverb placement in gap constructions 

By contrast, if the finite auxiliary has to precede the adverb either because
of a narrow scope reading of the adverb or because of subject-auxiliary inver-
sion, it depends on the type of adverb and the type of gap whether adverb
placement in front of the gap is acceptable. While all types of adverbs may oc-
cur in front of a wh-gap (15) to (19), epistemic (20), temporal (21, 22, 23), and
frequency adverbs (24, 25), but not subject-oriented (26, 27) and evidential
ones (28, 29, 30) may precede a gap induced by VP Ellipsis or Fronting.2,3 Note
that the order adverb – finite auxiliary in (some of) the sentences below is ac-
ceptable under a wide scope reading of the adverb (indicated by ADV > MOD
or ADV > NEG), giving rise, for example, to the interpretation that John is reg-
ularly obliged to take his medicine in (25a); however, a narrow scope reading
of the adverb is intended in these sentences (e.g. in (25) ‘John is obliged to take
his medicine regularly’) which is not accessible in pre-auxiliary position.
(15) [Mary told me that John was at the movies yesterday, but I don’t believe her.]
Where was he really __?
(16) [John plans to go on holiday with his mother who travels a lot.]
a. Where was she already __?
b. Where wasn’t she already __?
He asks his father where
c. *she already wasn’t __.
d. she wasn’t already __.
(17) [John wants to go to a pub, but he fears running into his ex-girlfriend who
hangs out in pubs most of the time.]
a. Where / In which pub is she often __?
b. Where / In which pub isn’t she often __?
He wonders where / in which pub
c. *she often isn’t __. (o.k. for ADV > NEG)
d. she isn’t often __.
(18) [John wanted to meet his ex-girlfriend.]
a. Where / In which pub was he wisely __?
[Despite all precautions, John ran into his ex-girlfriend.]
b. Where / In which pub wasn’t he wisely __?
I wonder where / In which pub
c. *he wisely wasn’t __. (o.k. for ADV > NEG)
d. he wasn’t wisely __.
(19) [John’s fingerprints were found at most, but not at all crime scenes.]
 Eva Engels

a. Where / At which crime scene was he evidently __?

b. Where / At which crime scene wasn’t he evidently __?
John’s advocate inquired of the prosecutor’s secretary where / at which crime
scene
c. *he evidently wasn’t __. (o.k. for ADV > NEG)
d. he wasn’t evidently __.
(20) Terry knows how to build an H-bomb. (Ernst 1983: 548)
No – does he really __?
(21) John hasn’t gotten along well with Grandpa lately.
a. Has he ever __? (Baker 1981: 313)
John has gotten along well with Grandpa lately.
b. Hasn’t he always __?
(22) John has never gotten along well with Fred, and
a. *Mary always hasn’t __.
b. Mary hasn’t always __.
(23) John performed excellently on the exam, but such a good student,
a. *he always wasn’t __.
b. he wasn’t always __.
(24) a. John had to go to the doctor yesterday. Must he often?
b. If it was so much fun to go to the gym yesterday, why don’t you regularly?
(25) [John is ill and can only be cured if he takes his medicine twice a day.] John
has to take his medicine, and (take his medicine,)
a. *he regularly must __. (o.k. for ADV > MOD)
b. he must regularly __.
(26) John intended to finish all he set out to do.
a. *Has he wisely? (Ernst 1983: 548)
John wants to move to London.
b. *Should he wisely?
(27) John intends to visit his ex-girlfriend, and (visit his ex-girlfriend,)
a. *he wisely should __.
b. *he should wisely __.
(28) John intended to finish all he set out to do.
a. *Has he clearly? (Ernst 1983: 548)
I’m not quite sure whether John has done his work.
b. *Hasn’t he clearly?
 Optimizing adverb placement in gap constructions 

(29) [There is evidence that John stole the jewelry, but there is no clue who killed
the jeweller.] The police suspect John of having killed the jeweller, but (killed
the jeweller,)
a. *he clearly hasn’t __. (o.k. for ADV > NEG)
b. *he hasn’t clearly __.
(30) John is a burglar, but a murderer,
a. *he clearly isn’t __. (o.k. for ADV > NEG)
b. *he isn’t clearly __.

In summary, the distribution of adverbs in gap constructions depends on the

type of adverb, its scope, and the syntactic construction. For the latter, it is
important to distinguish the type of gap it involves (wh-Movement vs. VP
Ellipsis / Fronting) and the position of the finite auxiliary it requires (subject-
auxiliary inversion vs. non-inversion). The acceptability of adverb placement
in front of a gap is listed in Figure 1. Some element must intervene between
an adverb with a wide scope reading and a gap in non-inversion construc-
tions; i.e. wide scope adverbs cannot precede any type of gap. However, if the
adverb is to be interpreted as having narrow scope or subject-auxiliary inver-
sion takes place, all types of adverbs may occur in front of a wh-gap, whereas
only epistemic, temporal, and frequency adverbs, but not subject-oriented and
evidential ones permit a subsequent gap caused by VP Ellipsis or Fronting.

adverb type scope syntactic construction

Fronting VP Ellipsis wh-Movement

non- non- non-
inversion inversion inversion inversion inversion
epistemic wide * *
*

temporal wide * *
*

narrow





frequency wide * *
*

narrow





evidential wide * * * *

narrow * * *


subject- wide * * * *

oriented narrow * * *

Figure 1. Adverb occurrence in front of a gap

 Eva Engels

. Previous approaches

. Baker (1971, 1981)

Baker (1971, 1981) accounts for the flexibility of the order of adverb and finite
auxiliary and the restrictions on it by assuming various auxiliary movement
rules applying at different stages in the derivation:
(31) a. Not Placement / Subject-Auxiliary Inversion
b. VP Deletion4
c. Stress Reduction (blocked by deletion sites)
d. Auxiliary Shift / Auxiliary Contraction (applying only to stress-
reduced auxiliaries)

Auxiliary Shift is considered to optionally move an unstressed auxiliary across

an adverb, giving rise to the variability in the order of adverb and finite auxil-
iary. The occurrence of a contracted auxiliary or an adverb in front of a gap is
ruled out by the sensitivity of Stress Reduction to deletion sites: applying only
to stress-reduced auxiliaries, Auxiliary Shift and Auxiliary Contraction can-
not affect an auxiliary preceding a gap; hence, the auxiliary may neither move
across the adverb, nor may it contract.
By contrast, according to Baker (1981: 312), auxiliary movement in inter-
rogative and negative sentences does not involve Auxiliary Shift, but rather
Subject-Auxiliary Inversion and Not Placement, respectively –, rules that force
the finite auxiliary to move in front of the subject or sentential not prior to
deletion. Consequently, the auxiliary may occupy a position in which it can
be affected by Stress Reduction and Auxiliary Contraction. Auxiliary contrac-
tion is thus predicted to be possible in gap constructions unless the auxiliary
immediately precedes the gap.
With respect to adverb placement, moving a finite auxiliary by the appli-
cation of Subject-Auxiliary Inversion or Not Placement may result in adverb
occurrence in front of a gap as in (20) and (21). However, not all types of
adverbs may precede all types of gaps even if they have narrow scope or subject-
auxiliary inversion takes place. Consequently, the auxiliary movement rules
cannot explain the dependence of adverb occurrence in front of a gap on the
type of adverb and the type of gap: all types of adverbs allow for a following
wh-gap; yet, while epistemic, temporal, and frequency adverbs can precede a
gap induced by VP Ellipsis or Fronting, subject-oriented and evidential ones
cannot. As Ernst (1983: 548) mentions, making Baker’s rules sensitive to dif-
ferent types of adverbs does not solve the problem. If the presence of e.g. a
 Optimizing adverb placement in gap constructions 

subject-oriented adverb would block Subject-Auxiliary Inversion, (32a) could

not be derived. However, the corresponding sentence without an elided VP
(32b) would also be ruled out:
(32) a. *Has he wisely? (Ernst 1983: 548)
b. Has he wisely finished all he set out to do?

What would need to be sensitive to different types of adverbs are the rules that
trigger the occurrence of the various types of gaps: subject-oriented and ev-
idential adverbs block VP Ellipsis and Fronting when no phonetic material
follows them because of prior auxiliary movement, while epistemic, frequency,
and temporal ones do not. By contrast, the presence of an adverb should never
block wh-Movement. Hence, the distributional pattern of adverbs in gap con-
structions cannot solely be accounted for by distinguishing various auxiliary
movement rules. At the very least, highly specific rules responsible for the
occurrence of gaps would also be necessary.

. Sag (1978, 1980), Sag and Fodor (1995), Kim and Sag (1996, 2002)

Sag (1978, 1980) accounts for the restrictions on adverb occurrence in gap
constructions by supposing the filter in (33):
(33) No surface structure is well-formed if it contains a sequence of the form
Adverb – Extraction site. (Sag 1978: 149)

The filter correctly predicts the unacceptability of pre-gap placement of an ad-

verb with wide scope reading in non-inversion constructions. However, it is
too strong, ruling out all adverb occurrences in front of a gap: some types of
adverbs may precede some types of gaps if they have narrow scope or subject-
auxiliary inversion takes place. Again, making the filter sensitive to different
types of adverbs, i.e. applying only to subject-oriented and evidential adverbs,
but not to epistemic, temporal, and frequency ones, would not suffice. On the
one hand, subject-oriented and evidential adverbs would still be wrongly pre-
dicted to be unacceptable in front of a wh-gap. On the other hand, epistemic,
temporal, and frequency adverbs with wide scope interpretation would be ex-
pected to be able to precede a gap in non-inversion constructions also, contrary
to fact. Hence, in order to account for the whole distributional pattern, Sag’s
filter would have to be sensitive not only to different types of adverbs, but also
to their scope and the syntactic constructions.
Later HPSG works, such as Sag and Fodor (1995) and Kim and Sag (1996,
2002), assume that a gap is literally empty. Supposing that an adverb adjoins to
 Eva Engels

the constituent it modifies, adverb placement in front of a gap is again wrongly

predicted to always be ungrammatical: no structure being projected at the gap
site, there is nothing the adverb could adjoin to. If the radical emptiness of gaps
is assumed, there seems to be no way to account for why placement of (some
types of) adverbs in front of (some types of) gaps is acceptable when the adverb
has narrow scope or subject-auxiliary inversion takes place.

. Empty Category Principle approaches

ECP approaches (e.g. Lobeck 1987, 1995; Potsdam 1997; Zagona 1988) fail to
account for the restrictions on adverb placement in gap constructions as well.
By requiring an elliptic VP to be properly governed, they ensure the occurrence
of an auxiliary, accounting for the contrast in (34):5
(34) Pete isn’t singing even though
a. most of his friends are __.
b. *most of his friends __. (Lobeck 1995: 47)

However, the licensing conditions on VP Ellipsis cannot even explain the

ungrammaticality of wide scope adverb placement in front of a gap in non-
inversion constructions. The Generalized Government Transparency Corol-
lary6 correctly predicts VP Ellipsis to be possible in inversion constructions;
the finite auxiliary governs its trace and thus licenses the elliptic VP:
(35) Why is Mary leaving, and why is John?

The fact that an adverb in post-auxiliary position can be interpreted as having

wide scope indicates that the order finite auxiliary – wide scope adverb is derived
by movement of the auxiliary across the adverb: the adverb c-commands the
trace of the finite auxiliary and thus is predicted to be able to take wide scope by
Ernst’s (1992) version of the Scope Principle.7 Consequently, auxiliary move-
ment as in (36a) should also be possible in a VP Ellipsis construction like (36b)
unless the intervening adverb blocks licensing of the elliptic VP. However, if it
is assumed to do so, the sentences in (37) are predicted to be unacceptable as
well, contrary to fact:
(36) Mary cannot lift 200 pounds, but
a. [IP John cani [AuxP probably [AuxP ti [VP lift 200 pounds]]]]
b. *[IP John cani [AuxP probably [AuxP t i [VP e]]]]
‘John is probably able to lift 200 pounds.’
 Optimizing adverb placement in gap constructions 

(37) [John has gotten along well with Fred recently, ...]
a. but [IP he hasi [NegP not [AuxP always [AuxP t i [VP e]]]]]
b. [CP Hasi [IP he t i ’ [AuxP really [AuxP t i [VP e]]]]]

Hence, relying exclusively on requirements on the auxiliary, ECP approaches

will not be able to account for the dependence of adverb occurrence in front of
a gap on adverb type, scope, and syntactic construction.

. Summary

Neither conditions on auxiliaries (Baker 1971, 1981; Lobeck 1987, 1995; Pots-
dam 1997; Zagona 1988) nor restrictions on adverb placement (Sag 1978, 1980;
Sag & Fodor 1995; Kim & Sag 1996, 2002) alone can explain the influence of
adverb type, scope, and syntactic construction on the grammaticality of ad-
verb occurrence in front of a gap. In the next section, an Optimality Theoretic
approach to adverb placement will be developed. Assuming that a hierarchy
of violable constraints determines the grammaticality of syntactic structures,
the interrelation of the factors affecting the distribution of adverbs in gap
constructions can be accounted for.

. An Optimality Theoretic approach

. Adverb placement: flexibility and restrictions

As illustrated in (2), adverb placement is restricted by semantic factors: under

a wide scope reading adverbs may occur on either side of the finite auxiliary
whereas narrow scope interpretation of an adverb is restricted to the post-
auxiliary position.
Ernst (1998, 2002) suggests that adverbs and auxiliaries are lexically spec-
ified for the type of semantic argument they select. The semantic types are
considered to be hierarchically ordered: SPEECH-ACT > FACT > PROPOSI-
TION > EVENT > SPECIFIED EVENT. A semantic (sub-)type may be freely
converted to a higher (sub-)type. According to Ernst, semantic structure is not
directly anchored in syntactic structure, i.e. a specific syntactic category does
not correspond to a specific semantic type; rather, it depends on the items in-
volved which semantic type a syntactic constituent may be associated with.
Assuming that an adverb semantically combines with its sister constituent, the
flexibility of and restrictions on the order of adverbs and auxiliaries (as well as
 Eva Engels

among adverbs) follow: an adverb may occur in any position in which its lexical
requirements can be fulfilled. For example, both EVENT-selecting frequency
adverbs and PROPOSITION-selecting epistemic ones can precede or follow a
finite EVENT-selecting aspectual auxiliary. The sister constituent of the adverb
(i.e. I’ or AuxP) can be associated with the semantic type the adverbs select for,
EVENT or PROPOSITION, respectively.
(38) a. [IP They [I’ frequently [I’ havei [AuxP2 t i [AuxP1 been
[E”’ [E” [E’
[VP knocked off their feet]]]]]]
[E
b. [IP They [I’ probably [I’ havei [AuxP2 t i [AuxP1 been
[PROP’ [PROP [E” [E’
[VP knocked off their feet]]]]]]
[E
(39) a. [IP They havei [AuxP2 frequently [AuxP2 t i [AuxP1 been
[E”’ [E” [E’
[VP knocked off their feet]]]]]
[E
b. [IP They havei [AuxP2 probably [AuxP2 ti [AuxP1 been
[PROP’ [PROP [E” [E’
[VP knocked off their feet]]]]]
[E
In addition, frequency adverbs may follow non-finite auxiliaries whereas epis-
temic ones cannot. This contrast is due to the different selectional properties
of the adverbs: the VP in (40) may only be of the semantic type EVENT since
the higher auxiliaries select for EVENTs. Consequently, the lexical requirement
of a frequency adverb, but not the one of an epistemic adverb can be met in
VP-adjoined position:
(40) a. [IP They havei [AuxP2 t i [AuxP1 been (Ernst 2002: 349)
[E”’ [E”
[VP frequently [VP knocked off their feet]]]]]
[E’ [E
b. *[IP They havei [AuxP2 t i [AuxP1 been [VP probably
[E”’ [E” [E’
[VP knocked off their feet]]]]]
[E
 Optimizing adverb placement in gap constructions 

Based on Ernst’s analysis, an Optimality Theoretic approach to the semantic

restrictions on adverb placement is outlined in Engels (2002): the input to a
syntactic competition is considered to be a semantic representation specifying
the type and content of an adverb’s semantic argument and, thus, its scope. For
some input I, GEN generates a set of syntactic forms {cand1 , cand2 , ..., candn }.
The inviolable constraint Modifier in GEN guarantees that in all output can-
didates, an adverb is generated as sister to its semantic argument specified in
the input.
(41) Modifier: The foot of an adverb chain is the sister of the foot of the chain
of its semantic argument.

By referring to the feet of chains, Modifier permits GEN to produce candi-

dates in which an adverb does not surface as sister to its semantic argument
due to movement of the adverb or its semantic argument. For example, (42a,
b) will be included in the candidate set for an input that specifies the adverb as
taking narrow scope with respect to negation, while (42c) will not since adver-
bial scope is not reflected correctly in syntactic structure: the sister constituent
of the adverb in its base position contains not which does not belong to the
adverb’s semantic argument.
(42) a. [Read a book]j John hasi not always t i t j .
b. John hasi alwaysj not t j t i read a book.
c. John hasi always not t i read a book.

Considering semantic types not to be connected to specific syntactic categories,

GEN can (base-)adjoin an adverb to any projection which may represent its se-
mantic argument. Consequently, different sets of candidates are produced de-
pending on the scope of an adverb.8 The function EVAL chooses the candidate
that optimally satisfies the constraint hierarchy as output O.
Assuming that auxiliaries are generated in their own projections and that
the finite auxiliary has to move to the highest head position, the variability of
the order of finite auxiliary and wide scope adverb in non-inversion construc-
tions results from the freedom to adjoin an adverb with wide scope reading
somewhere above the base position of the finite auxiliary: adjoined to I’, the
adverb precedes the finite auxiliary in I, while adjoined to AuxP, the adverb fol-
lows it.9 Auxiliary movement to I is ensured by the dominance of ObHd over
Stay, as shown in Tableau 1 (see Grimshaw 1997).
(43) a. Obligatory Heads (ObHd): A projection has a head.10
b. Stay: Trace is not allowed. (Grimshaw 1997: 374)
 Eva Engels

Tableau 1. Variable order wide scope adverb + finite auxiliary

ObHd Stay

F (a) [IP John [I’ musti [AuxP frequently [AuxP ti [VP see the doctor]]]]] *

F (b) [IP John [I’ frequently [I’ musti [AuxP ti [VP see the doctor]]]]] *
(c) [IP John [I’ e [AuxP frequently [AuxP must [VP see the doctor]]]]] *!
(d) [IP John [I’ frequently [I’ e [AuxP must [VP see the doctor]]]]] *!

Tableau 2. Strict order finite auxiliary – narrow scope adverb

ObHd Stay

F (a) [IP John [I’ musti [AuxP ti [VP frequently [VP see the doctor]]]]] *
(b) [IP John [I’ frequentlyk [I’ musti [AuxP ti [VP tk [VP see the doctor]]]]]] **!

Note that for ease of exposition, the constraint forcing movement of the subject
from Spec,VP to Spec,IP, the trace left behind, and the violation of Stay caused
by this movement is omitted in all Tableaux; additionally, any constraints will
be left out of Tableaux that are irrelevant for the phenomenon at hand in the
sense that they do not affect any of the competing candidates.
The frequency adverb with wide scope interpretation in Tableau 1 modifies
the event of ‘John being obliged to see the doctor’. Adjunction of the adverb to
I’ or AuxP reflects this semantic relation syntactically (Modifier); (the traces
of) the elements which belong to the adverb’s semantic argument are included
in the sister constituent of the adverb. When ObHd is ranked above Stay, the
finite auxiliary has to move to I. Candidate (c) and (d) are ruled out because I is
empty, violating ObHd. Candidate (a) and (b) tie on the constraint profile and,
consequently, arise as variants: adjoined to I’ or AuxP, the wide scope adverb
precedes or follows the finite auxiliary, respectively.
By contrast, if the frequency adverb takes narrow scope with respect to the
modal, expressing the meaning that John is obliged to see the doctor several
times, it may only occur in post-auxiliary position (see Tableau 2). Because
of Modifier, only structures in which the adverb is merged below the modal
enter the syntactic competition. The extra violation of Stay rules out move-
ment of the adverb in front of must as in candidate (b). Candidate (a) is the
sole output: the adverb has to occur in its base position following the finite
auxiliary.11
Summing up, the flexibility of and restrictions on adverb placement result
from the interplay of GEN and the syntactic competition. GEN is restrained
 Optimizing adverb placement in gap constructions 

in that only syntactic structures are produced in which an adverb is generated

as sister to its semantic argument (specified in the input). As a consequence,
under a wide scope reading an adverb may adjoin somewhere above the base
position of the finite auxiliary, whereas the adverb must be merged below the
wide scope element (e.g. modal verb or negation) if it is to take narrow scope.
Due to ObHd >> Stay, the finite auxiliary obligatorily moves to I and adverbs
occur in their base positions: a wide scope adverb may thus precede or follow
the finite auxiliary while a narrow scope adverb must follow it.

. Wh-Movement, Fronting, and VP Ellipsis

Since Stay penalizes movement, Wh-Movement and Fronting must be trig-

gered by some constraint outranking Stay:
(44) a. Wh-in-Specifier-Position (WhSp): A wh-operator occurs in speci-
fier position. (Müller 1997: 263)
b. Operator-in-Scope-Position (OpSc): An operator occurs in scope
position.12

Because of WhSp, OpSc >> Stay, wh-phrases and topic constituents occur in
the clause-initial position in an optimal output. Unlike matrix wh-Movement,
Fronting does not involve subject-auxiliary inversion. This is due to the fact
that WhSp, but not OpSc explicitly requires placement of a phrase in Spec,CP:
projecting CP introduces a further head position; as a consequence, either Stay
or ObHd is (additionally) violated, depending on whether or not the finite aux-
iliary moves to C. Therefore, if movement to Spec,CP is not explicitly required,
adjunction to IP will be preferred (compare Tableau 3 and 4).
Although CP is also projected in embedded wh-questions due to WhSp,
subject-auxiliary inversion does not take place. Grimshaw (1997) accounts for
this contrast between matrix and embedded questions by the constraint Pu-
rity of Extended Projection (PureEP)13 which prohibits movement into

Tableau 3. Matrix wh-Movement

WhSp OpSc ObHd Stay

F (a) [CP whatj hasi [IP John ti ’ [AuxP ti [VP read tj ]]]] ***

(b) [CP whatj e [IP John hasi [AuxP ti [VP read tj ]]]] *! **

(c) [IP whatj [IP John hasi [AuxP ti [VP read tj ]]]] *! **

(d) [IP John hasi [AuxP ti [VP read what]]] *! * *

 Eva Engels

Tableau 4. Fronting

OpSc ObHd Stay

(a) [CP [this booktop ]j hasi [IP John ti ’ [AuxP ti [VP read tj ]]]] ***!
(b) [CP [this booktop ]j e [IP John hasi [AuxP ti [VP read tj ]]]] *! **

F (c) [IP [this booktop ]j [IP John hasi [AuxP ti [VP read tj ]]]] **
(d) [IP John hasi [AuxP ti [VP read this booktop ]]] *! *

Tableau 5. Embedded wh-Movement

Wh Op Pure Ob Stay
Sp Sc EP Hd
(a) [CP whatj hasi [IP John ti ’ [AuxP ti [VP read tj ]]]] *! ***

F (b) [CP whatj e [IP John hasi [AuxP ti [VP read tj ]]]] * **

(c) [IP whatj [IP John hasi [AuxP ti [VP read tj ]]]] *! ** **
(d) [IP John hasi [AuxP ti [VP read what]]] *! * *

the highest head position of a subordinate clause. Outranking ObHd, PureEP

blocks subject-auxiliary inversion in embedded questions (see Tableau 5).
While in wh-Movement and Fronting constructions some phrase is pro-
nounced in a derived position, an elliptic VP is not phonetically realized at
all, obscuring its actual position. However, since infinitival to may license VP
Ellipsis unless the infinitival clause is an island, Johnson (2001) suggests that
VP Ellipsis involves VP Fronting:
(45) a. Mag wants to read Fred’s story, and I also want to __.
(Johnson 2001: 440)
b. *Mag came to read Fred’s story, and I also came to __.
(46) a. You shouldn’t play with rifles because it’s dangerous to __.
(Johnson 2001: 442)
b. *You shouldn’t play with rifles because to __ is dangerous.

Under the assumption that the VP has to topicalize before its phonetic material
is elided, the ungrammaticality of the (b)-sentences can be explained: the VP
cannot land within the infinitival clause, nor may it move out of the infinitival
clause because of its islandhood.
I do not know of any Optimality Theoretic approach to VP Ellipsis. As
in e.g. Hankamer and Sag (1976), Chomsky (1995), and Lasnik (1995), VP
 Optimizing adverb placement in gap constructions 

Tableau 6. VP Ellipsis

VP-E OpSc ObHd Stay

F (a) [IP [VP read the booktop ]j [IP John hasi [AuxP ti tj ]]] **

F (b) [IP [VP read the book top]j [IP John hasi [AuxP ti tj ]]] **
(c) [IP John hasi [AuxP ti [VP read the booktop ]]] *! *
(d) [IP John hasi [AuxP ti [VP read the book top]]] *! * *

Ellipsis will be considered to be a PF phenomenon: the phonetic material of

VP is deleted at PF for economy of pronunciation.14 The constraint in (47)
ensures that VP Fronting takes place prior to deletion:
(47) VP-Ellipsis (VP-E): If a VP α is deleted, α occurs in topic position.

Since Stay penalizes movement, an elliptic VP must have a topic feature;

OpSc forces its movement to the topic position in which it can be elided (see
Tableau 6).
Hence, in VP Ellipsis constructions, the ellipsis site occurs in the clause-
initial position and a trace occupies the base position of VP. Concerning adverb
placement, the question is thus not whether an adverb may precede the ellipsis
site, but whether it may precede the trace left behind by VP Fronting.
To sum up, wh-Movement, Fronting, and VP Ellipsis are similar in that
they involve movement of some constituent to the clause-initial position and a
trace occupies the corresponding base position. While the moved constituent
adjoins to IP in VP Ellipsis and Fronting constructions (VP-E, OpSc >> Stay),
wh-questions involve movement to Spec,CP (WhSp >> Stay); depending on
the type of question, matrix vs. embedded, subject-auxiliary inversion takes
place (PureEP >> ObHd >> Stay).

. Adverb placement in gap constructions

.. Wide scope adverb placement in non-inversion gap constructions

In gap constructions, the variability in the order of the finite auxiliary and an
adverb with wide scope reading is limited. Unless some other overt element
intervenes between the adverb and the gap, the order wide scope adverb – finite
auxiliary is obligatory. Since adverbs and contracted auxiliaries are similar in
their avoidance of the pre-gap position and the acceptability of auxiliary con-
traction is clearly dependent on the phonological context to the left of a gap,
 Eva Engels

it will be assumed that adverbs also require phonetic material that separates
them from a gap:
(48) Adjunct-Content (AdjCon): The sister constituent of an adjunct in-
cludes phonetic material.

Movement of a phrase out of the constituent the adverb adjoins to deprives this
constituent of phonetic material. In case AuxP does not dominate any overt el-
ement due to Aux-to-I movement and, for example, VP Ellipsis, post-auxiliary
adverb placement (i.e. AuxP-adjunction) is suboptimal to pre-auxiliary ad-
verb placement (i.e. I’-adjunction) because of AdjCon (see Tableau 7). AuxP-
adjunction fatally violates AdjCon whereas I’-adjunction satisfies the con-
straint: I’ includes the finite auxiliary and thus an overt element. Therefore,
the adverb must adjoin to I’ and, consequently, precedes the finite auxiliary.15
However, by requiring some phonetic material in the constituent the ad-
verb adjoins to, AdjCon does not rule out the order finite auxiliary – wide scope
adverb in gap constructions per se. Satisfying AdjCon, this order is correctly

Tableau 7. Strict order wide scope adverb – finite auxiliary in gap constructions

VP- Op Ob Stay Adj

E Sc Hd Con
(a) [IP [VP read the book top]j [IP she hasi [AuxP ** *!
wisely [AuxP ti tj ]]]]

F (b) [IP [VP read the booktop]j [IP she [I’ wisely [I’ hasi **
[AuxP ti tj ]]]]]
(c) [IP she hasi [AuxP wisely [AuxP ti [VP read the *! *
booktop ]]]]
(d) [IP she [I’ wisely [I’ hasi [AuxP ti [VP read the *! *
booktop ]]]]]

Tableau 8. Variable order wide scope adverb + finite auxiliary in gap constructions with
stranded participial auxiliary

VP- Op Ob Stay Adj

E Sc Hd Con

F (a) [IP [VP reading the booktop]j [IP she hasi **

[AuxP1 frequently [AuxP1 ti [AuxP2 been tj ]]]]]

F (b) [IP [VP reading the book top]j [IP she [I’ fre- **
quently [I’ hasi [AuxP1 ti [AuxP2 been tj ]]]]]]
 Optimizing adverb placement in gap constructions 

predicted to be acceptable if the sister constituent of the adverb contains some

overt element (see Tableau 8). Hence, the order of an adverb with wide scope
interpretation and the finite auxiliary is optional as long as some phonetic
material follows this sequence due to the requirement of AdjCon.

.. Narrow scope adverb placement in gap constructions

Recall that under the narrow scope reading, an adverb is merged below the el-
ement that takes scope over it and that adverbs occur in their base position, i.e.
narrow scope adverbs follow the finite auxiliary (Modifier, ObHd >> Stay).
They cannot even move to a higher position in order to escape their placement
in front of a gap due to Stay >> AdjCon. Hence, while pre-auxiliary posi-
tioning may prevent an adverb with wide scope reading from preceding a gap
in non-inversion constructions, this is not an option for adverbs with narrow
scope interpretation. The acceptability of their occurrence in front of a gap
depends on the type of adverb and the type of gap: temporal and frequency
adverbs may precede any type of gap. Narrow scope evidential and subject-
oriented adverbs, by contrast, may occur in front of a wh-gap, while they seem
to prohibit VP Ellipsis and Fronting.
First, since it depends on the type of adverb whether or not it may precede
a Fronting gap, assume that AdjCon is sensitive to different types of adverbs;
i.e. AdjCon is considered to constitute a family of subconstraints relating to the
various types of adverbs, AdjConfrequency, AdjConsubject–oriented etc. Secondly, fol-
lowing the Copy and Deletion Theory of Movement (Chomsky 1995), a trace
will be regarded as a copy of a moved constituent from which the phonetic
features are deleted. Because of the interaction of Pronunciation Economy
(requiring the deletion of phonetic matrices) and Recoverability (requir-
ing unpronounced elements to be recoverable), at most one copy of some
constituent will normally be phonetically realized (see Groat & O’Neil 1996;
Pesetsky 1998; Nunes 2001). The constraint in (49) may regulate which copy is
pronounced:
(49) Phonetic Realization (PhRe): Don’t pronounce any non-highest mem-
ber of a chain.

To satisfy this constraint, a moved constituent is not spelled out in its base
(or some intermediate) position, as indicated by striking through the unpro-
nounced material. Since the various types of gaps contrast in whether or not
subject-oriented and evidential adverbs may precede them, it will be assumed
that PhRe is subdivided according to the type of movement, PhRewh , PhRefront
etc.16 The dependence of narrow scope adverb occurrence in front of a gap on
 Eva Engels

Tableau 9. Acceptability of the order narrow scope frequency adverb – Fronting gap

VP Op Ob Stay Ph Adj
E Sc Hd Refront Confrequ

F (a) [IP [VP see ...top ] [IP John [I’ must [AuxP ** *
must [VP frequently [VP see ...top ]]]]]]
(b) [IP [VP see ...top ] [IP John [I’ must [AuxP ** *!
must [VP frequently [VP see ...top ]]]]]]
(c) [IP [VP see ...top ] [IP John [I’ frequently ***!
[I’ must [AuxP must [VP frequently [VP
see ...top ]]]]]]]

the type of adverb and the type of gap can now be accounted for by the ranking
of the AdjCon-subconstraints with respect to the PhRe-subconstraints.
Since frequency and temporal adverbs may precede all types of gaps, both
PhRe-subconstraints must outrank their AdjCon-subconstraints (PhRewh >>
PhRefront >> AdjConfrequ , AdjContemp ). Consequently, deletion of the phonetic
material in the foot of the chain of a wh-moved or fronted constituent may take
place even though this results in adjoining an adverb to a phonetically empty
constituent, as shown for VP Ellipsis in Tableau 9.
Hence, while a temporal or frequency adverb may precede a gap if it takes
narrow scope, it cannot do so if it takes wide scope; i.e. while placement of
a narrow scope adverb in front of a gap is optimal in spite of the violation
of AdjCon, pre-gap placement of a wide scope adverb is excluded just be-
cause of the violation of AdjCon: pre-auxiliary positioning may prevent a wide
scope adverb from adjoining to an empty constituent and thus from violating
AdjCon.
Unlike temporal and frequency adverbs, evidential and subject-oriented
ones seem to prohibit VP Ellipsis and Fronting under a narrow scope read-
ing, while they do permit a following wh-gap: their AdjCon-subconstraints
outrank PhRefront whereas they are dominated by PhRewh (PhRewh >> Adj-
Consub-ori , AdjConevid >> PhRefront ). Consequently, the phonetic matrices are
retained in the foot of the chain of a fronted constituent preventing the ad-
verb from preceding a gap, while a wh-moved phrase is spelled out in the
head position of the chain resulting in adverb placement in front of a gap (see
Tableau 10 and 11). Note that in some sense, VP Ellipsis takes place in the out-
put of Tableau 10; the output satisfies OpSc and Vp-E: a copy of VP occurs in
clause-initial position and its phonetic material is deleted there. However, this
is obscured by the phonetic realization of VP in its base position.17
 Optimizing adverb placement in gap constructions 

Tableau 10. Unacceptability of the order narrow scope subject-oriented adverb –

Fronting gap

VP Op Ob Stay Adj Ph
E Sc Hd Consub-ori Refront
(a) [IP [VP see ...top ] [IP John [I’ must [AuxP ** *!
must [VP wisely [VP see ...top ]]]]]]

F (b) [IP [VP see ...top ] [IP John [I’ must [AuxP ** *
must [VP wisely [VP see ...top ]]]]]]

(c) [IP [VP see ...top ] [IP John [I’ wisely ***!
[I’ must [AuxP must [VP wisely [VP see
...top ]]]]]]]

Tableau 11. Acceptability of the order narrow scope subject-oriented adverbs – wh-gap
(embedded)

Wh Pure Ob Stay Ph Adj

Spec EP Hd Rewh Consub-ori

F (a) [CP where e [IP John [I’ was [NegP not [VP * ** *
wisely [VP was where]]]]]]

(b) [CP where e [IP John [I’ was [NegP not [VP * ** *!
wisely [VP was where]]]]]]

(c) [CP where e [IP John [I’ wisely [I’ * ***!

was [NegP not [VP wisely [VP was
where]]]]]]]

Summing up, in the narrow scope reading adverbs have to follow the fi-
nite auxiliary (Modifier, ObHd >> Stay >> AdjCon). The acceptability of
their occurrence in front of a gap depends on the type of adverb and the
type of gap, as captured by the ranking of the AdjCon-subconstraints and the
PhRe-subconstraints: PhRewh >> AdjConsub-ori , AdjConevid >> PhRefront >>
AdjConfrequ , AdjContemp .
Wh-Movement is obligatory. Due to the dominance of PhRewh over all
AdjCon-subconstraints, a wh-phrase is spelled out in the clause-initial posi-
tion; the phonetic material in its base position is deleted even if this leads to
adjunction of an adverb to a phonetically empty constituent. Fronting and VP
Ellipsis, by contrast, are optional. However, they seem to be prohibited if they
would result in placement of a subject-oriented or evidential adverb in front of
a gap: the AdjCon-subconstraints of these types of adverbs outrank PhRefront
blocking the deletion of the phonetic matrices in the foot of a Fronting-chain in
 Eva Engels

order to prevent these adverbs from preceding a gap. VP Ellipsis and Fronting
may take place if the adverb has wide scope since adverb placement in front of
a gap can be avoided by pre-auxiliary positioning.
The contrast between temporal / frequency adverbs and subject-oriented /
evidential adverbs in permitting a subsequent Fronting gap results from the
difference in the ranking of their AdjCon-subconstraints with respect to
PhRefront . Note that the hierarchical ranking of the AdjCon-subconstraints
does not correspond to the hierarchy of semantic types, i.e. the ability to pre-
cede a Fronting gap is not influenced by the type of semantic argument an
adverb selects for: while the AdjCon-subconstraints of FACT-selecting ev-
idential and EVENT-selecting subject-oriented adverbs dominate PhRefront ,
PhRefront outranks the AdjCon-subconstraints of EVENT-selecting frequency
and PROPOSITION-selecting epistemic adverbs, as shown in the next section.
Why PhRefront splits the AdjCon-subconstraints in the way it does is a question
for further research.

.. Adverb placement in gap constructions with

subject-auxiliary inversion
The distributional pattern of adverbs in gap constructions with subject-
auxiliary inversion parallels the one in constructions in which the adverb takes
narrow scope. This similarity is not surprising: in both cases the finite auxil-
iary precedes the adverb because of either base generation above it (Modifier)
or movement across it (ObHd >> Stay). Matrix CP being projected, the fi-
nite auxiliary moves to C, and, in contrast to non-inversion constructions, the
occurrence of a wide scope adverb in front of a gap cannot be avoided by its
adjunction to I’. Assuming that C is associated with an interrogative operator,
the base generation of an adverb as adjunct to C’ or CP is ruled out by GEN
since the scope of the adverb is not reflected correctly (see Ernst 2002).18 Ad-
ditionally, Stay >> AdjCon blocks movement of an adverb in front of some
overt element. The ungrammaticality of the sentences in (50) is thus expected:
(50) a. *Really where is he?
b. *Where really is he?
c. *Where is really he?

Hence, the ranking of PhRewh and PhRefront with respect to the AdjCon-
subconstraints will be decisive for the acceptability of adverb placement in
front of a gap in subject-auxiliary inversion constructions, predicting it to de-
pend on the type of adverb and the type of gap and thus to mirror the pattern of
narrow scope adverbs. Those adverbs whose AdjCon-subconstraints are dom-
 Optimizing adverb placement in gap constructions 

Tableau 12. Acceptability of the order epistemic adverb – Fronting gap in inversion
constructions
VP Wh Op Ob Stay Ph Adj
E Spec Sc Hd Refront Conepist

F (a) [CP Q has [IP [VP gotten ...top ] [IP he *** *

has [AuxP really [AuxP has [VP gotten ...
top ]]]]]]

F (b) [CP Q has [IP [VP gotten ... top ] [IP he *** *
[I’ really [I’ has [AuxP has [VP gotten ...
top ]]]]]]]

(c) [CP Q has [IP [VP gotten ... top ] [IP re- ****!
ally [IP he has [AuxP really [AuxP has [VP
gotten ... top ]]]]]]]
(d) [CP Q has [IP [VP gotten ... top ] [IP re- ****!
ally [IP he [I’ really [I’ has [AuxP has [VP
gotten ... top ]]]]]]]]

(e) [CP Q has [IP [VP gotten ... top ] [IP he *** *!
has [AuxP really [AuxP has [VP gotten ...
top ]]]]]]

(f) [CP Q has [IP [VP gotten ... top ] [IP he *** *!
[I’ really [I’ has [AuxP has [VP gotten ...
top ]]]]]]]

inated by a PhRe-subconstraint may occur in front of the corresponding type

of gap: phonetic deletion in the base position of the moved constituent takes
place even if it results in adjunction of the adverb to a phonetically empty con-
stituent, as illustrated for an epistemic adverb in a yes/no-question with an
elliptic VP in Tableau 12 and for a subject-oriented adverb in a wh-question
in Tableau 13. Note that since I’-adjunction and AuxP-adjunction of an adverb
with wide scope interpretation do equally well on AdjCon and PhRe, the ad-
junction site is optional in gap constructions with subject-auxiliary inversion.
By contrast, those types of adverbs whose AdjCon-subconstraints out-
rank a PhRe-subconstraint block deletion of the phonetic material in the
foot position of the corresponding movement chain in order to escape their
placement in front of a gap. Inversion and non-inversion constructions thus
seem to contrast in permitting VP Ellipsis in sentences containing evidential or
subject-oriented adverbs (compare Tableau 7 and 14).
 Eva Engels

Tableau 13. Acceptability of the order subject-oriented adverb – wh-gap in inversion

constructions
Wh Ob Stay Ph Adj
Spec Hd Rewh Consub-ori

F (a) [CP where was [IP John was [VP wisely *** *
[VP was where]]]]

F (b) [CP where was [IP John [I’ wisely [I’ was *** *
[VP was where]]]]]
(c) [CP where was [IP wisely [IP John was ****!
[VP wisely [VP was where]]]]]
(d) [CP where was [IP wisely [IP John [I’ ****!
wisely [I’ was [VP was where]]]]]]

(e) [CP where was [IP John was [VP wisely *** *!
[VP was where]]]]
(f) [CP where was [IP John [I’ wisely [I’ was *** *!
[VP was where]]]]]

To sum up, both narrow scope of an adverb and projection of matrix CP

enforce placement of the finite auxiliary in front of an adverb (Modifier,
ObHd >> Stay). The finite auxiliary thus may not serve to prevent an adverb
from adjoining to a phonetically empty constituent as it does in non-inversion
gap constructions with wide scope adverbs (Modifier, Stay >> AdjCon).
Depending on the type of adverb and the type of gap, the adverb may pre-
cede the gap. These dependencies are accounted for by the relative ranking
of the AdjCon-subconstraints and the PhRe-subconstraints which determines
whether or not deletion of the phonetic material in the base position of some
moved constituent may take place, leaving an adverb in front of a gap.

. Adverb placement in French gap constructions

Evidence in favor of the analysis presented above comes from restrictions on

pre-gap adverb placement in French which can only be discussed briefly here
for reasons of space.
In contrast to English, adverbs cannot intervene between the subject and
the finite verb in French; the sentences in (51) indicate that adjunction of an
adverb to I’ is impossible in French and that both finite auxiliaries and finite
lexical verbs undergo movement to I.19 In addition, even e.g. PROPOSITION-
 Optimizing adverb placement in gap constructions 

Tableau 14. Unacceptability of the order evidential adverb – Fronting gap in inversion
constructions
VP Wh Op Ob Stay Adj Ph
E Spec Sc Hd Conevid Refront
(a) [CP Q has [IP [VP finished ...top ] [IP he *** *!
has [AuxP clearly [AuxP has [VP finished
... top ]]]]]]
(b) [CP Q has [IP [VP finished ... top ] [IP he *** *!
[I’ clearly [I’ has [AuxP has [VP finished ...
top ]]]]]]]

(c) [CP Q has [IP [VP finished ... top ] [IP ****!
clearly [IP he has [AuxP clearly [AuxP has
[VP finished ... top ]]]]]]]
(d) [CP Q has [IP [VP finished ... top ] [IP ****!
clearly [IP he [I’ clearly [I’ has [AuxP has
[VP finished ... top ]]]]]]]]

F (e) [CP Q has [IP [VP finished ... top ] [IP he *** *
has [AuxP clearly [AuxP has [VP finished
... top ]]]]]]

F (f) [CP Q has [IP [VP finished ... top ] [IP he *** *
[I’ clearly [I’ has [AuxP has [VP finished ...
top ]]]]]]]

selecting epistemic adverbs can follow participial verbs in French (52); i.e. non-
finite verb forms are able to move across an adverb:
(51) a. Jean (*souvent) a (souvent) embrassé Marie.
Jean (often) has (often) kissed Marie
‘Jean has often kissed Marie.’
b. Jean (*souvent) embrasse (souvent) Marie.
Jean (often) kisses (often) Marie
‘Jean often kisses Marie.’
(52) Jean a résolu probablement tes problèmes.
Jean has resolved probably your problems
‘Jean has probably resolved your problems.’ (Ernst 2002: 377)

Due to the ban on I’-adjunction, an adverb cannot escape its placement in

front of the gap by pre-auxiliary positioning in French; all types of adverbs
may precede a gap induced by VP Fronting:
 Eva Engels

(53) Soulever 200 kilos Jean pouvait {probablement / évidemment /

Lift 200 kilos Jean was.able.to {probably / evidently /
récemment / souvent / facilement} __.
recently / often / easily}
‘Jean could {probably / evidently / recently / often / easily} lift 200 kilos.’

However, as in parenthetical clause-final position (54a), epistemic and eviden-

tial adverbs following a non-finite modal in VP Fronting constructions strongly
tend to be set off from the clause by comma intonation; i.e. movement of the
non-finite modal (or rather, phonetic deletion in the foot of its chain) seems
not to be possible if it involves pre-gap placement of an epistemic or evidential
adverb (54b). By contrast, temporal, frequency, and subject-oriented adverbs
may follow the non-finite modal in VP Fronting constructions without comma
intonation. Note that subject-oriented adverbs should have narrow scope with
respect to a modal (see Note 3), while temporal and frequency adverbs may
take wide or narrow scope in post-modal position. Because of Modifier, the
modal must have crossed the adverb with wide scope reading in (54b) whereas
the order modal – adverb is base-generated under the narrow scope reading of
an adverb as in (54c).
(54) a. Jean a embrassé Marie *(,) {probablement / évidemment}.
Jean has kissed Marie , {probably / evidently}.
‘Jean has kissed Marie, {probably / evidently}.’
b. [Jean has been in an accident and is now paraplegic.]
[Courir 100 mètres en 6 secondes]j Jean a pui
Run 100 meters in 6 seconds Jean has been.able.to
{*probablement / *évidemment / auparavant / souvent} ti tj
{probably / evidently / previously / often}
‘Jean has {probably / evidently / previously / often} been able to run
100 meters in 6 seconds.’
c. [Jean has not only managed to run a marathon...]
[Courir 100 mètres en 6 secondes]i Jean a pu {juste
Run 100 meters in 6 seconds Jean has been.able.to {just
avant / souvent / facilement} ti .
before / often / easily}
‘Jean has been able to {often / easily} run 100 meters in 6 seconds {just
before that.}’

Summing up, adjunction of any type of adverb to a phonetically empty con-

stituent is possible in French VP Fronting constructions; i.e. the constraint
requiring spell-out of the fronted constituent in the head position of its
 Optimizing adverb placement in gap constructions 

chain (PhRefront ) dominates all AdjCon-subconstraints. However, epistemic

and evidential adverbs seem to block participle movement: their AdjCon-
subconstraints outrank the constraint that requires phonetic deletion in the
base position of the modal (PhReparticiple ), preventing these adverbs from pre-
ceding a gap. By contrast, temporal and frequency adverbs may occur in pre-
gap position following the non-finite modal even if they take wide scope: their
AdjCon-subconstraints are outranked by PhRepart .
Since subject-oriented adverbs must be generated below a modal verb,
the ranking of their AdjCon-subconstraint with respect to PhReparticiple can-
not be determined. However, c’est...que Clefts indicate that subject-oriented
adverbs are rather to be grouped with temporal and frequency adverbs than
with epistemic and evidential ones. Clefting may result in pre-gap placement
of temporal, frequency, and subject-oriented adverbs, but not of epistemic or
evidential ones (55); i.e. to avoid adjunction of an epistemic or evidential ad-
verb to a phonetically empty constituent, Clefting (or spell-out of the VP in
the c’est clause) is prevented from violating the constraints that requires this
(Cleft, for instance).
(55) C’est soulever 200 kilos que Jean pouvait {*probablement /
It is lift 200 kilos that Jean was.able.to {probably /
*évidemment / récemment / souvent / facilement}
evidently / recently / often / easily}
‘Jean {probably / evidently / recently / often} could {often / easily} lift 200
kilos.’

The (un-)acceptability of pre-gap adverb placement in French c’est...que Cleft

and VP Fronting constructions may be accounted for by the constraint
hierarchy in (56):
(56) PhRefront >> AdjConepist , AdjConevid >> Cleft >> PhReparticiple
(AdjConsub-ori) >> (AdjConsub-ori,) AdjContemp , AdjConfrequ

In summary, similar to English, adverb occurrence in French gap constructions

is sensitive to adverb type, gap type (VP Fronting vs. c’est...que Cleft), and verb
positioning (finite verb movement vs. non-finite verb movement). However,
English and French differ in which types of adverbs block phonetic deletion in
the foot position of some movement chain to escape pre-gap placement. This
cross-linguistic variation can be traced back to the difference in the ranking
of the AdjCon-subconstraints to each other (compare (57) below): while Adj-
Consub-ori outranks AdjConepist in English, the reverse ranking holds in French.
 Eva Engels

Consequently, the subconstraints of the AdjCon-family seem not to be subject

to an internal subhierarchy.

. Conclusion

Adverb placement in English gap constructions was shown to be influenced

by three factors: the type of adverb, its scope, and the syntactic construc-
tion (wh-Movement vs. Fronting / VP Ellipsis and subject-auxiliary inversion
vs. non-inversion). As argued in Section 2, neither requirements on auxil-
iaries, nor requirements on adverb placement alone may account for the whole
distributional pattern, predicting adverb occurrence in front of a gap to be
(un)acceptable independent (of some) of these factors. Considering grammat-
icality to be optimal satisfaction of a hierarchy of violable constraints, the
interrelation of these factors follows from the constraint hierarchy in (57) in
an Optimality Theoretic framework:
(57) VP-E, WhSp, OpSc, PureEP >> ObHd >> Stay >> PhRewh >>
AdjConsub-ori, AdjConevid >> PHREfront >> AdjConepist , AdjConfrequ,
AdjContemp

The inviolable constraint Modifier ensures that GEN only produces struc-
tures in which the adverbial scope specified in the input is reflected by sister-
hood: under a wide scope reading adverbs are merged somewhere above the
base position of the finite auxiliary in all output candidates, whereas under a
narrow scope reading, adverbs are generated below the wide scope element.
Due to the requirement of Stay, adverbs occur in their base positions.
The dominance of WhSp, OpSc, and VP-E over Stay triggers wh-Movement,
Fronting, and VP Ellipsis, respectively. Remember that VP Ellipsis is consid-
ered to involve VP Fronting so that a trace occupies the base position of VP.
The contrast between matrix questions, subordinate questions, and Fronting
with respect to subject-auxiliary inversion results from the ranking of PureEP,
ObHd, and Stay.
The obligatory movement of the finite auxiliary to the highest head po-
sition (ObHd >> Stay) and the freedom to adjoin an adverb with wide scope
interpretation to I’ or AuxP account for the fact that the order of finite auxiliary
and wide scope adverb in non-inversion constructions is optional as long as it
does not lead to adverb placement in front of a gap (AdjCon); hence, the order
finite auxiliary – adverb is ruled out if no overt element follows the adverb.
 Optimizing adverb placement in gap constructions 

Modifier and ObHd >> Stay enforce positioning of the finite auxiliary
above the adverb if the adverb takes narrow scope or matrix CP is projected.
Adverb placement in front of a gap cannot be circumvented by moving the ad-
verb across an overt element (Stay >> AdjCon). The various types of adverbs
differ in which types of gaps may follow them. The influence of adverb type
and gap type on the grammaticality of adverb occurrence in front of a gap can
be traced back to the AdjCon-subconstraints and the PhRe-subconstraints;
their relative ranking determines whether or not the phonetic material in the
foot position of some movement chain may be deleted, resulting in adverb ad-
junction to a phonetically empty constituent. Wh-Movement is obligatory; a
wh-phrase must be spelled out in the clause-initial position allowing all types of
adverbs to occur in front of a wh-gap (PhRewh >> AdjCon). Additionally, epis-
temic, frequency, and temporal adverbs can precede a gap induced by Fronting
(PHREfront >> AdjConepist , AdjConfrequ , AdjContemp ). By contrast, the pho-
netic matrices are retained in the base position of some fronted constituent
if this prevents a subject-oriented or evidential adverb from preceding a gap
(AdjConsub-ori , AdjConevid >> PHREfront ). VP Ellipsis and Fronting thus seem
to be permitted only if they do not involve adjunction of a subject-oriented or
evidential adverb to a phonetically empty constituent: they may take place in
non-inversion constructions in which the adverb takes wide scope since adverb
occurrence in front of the gap can be avoided by pre-auxiliary positioning.
Phenomena of adverb distribution in French c’est...que Cleft and VP
Fronting constructions support the Optimality Theoretic approach and the
hypothesis that AdjCon is differentiated according to adverb type.

Notes

* I thank Gisbert Fanselow and Artemis Alexiadou for helpful discussions and suggestions;
any errors in this paper are my responsibility. This research was financed by the German
Research Foundation (DFG), Graduiertenkolleg “275: Ökonomie und Komplexität in der
Sprache”, project 5220 8303, University of Potsdam.
. Other types of adverbs, such as discourse-oriented and evaluative ones, show the same
distributional pattern. However, they are not examined here since they may neither take
narrow scope nor occur in the scope of questions for semantic reasons (see Ernst 2002).
. Note that epistemic adverbs cannot take narrow scope and only few of them (ones that
carry some implication) may occur in questions (see Bellert 1977; Ernst 2002).
. Since the adverb in (15)–(25) is the final overt element in the clause, its position relative
to the gap cannot be deduced from surface order. However, there is evidence that the adverb
does precede the gap in these sentences. First, these adverbs cannot occur postverbally:
 Eva Engels

(i) *Has he [gotten along well with Fred] ever?

(ii) *... but he hasn’t [gotten along well with Fred] always.

Secondly, auxiliary contraction, requiring a stressed syllable to the left of the gap, is possible:

(iii) Where’s he usually (when Mary is in London)?

(iv) (John hasn’t gotten along with Grandfather lately.) – Who’s ever?

In addition, note that it is the pre-gap placement of a subject-oriented or evidential adverb

that is responsible for the ungrammaticality of (26)–(30). According to Ernst (2000, 2002),
negation may take scope over evidential adverbs (v), modal verbs must take scope over
subject-oriented adverbs (vi) and (vii), and both types of adverbs may occur in questions
(viii) and (ix):

(v) They haven’t clearly finished all their work yet. (Ernst 2002: 104)

(vi) The protagonist in your novel must cleverly solve the mystery by herself. (Ernst 2000: 82)

(vii) *The protagonist in your novel cleverly must solve the mystery by herself.

(viii) Has she obviously finished her work? (Ernst 2002: 104)

(ix) Did Frank easily beat all his opponents? (Travis 1988: 302)

. Baker (1981) only discusses VP Ellipsis constructions. However, his analysis may be ex-
tended to wh-Movement and Fronting constructions by assuming that the gaps left behind
by these operations also block Stress Reduction.
. These papers only discuss VP Ellipsis constructions. Under the assumption that traces
have to be head-governed, the corresponding pattern in VP Fronting constructions can
be accounted for. Additionally, considering contracted auxiliaries not to be proper gover-
nors, the ungrammaticality of their occurrence in front of a gap follows. Note, however, that
contracted auxiliaries may appear in gap constructions as long as some appropriate overt
element precedes the gap.
. Generalized Government Transparency Corollary:

An X0 which is coindexed with and governs an empty head governs everything that
head would govern. (Lobeck 1995: 146)

. Scope Principle:

A scopal element A has scope over a scopal element B in case A c-commands a member
of the chain containing B. (Ernst 1992: 139)

. Engels (2002) differs from Ernst (2002) in accounting for the reading of a post-auxiliary
adverb: Ernst assumes that the lexical requirements have to be met at S-structure, permitting
overt head movement to be reconstructed for interpretation. Consequently, independent of
the scope of the adverb, the same structure underlies the sequence finite auxiliary – adverb:
 Optimizing adverb placement in gap constructions 

(i) John [I’ must i [AuxP frequently [AuxP t i [VP run]]]]

[E” MUSTdeont [E’ FREQUENTLY [E R(e) & Agt(e,j)]]]
[E” FREQUENTLY [E’ MUSTdeont [E R(e) & Agt(e,j)]]]
By contrast, in Engels (2002) I propose that Modifier restricts GEN insofar as an adverb
must be merged as sister to its semantic argument. Hence, the difference in the reading of a
post-auxiliary adverb corresponds to a difference in the structural position of the adverb:

(ii) John [I’ must i [AuxP t i [VP frequently [VP run]]]]

[E” MUSTdeont [E’ FREQUENTLY [E R(e) & Agt(e,j)]]]

(iii) John [I’ must i [AuxP frequently [AuxP t i [VP run]]]]

[E” FREQUENTLY [E’ MUSTdeont [E R(e) & Agt(e,j)]]]
. The order variability may also be accounted for by considering movement of the finite
auxiliary to be optional. Adjoined to AuxP, the adverb precedes or follows the finite auxiliary
depending on whether or not auxiliary movement takes place (see e.g. Baker 1981). For
reasons laid out in Engels (2002), this view is not pursued here.
. Like Vikner (2001), I take this constraint to be satisfied if a head position is filled by an
overt element or a trace.
. Note that in case negation takes scope over an adverb, the adverb will follow the finite
auxiliary in an optimal output even if it is merged above the auxiliary. Because of ObHd
>> Stay, the finite auxiliary has to move across the adverb which has to occur in its base
position.

Tableau (i). Strict order finite auxiliary – narrow scope adverb

ObHd Stay

F (a) [IP John hasi [NegP not [AuxP frequently [AuxP ti [VP seen the *
doctor]]]]]
(b) [IP John e [NegP not [AuxP frequently [AuxP has [VP seen the doc- *!
tor]]]]]
(c) [IP John [I’ frequentlyj [I’ hasi [NegP not [AuxP tj [AuxP ti [VP seen **!
the doctor]]]]]]]

. Operators include wh-phrases and topicalized constituents; any position c-commanding
the IP counts as a scope position (Vikner 2001: 228).
. Purity of Extended Projection (PureEP)

No adjunction takes place to the highest node in a subordinate extended projec-

tion; and no movement takes place into the highest head of a subordinate extended
projection. (Grimshaw 1997: 374)

. To account for which VP can be elided and how an elliptic VP is interpreted, an addi-
tional condition is necessary:
 Eva Engels

Focus condition on VP-ellipsis:

A VP α can be deleted only if α is e-given. (Merchant 2001: 26)

e-givenness:

An expression E counts as e-given iff E has a salient antecedent A and, modulo ∃-type
shifting,
(i) A entails F-clo(E), and
(ii) E entails F-clo(A). (Merchant 2001: 26)

F-closure:

The F-closure of α, written F-clo(α), is the result of replacing F-marked parts

of α with ∃-bound variables of the appropriate type (modulo ∃-type shifting).
(Merchant 2001: 14)

E-Givenness implies that the constituent containing the elided VP includes a focussed el-
ement that contrasts with some element in the constituent containing the antecedent VP;
thereby, it accounts for the necessity of syntactic identity of antecedent and elliptic VP as
well as for the availability of sloppy identity interpretation of pronouns (see Rooth 1992a
and b; Fiengo and May 1994; Schwarzschild 1999; Fox 2000; Johnson 2001; Merchant 2001).
. Assuming that VP Ellipsis and VP Fronting may only affect a constituent, a stranded
adjunct in sentence-final position must be right-adjoined. Since AdjCon also pertains to
stranded adjuncts, the pattern in (i) is expected:

(i) Mary has never met John in London, but

a. she always has __ in New York.
b. *she has always __ in New York.

The stranded adjunct has to adjoin above the finite auxiliary in order to satisfy AdjCon.
Hence, AuxP does not dominate any overt element and I’-adjunction of the medial adverb
(ia) is preferred over AuxP-adjunction (ib).

Tableau (ii). I’-adjunction of stranded adjuncts

VP-E OpSc AdjCon

F (a) [IP [VP met ...top ]j [IP she [I’ always [I’ [I’ hasi
[AuxP ti tj ]] in NY]]]]
(b) [IP [VP met ...top ]j [IP she [I’ always [I’ hasi *!
[AuxP [AuxP ti tj ] in NY]]]]]
(c) [IP [VP met ...top]j [IP she hasi [AuxP always *!
[AuxP [AuxP ti tj ] in NY]]]]
(d) [IP she hasi [AuxP always [AuxP [AuxP ti [VP *!
met ...top ]] in NY]]]
 Optimizing adverb placement in gap constructions 

. Correspondingly, a constraint PhRehead may be assumed which, by its dominance over
all AdjCon-subconstraints, guarantees that the highest member of the auxiliary chain is
phonetically realized in the output. Hence, placement of a wide scope adverb in front of a
gap in an inversion construction cannot be avoided by pronouncing the finite auxiliary in
Aux or I.
. VP Ellipsis and Fronting resulting in placement of a subject-oriented or evidential ad-
verb in front of a gap cannot be ruled out directly by prohibiting movement in these
contexts; such an analysis would require that the respective AdjCon-subconstraints out-
rank VP-E and OpSc, contrary to what has been established here: VP-E and OpSc have to
outrank Stay in order to allow for Fronting and VP Ellipsis. Stay, in turn, must outrank
AdjCon to block adverb movement to a higher position. Consequently, VP-E and OpSc
dominate AdjCon. Hence, the existence of PhRe and the analysis of (blocking of) phonetic
deletion in the base position of a moved constituent is called for.
. Base generation of a wide scope adverb as an IP-adjunct cannot be excluded on semantic
grounds. Yet, in contrast to non-inversion constructions, an adverb may not precede the
subject in subject-auxiliary-inversion constructions:

(i) a. Probably John has read the book.

b. *Has really John read the book?

Hence, base-adjunction of the adverb to IP in (ib) has to be ruled out by a constraint that
cannot be discussed here. But note that this constraint must outrank AdjCon: an adverb
may thus not escape its placement in front of a gap by pre-subject positioning.
. *LxMv >> ObHd >> Stay in English accounts for the fact that finite auxiliaries, but
not finite lexical verbs, move to I, while the ranking ObHd >> *LxMv, Stay in French
guarantees that V-to-I-movement takes place independent of whether the finite verb is an
auxiliary or a lexical verb.

*LxMV: A lexical head cannot move. (Grimshaw 1997: 374)

References

Baker, C. L. (1971). “Stress level and auxiliary behavior in English.” Linguistic Inquiry, 2,
167–181.
Baker, C. L. (1981). “Auxiliary-Adverb word order.” Linguistic Inquiry, 12, 309–315.
Bellert, I. (1977). “On semantic and distributional properties of sentential adverbs.”
Linguistic Inquiry, 8, 337–351.
Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press.
Engels, E. (2002). “Preverbal adverb placement: An optimality theoretic account.” Ms.,
University of Potsdam.
Ernst, T. (1983). “More on adverbs and stressed auxiliaries.” Linguistic Inquiry, 14, 542–549.
Ernst, T. (1992). “The phrase structure of English negation.” The Linguistic Review, 9,
109–144.
 Eva Engels

Ernst, T. (1998). “The scopal basis of adverb licensing.” In P. Tamanji & K. Kusumoto
(Eds.), Proceedings of the 28th Meeting of the North East Linguistic Society (pp. 127–142).
Amherst, MA: GLSA University of Massachusetts.
Ernst, T. (2000). “Semantic features and the distribution of adverbs.” ZAS Papers in
Linguistics, 17, 79–97.
Ernst, T. (2002). The Syntax of Adjuncts [Cambridge Studies in Linguistics 96]. Cambridge:
Cambridge University Press.
Fiengo, R., & May, R. (1994). Indices and Identity [Linguistic Inquiry Monographs 24].
Cambridge, MA: MIT Press
Fox, D. (2000). Economy and Semantic Interpretation [Linguistic Inquiry Monograph 35].
Cambridge, MA: MIT Press.
Grimshaw, J. (1997). “Projection, heads, and optimality.” Linguistic Inquiry, 28, 373–422.
Groat, E., & O’Neil, J. (1996). “Spell-Out at the LF interface.” In W. Abraham, S. D.
Epstein, H. Thráinsson, & C. J.-W. Zwart (Eds.), Minimal Ideas: Syntactic Studies in
the Minimalist Framework (pp. 113–139). Amsterdam and Philadelphia: Benjamins.
Hankamer, J., & Sag, I. (1976). “Deep and surface anaphora.” Linguistic Inquiry, 7, 391–426.
Johnson, K. (2001). “What VP Ellipsis can do, and what it can’t, but not why.” In M. Baltin
& C. Collins (Eds.), The Handbook of Contemporary Syntactic Theory (pp. 439–479).
Oxford: Blackwell.
Kim, J.-B., & Sag, I. (1996). “The parametric variation of English and French negation.” In
J. Camacho, L. Choueiri, & M. Watanabe (Eds.), Proceedings of the 14th Annual Meeting
of the West Coast Conference on Formal Linguistics (pp. 303–317). Stanford: CSLI.
Kim, J.-B., & Sag, I. (2002). “Negation without head movement.” Natural Language and
Linguistic Theory, 20, 339–412.
Lasnik, H. (1995). “Verbal morphology: Syntactic Structures meets the Minimalist
Program.” In P. Kempchinsky & H. Campos (Eds.), Evolution and Revolution in
Linguistic Theory: Essays in Honor of Carlos Otero (pp. 251–275). Washington D.C.:
Georgetown University Press.
Lobeck, A. (1987). Syntactic Constraints on VP Ellipsis. Bloomington, IN: Indiana University
Linguistics Club.
Lobeck, A. (1995). Ellipsis: Functional Heads, Licensing, and Identification. New York: Oxford
University Press.
Merchant, J. (2001). The Syntax of Silence: Sluicing, Islands, and the Theory of Ellipsis [Oxford
Studies in Theoretical Linguistics 1]. New York: Oxford University Press.
Müller, G. (1997). “Partial Wh-Movement and Optimality Theory.” The Linguistic Review,
14, 249–306.
Nunes, J. (2001). “Sideward movement.” Linguistic Inquiry, 32, 303–344.
Pesetsky, D. (1998). “Some optimality principles of sentence pronunciation.” In P. Barbosa,
D. Fox, P. Hagstrom, M. McGinnis, & D. Pesetsky (Eds.), Is the Best Good Enough?
(pp. 337–383). Cambridge, MA: MIT Press.
Potsdam, E. (1997). “NegP and subjunctive complements in English.” Linguistic Inquiry, 28,
533–541.
Rooth, M. (1992a). “Ellipsis redundancy and reduction redundancy.” In S. Berman &
A. Hestvik (Eds.), Proceedings of the Stuttgart Ellipsis Workshop. Stuttgart: University
of Stuttgart.
 Optimizing adverb placement in gap constructions 

Rooth, M. (1992b). “A theory of focus interpretation.” Natural Language Semantics, 1, 75–

116.
Sag, I. (1978). “Floated quantifiers, adverbs, and extraction sites.” Linguistic Inquiry, 9, 146–
150.
Sag, I. (1980). “A further note on floated quantifiers, adverbs, and extraction sites.” Linguistic
Inquiry, 11, 255–257.
Sag, I., & Fodor, J. (1995). “Extraction without traces.” In R. Aranovich, W. Byrne, S. Preuss,
& M. Senturia (Eds.), Proceedings of the 13th Annual Meeting of the West Coast
Conference on Formal Linguistics (pp. 365–384). Stanford: CSLI.
Schwarzschild, R. (1999). “GIVENness, AvoidF and other constraints on the placement of
accent.” Natural Language Semantics, 7, 141–177.
Travis, L. (1988). “The syntax of adverbs.” In McGill Working Papers in Linguistics. Special
Issue on Comparative Germanic Syntax (pp. 280–310).
Vikner, S. (2001). “Verb movement variation in Germanic and Optimality Theory.”
Habilitation Thesis, University of Stuttgart.
Wilder, C. (1997). “English finite auxiliaries in syntax and phonology.” In J. Black &
V. Motapanyane (Eds.), Clitics, Pronouns and Movement [Current Issues in Linguistic
Theory 140] (pp. 321–362). Amsterdam and Philadelphia: Benjamins.
Zagona, K. (1988). Verb Phrase Syntax: A Parametric Study of English and Spanish [Studies
in Natural Language and Linguistic Theory 13]. Dordrecht: Kluwer.
 Domain adverbs and the syntax of adjuncts*

Thomas Ernst
University of Massachusetts at Amherst and Marlboro College

This paper proposes a semantic analysis of domain adverbs, such as

financially or physically, that improves on previous analyses, and provides
evidence for the view that adjunct distribution is based primarily on
semantics, i.e. that adjuncts are largely free to adjoin anywhere in a clause as
long as they receive a well-formed interpretation. Domain adverbs indicate
that some contextually salient entity within their c-command domain is “in”
the relevant domain; this explains their relatively free distribution with
respect to other adverbs. Sentence-initial “topic readings” provide important
evidence for a semantically-based theory: interpreting them by means of a
general topic rule is superior to positing a separate “homonym” for each
adverb’s topic reading, as would be required on many syntax-based theories.

. Introduction

Domain adverbs designate real-world domains such as politics, economics,

sports, and so on, and are fairly common in speech and writing. (1a–e) provide
illustrative examples:
(1) a. Linguistically, these examples are interesting.
b. His statements have been misrepresenting our position economically.
c. At that point the climb physically had become more difficult.
d. He is a midget politically.
e. Chemically, why would this pose a problem?

They have several properties that may be important in advancing our knowl-
edge of how adjuncts work. First, they often function to restrict our under-
standing of some general predicate, like be interesting in (1a), to a specific
domain. This is not always necessary, of course, since These examples are in-
teresting might be uttered in the midst of a syntax class where the domain (i.e.,
syntax, or at least linguistics) is clear from context. This is parallel to the well-
 Thomas Ernst

known cases of contextual domain-restriction – in the formal-semantics sense

of domain – such as in Everyone left, where the set of people involved is of course
not every person in the world, but every person, say, in our bowling party.
Second, domain adverbs have an unusually wide syntactic distribution
(Ernst 1984, 1985). (1c), for example, has physically between the subject and
finite auxiliary, but the adverb can go anywhere with no difference in meaning,
as (2) shows (deleting the initial PP in (1c) for clarity):
(2) a. Physically, the climb had become more difficult.
b. The climb physically had become more difficult.
c. The climb had physically become more difficult.
d. The climb had become physically more difficult.
e. The climb had become more difficult physically.

This contrasts with many other adverbs like probably or badly, whose distribu-
tion is far more restricted; compare (3a, b) to (4a, b):1
(3) a. (Probably,) these examples are (probably) interesting (*probably).
b. (*Badly,) his statements (*badly) have (*badly) been (badly) misrepre-
senting our position (badly).
(4) a. (Linguistically,) these examples are (linguistically) interesting (linguis-
tically).
b. (Economically,) his statements (?economically) have (economically)
been (economically) misrepresenting our position (economically).

Third, domain adverbs are one type of adverb that has a “framing” function in
sentence-initial position, as in (1a) (and including, notably, the position to the
left of a fronted wh-phrase, as in (1e)), where it sets the scene for the rest of the
sentence, restricting the addressee’s attention to the domain in question. Note,
in particular, that sentence-initial domain adverbs in their framing function
not only restrict material below them as in (1a–e), but may also have a topic
reading. On this reading they serve to further specify a topic given previously,
either overtly or in context; the rest of the sentence then provides a comment
for this more specified topic. Observe (5):
(5) – What have they done in their last two years in office?
– Well, economically, they have passed new tax legislation; politically, they
have raised far more money for the party than was expected.

In this example, the question establishes the topic (things they have done in
their two years in office), and the adverbs in the response divide up this topic
 Domain adverbs and the syntax of adjuncts 

into economic and political accomplishments, each with one example given in
the comment.
In this paper I use these properties of domain adverbs to sharpen our the-
ory of adverbial syntax and semantics. The main goals are (i) to provide a better
basic semantic analysis for domain adverbs than that of Ernst (2002); and (ii)
in doing so, to provide further evidence for the semantically based conception
of adjunct syntax. In particular, the analysis supports the view that, for the
most part, adjuncts are free to adjoin anywhere in a clause that permits a well-
formed semantic representation, and that, as a part of this, the two readings
represented by (1) and (5) do not need separate lexical entries, but instead are
derived from a unified lexical entry plus rules of semantic composition keyed
to points of syntactic structure.
Before getting started, we must first dispose of two genuinely different
readings of domain adverbs, illustrated in (6)–(7):
(6) a. His knee will have to be surgically replaced.
b. The aliens expressed themselves telepathically.
(7) a. She always reacts very theatrically / dramatically.
b. They have managed to live very economically.

(6) shows the means-domain reading, where the adverb designates the means
by which an action is taken.2 (7) gives manner readings. These can be iden-
tified easily, since they can be modified by degree adverbs, while regular (and
means-) domain adverbs cannot; cf. (8), with regular readings, where this is
impossible:
(8) a. These themes are parallel (*very) theatrically / dramatically.
b. This is a significant result (*very) economically.

I will refer to the domain readings in (1) and (5), as opposed to the means-
domain and manner readings in (6) to (7), as pure domain readings. It should
also be noted that a range of PPs and longer phrases can also represent do-
main semantics, as (9) illustrates, though we will restrict our attention to
adverbs here.
(9) a. From a financial point of view, things could be a lot worse.
b. At that stage, England was of little consequence in global terms.
c. “By 1951, Elmer’s Glue-All was ready-mixed and packaged in squeeze
plastic bottles. Little has changed, Glue-All-wise, over the years.”
(Philadelphia Inquirer, 8/23/97:A4)
d. They were quite impressed on the scientific level.
 Thomas Ernst

e. Psychologically speaking, there has been little progress.

This paper is organized as follows. In Section 2, I provide a semantic analysis

of the regular and topic readings of regular domain adverbs. This is followed
in Section 3 by a brief discussion of two approaches to adverbial licensing, one
of them more semantically based and the other more syntactically based, and
then it is shown, in Section 4, how domain adverbs provide evidence for the
former. I close with a brief summary and conclusion.

. The semantics of domain adverbs

. Regular readings

Let us first examine the regular readings of the domain adverbs in (1a–e), given
again here:
(1) a. Linguistically, these examples are interesting.
b. His statements have been misrepresenting our position economically.
c. At that point the climb physically had become more difficult.
d. He is a midget politically.
e. Chemically, why would this pose a problem?

In (1a), the predicate be interesting is relatively unspecified (vague); something

can be interesting from many different points of view: linguistically, socially,
financially, demographically, and so on. The domain adverb serves to narrow
down the wide range of possible domains. Generally, it seems possible to say
that the predicate is “interpreted with respect to” or is “relativized to” some
domain; alternatively, that it denotes an eventuality that is “in” a particular do-
main. If we try to be more precise about exactly what is understood as “in”
the linguistic domain for (1a), we might say that the implications of the ex-
amples are linguistic (we become interested in something to the extent that it
has implications for some domain that we care about). (1c) is similar in that
be difficult is a general predicate applicable to any domain; for this predicate,
what is “in” the domain is either the implied difficulties, or perhaps criteria for
being easy or difficult. It is perhaps easier to point to an entity in a domain for
(1b) and (1e): in the former, his statements misrepresent some aspect of our
position (the parts of it pertaining to economics), and in the latter we can say
the problem itself is characterized as in the chemical domain.
In all four of these sentences the adverbs narrow the domain of some entity
within a relatively unspecified predicate (i.e. one applicable to many domains),
 Domain adverbs and the syntax of adjuncts 

and thus they can be termed specify cases (Ernst 1984). (1d) is different, how-
ever, since midget is normally taken as belonging to the physical domain. This
sentence is thus normally interpreted metaphorically, with the domain shifted
from the physical to the political domain, in what we may term a shift reading
(Ernst 1984). I will not attempt to characterize this reading in further detail, ex-
cept to say that I adopt the common view of metaphor that it can be triggered
by a clash of domains and involves trying to find similarities between structures
(schemas, images) in the two domains (see Ortony 1979 and references cited
there for discussions of metaphor). (10) provides some more examples:3
(10) a. Financially, he’s an octopus, with tentacles everywhere.
b. After their article was so well received, they reinvested the profit profes-
sionally by writing three follow-up articles.
c. The party was decapitated politically in the wake of their longtime
leader’s death.

Ernst (2002) proposed that regular domain readings be analyzed in terms of a

restriction on a universally available, covert predicate UNDER associated with
any predicate, and situated at the lowest level of the representation, the basic
event. For (1a), for example, the semantic representation would be as shown in
(11):4
(11) CR(c*, linguistic) [I(e) & Th(e, examples) & UNDER(e, c*)]

The intent of this representation was that the variable c* serve as a contextual
restriction on the predicate, imposing a further condition ‘UNDER(e, c*)’ on
the eventuality. The domain adverb linguistically would represent an operator
‘CR(c*, linguistic)’, binding c* below; the set of eventualities denoted by I (=‘be
interesting’) would thus be restricted to the subset of eventualities of being
interesting in the linguistic domain. The same representation might obtain in
a context where linguistics is being discussed, so that ‘CR(c*, linguistic)’ is not
overtly expressed in a sentence; or, other contextual restrictions might hold,
restricting the set of eventualities in a similar way.
One reason for having this type of representation was to capture the fact
that domain adverbs of this sort seemed to have the same semantic effect re-
gardless of their position in the sentence (see (2) and (4)). Positing a semantic
representation like (11) predicted this pattern, because what was important for
semantically licensing the adverb was not its position in the sentence, but the
position of UNDER – always down in the basic event – so that the operator rep-
resented by the adverb could bind the variable c* regardless of its hierarchical
position. In essence, then, the adverb acted as if it had low scope even though it
 Thomas Ernst

could appear anywhere in a clause. This contrasted with adverbs like probably
and badly (see (3)), which take scope determined strictly by their position in
a clause, ultimately restricting the adverb’s distribution. On a theory where an
adverb’s syntactic distribution is to be predicted as much as possible from its
semantic representation, this is a desirable result.
However, there is reason to reject this analysis: some sentences show that
domain adverbs can modify items higher in the tree than just the predicate and
its arguments. Observe (12):
(12) a. Philosophically, Bill did not clearly start the car (himself ).
b. Statistically, the bus probably won’t break down.

Imagine, for (12a), a debate in a philosophy class about immediate causation,

with an example where Bill turns the key in his car’s ignition, which starts a flow
of electricity to the starter motor, which in turn engages the various systems of
the engine. From a philosophical point of view, one can argue about whether
Bill actually starts the car, and so from this point of view things are not clear.
In this context, philosophically modifies (not) clearly. That this is so can be seen
by comparing this sentence with (13a), which is unacceptable under normal
assumptions (and avoiding the manner reading where Bill is, say, philosophical
about doing a task with dreaded but unavoidable consequences).
(13) a. *Bill did not clearly start the car philosophically.
b. ??The bus probably won’t break down statistically.

(We must also avoid a prosodic break before the adverb, i.e. comma into-
nation.) The same point can be made with (12b), given a scenario where a
nervous passenger is reassured before a bus trip by someone who knows that
(say) only 3% of this type of bus on this route have broken down in the last
20 years. However, if the speaker knows that this particular bus had a crack in
its driveshaft and that the bus company has a reputation for shoddy repairs, it
might be that, other than in the statistical domain, the bus probably will break
down. So, again, the domain adverb appears to modify probably (not); one can
use the paraphrase ‘It is a statistical probability that...’.5 The oddness of (13b)
confirms the point, in a way parallel to the previous example: in sentence-
final position the adverb must modify break down, which makes little sense
in context; therefore in (12b) it must modify something higher.
These cases show that a representation along the lines of (11) cannot be
right, because it does not allow for the adverb to modify anything above the
basic event. Instead, the adverb must be able to take the scope indicated by its
surface position in the sentence. Thus we need something more along the lines
 Domain adverbs and the syntax of adjuncts 

of (14a) for (1a) and (14b) for (12b) (ignoring tense), where a formula ‘IN-X-
DOM [...]’ is to be interpreted as ‘some contextually salient entity within [...]
is in the X domain’:6
(14) a. IN-LING-DOM [I(e) & Th(e, examples)]
b. IN-STAT-DOM [PROB [ ∼ [B(e) & Th(e,bus)]]]

(14a) indicates that these examples are interesting and this holds in the lin-
guistics domain; for (14b), it is probable that the bus will not break down and
something about this holds in the statistical domain. For (1a) and simple sen-
tences like it, this new formulation changes nothing, since (11) and (14a) are
equivalent: the eventuality e (examples being interesting), or potentially some
part of it, is taken as being in the linguistic domain. For a more complex sen-
tence like (12b) the domain adverb is free to home in on some higher element
in the sentence, here the element PROB(ABLY) in (14b), so that we are es-
sentially talking about statistical probabilities of the basic event occurring. We
account for the differences between (12a–b) and (13a–b) by saying that only
in the former does the domain adverb c-command, and thus take scope over,
the epistemic adverbs and negation. Therefore, in (13a-b) it is forced to mod-
ify the basic event (represented by the main verb and its argument), which is
anomalous in these particular cases.
Treating the choice of entity actually “in” the domain as a pragmatic one
is supported by the following examples, which demonstrate the wide range of
these focusing possibilities.
(15) a. What is going on in this story conceptually?
b. Prices for new homes have climbed nationally.
c. These sentences say nothing syntactically about coreference.
d. They’ve accomplished a lot artistically.
e. Everyone knew that Terri came from a family that had been targeted
politically. (Inquirer Magazine 6/27/99 p. 12)
f. They acted in concert politically.
g. These values once divided people philosophically.

My sense of the entity that counts as “in” the indicated domain is given in (16),
for each sentence in (15):
(16) a. important aspects of theme or plot
b. rise in prices
c. factors used in determining coreference
d. accomplishment-events done by them
e. reasons for/methods of ‘aiming at’ or ‘focusing on’ the family
 Thomas Ernst

f. their actions
g. division between types of (people’s) thinking

If the estimations in (16a–g) are close to the mark (or at least a majority of them
are), it can be seen that the eventuality represented by the verb can sometimes
be taken as “in” the domain, as for (16d) or (16f–g), but in other cases it is some
entity represented by the subject (prices in (16b)), implied reasons for or con-
tributing factors in the event ((16c), (16e)), or aspects of some salient object
(the story in (16a), part of an adjunct PP). Along with the examples in (12),
it seems that we should leave the semantics mostly open, as long as we specify
reference to a contextually salient entity associated with the eventuality.7
There does seem to be at least one restriction: the entity described as X
must be characterized as such in terms of the domain it is in, and it must be
possible, in principle, for it to be in some other domain. Examine (17):
(17) a. Politically, they have evolved.
b. *Politically, they have voted.
c. *Politically, they are workers.

In (17a), the event described by the sentence counts as an evolving according to

political criteria, and there is a possible contrast, as someone may evolve either
politically or morally. In (17b), by contrast, while voting is defined as a political
act, this is the only way to define it (at least in the apparent absence of an easy
metaphorical interpretation). Since there is no other domain that can define
voting, this sentence is unacceptable. (While vote electronically is possible, this
involves a means-domain reading where electronically does not supply criteria
for an event being one of voting.) As for (17c), note that workers cannot be
taken in the sense of (18):
(18) They are political workers.

Workers is apparently defined in an economic sense, and in (18) the adjective

political conveys the idea that these people do political work. However, they are
not characterized as workers by political criteria; only after defined indepen-
dently are they described as political. This reading is impossible with the adverb
in (17c), so the sentence is unacceptable. Consider the further contrast in (19):
(19) a. Socially, she considered it a setback.
b. *Socially, she described the setback.

(19a) could be used where a woman trying to establish herself in a new town is
falsely accused of some wrongdoing; a setback is “in” the social domain in that
 Domain adverbs and the syntax of adjuncts 

this domain provides the criteria for taking the event as a setback. In (19b),
however, the use of the indicates that the setback has already been recognized
as such by the speaker, according to some independent criteria. Thus socially in
this case cannot be part of what defines the event as a setback. Again, parallel
to (17) to (18), it is possible to use an adjective for this meaning (She described
her social setback), but not the adverb.
To summarize, regular readings of domain adverbs indicate that some con-
textually salient entity in the adverb’s c-command domain is “in” the relevant
domain, such that the domain supplies criteria for taking the entity as de-
scribed, and there is an implicit contrast with other domains it could be in.
Now we turn to the topic readings.

. Topic readings

An example of topic readings of sentence-initial domain adverbs was given

above in (5) (repeated here), and (20) to (22) provide three further examples.
In each of these, the first part sets up the topic, and the second part features
one or more domain adverbs restricting that topic.
(5) – What have they done in their last two years in office?
– Well, economically, they have passed new tax legislation; politically, they
have raised far more money for the party than was expected.
(20) – “The Principle of Compositionality seems to us to be supported by con-
siderations much like those which reveal the existence of phrase structure
in syntax.
– Syntactically, one assigns strings like most dentists, the dogcatcher, and
John’s teacher to a common grammatical category [NP] because of the fact
that... Semantically, it appears that the meanings of (1–1) through (1–3) de-
pend in a regular way on the noun phrases the sentences contain.”
(Dowty 1979)
(21) – “Unlike ‘Good Morning Vietnam’ [...] the Afghan program is by Afghans,
for Afghans, about Afghan issues.
– Editorially, the show has been helped by advisers with experience work-
ing for the British Broadcasting Corporation and the Voice of America, ...”
(N. Y. Times website, 9/4/02)
(22) The governor’s education program is a sham: politically, it gives her cover, but
educationally it is a disaster.
 Thomas Ernst

These examples differ from the regular readings seen above in (1) and (15) pri-
marily in that the domain adverb does not restrict the predicate or anything
else within the sentence. Rather, it seems to restrict the topic, whether it be a
purely contextual topic, a topic provided by a question, or an overt topic in the
same sentence. (5) is a case of a topic provided by a question; the topic here
is clearly ‘things they’ve done in two years’, with the answer providing exam-
ples of things in this set. The adverb economically thus further restricts the set
of events mentioned by the first conjunct, to things in the economic domain
that they’ve done in two years, and politically likewise restricts the second con-
junct to considering only things in the political domain that they’ve done in
two years. (20) works in a similar way: the first sentence sets up a topic, i.e.
considerations that support the existence of certain theoretical constructs. The
second part of (20) goes on to consider first the syntactic domain, and how
evidence there supports the existence of phrase structure, and then the seman-
tic domain, and how evidence there supports the existence of the Principle of
Compositionality. Again, the domain adverbs carve up the entities given in the
topic into two subsets, with the comment filling in more specific information
about these subsets.
We may take topics of this sort to represent a common ground (CG), made
up of a set of pairs of worlds and individuals; individuals in this set may be
either objects, eventualities, or propositions (see Portner & Yabushita 1998 for
one formal treatment of this sort, including discussion of the use of worlds).
This CG acts as the universe for individuals figuring in the comment that fol-
lows, like a domain for quantification, and it can be added to in a dynamic
semantics. The meaning of (5) might thus be put informally in (23a), with the
domain adverb economically restricting the CG as indicated in (23b), according
to the rule given in (24):
(23) a. CG includes [things they’ve done in the last 2 years]
b. CG includes [things they’ve done in the last 2 years that are within the
economic domain]
(24) Rule for topic readings:
Given a topic T in the common ground CG, an adverb denoting domain
X restricts T: IN-X-DOM (T).

Taking the topic in (5) to be a set of events (omitting the expression of paired
worlds, to keep the exposition simple), the first conjunct of the answer in (5)
can be represented in (25b), with (25a) provided as the sentence’s representa-
tion without economically:
 Domain adverbs and the syntax of adjuncts 

(25) a. [CG {e|D(e) & Agt(e, they) & IN (e, lty)}] [P(e) & Agt(e, they) &
Th(e, ntl)]
b. [CG {e|D(e) & Agt(e, they) & IN (e, lty) & IN-ECON-DOM (e)}] [P(e)
& Agt(e, they) & Th(e, ntl)]
(D = do, P = pass, lty = the last two years, etc.)

(25b) says that the topic CG is the set of events done by them in the last two
years that count as being in the economic domain, and that within this set is
an event of them passing new tax legislation.
The type of restriction represented by (24) is a general one, relevant for
more than domain modification. As (26a–c) illustrate, respectively, the same
sort of topic construction can be made with locative or temporal PPs, and NPs
denoting objects. The initial parenthesized PP can be taken either as overt, or
as given by the preceding context:
(26) a. (In the US,) Within most regions that were devastated by fire, aid from
the government can be substantial.
b. (Last year,) In May, we found out that we’d gotten the grant.
c. (For woodwind players,) (At least) as far as bassoonists and oboists are
concerned, the reed is a huge focus of one’s attention.

(Presumably, times also count as individuals for the CG, for cases like (26b)).
Thus we suppose that there is some general interpretation rule for topics along
the lines of (27):
(27) General rule for topics:
Add an S-initial topic to the list of statements characterizing the set of
individuals in CG.

As an example of applying (27) to the non-domain cases in (26), consider the

analysis of locative PPs in Maienborn (2001). Maienborn proposes an analysis
of locative topics by which the framing and external-locative readings illus-
trated in (28) have representations along the lines of (29), where the angled
brackets enclose a comment-topic pairing, and vx is a contextually salient en-
tity within the comment, asserted to be located in the place denoted by the
topic (Hollywood):
(28) a. In Hollywood they make a lot of action-adventure films. (framing loca-
tive)
b. They made the film in Hollywood. (external locative)
(29) a. <[M(e) & Agt(e, they) & Th(e, f)], Hollywood> & LOC(vx , Holly-
wood)
 Thomas Ernst

b. [M(e) & Agt(e, they) & Th(e, f) & LOC(e, Hollywood)]

It seems possible to recast (29a) as (30), where the common ground topic is
the set of events for which some contextually salient entity is in Hollywood (xe
now standing for that entity):
(30) [CG {e|LOC (xe , Hollywood}] [M(e) & Agt(e, they) & Th(e, f)]]

(26a) could thus be represented as (31) (with r standing for most regions that
were devastated by fire):
(31) [CG {e|LOC (xe , US) & LOC(xe , r)}] [Be-S(e) & Th(e, aid-from-govt)]

Though many issues are perforce left aside here (among them certain details
of how topics are added to the CG), the adaptability of Maienborn’s analysis
suggests that something along the lines of (27) should serve as a general rule
for topics of this sort.
Note that the present treatment of topic and regular readings allows ex-
plaining the interaction of a predicate’s presupposed domain and the domain
overtly expressed by the adverb, and thus the oddness or unacceptability of cer-
tain sequences. In (32a), for example, where economically has a topic reading
as in (5), the entire event of their passing tax legislation is asserted to be in a set
of economic events.
(32) a. Economically, they have passed tax legislation.
b. *They have passed new tax legislation economically.

Recall that the entity taken as “in” a domain must be characterized in terms
of that domain, and be able to be contrasted with a characterization in terms
of some other domain. (32a) is therefore fine, since on the topic reading as in
(5), the domain adverb characterizes and allows contrasting domains for the
topic, i.e. things they have done in their two years in office. However, (32b) is
odd because the low position of the adverb requires it to have a regular reading
with scope over only the basic legislation-passing event, restricting the latter
to the economic domain. Passing legislation is defined as something political,
not economic; there does not seem to be a sense in which political passing of
legislation can be contrasted with economic passing of legislation. Thus (32b)
is unacceptable. (32a) does not have this problem because economically restricts
the topic (things they’ve done in two years), not pass new tax legislation.
We have examined both regular and topic domain readings. Both of these
are possible in sentence-initial position, as (33) illustrates:
(33) Economically, the province is very powerful.
 Domain adverbs and the syntax of adjuncts 

If (33) is part of a discussion of reasons why politicians respect the province’s

importance, it could be part of (34), with the two domain adverbs having
topic readings:
(34) Economically, the province is very powerful; politically, many of the top min-
isters in the government are from there.

(35) gives informal versions of the topic of (34) ((35a)), and the semantic
representation for its first conjunct (35b):
(35) a. [ CG Reasons to respect the province]
b. [ CG Economic reasons to respect the province] [Powerful(e) &
Th(e, province)]

A regular reading for (34) might surface in a context where we are focusing on
how important the province is. This is illustrated in (36), with a rough semantic
representation for the first conjunct in (37):
(36) Economically, the province is very powerful, but politically it has little clout.
(37) [ CG Relating-to-Importance(e) & Th(e, province)]
[IN-ECON-DOM [Powerful(e) & Th(e, province)]]

. Conclusion

We have proposed a semantic analysis of domain adverbs in which a given

adverb modifies not a covert, ubiquitous UNDER predicate within the basic
event, but instead takes the scope indicated by its surface hierarchical position,
and says that some contextually salient entity within its scope is “in” its do-
main. Data like (12) show that this is superior to the UNDER analysis of Ernst
(2002), since only it can predict the possibility of a domain adverb modifying
another adverb within its scope. The original facts are still accounted for, as
modifying the basic predicate is always one option. We have also suggested that
the topic reading be treated by allowing the restriction represented by the do-
main adverb to be incorporated into a common-ground topic. The semantic
content of the adverb is the same for the two readings; the only difference in
the two readings is the application of the topic rule to create the topic reading.
 Thomas Ernst

. Syntax

. Theories of adjunct syntax

The semantic analysis presented above is meant as part of a semantically based

approach to the syntax of adverbial adjuncts, where one tries to predict as
much as possible of a given adjunct’s distribution from its semantics and the
interaction with other elements in its clause. To this end, a leading hypothe-
sis adopted here is that adjuncts almost always are freely adjoined, not in Spec
positions (as claimed by Kayne 1994; Laenzlinger 1998; Cinque 1999, for ex-
ample), and there is no use of feature mechanisms or reference to particular
phrasal heads to license them.8 In general, following the line put forth in Jack-
endoff (1972), an adverbial is allowed in any position where an appropriate
semantic rule can interpret it, and where the resulting clause is semantically
well-formed in general.
This is not to say that there are no constraints at all on adverbial distribu-
tion, or that there is no reference at all to syntactic structure. The semantically
based approach (at least in its incarnation in Ernst 2002) claims rather that
(a) rules of composition applying to a broad range of adverb types (as op-
posed to specific adverbs) may sometimes be pegged to certain areas of the
clause, and that (b) certain aspects of an adverbial’s distribution are best cap-
tured by means of selectional restrictions, expressed in the lexicon as semantic
conditions. Let us examine these two mechanisms in turn.
An example of semantic rules operating over specific syntactic ranges is
the manner rule, which produces the manner reading of predicational adverbs
like clearly, oddly, wisely, similarly, and willingly.9 Thus in (38) to (39), with
(respectively) evaluative and agent-oriented adverbs, the first sentence gives a
clausal reading and the second sentence a manner reading:
(38) a. Appropriately, Jennie planned the meal.
b. Jennie planned the meal appropriately.
(39) a. She cleverly has been opening the boxes.
b. She has been opening the boxes cleverly.

In (38a), it is appropriate that Jennie planned the meal (maybe it was her turn)
even if the dishes she chose might have been inappropriate; in (38b) her choices
were fitting, even if it was not right for her to do the choosing. Similarly, in
(39a) cleverly has a clausal reading along the lines of ‘She was clever to open
the boxes’ (though the way she opened them might have been stupid), but a
manner reading in (39b) (‘She opened the boxes in a clever manner’, though
 Domain adverbs and the syntax of adjuncts 

she might have been unwise to open them at all). Many other adverb classes
show this pattern, as (40) to (41) illustrate:
(40) a. Bob clearly must speak to her about her role in this.
b. Bob must speak to her clearly about her role in this.
(41) a. Similarly, these two machines rarely require maintenance.
b. These two machines work similarly.

Rather than positing two homonymous lexical entries for cases like these, it is
simpler to assume a unified lexical entry for a given predicational adverb, with
a general rule to derive one of the readings. By this rule, for example, the result
for (39) is the clausal reading in (42a) for (39a), along with the manner reading
in (42b) for (39b):10
(42) a. [O(e) & Agt(e, she) & Th(e, b) & CLEVER (e, she)]
where e is mapped onto a scale of cleverness evaluated with respect to
the comparison class of all events
b. [O(e) & Agt(e, she) & Th(e, b) & CLEVER(e*, she)]
where e* is mapped onto a scale of cleverness evaluated with respect
to the comparison class of all events of opening

On this analysis, the manner reading says in essence that this event of Jennie
opening the boxes (perhaps by using her teeth) is above the norm for clever-
ness, compared to other box-opening events (say, using a hammer, tearing at it
with her hands, etc.).
The rule creating (42b) applies only to adverbs adjoined to PredP, assum-
ing the clause structure in (43), where V always moves to Pred (thus PredP
roughly corresponds to what traditionally has been called the lexical VP; cf.
Bowers 1993; Hale & Keyser 1993). ModalP and AspectP only appear when
modals and/or aspectual have or be are present in a clause, and direct objects
are assumed to be in Spec,VP.
 Thomas Ernst

(43) TP

Tense ModalP

Modal AspectP

Aspect PredP

Pred VP

Vi DP V'

ti

Adverbs can adjoin anywhere in this tree, in principle.11 When a predicational

adverb like appropriately or wisely adjoins above PredP it will take its clausal
meaning, but only if it is adjoined within PredP will it have a manner reading.
In this way we capture the fact that manner readings only occur low in a clause,
and all that is needed is the one compositional rule with a phrase structural
restriction on its application.
Now let us turn to point (b) above, the matter of certain restrictions on
adverbial distribution being expressed via semantic selection in the lexicon.
(44) a. (*Often,) Why should these phenomena (often) matter?
b. (*Pointedly,) Can he ignore them (pointedly) and still succeed?
(45) a. (Linguistically,) Why should these phenomena matter (linguistically)?
b. (Politically,) Can he ignore them and still succeed (politically)?

The ungrammaticality of (44a–b) with initial adverbs is treated in Ernst (2002)

by saying that often and pointedly must combine with syntactic constituents
representing eventualities, as they can when the adverbs are in a lower position,
as shown. In initial position, however, they run afoul of the fact that questions
feature an initial interrogative speech-act operator in Comp (or Spec,CP). Thus
the pattern shown in (44) is accounted for if the two adverbs select semantically
for events but not speech act operators; when in initial position, their require-
ments cannot be fulfilled. On the other hand, the domain adverbs shown in
(45) have no such selectional requirement, and so are fine.
Thus the semantically based approach to adverbial distribution claims that
the use of syntactic mechanisms is minimal. Primarily, distributional patterns
are accounted for by the selectional requirements of lexical items, in interaction
 Domain adverbs and the syntax of adjuncts 

with various compositional rules applying broadly across adverb classes and
sometimes restricted to certain parts of a clause.

. Topic readings of domain adverbs

The semantic rule deriving topic readings for domain adverbs is another ex-
ample of a rule restricted to some part of phrase structure. Since these readings
are found only with adverbs in sentence-initial position, it must be that the rule
applies for items adjoined to CP. (46) illustrates this, given that wh-phrases are
in Spec,CP:
(46) – What are some of the committee’s main concerns?
– Well, domestically, why is the economy still so sluggish? And internationally,
how can we maximize the effect of our foreign aid?

I assume that application of the rule is optional, since the regular domain read-
ing is also possible in this position (see (1e)). For sentence-initial occurrences
in declarative sentences, like (1a), we may assume a covert CP with an empty
head.12

. A Note on alternative orders

It was claimed in Ernst (2002) that alternative orders of a domain adverb and
some other event-internal modifier, like a measure or manner adverb, should
be uniformly grammatical in principle, since there is essentially no semantic
interaction between the two types. Pairs like (47) to (48) were provided as
support for this assertion.
(47) a. Emerging democracies evolve rather slowly politically.
b. Emerging democracies evolve politically rather slowly.
(48) a. These effects have been shown very clearly statistically.
b. These effects have been shown statistically very clearly.

(Speakers differ on whether sentences like these are perfectly fine or slightly
awkward, or sometimes require a particular intonation, but all agree that they
are acceptable.) On the analysis offered above, there are now potential semantic
interactions between domain and manner adverbs, since if a domain adverb
c-commands the manner adverb as in the (a) sentences above, it can focus the
latter. (See (49), representing the VPs in (47) to (48) containing the main verb
and two following AdvPs.)
 Thomas Ernst

(49) VP

VP AdvP

V AdvP

Thus, for example, (48a) has a reading where it is not a matter of being shown
statistically, but that it is statistically clear. (47a) shows that this reading is
not obligatory, since politically modifies evolve, not the manner adverb slowly
(presumably because it does not make pragmatic sense for something to be po-
litically slow, as opposed to being slow in some other domain). Note, however,
that (48b) can have this same reading, even though very clearly would appear to
c-command the domain adverb. This reading can be explained by the fact that
there is an alternative syntactic structure, where statistically is part of a single
AdvP statistically very clearly, within which the statistically again modifes the
manner adverb. Thus the new analysis given here makes the same correct pre-
diction as did Ernst (2002) about the grammaticality of both orderings and
both readings. But in strictly semantic terms, it is superior in allowing for the
reading where the domain adverb specifically modifies the manner adverb.

. Theoretical implications

. An argument for the semantically based account of adverb distribution

One goal of this paper has been to provide a semantic analysis of domain ad-
verbs that improves on that of Ernst (2002); given (a) the data shown in (12)
to (13), and (b) the ability of the analysis proposed above to extend easily
to topic readings (which Ernst 2002 could not do), it would seem that this
goal has been accomplished. The second goal was to support the semantically-
based approach to the syntax of adverbs. In what follows, I will show that this
holds as well. To do so, we must first examine how this theory and the more
syntactically-based theories account for an adverb’s distribution.
As noted earlier, the semantically based theory of adverb distribution holds
that syntactic constraints on adverbs are minimal, and that instead adverbs can
be adjoined largely anywhere in a clause, with unacceptable sentences being
(for the most part) the result of semantic ill-formedness.
(50) a. Harold probably had cleverly invited Amy over for tea.
b. *Harold cleverly had probably invited Amy over for tea.
 Domain adverbs and the syntax of adjuncts 

In (50) the order of probably and cleverly must be as shown in (50a), and
not reversed as in (50b). A semantically based theory using the sort of selec-
tion mentioned above can account for this by requiring that epistemic adverbs
like probably combine with phrases representing propositions, while agent-
oriented adverbs such as cleverly require event-denoting constituents (see Ernst
1998, 2002 for discussion). On this view, (50a) is well-formed because both
adverbs’ selectional requirements are met, while in (50b) is ill-formed because
CLEVER is forced to combine with a proposition.13
(51) a. [P PROB [P CLEVER [E PERF [E I(e) & Agt(e, h) & Th (e, a) &
For(e, t)]]]]
b. [P CLEVER [P PROB [P [E PERF [E I(e) & Agt(e, h) & Th (e, a) &
For(e, t)]]]]]

Crucially, this account assumes a mechanism by which events can be freely con-
verted to propositions – as indicated by the two subscripted brackets right after
PROB in (51b) – but that the reverse is banned, so that (51b) cannot be saved by
“lowering” a proposition to an event so that the requirements of CLEVER are
satisfied. The same explanation holds for a possible base-generation analysis
of (52), so that this sentence is predicted to be ungrammatical given an addi-
tional constraint (however it is to be stated) that topicalization of one adverb
over another is blocked:14
(52) *Cleverly, Harold had probably invited Amy over for tea.

In a system based on syntactic feature licensing, the contrast in (50) can be

accounted for if a feature on some functional head licenses each adverb (pos-
sibly according to its semantic type), e.g. [+Modal] for probably and [+Agent-
Oriented] for cleverly. If the first of these must be licensed by a Modal head
and the second by some head (Aspect, Agt-OrP, etc.) that necessarily is lower
in clausal structure, then the facts in (50a–b) are predicted; (53) illustrates this
(ignoring much irrelevant detail).
 Thomas Ernst

(53) ModalP

AdvP Modal'

Modal AgtOrP

AdvP AgtOr'

AgtOr PerfP

Perf VP

...
probably ø cleverly ø have V

Note that on this account, where multiple positions for one adverb are possi-
ble with no change in meaning (see (54)) one must say either that (i) different
heads may license the same adverb, (ii) an adverb can move to various posi-
tions after being licensed by its unique licensing head, or (iii) other elements,
especially auxiliary verbs, may move around the adverb in its base position (or
some combination of (i–iii)).
(54) (Probably,) Harold (probably) would (probably) have (probably) been ar-
rested if not for Amy’s quick thinking.

Ernst (2002) argues against all of (i–iii), on the basis of adverbs like probably,
cleverly, and often, mostly on the grounds that they require very complicated
and otherwise unnecessary additions of syntactic mechanisms to UG, and that
the semantically based alternative account can get by with fewer and sim-
pler theoretical devices, needed independently. With a fairly simple semantic
analysis in place, we can now make the same sort of argument on the ba-
sis of domain adverbs. Consider the fact that domain adverbs occur in every
generally available position in a clause in English, as shown by (4):
(4) a. (Linguistically,) these examples are (linguistically) interesting (linguisti-
cally).
b. (Economically,) his statements (?economically) have (economically) been
(economically) misrepresenting our position (economically).

This follows from their semantics, which allows them to “focus” some contex-
tually salient entity within their sister (the rest of the clause below). If this entity
is the predicate, as in (1), then the semantic representation will be well-formed
 Domain adverbs and the syntax of adjuncts 

regardless of where the adverb is placed; no significant semantic interactions

with other adverbs are possible, and so there are no anomalies like that which
rules out (50b).
What are the alternatives for domain adverbs’ multiple positions on the
syntactically-based analysis, according to (i–iii)? Different heads could license
a domain adverb in each position, as in (i), but this is unnecessary if we have
a semantic analysis as outlined above, and far more complicated. As for (ii), it
is implausible to suggest that other elements can all move around a single base
position for domain adverbs, since this base position would have to be above
Spec,CP (as shown by (1e), assuming only upward movement, as is standard),
entailing in essence that every single element of a clause can relocate to puta-
tive positions above CP while preserving the same base order; also the added
structure needed to supply landing sites adds enormous complication to the
grammar. Finally, moving the adverb on option (iii) is perhaps a more plausible
solution, but it would require conditioning numerous differences in the move-
ment possibilities of different adverb classes. For example, an agent-oriented
adverb like cleverly can topicalize, but not over probably, as (52) showed. Yet
a domain adverb like musically would have to be taken as topicalizing from a
postverbal base position across probably, which is fine (see (55)).
(55) (Musically,) Jocelyn was probably improving (musically).

Also, most adverbs cannot appear before wh-phrases or inverted auxiliaries

in questions, while domain adverbs can, as noted in (44) to (45). More such
examples can be given. An adverb-movement approach must be complicated
to account for these differences in conditioning the movements, while the
semantically-based analysis handles these the facts naturally: (44a–b) are im-
possible because often and pointedly require events as their operands, while in
clause-initial position in questions they can only modify a speech act operator,
which is not an event description. Domain adverbs have no requirements of
this sort, and so are acceptable in (45).
The conclusion is that domain adverbs provide evidence for the semanti-
cally based theory of adverb distribution. On the latter, their fairly free dis-
tribution is accounted for easily, given the semantics outlined above, since
they are well-formed on their regular reading regardless of where they occur
in a sentence. Other adverbs, like probably, cleverly, often, and pointedly, have
more stringent semantic requirements which rule them out in certain posi-
tions. Most if not all of these semantic requirements, such as the event-taking
requirement on cleverly, are needed even on a syntactic account of their dis-
tribution. Thus the high degree of complication necessary for the syntactic
 Thomas Ernst

account, regardless of which of the options (i–iii) is taken, makes it inferior

to the semantic approach.

. A second argument

Among pure domain adverbs, we have seen that there are two readings, the reg-
ular reading and the topic reading, analyzed above as the result of one lexical
entry per adverb interacting with two rules of semantic composition. Signifi-
cantly, the rule for topic readings is very general, applying to at least three other
kinds of modifiers as well. This helps to provide evidence against the strict form
of the syntactic approach to adverb distribution advanced by Cinque (1999).
Cinque claims that every instance of a given adverb occurring in two po-
sitions with different meanings must represent two homonymous versions of
the adverb, licensed by two distinct functional heads. If nothing else, this re-
quires two functional projections to syntactically license pure domain adverbs,
and two lexical entries for each adverb. For example, this approach would re-
quire one head at the beginning of a clause (say, DomPT ) for topic readings,
and another one somewhat lower down for regular readings (e.g. DomPR ). The
first head would license adverbs meaning roughly ‘As far as entities in the X do-
main are concerned’, and the second one would license adverbs meaning ‘Some
contextually salient entity is in the X domain’.
Now, it is true that the semantic analysis provided above requires two rules
of composition, which do the work of the semantics represented by the two dif-
ferent DomP heads on a Cinque-style theory. But the semantically based theory
has the advantage of generality. The syntactic approach would require several
separate topic-projections, one for each type of adverb. The beginning of a
clause would have to have a series of empty functional heads to license adver-
bial topic expressions, each one keyed to a different type of adverbial meaning,
e.g. a sequence like LocPT – TempPT – DomPT – ... On the other hand, on
the semantically based analysis, invoking the general rule in (27) adds noth-
ing more to the grammar, thus making it simpler and also directly capturing a
generalization about topics. Since this same argument can be repeated for other
sets of adverbial modifiers that seem treatable by a single semantic rule, as for
the clausal/manner ambiguity discussed earlier, domain adverbs thus support
the general point that the semantically-oriented theory of adjunct distribution
is superior to the syntactic, feature-based theory.
 Domain adverbs and the syntax of adjuncts 

. Conclusion

Domain adverbs have two types of pure domain readings, a regular reading as
in (1) and a topic reading like that in (5). We have shown that the analysis of
Ernst (2002) can be improved on if regular readings are treated as taking scope
as indicated by their hierarchical position in a clause and as modifying some
contextually salient, predicational element within that scope. Moreover, topic
readings can be handled by means of a general topic interpretation rule adding
the restriction represented by the adverb to the common ground.
This improved semantic analysis still allows us to account adequately for
the distributional freedom of domain adverbs and their alternate orderings
with respect to manner (and other event-internal) adverbs; thus, one may
ask what the implications are for syntactic theory. We concluded that domain
adverbs provide two arguments for the semantically based view of adjunct syn-
tax. First, their syntax is accounted for with nothing more than (mostly) free
adjunction plus independently needed semantic rules, while feature-based the-
ories in which designated functional heads license adverbs must posit many
features, extra heads, additional structure, and/or movement rules to get the
same result. Thus the approach advocated here is far simpler. Second, the exis-
tence of a general topic rule for the topic domain reading permits a simple anal-
ysis with one (independently needed) head and semantic rule, whereas a strict
version of feature theories like that of Cinque (1999) would require at least
three empty heads, features, and semantic specifications to get the same result.
Again, the semantically based approach is simpler, and thus receives support.

Notes

* I owe thanks to Angelika Kratzer and Barbara Partee for very helpful discussion of the
material herein, and to an anonymous reviewer for pointing out several problems in an
earlier draft. Alas, all errors remain my own.
. Domain adverbs are like those of other subclasses in sometimes being degraded right after
subjects, but speakers still generally accept them if there is reason to foreground the domain.
. The main reason for treating means-domain adverbs separately is that they seem to be
impossible outside of the lower part of the clause. If the means-domain reading were simply
one contextually-determined possibility within a vague, general domain semantics, as in (2)
or (4), then all the positions in (i–ii) should be possible, contrary to fact.

(i) They most often had (*electronically) been communicating (electronically) with each
other (electronically). (on means-domain reading)
 Thomas Ernst

(ii) The team (*morphologically) was (morphologically) classifying all the specimens (mor-
phologically).

. It appears that there is a continuum between specify and shift readings, depending on the
degree to which there is already a presupposed domain for the predicate, and on whether
or not the latter can be used easily (or conventionally) for metaphor. In (i), for example,
sing is fairly well rooted in the musical (or physical, action, etc.) domain, so that (i) is hard
to interpret; (ii)–(iii) are progressively easier to interpret metaphorically, with be a giant
conventionally used in this way. But in context one might use sing metaphorically, as in (iv),
meant as roughly synonymous to (v). A predicate without a strongly presupposed domain,
such as be outstanding in (v), needs no metaphorization (a specify case), and so stands with
(iii) at one end of the continuum.

(i) *Sociologically, she sings.

(ii) ?Sociologically, she is ten feet tall.

(iii) Sociologically, she is a giant.

(iv) ?Sociologically, she just sings.

(v) Sociologically, she is outstanding.

For further discussion, see Ernst (1984).

. One line of current thinking holds that states like (be) interesting are not properly rep-
resented by an event variable in the narrow sense, but instead should either have a state
variable s in place of e in (11) and other such representations, or possibly do without any
‘Neo-Davidsonian’ variable of this general sort. Going into this issue in depth would take us
too far afield, and is not directly relevant to the point at hand, so I continue to use e for even-
tualities including states. However, two points are worth noting briefly. First, even for some
of those who deny that states need a variable at the level of the predicate, an eventuality
variable may be introduced at a higher level of clausal structure, such as Infl (see Maien-
born 2003). Second, it may be that the evidence against an eventuality variable (or at least
a distinct state variable) is not as strong as has been claimed, since much of it rests on the
purported inability of states to occur with locative or manner modification. The following
examples provide reason for doubt on this score, where (i)–(ii) show manner modification
with states, and (iii)–(iv) show locative modification with states:

(i) Yes, Jay was quiet, but he was quiet in a way that made us all very nervous.

(ii) Paul was oddly tired, without sleepiness and yet with no desire to move.

(iii) The baby was pretty quiet on the train this afternoon, don’t you think?

(iv) (Soccer coach to players before a match:) The grass is slippery on the left side of the field,
so be careful.

. I take probably, like other epistemic adverbs, as being a predicate in semantic representa-
tion, not an operator as is often assumed. See Ernst (2002, Chapter 2) for discussion.
 Domain adverbs and the syntax of adjuncts 

. I adopt here a notation where square brackets can enclose descriptions of eventualities
(including states) and propositions; I assume that these entities are introduced in the fashion
of Discourse Representation Theory (Kamp & Reyle 1993) but do not show this graphically
(i.e. the ‘universe’ line at the top of boxes in DRT representations) for the sake of clarity and
brevity; see Ernst (2002) for discussion. Thus, (14b) could be noted more precisely as (i),
where S abbreviates ‘state’ and E ‘event’:

(i) IN-STAT-DOM [S PROB [S ∼ [E B(e) & Th(e,bus)]]]

‘[E B(e) & Th(e,bus)]’ is thus an event (description) characterized by two statements about
its event-type and its theme, and ‘[S ∼ [E B(e) & Th(e,bus)]]’ is a state formed by negating
the basic event. Note that this represents a commitment to eventualities and propositions
as interconvertible entities, with possible subclassifications, if only for states as a subtype of
eventuality; it also represents a commitment to the existence of negative states (on which
see, among others, Moltmann 1991; Asher 1993: 52; Bartsch 1995: 31).
. See Maienborn (2001) for a discussion of such pragmatic inferencing with locative mod-
ifiers; I will assume that something similar is possible for domain adverbs. As for a more
precise notion of ‘in a domain’, one way of capturing the relation between eventualities and
domains is to relate it to the notion of aboutness, as discussed in Partee (1989) (though
Partee denies originating the notion): a proposition is ‘about’ an individual, property, etc.
if the truth of the proposition varies systematically with the properties of that individual
(property, etc.) in possible worlds. We could then say that an eventuality is in a domain
if the associated proposition is ‘about’ that domain: the truth of this proposition varies
systematically with properties of the domain.
. There may be isolated exceptions, e.g. for the adverb not if it is in Spec position, as
proposed in Ernst (1992).
. Predicational adverbs typically (i) come from open classes, (ii) take a proposition, fact,
or event as one of their arguments, (iii) show the clausal/manner pattern of ‘homonymous’
readings in most cases (see below for discussion), and (iv) in English, are composed of an
adjectival stem plus -ly (similarly with -ment in French, -mente in Italian and Spanish, etc.).
. See Ernst (2002, Chapter 2) for extensive discussion of the difference between clausal
and manner readings, and of the approach to their semantics based on comparison classes.
. Although this statement holds true for universal grammar, I also assume a principled
restriction on left-adjunction to VP in head-initial languages, accounting for the ‘adjacency
requirement’ between verbs and direct objects, in English and other VO languages lacking
head-movement to Tense; see Ernst (2002, Chapter 4).
. I see no great problem with the alternative of allowing the rule to apply in IP as well, as
long as the lack of topic readings after C in subordinate clauses (see (i), for example, which
allows only the regular reading in the absence of strong parenthetical intonation for econom-
ically) can be chalked up to a separate semantic effect, i.e. presumably the impossibility of
mapping contextual topics into subordinate clauses.

(i) – What have they done in their two years in office?

– ?Well, I believe that economically they have passed new tax legislation...
 Thomas Ernst

. I assume that have is represented by an event-taking aspect operator PERF, as proposed
in Kamp & Reyle (1993), de Swart (1998), among others, and that it can be generated and
interpreted below the two adverbs; again, tense is ignored here for simplicity. Also, I adopt
the VP-internal subject hypothesis, by which subjects originate in VP and are interpreted in
their base positions.
. An alternative explanation would hold that cleverly must operate on some ‘controllable’
entity, in the sense that the agent of the action must be able to do or not do the action; in this
case, it cannot occur above probably because a proposition is not something controllable. For
further discussion of this and other issues surrounding (52), see Ernst (2002).

References

Asher, N. (1993). Reference to Abstract Objects in Discourse. Dordrecht: Kluwer.

Bartsch, R. (1995). Situations, Tense, and Aspect. Berlin: Mouton de Gruyter.
Bowers, J. (1993). “The syntax of predication.” Linguistic Inquiry, 24, 591–656.
Cinque, G. (1999). Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford:
Oxford University Press.
Dowty, D. R. (1979). Word Meaning and Montague Grammar: The Semantics of Verbs and
Times in Generative Semantics and in Montague’s PTQ. Dordrecht: Reidel.
Ernst, T. (1984). Towards an Integrated Theory of Adverb Position in English. Bloomington,
IN: Indiana University Linguistics Club.
Ernst, T. (1985). “On the odd syntax of domain adverbs.” Language Research, 21 (2), 167–
177.
Ernst, T. (1992). “The phrase structure of English negation.” The Linguistic Review, 9 (2),
109–144.
Ernst, T. (1998). “The scopal basis of adverb licensing.” In P. Tamanji & K. Kusumoto
(Eds.), Proceedings of the 28th Meeting of the North East Linguistic Society (pp. 127–142).
Amherst MA: GLSA University of Massachusetts.
Ernst, T. (2002). The Syntax of Adjuncts. Cambridge: Cambridge University Press.
Hale, K., & Keyser, S. J. (1993). “On argument structure and the lexical expression of
syntactic relations.” In K. Hale & S. J. Keyser (Eds.), The View from Building 20 (pp. 53–
109). Cambridge, MA: MIT Press.
Jackendoff, R. (1972). Semantic Interpretation in Generative Grammar. Cambridge MA: MIT
Press.
Kamp, H., & Reyle, U. (1993). From Discourse to Logic. Dordrecht: Kluwer.
Kayne, R. (1994). The Antisymmetry of Syntax. Cambridge, MA: MIT Press.
Laenzlinger, C. (1998). Comparative Studies in Word Order Variation. Adverbs, Pronouns, and
Clause Structure in Romance and Germanic. Amsterdam and Philadelphia: Benjamins.
Maienborn, C. (2001). “The position and interpretation of locative modifiers.” Natural
Language Semantics, 9, 191–240.
Maienborn, C. (2003). “Against a Davidsonian analysis of copula sentences.” In M. Kadowaki
& S. Kawahara (Eds.), Proceedings of the 33rd Meeting of the North East Linguistic Society
(pp. 167–186). Amherst, MA: GLSA University of Massachusetts.
 Domain adverbs and the syntax of adjuncts 

Moltmann, F. (1991). “Measure adverbials.” Linguistics and Philosophy, 14, 629–660.

Ortony, A. (1979). Metaphor. Cambridge: Cambridge University Press.
Partee, B. (1989). “Speaker’s reply.” In S. Allén (Ed.), Possible Worlds in Humanities, Arts and
Sciences (pp. 152–161). Berlin: Walter de Gruyter.
Portner, P., & Yabushita, K. (1998). “The semantics and pragmatics of topic phrases.”
Linguistics and Philosophy, 21, 117–157.
de Swart, H. (1998). “Aspect shift and coercion.” Natural Language and Linguistic Theory, 16,
347–385.
 Depictives and transparent adverbs

Wilhelm Geuder
University of Constance

This paper investigates the contrast between VP-adjoined adjectives

(so-called depictives) and adverbial forms in English, which manifests itself
in minimal pairs like John left sad vs. John left sadly. In spite of expectations as
to clear differences in semantic types, the two forms of adjuncts are often
difficult to tell apart on semantic grounds, and the distinction even seems
altogether absent in languages without adverbial morphology like German.
The paper argues that depictives and adverbs are indeed more closely related
than often assumed, due to the existence of a class of “transparent” adverbs
(to be distinguished from manner adverbs), which share with depictives the
property of denoting states and predicating of an individual. The
morphological distinction in English is argued to be tied to a fairly subtle
semantic parameter that is identified as the “event-dependence” of a state.

. Introduction

Depictives are adjectival adjuncts that stand in a predication relation with ei-
ther the subject or the object of the clause. Sentence (1) with the indexing in
(1a) means that John was sad while leaving Mary (a subject depictive) whereas
in (1b) it was Mary who was sad while being left by John (an object depictive
construction):
(1) a. Johni left Mary sadi .
b. John left Maryj sadj .

In this paper, I investigate the status of the distinction between depictives and
adverbial adjuncts in English. The depictive-adverb distinction can give rise to
true minimal pairs because many adjectives that can be used as depictives can
also take the adverbial affix -ly:
(1) c. John left Mary sadly.
 Wilhelm Geuder

Since the two forms in (1a, b) and (1c) cannot be used interchangeably, there
must be some semantic factor that governs the choice. There is surprisingly
little discussion of this question in the literature, and I suspect the reason for
this lies in a prevailing assumption that the answer would be very straightfor-
ward. Given the widely accepted analysis of manner adverbs as predicates of
events (e.g. Parsons 1990), and also given a large body of syntactic literature on
the depictive construction that is centred on issues of control and predication
with depictives (e.g. Williams 1980, 1994; Roberts 1987, 1988; Demonte 1988;
Déchaine 1993; Winkler 1997), it might seem that depictives and adverbs can
be characterised via the fact that they predicate of different entities (such a view
has explicitly been proposed e.g. in Déchaine 1993 or Larson and Segal 1995).
Hence, one might try to state the distinction as follows:
(2) a. leave Mary sad
leave(e, x, Mary) & sad(x) (depictive)
b. leave Mary sadly
leave(e, x, Mary) & sad(e) (adverb)

In this paper, it will be argued that the representation (2) is actually not a cor-
rect account of minimal pairs like (1a, b) vs. (1c). I do not want to deny that
(2b) is a correct analysis of manner adverbs. However, manner adverbs are not
the only case that has to be considered in the analysis of the depictive-adverb
contrast. As will be shown, another type of adverbial reading comes into play
that differs from manner adverbs but resembles depictives in that it refers to
a state of an individual. This type of adverb, which is rarely recognised in the
literature, has been termed “transparent adverb” in Geuder (2000).
Moreover, another factor contributes to the difficulties with the adverb-
depictive distinction: the event variable has to be factored not only into the
semantics of adverbial forms but into the semantics of depictives as well. How-
ever, if depictives make reference to the event variable of the clause and if,
conversely, there are adverbs that denote states of individuals, we have to deal
with minimal contrasts between two forms that each target both an individual
and an event. Indeed, this would provide an explanation for the fact that it is of-
ten very difficult to distinguish “adverbial” and “depictive” readings on purely
intuitive grounds, i.e. in the absence of morphological marking. Note that in
English we have to reckon with the possibility of adverbs that do not carry spe-
cial adverbial morphology. If the distinction between adverbs and depictives is
to be maintained for such cases, it cannot be read off the grammar but must be
derived from a semantic analysis. As an example for how difficult this can be,
consider the following case mentioned in Roberts (1987: 173, Note 2):
 Depictives and transparent adverbs 

(3) The game was played barefoot.

The problem for Roberts (1987) is that there seems to be a generalisation

according to which depictives cannot predicate of implicit arguments:
(4) a. Hei left the room sadi .
b. ??The room was left sad.

If (3) had to be viewed as a depictive construction, it would constitute a coun-

terexample. To explain the acceptability of (3), Roberts speculates that barefoot
is actually an adverb, not a depictive. Since barefoot is etymologically an adver-
bial NP, not an adjective, this might be a reason against affixing the ending -ly
even though it is an adverb. However, Roberts is not able to provide any explicit
criteria that could decide the question.1 Maybe one would indeed want to say
that barefoot somehow characterises the event of playing, but on the other hand
it is hard to say what it might assert about the event that goes beyond predicat-
ing the property barefoot of the players. Conversely, one might propose that, in
view of these difficulties, we should not posit a strict separation between the
two categories and consider depictives as a kind of adverbial modifier, too. But
this leaves the morphological contrast in (2) unexplained.
The analysis in (2) above posits a clear-cut semantic difference between
the two types of adjuncts. As we have seen, this clear distinction is not always
matched by clear semantic intuitions, and as soon as there are no unambiguous
morphological clues, the distinction between adverb and depictive suddenly
seems to be blurred. The semantic distinction we are looking for must there-
fore be fairly subtle, or to put it differently, there must be a great deal of
semantic similarity between depictives and certain adverbial modifiers, which
is not brought out by the representation in (2). In the next sections, I will
first examine the factors that lead to the perception of a strong similarity be-
tween depictives and adverbs, and then delineate their proper distinction on
this basis.
The paper is structured as follows: in Section 2, I discuss the semantic
interpretation of depictives in order to demonstrate that the event variable in-
troduced by the main verb of a clause is needed to anchor depictives. Hence,
they share a fundamental property with adverbial constructions, namely that
of making reference to an event. Next, I argue that there is a class of “transpar-
ent” adverbs, which are not simply predicates of events (as manner adverbs are)
but rather denote states of their own and predicate of an individual (which is
the holder of that state). The issue of the minimal semantic difference between
a depictive and an adverbial form, and how to choose between them, is treated
 Wilhelm Geuder

in Section 4. The choice is shown to depend on a fairly subtle semantic factor,

namely the type of relation that connects event and state. Finally, by way of an
outlook, Section 5 presents some considerations as to how the remaining dif-
ferences in the behaviour of adverbs and depictives, especially with respect to
predication, might be explained even if they are viewed as spanning a semantic
continuum. A conclusion is provided in Section 6.

. Depictives and events

As already pointed out in the introductory section, the conventional view on

depictives treats them exclusively as predicates of individuals. In this section, I
set out to show that they additionally have to be anchored to the event variable
of the clause. There is a fairly simple consideration which already indicates that
depictives can’t be just predicates of individuals. As is well-known, individual-
level adjectives are in most cases excluded from depictive constructions (like
e.g. blond, intelligent – although there are also some examples that work). The
problem is that they would perfectly fit in with the structure (2a), and even
more so from the point of view of Kratzer’s (1988/1995) influential analysis
according to which IL-predicates lack an additional event argument. Whether
Kratzer’s theory is adopted or not, IL-adjectives undeniably show a tendency
to enter into predication structures that only concern the individual (e.g. with
unselective binders like usually, as in Cats are usually intelligent). Their David-
sonian argument, if there is one, is at least more elusive. Since it is precisely the
IL-adjectives that are excluded from depictive constructions, one would sus-
pect that a crucial property of depictive constructions is the involvement of the
event arguments of verb and adjective.
Clearly, depictives do not fit the standard picture of event predication as
found with manner adverbs. However, their syntactic position and the type
of temporal interpretation they receive seem to betray a dependence on the
event argument. The syntax indirectly bears on the semantic analysis in so far
as it turns out that depictives are always adjoined at the VP level, where event
adverbs are found as well. So let us first take a look at the relevant syntactic
observations. (The material in this and the next section is largely drawn from
Geuder 2000).
 Depictives and transparent adverbs 

. The syntax of depictives

Depictives invariably occur in postverbal position in English. When we con-

sider the ordering of depictives relative to other postverbal elements we find
that depictives have to follow resultative adjectives and that subject depictives
follow object depictives (examples from Winkler 1997):
(5) John kicked the door open tired. (resultative < subj. dep.)
(6) Murphy hammered the coin flat hot. (resultative < obj. dep.)
(7) John ate the meat raw tired. (obj. dep < subj. dep.)

This ordering indicates that both subject and object depictives in English have
to be right-adjoined. Resultatives then occupy the immediate complement po-
sition next to the base position of the English verb before it undergoes short
movement which puts it before the object (cf. Bowers 1993, and others). Con-
sequently, in a language like German, which for the most part disallows right-
adjunction of modifiers, depictives are left-adjoined and precede resultatives:
(8) a. dass er müde seinen Teller leer aß
that he tired his plate empty ate
b. dass er seinen Teller müde leer aß
c. *dass er seinen Teller leer müde aß

The ordering in (8c) is strongly ungrammatical; it is not just the mild effect
that would result from infelicitous scrambling. The fact that German presents
the mirror-image of the English ordering lends additional support to the claim
that depictives are right-adjoined in English.
In principle, the sentence-final position in English is structurally ambigu-
ous and the phrasal level of adjunction remains to be determined. Several
authors have proposed that at least subject depictives should be adjoined at
the IP-level (e.g. Déchaine 1993). However, Roberts (1988) already presented
evidence to the effect that even subject-predicating depictives must be attached
at the VP-level, adducing a number of standard “VP tests” like VP fronting,
though-movement, and pseudoclefts. The result is that depictives cannot be
stranded by processes that affect VPs. The following example illustrates the
point for pseudoclefts:
(9) a. What John did was leave the room happy.
b. ?What John did happy was leave the room.

Moreover, Roberts (1988) observed that depictives invariably go with the main
verb under negation:
 Wilhelm Geuder

(10) Bill didn’t leave angry at John.

This sentence cannot mean that Bill, being angry at John, didn’t leave. If ad-
junction at the IP-level were possible, or even obligatory, this would not be
expected. In this respect, depictives differ from certain other right-adjoined
elements (Roberts’ examples):
(11) a. John didn’t kiss his wife [because he loves her].
b. John didn’t kiss his wife(,) deliberately.

The above examples are ambiguous in that the adjuncts can be interpreted
inside or outside the scope of negation. Disambiguation is provided by intona-
tion: in a structure with right adjunction, in which the modifier is outside the
scope of negation, the VP kiss his wife receives the unmarked sentence accent,
and the adjunct receives an accent of its own. When the sentence accent is dis-
tributed over the postverbal elements, this is standardly taken to indicate that
they are contained in VP. Interestingly, depictives always have to be integrated
into the VP intonationally, as shown in (12a):
(12) a. John didn’t leave háppy.
b. John didn’t léave happy.

A stress pattern as in example (12b) is possible in principle, but can only be

understood as involving contrastive focus on the verb (and so cannot be used
as an indicator of syntactic structure).
As another piece of evidence, Ernst (2002: 286) observes that manner ad-
verbs can in principle follow depictives, although Ernst judges such cases to
be somewhat marginal and in need of contextual support (e.g. speaking about
work in a painter’s studio):
(13) Al sits clothed quietly, but is often agitated when he has to be nude.

Since manner adverbs cannot be adjoined higher than VP when they are on
a left branch, it must be assumed that quietly is also a VP-adjunct in (13),
consequently the depictive must also be a VP-adjunct.

. The temporal interpretation of depictives

It is a natural assumption that VP is the syntactic domain in which all material

is generated that requires the event variable for its interpretation, i.e. manner
adverbs and other event-related adverbials and of course the verb’s arguments.
Intuitively, depictive constructions have the meaning that some state holds at
 Depictives and transparent adverbs 

the same time as the event. For a sentence like (14a), however, there would
be two ways to implement this. One way to build up a cotemporal reading
would be to link both situations to the same reference time, as is done by the
when-construction in (14b).
(14) a. John left tired.
b. When John left, he was tired.

The interpretation of (14b) can be formalised as follows:2

(15) ∃t [∃e [leave(e, John) & t = τ (e)] & t < t0 ]
& [∃s [tired(s, he) & τ (s) ⊇ t] & t < t0 ]

When-clauses are often used to paraphrase the meanings of depictive con-

structions. However, the two constructions are not semantically equivalent.
The parallelism breaks down as soon as we use a pluperfect. The difference
that emerges in such examples shows that depictives are not anchored to the
reference time:
(16) a. When we had eaten, we were tired.
b. We had eaten tired.

The pluperfect shifts the reference time to an interval in the “aftermath” of

some event (as Johnston 1994 called it). The interpretation of (16a) can then
be represented as in (17), which indicates that we were tired at some time after
the eating. (For simplicity, the perfect is rendered as the function “after”; the
pluperfect arises by combining this with a reference time shifted to the past by
the tense component).
(17) ∃t [∃e [eat(e, we) & t = after(e)] & t < t0 ]
& [∃s [tired(s, we) & τ (s) ⊇ t] & t < t0 ]

The meaning of the depictive construction (16b), however, is different from

what (17) asserts. The sentence We had eaten tired can only mean that we were
tired during the eating, independently of the way this event is temporally lo-
cated. So the depictive must be interpreted as a situation that overlaps the event
itself, not the reference time of the clause. Hence the interpretation of (16b)
should be formulated by using an overlap relation “o” between the depictive
state and the event itself.3 In (18) I show the denotation of the VP with a de-
pictive that is contained in the example (16b), before combination with the
subject and tense.4
(18) λx [∃e [eat(e, x) & ∃s [tired(s, x) & s ◦ e] & t = after(e)]]
 Wilhelm Geuder

It may be noted that the representation of the depictive part “∃s [tired(s, x)
& s ◦ e]” contains a free event variable; in this respect it resembles a neo-
Davidsonian manner adverb, though it differs greatly in lexical content. In-
deed, the semantics given here is extremely weak. The overlap relation simply
requires that the runtimes of the event and the state have some part in com-
mon. If taken in context, sentences with depictives are usually understood to
have a more specific interpretation: the state in question will usually be as-
sumed to have held before the event; so the overlap would always have to be
with a non-initial part of the state. However, we will see later in Section 4 that
this need not be specifically encoded into the semantics of depictives, but falls
out from the contrast with adverbial forms.

. Depictive constructions as complex events?

The position that depictives relate to the event variable of the clause can already
be found in a few papers that raised the question of a semantics for depictives,
such as McNally (1994) or Wunderlich (1997). However, these authors did not
go beyond the simple assertion that the state described by the depictive ad-
jective is a “part” of the situation (McNally 1994: 568) or that it is “relativised”
(Wunderlich 1997: 127) to the situation described by the verb, and holds simul-
taneously with it. More recently, however, Rothstein (2001, 2003) has proposed
an explicit event-semantic account of depictives that analyses them via an oper-
ation of sum formation. One example that Rothstein (2003) discusses is (19a),
which she assigns the analysis in (19b):
(19) a. John drove the car drunk.
b. ∃e* ∃e1 ∃e2 [e* =S (e1 ⊕ e2) & drive(e1, J., the car) & drunk(e2, J.)]

As can be seen, no explicit distinction between states and events is being made
here; the depictive is assumed to denote an eventuality e2 which is of a sort
sufficiently similar to e1 (the denotation of the verb) so as to allow the sum
formation e1 ⊕ e2. Rothstein’s superscript “S” on the equation sign is meant to
express that the resulting eventuality is not a plurality of events but a singular
event. I do not want to go into a discussion of problems with the algebraic
aspects of this proposal here, although the proposal of summing up a telic
eventuality (the verb’s denotation) with an atelic, homogeneous one (a state)
certainly invites questions. We somehow have to make sure that the outcome is
telic in case the main verb is telic and atelic if the verb is atelic; obviously, the
second summand must have no influence on the aktionsart.
 Depictives and transparent adverbs 

Crucially (as Rothstein does acknowledge), sum formation as such allows

no predictions as to the temporal ordering of the two events. Note, for instance,
that in Krifka (1992), one of the standard works on the algebraic semantics
of aktionsart, a mereological structuring of events is assumed in which the
parts are always understood as occurring in temporal succession. Precisely this
must be excluded in our case, and the algebraic part structure is indeed neutral
with respect to this issue. Rothstein (2003) therefore formulates the additional
condition that the two predicates must be “time-participant connected”:
(20) TPCONNECT (e1, e2, y) iff:
i. τ(e1) = τ(e2);
ii. e1 and e2 share a participant y

Obviously, it is only this stipulation that allows us to capture the empirical

properties that depictive constructions exhibit. For the purpose of the present
paper, this is the most important finding and we can actually remain fairly
agnostic as to whether a sum event is formed or not. The notion of TP-
connectedness mirrors the analysis in terms of overlap proposed in Section 2.2
to a large extent. It is unproblematic that the formulation (20) requires the state
to cover exactly the same interval as the event. I believe that overlap is the more
realistic formulation, for there are cases in which a depictive marks the state of
a participant at the onset of the event but the state doesn’t last throughout the
event, as is drastically illustrated by the object depictive in (21):
(21) They dissected the frog alive.

The dissecting can go on when the frog is no longer alive. – Moreover, I be-
lieve that deciding about the exact runtime of the depictive state is a matter
of inference; overlap is just the minimal requirement. Later, I argue that our
interpretation of depictive constructions is driven by the pragmatic contrast
with adverbial constructions and that the semantics of depictives need not
contain those specific features that can be obtained from inference. If this is
correct, it could easily be integrated in Rothstein’s model: it would be entirely
unproblematic to change the condition “τ(e1) = τ(e2)” to “τ(e1) ⊇ τ(e2)”.
Apart from TP-connectedness, it remains to be seen whether the existence
of the sum event e* in (19) can be demonstrated. Rothstein makes the in-
teresting suggestion that manner adverbs should be able to predicate of e*.5
Hence, her analysis should allow a logical form such as (22a) while the overlap
approach could only deliver (22b):
(22) John drove drunk soberly.
 Wilhelm Geuder

a. ∃e* ∃e1 ∃e2 [e* =S (e1 ⊕ e2) & drive(e1, J.)

& drunk(e2, J.) & sober(e*)]
b. ∃e [drive(e, J.) & ∃s [drunk(s, J.) & s ◦ e] & sober(e)]

Since the adverb applies to different events in (22a) and (22b), it should theo-
retically be possible to decide which representation is correct. One would have
to demonstrate for the context of a particular situation that the interpretation
of sober has different truth-conditions depending on whether it applies to a
complex event e* (described by the property drunk-sober) or a simple event
e as expressed by the verb alone. In practice, however, the problem is that the
conceptual interpretation of manner adverbs is still an ill-understood area, and
we have to await a framework in which the relevant conditions can be laid out
explicitly. Moreover, it would have to be shown that this effect arises regularly
as an effect of the semantic form of depictive constructions. A fact that invites
scepticism is that Rothstein’s only example involves the depictive drunk, which
possibly has exceptional properties. Guémann (1990) observes that certain de-
pictives, but not others, have what he calls a “manner” reading. Their special
status is evidenced by the fact that they can be combined with nominalisations
in the very same way (Guémann 1990: 147):
(23) a. John drove drunk.
b. John’s drunk driving
(24) a. John departed furious.
b. #John’s furious departure

As (24b) shows, “normal” depictives cannot be transferred to nominalisations

in the same sense. Rather than taking some depictives to be manner modifiers,
I suspect that the collocation drive drunk might have developed the lexical sta-
tus of a compound property, hence the impression that drunk modulates the
event property drive in a way similar to a manner modifier. So actually, Roth-
stein’s analysis could be correct for this example, but not carry over to depictive
constructions in general.
Note, finally, that it does not seem entirely impossible to coordinate depic-
tives and manner adverbs:
(25) She ate her meal cold and hurriedly. (Guémann 1990: 148)

Such examples – even if marginal – are completely unexpected under Roth-

stein’s analysis but are compatible with the “intersective” representation that I
have proposed in (18) above (Section 2.2).
 Depictives and transparent adverbs 

To conclude this section, let me point out that Rothstein’s (2003) analysis,
although it clearly contradicts the proposal made here, actually differs only in
a small aspect. Comparing the two approaches, it can be noted that the sum-
mation approach has to stipulate exactly the same kind of temporal relation
that is already present in an analysis in terms of temporal overlap. In any case,
some kind of operation is needed which conjoins the event arguments of verb
and depictive; the difference is only whether this operation is seen as giving rise
to an additional individual (e1 ⊕ e2 = e*) or not (e1 ◦ e2). Moreover, I don’t
think that Rothstein’s proposal can throw very much light on the distinction
between adverbs and depictives: if confronted with cases in which it is diffi-
cult to distinguish adverbs and depictives semantically, one would simply have
to consider extending the summation analysis to certain adverbs. The central
question is still whether depictives can as easily be distinguished from adverbs
as Rothstein (2003: 559–561), along with many others, supposes. This is the
issue I turn to now.

. Manner adverbs and transparent adverbs

This section provides a closer inspection of the semantics of those adverbs

which contrast with depictive constructions. A particular lexical class of ad-
jectives causes the problem of minimal contrasts, namely those adverbs that
are derived from stative predicates of individuals like sad, angry, etc. Adjectives
that directly qualify properties of events by virtue of their underlying lexical
meaning, like quick, don’t occur in depictive constructions.
For adjectives like sad and angry, the distinction between depictive and
manner uses is usually quite sharp because manner adverbs of this type involve
a lexical shift from individual to event predication. Saying that the manner of
some action is “angry” is not tantamount to ascribing this state to an individual
in the event. For example, the second sentence of (26) below contains a manner
adverb saying that John’s shouting is marked with anger, but the context leads
one to expect that the predicate angry is not true of John in this situation:
(26) – How did you manage to make them believe you were a real officer?
– Well, I kept shouting at them all the time real angrily.

The property of individuals angry is something that we can infer from the use
of a manner adverb angrily in certain contexts, but not in others. So the asser-
tion of a manner adverb concerns a different thing: the type of shouting – a
true property of the event itself. The manner predication that is derived from
 Wilhelm Geuder

the adjective angry exploits our knowledge about typical correlations between
emotional states of the agent and properties of the action that is performed,
but, as I see it, this merely defines a relationship between two word meanings.
Likewise, the individual-level adjective intelligent can be shifted to a man-
ner reading which again makes a different kind of predication. It can be noted
that the adverb is no longer “individual-level” in any sense, since it makes an
assertion that is entirely limited to the episode of problem solving:
(27) a. John is intelligent.
b. John solved the problem intelligently.

The statement (27a) can serve as an explanation for why the event in (27b)
could proceed the way it did, but this again expresses our knowledge about
possible relationships between a property of John and a property of the event.
Therefore, the manner adverb intelligently (e) is again distinct from the predi-
cation intelligent(x). In general, we can say that manner readings are “opaque”
with respect to the property of individuals denoted by the underlying adjective.
When we have to distinguish adverbial from depictive constructions, the
difficulty is often that the adverb in question does actually seem to strictly en-
tail the property of individuals denoted by its adjectival base. By definition,
such adverbs could not be called manner adverbs. The existence of such a
second type of adverb often goes unrecognised in the literature but has been
described in Geuder (2000) and Ernst (2002). I propose calling them “trans-
parent adverbs”; in my opinion, Ernst’s classification of these adverbs as the
“state reading” of “mental attitude adverbs” is not entirely correct.
A number of diagnostics show the distinct behaviour of this second type
of adverb. Ernst (2002: 67) demonstrates a difference in entailments between
(28a) and (28b):
(28) Though her emotions were in a turmoil. . .
a. . . .she managed to leave the room calmly.
b. ??. . .she calmly had left the room.

Example (28a) shows the opaqueness of manner adverbs with respect to their
adjectival base that has already been discussed: the manner adverb calmly serves
to pick out that manner of the event that is typically connected to calmness on
the part of the agent. The manner adverb is compatible with a different actual
state of the agent – but not the preverbal occurrence of calmly in (28b). Hence,
this adverb must be taken to assert the state calm as of an individual. It is also
telling that the paraphrase in a calm manner would not be used here.
 Depictives and transparent adverbs 

The difference in the readings appears correlated with a difference in syn-

tactic position, but we cannot simply say that transparent adverbs are the ones
in preverbal position and manner adverbs the ones in postverbal position; the
picture is more complex. Manner adverbs can in principle precede the verb as
well, although they preferentially go into the postverbal position. Most speak-
ers would prepose them only if there is enough heavy material following the
verb6 or if the verb is passive.
(29) a. She walked carefully on the ice.
b. She carefully walked on the ice.
c. She walked carefully.
d. ?She carefully walked.

However, transparent adverbs never degrade in acceptability when they occur

in preverbal position and they don’t require heavy material after the verb (or
passive morphology) in order to be allowed there:
(30) a. He sadly discovered a mistake.
b. He angrily forwarded me the letter.

Also, we do find postverbal adverbs that have to be interpreted as transparent.

In (31) below, the adverb does not serve to characterise a subtype of the event in
question, but rather a state of the agent; the sentences could not be continued
with a statement that contradicts the sense of the underlying adjective:
(31) a. He discovered sadly that the solution was incorrect.7
??(. . . but actually
he was glad that this had happened.)
b. He forwarded the letter angrily to his solicitor.
??(. . . but he didn’t really feel angry.)

Sometimes transparent adverbs are unacceptable in the absolute final position,

but there are also examples that are fine:
(32) a. ??He discovered the mistake sadly.
b. He left the meeting disappointedly.

Note that the contrast in (28), already cited above from Ernst (2002),
does not exclude the possibility that transparent adverbs can occur in VP-
internal position:
[28] Though her emotions were in a turmoil. . .
a. . . .she managed to leave the room calmly.
b. ??. . .she calmly had left the room.
 Wilhelm Geuder

Example (28a) is acceptable under a manner reading of calmly; a transparent

reading would lead to unacceptability, but since an acceptable construal has
already been found, no effect of this other construal will be detected in a gram-
maticality judgement, even if it is available in principle. The word order in
(28b) is chosen in a way so as to make a manner reading impossible, and only
here do we find a clash.
So in sum, we can say that transparent adverbs differ from manner adverbs
mainly in that they have different positional preferences, although they often
do not differ in terms of the absolute positions they are allowed to occupy (with
the exception of the pre-auxiliary position in (28b)). Patterns like (30a), (31a),
(32a) moreover suggest that transparent adverbs are sometimes required to be
unstressed, whereas it is typical of manner adverbs (when in final position) to
attract the sentence accent.
It is still often difficult to test which type of adverb we are dealing with
in a particular example. As we have seen, the syntactic criteria are subtle and
there is a good deal of variability in the grammatical behaviour of adverbs. The
semantic distinction is blurred by the fact that the semantics of manner adverbs
is often compatible with the additional assumption that the state expressed by
the adjectival stem also holds. Except for special contexts, this may even be
the default assumption – it is just not what the adverb asserts. Indeed, if we
acknowledge the fact that, in interpreting a sentence, the semantic form will
be enriched by inferences, it seems that we can often derive similar results on
different grounds:
(33) She walked out calmly.
a. ∃e [walk-out(e, she) & calm(e)]
+ inference: there is (visual) evidence, [calm(she)] may well hold
b. ∃e [walk-out(e, she) & [ . . . ∃s calm(s, she) . . .]
+ inference: her calmness may well influence the way the event un-
folds

(In (33b), the transparent adverb has been provisionally represented by the
state component alone; I’ll turn to the necessary refinements immediately). My
position is that such convergences as in (33) do not undermine the assumption
of clear-cut differences in semantic form. There is a fair amount of evidence
that a difference between transparent and manner adverbs exists in principle.8
It has now become clear that the minimal contrast between depictives and
adverbs involves not so much a contrast with manner adverbs, but a con-
trast with transparent adverbs. These share many more semantic properties
with depictives than manner adverbs do: both depictives and transparent ad-
 Depictives and transparent adverbs 

verbs describe a state of an individual that is related to the event described

by the verb.

. The semantics of transparent adverbs and depictives

. Ernst (2002)

Ernst (2002) is the only other work in the literature I am aware of that discusses
the existence of adverbs which denote states of their own. An assortment of
interpretation rules that Ernst proposes for the relevant adverb types is given
in (34). Ernst’s categories “State” and “Intentional” are the ones that would fall
into my class of transparent adverbs:
(34) a. Manner: ADV(e) = e [manifests] ADJ(x), with x = Agent(e)
b. State: ADV(e) = e [is accompanied by] ADJ(x), with x = Agent(e)
c. Intentional: ADV(e) =e [is intended with] ADJ(x), with x = Agent(e)
(from Ernst 2002: 63–66)

The two fundamental problems that I see with Ernst’s account as a whole are,
first, that the difference between manner and what I call transparent adverbs
is not represented – (34a) vs. (34b, c) do not indicate a difference in type or a
difference in the actual presence of a state. Secondly, Ernst assumes no semantic
difference between depictives and those adverbial forms he subsumes under
“state” in (34b); rather, he explicitly states (Ernst 2002: 67) that he considers
depictives to have the same representation. This is a serious shortcoming, for
depictives and adverbial forms cannot be used interchangeably.
Let us, however, first focus on the semantic analysis that Ernst assigns to
those adverbs that entail a state, i.e., the types (34b) and (34c). He summarily
refers to them as “mental-attitude adverbs” (which, actually, does not seem to
express exactly the right generalisation) with the shared property that “the sub-
ject experiences a certain mental state during the time that the event holds” (p.
63). The borders between the “intentional” type (e.g. reluctantly, willingly) and
the “state” type (with delightedly, calmly as typical examples) are fluid because
either lexical type of adverb is thought to be possible in either interpretation.
As an example for the state / intentional variation, consider again Ernst’s pair:
(35) a. She calmly had left the room.
b. She had left the room calmly.
 Wilhelm Geuder

Ernst observes (p. 64): “[a] also seems to have an intentional reading. . . it is
her decision to leave that is calm; her mental state precedes the event described
(while for the state reading it is simultaneous)”. It is also to be noted that the
word order in (35a) points to calmly as a sentential adverb, since the position
in front of the finite auxiliary would not be available for a manner adverb. On
the other hand, (35b) displays an adverb calmly which has the same word order
and intonational properties as a manner adverb. With respect to this, it seems
as if “state” and “intentional” variants were a case of true ambiguity. However,
this position does not seem to be tenable. I rather conclude from the facts that
there is only a “circumstantial” semantics that underlies both uses of calmly
above. This becomes clear as soon as we compare the “intentional reading” of
calmly with the group of “intentional adverbs” like reluctantly, willingly, and
intentionally (the latter is surprisingly absent from Ernst’s discussion). These
latter adverbs denote attitudes by virtue of their lexical content, and therefore
they also have an intentional reading in VP-internal positions:
(36) John kicked me intentionally / reluctantly.9

However, with a mental state adjective, the “intentional” semantics behaves

as an additional flavour, not as a distinct reading. Sometimes an intentional
component is simply too weak and elusive to count as the expression of a
propositional attitude. This is already evident in Ernst’s own example (repeated
as (37a)) as well as in many other cases:
(37) a. She calmly left the room.
b. The bartender cheerfully opened a bottle.

Note also that the semantics adumbrated in Ernst (2002) for (37a) states that
the “decision” to do something, rather than the execution of an action, is ac-
companied by the emotion in question. However, this would seem to be the
same notion of “accompaniment” then: the decision can be said to “accom-
pany” the action, and the adverb accompanies the decision. I am not convinced
that there is a principled distinction to the state variant. Furthermore, note the
marked difference between an attitude and “circumstances of action” with the
adverb angrily in (38a):
(38) a. I angrily forwarded the letter to my solicitor.
b. I reluctantly forwarded the letter to my solicitor.

The concept “anger” may have an attitude component, but the event descrip-
tion under the adverb angrily does not represent the objective of such an
attitude: the thing that angers me is not my forwarding the letter. In contrast to
 Depictives and transparent adverbs 

this, the forwarding is indeed what constitutes the objective of my reluctance

in (38b). So what is meant in connection with adverbs such as angrily is really
only the emotional state of anger that accompanies the action in some sort of
way. Details of the syntactic position might influence the exact way this hap-
pens. Ernst seems to feel that an “intentional” component is present only with
mental attitude adverbs in preverbal position. In this position, a background-
ing effect appears (as is the case with preverbal manner adverbs), the adverb
cannot be stressed and is often infelicitous under negation. It is not inconceiv-
able that such factors lead to additional subtle shades of meaning of an adverb
with an essentially “circumstantial” semantics, given that the “state” reading as
such has a very weak semantics on Ernst’s account.
The tentative conclusion that I draw from this is that there is no more than
one single semantic type of transparent adverb present in Ernst’s (2002) dis-
cussion of the “intentional” / “stative” distinction (apart from adverbs like
reluctantly that express an attitude already by virtue of their underlying lexi-
cal meaning; these are not relevant to the objective of this paper). However,
Ernst’s characterisation of the adverbial meaning as a state that “accompanies”
the event does not seem to provide the right kind of clue to the contrast with
depictives, and it also seems that Ernst’s treatment does not exhaust all the in-
terpretational possibilities. In the following section (again taking up material
from Geuder 2000), I show that the semantic parameter we are looking for can
best be detected by analysing the minimal contrasts between depictives and
adverbial forms.

. Minimal pairs

The existence of a meaning difference between depictives and transparent ad-

verbs can be established very clearly by considering minimal pairs in which
only one of them is permitted. Let us consider cases in which a depictive is not
acceptable, but the corresponding adverbial form must be used, such as (39).
Here, we see that the use of a transparent adverb saves the construction.
(39) He discovered {?? sad / ok sadly} that the solution was incorrect.

From what we know so far, it is not obvious what kind of factor could ever
prohibit the use of a depictive. It is true that for the case of individual-level (IL)
adjectives, the explanation suggests itself that they would make the assertion of
an overlap with the event uninformative. But even this cannot be the whole
story, for in certain contexts, IL-depictives turn out to be acceptable:
 Wilhelm Geuder

(40) a. My dad was born compulsive and will die compulsive.

b. The tablecloth went to the cleaners white and came back yellowish.
(from McNally 1994)

Moreover, it is certainly not sufficient to invoke the ban on IL-adjectives

as a general explanation for ill-formed depictive constructions (as Déchaine
1993: 219 seems to do):
(41) ??She gave her paper aggressive / nervous.

These adjectives clearly describe passing phases, so whether or not they are IL-
level in some sense, we can’t derive an explanation for their unacceptability in
the usual way.
This leads to the general point that an explanation for the unacceptabil-
ity of certain depictives cannot be given in terms of lexical types of adjectives
alone. Rather, it must be the interplay of verb meaning and adjective type that
decides on the acceptability of depictive constructions. Let us now consider a
few minimal pairs with the same stage-level adjective:
(42) a. He left sad.
b. He discovered {?? sad / ok sadly} that the solution was incorrect.
(43) a. He returned angry.
b. He read the review of his book {? angry / ok angrily}.
(44) a. She returned from the meeting very proud.
b. She showed us the pictures very {?? proud / ok proudly}.

The (attempts at) depictive constructions in the (b)-examples may not be

strictly speaking ungrammatical; they might rather be taken to form intel-
ligible statements, but statements “one simply wouldn’t want to make”. The
difference seems to be connected to the choice of verbs, for the verbs return,
come, leave in the (a) examples seem to be especially well-suited for depictive
adjuncts. These are fairly neutral verbs with a kind of presentational effect:
a quality of the subject becomes visible at a certain point. There is no fur-
ther interaction between the state and the event. This is different in the (b)
cases above. We can’t help assuming that there must be a connection between,
say, the discovery of the mistake and the sadness of the observer. I would like
to propose that precisely this type of inference, which is added involuntarily,
is what makes depictives unacceptable. When the emotional state arises as an
effect of the event described by the verb, the adverbial form has to appear. Con-
versely, where the depictive is used, the mental state is understood as holding
independently of the event. Sometimes this conceptualisation of the situation
 Depictives and transparent adverbs 

is intuitively rejected, and this is why depictives are not tolerated in the (b)
sentences above.
In cases in which both forms are in principle allowed, a meaning difference
along these lines can still be detected. Consider the example:
(45) a. John left the meeting angry.
b. John angrily left the meeting.

In (45a) one would infer a scenario in which a meeting ended as previously

scheduled, and at that time John left and was in an angry state. For (45b) one
would infer a dependency relation; in particular, the leaving may be construed
as dependent on the emotion of anger and would perhaps not have occurred
under more pleasant circumstances.

. Transparent adverbs as event-dependent states

The exact nature of the semantic link between transparent adverbs and the
event described by the verb is obviously variable. So far, we have cases like
(42b), (43b), (44b), suggesting that the emotional state is brought about by
the event, and cases which suggest that the action is brought about by the
emotional state:
(46) a. I angrily forwarded the letter to my solicitor.
b. I hungrily opened the fridge.

So there must be at least these two ways of enriching the state predication. For
the latter type exemplified in (46), I posit a relation motive that is added on
top of the semantic representation of a depictive construction. The connecting
element cannot just be causation because there is no strict necessity that the
emotion has to result in the action; the action is just seen as an understandable
outcome. We get the following representation, built on the depictive semantics
given in (18) in Section 2.2:
(47) a. I angrily forwarded the letter to my solicitor.
b. ∃e [forward(e, I, the letter, to my solicitor)
& ∃s (angry(s, I) & s ◦ e & Rmotive (s,e))]

It should be noted that (47a) is not compatible with a state of affairs in which,
at some point, I felt anger and decided to consult my solicitor but sent off
the letter only the day after when my anger had already lessened. Rather, the
emotion of anger is required to be present in the situation, even if its onset is
prior to the forwarding. The fact that a motivational state, even if it precedes
 Wilhelm Geuder

an action, still has to overlap with it, too, creates a close parallelism with the
interpretation of depictives, which also minimally require overlap, yet allow the
assumption that the state has held for some time before the event. The depictive
state can have this temporal extension because state and event are completely
independent of each other. A motivational state also exists independently of
the corresponding action, but not the other way round, and this is what forces
us to use an adverbial form.
Let us return to the other examples discussed under (42) above, with (42b)
serving as another exemplary case.
[42] b. He discovered sadly that the solution was incorrect.

This sentence appears to exhibit a link between the event and the emotional
state which runs in the opposite direction: here the state appears to be caused
by the event. This causation relation has peculiar properties, though, that set it
apart from the usual billiard ball model of causation. In particular, we find that
the emotional state and its cause do not simply occur in succession, but show
considerable overlap. To see this, consider verbs which describe psychological
causation:
(48) a. The review of his book angered John.
b. Reading the review angered John.
(49) The TV set worried John.

Causative psych-verbs are a long-standing problem in lexical semantics because

they do not align themselves with causative accomplishment verbs in their tem-
poral properties. They rather represent the unusual case of stative causatives.
The simplest way of representing e.g. (48a) is by using a decomposition that
contains cause but no become, since in the system of Dowty (1979) it is only
become that introduces the temporal properties of a succession of states:
[48] c. x cause [angry(y)]

This atemporal interpretation of causality also seems to be reflected in the se-

mantics of transparent adverbs. Again, we can begin to capture their meaning
by building on the temporal properties already present in depictive construc-
tions, adding a causal component (I will have to add some qualifications later).
Consider the analysis of the VP in (42b) (=50a) in (50b):
(50) a. He discovered sadly that the solution was incorrect.
b. λx ∃e (discover(e, x, p) & ∃s [sad(s, x) & s ◦ e & e cause s])
 Depictives and transparent adverbs 

Comparing the “motivational” transparent adverbs to this causal type, there

is a slight difference in the way the temporal overlap relation has to be inter-
preted. In the present case, the state could not be assumed to extend to a time
before the event, while this was possible for adverbs of motivation. However,
I think this is a difference that need not be reflected in the semantic entries.
The exact nature of the overlap can be inferred from the properties of the sit-
uation. Choosing the precise type of relation that connects a state to the event
is already a matter of inference: we know that certain emotional states trigger
typical reactions (like anger, hunger) while others are typical effects of events
they accompany (like sadness, excitement). Some states can be contextualised
either way, as shown for instance by the fact that the transparent adverb angrily
can have both readings, cf. (43b) and (46a) above.
I have claimed that a causal connection between event and state is required
in the analysis of transparent adverbs like sadly, but it may be asked whether
this is already the whole story. At least in certain cases, the adverb might also
be thought to express an attitude. Distinguishing between the cause and the
content of an emotional attitude is of course a difficult philosophical question.
Sometimes paraphrase relations can be found that support a causal analysis:
there are prepositional phrases like to my delight, to my surprise etc. which,
being directionals, clearly present an emotional state as a consequence of an
event. On the other hand, there are paraphrases with an adjective plus infini-
tive that might be taken as expressing an attitude. As the examples in (51) and
(52) show, there is no entirely systematic correspondence between transparent
adverbs and such paraphrases, but they all seem to fall into the same range of
variants that describe the causation and/or content of an emotional state:
(51) a. I accepted the invitation delightedly.
b. I accepted the invitation with delight.
c. I was delighted to accept the invitation.
(52) I noticed {delightedly / to my delight} that the recital had attracted a large
audience.

Indeed, it seems that adverbs as well as their prepositional paraphrases never

include the possibility that an emotional state comes about by purely external
factors (like drugs), but also imply that it is mediated by a mental repre-
sentation.
For want of a better label, however, I will keep calling the non-motivational
interpretation of transparent adverbs a “causal” one because it underlines the
 Wilhelm Geuder

fact that the emotional state is dependent on the event, in contrast to the
first variant.
Not only is it hard to judge whether this second interpretation should be
analysed in terms of causation, or attitude, or content of mental representation,
but it is also quite difficult to say what the exact number of readings for trans-
parent adverbs is in general. Apparently, even mixed cases are not impossible.
Compare, for instance the following two examples:
(53) a. Following this insult, John angrily stood up and left the room.
b. Gene angrily read through the terms again. The contract was short, to
the point, and quite clear. “I want my lawyer to look this over.”10

In (53a), the state of anger is more clearly separated from the resulting action
than in (53b). The context of the latter example indicates that a state of anger
works as a motivation for the second reading event because the state existed al-
ready before it; nevertheless, we may assume that this state is reinforced during
the reading and so this usage seems to shade off into the reading of a caused
state to some extent. We are not dealing with an ambiguity between two adver-
bial readings, but with one continuing state that relates to the event in various
ways simultaneously.
Given that their interpretation in single instances is heavily influenced by
inferences triggered by the context at hand, it will be advisable to formulate
the semantics of transparent adverbs in somewhat underspecified terms. This
does not undermine the approach, however. What is essential is the finding
that transparent adverbs, while distinct from manner adverbs, are also more
than just a state that accompanies an event since they involve some kind of
additional link to the event, a fact that we can summarily refer to as their
“event-dependence” (for short; although it is strictly speaking co-dependence,
since it can go in either direction). The event-dependence of a state as a cover
notion for the semantics of transparent adverbs is a pretty vague concept, it
is true, but this is not necessarily a shortcoming, since many of the general
cognitive principles that structure language are of this underspecified sort.11
The semantics of transparent adverbs can thus be described via the tem-
plate in (54), with RD as a cover term for a whole family of notions that contains
a dependency relation between s and e (“V” and “A” represent the meaning of
the verb and the adjectival stem of the adverb, respectively).
(54) ∃e [V(e, x, y) & ∃s (A(s, x) & s ◦ e & RD (s,e))]

There is still some work left to do in delimiting the range of options for RD . The
intuitions reported by Ernst (2002: 64) on “intentional” interpretations of the
 Depictives and transparent adverbs 

adverb calmly certainly should also be integrated into this picture. As already
quoted above, Ernst observes that, in the sentence She calmly left the room, “it is
her decision to leave that is calm; her mental state precedes the event described”.
This description of the psychological state as a precursor of the action might
point to a slightly different variant of dependence between state and event, but
it still somehow aligns with the other cases of motivational adverbs.

. Choosing between adverbs and depictives

As we have seen, accounting for the various ways in which transparent ad-
verbs characterise a state as event-dependent is not an easy task. It is much
easier to characterise depictives: they are understood as states that exist entirely
independently of the event.
For an explanation of the semantic contrast between depictives and trans-
parent adverbs, it seems sufficient to represent depictives without explicitly
entering the event-independence of the state into the representation. If it is
granted that the adverbial morphology signals event-dependence of a state,
the absence of such morphology will trigger a Gricean implicature that there
is no such dependency relation. The very fact that the distinctive interpreta-
tion of depictives can only be characterised via negation (i.e., that they are
independent of the event) already points to an effect of implicature. Thus, the
independent existence of the state expressed by a depictive would not be an
aspect of its semantic form.
If the choice between adverbs and depictives were largely driven by prag-
matic considerations, we would predict that the notion of an independently
existing state need not be activated if for some reason no adverbial form is
available that could enter into a contrast with the adjective – say, if for purely
morphological reasons no such form could be derived. This opens up a way
of analysing the question of the adverb-depictive distinction in a language like
German, in which adverbial morphology is simply absent: the contrast between
depictives and transparent adverbs simply doesn’t apply (a full analysis of the
German case would exceed the limitations of this paper, but see Section 5.1
below for a brief illustration).
In English, there is one phenomenon that also might find an explanation
along the lines just sketched. Consider the apparently exceptional lack of a
contrast between depictive and adverbial forms in the following examples:
(55) a. She read the letter perplexed / perplexedly.
b. She read the letter delighted / delightedly.
 Wilhelm Geuder

Here, adjectival forms can appear although one would certainly wish to infer
that the states arise in reaction to the event. An adverbial form can be used
instead, it is true, but no strong meaning difference is felt (unless a manner
reading is taken into consideration, which is of course not the point here). All
such examples that I am aware of involve participial forms, and maybe this
points towards an explanation. As pointed out by Pesetsky (1995: 91), English
adverbial morphology cannot be applied productively to participles, so if af-
fixation to a participial form occurs, this must be due to lexicalisation of this
form as a simple adjective. The free variation of the forms in (55) could then be
attributed to the existence of doublets, i.e. the same form can be a derived de-
verbal participle (which cannot carry the affix -ly) or a lexical adjective (which
can carry the affix). If it is correct that the interpretation of depictive forms is
partly driven by the contrast to adverbs, one should indeed find that a bare par-
ticipial form can felicitously be used for event-dependent states as soon as there
is some factor that precludes the existence of an adverbial form for this lexical
entry. Hence, the bare forms in (55) should be deverbal participles, not the
lexicalised variants. What is certainly still true is that the adverbial form has
a narrower range of readings than the bare participle (putting aside manner
readings, as usual).

. Grammatical contrasts between adverbs and depictives

In a nutshell, the account presented in this paper describes transparent adverbs

as “depictives plus something”. What is added to the semantic form of the de-
pictive is a dependency relation that connects the depictive state to the event
expressed by the verb. If it is true that depictives have a semantics that simply
contains less restrictions than the adverbial semantics, the fact that depictives
and adverbs exhibit a few sharp distinctions in their grammatical properties
is in need of explanation. In this last section, I am going to discuss three phe-
nomena that appear to pose problems for the assumption of a strong semantic
relatedness between depictives and transparent adverbs. These phenomena are
differences with respect to complementation, orientation, and control. By and
large, we will see that the observations to be made in these areas do not threaten
the approach I have taken to the depictive-adverb distinction.
 Depictives and transparent adverbs 

. Complementation

Depictive adjectives can take prepositional complements, but this is not possi-
ble for (transparent) adverbs:
(56) a. John left the meeting angry at his colleagues.
b. *John left angrily at his colleagues.

Given the claim that the same state predication as in (56a) also underlies the
adverbial construction in (56b), one might expect that the same complements
should be able to appear. However, the contrast in (56) is not necessarily a
problem for the analysis I have proposed. It is not to be denied that depic-
tives and adverbs in English form two distinct morphosyntactic classes; my
analysis only claims that they are related with respect to their semantic inter-
pretation. Indeed, the differences in complementation seem to be tied to the
morphosyntactic categories of adjective and adverb in English, not to the se-
mantics. This can be shown from a comparison of the effect in (56) with the
German counterparts of the constructions at issue.
As already briefly mentioned, there is no special morphology for VP-
adverbs in German, so depictives and all types of VP-adverbs always look
the same and arguably belong to a single morphosyntactic category of adjec-
tive (even though these adjectives serve different semantic functions). While
a morphological distinction is lacking, there is a loose correspondence to the
depictive-adverb distinction in word order patterns. The (b) examples below
show the preferred rendering of the corresponding English sentences in (a) (as
usual, the word order in embedded clauses yields the most reliable evidence):
(57) a. She angrily left the room.
b. (dass) sie verärgert den Raum verließ
that she angry the room left
(58) a. She left the room angry.
b. (dass) sie den Raum verärgert verließ
that she the room angry left
The sentence accent in the German examples is marked in bold. In (58b) we see
that the sentence accent spreads over verb and depictive. It seems also possible
to get an “adverbial” (i.e., event-dependent) reading for (58b) if verärgert is de-
stressed and the verb gets the only stress instead. Now we can ask what happens
if a complement is added to the adjective verärgert in (57b), the sentence which
triggered a “motive” interpretation for the adjective:
 Wilhelm Geuder

(59) dass sie verärgert über ihre Kollegen den Raum verließ
that she angr(il)y at her colleagues the room left
Without a doubt, this structure is grammatically well-formed (with stress as
indicated), although the parallel sentence (57b) could not have the depictive in-
terpretation. We can conclude that there are no deep semantic reasons against
having event-dependent interpretations of adjuncts with complements. The
deviance of a construction like angrily at her colleagues is then just a fact about
the surface grammar of English.

. Orientation

Another striking difference between transparent adverbs and depictives is that

the former can only be predicated of the agent, not of the object. In contrast,
there exist subject-oriented as well as object-oriented depictives:
(60) a. Johni left Maryk sadi/k .
b. Johni left Maryk sadlyi/*k . (Déchaine 1993)

There are sometimes marginal instances of object orientation, especially with

adverbs like reluctantly:
(?)
(61) I watched how the police took a man reluctantly to the car.

– but in general transparent adverbs cannot be used this way.

It seems plausible, however, that this difference between adverbs and de-
pictives can be traced back to that part of their semantics that does distinguish
them: the presence of a dependency relation between state and event. Let us
first go back to the case of adverbs with a “motive” interpretation, as in (38),
repeated below:
[38] a. I angrily forwarded the letter.
b. ∃e ∃y letter(y) & [forward(e, I, y) & ∃s (angry(s, I) & s◦e &
Rmotive (s,e))]

It is indeed evident that the motive relation must imply the restriction that the
holder of the state and the agent of the event are identical; this follows from
the very notion of a motivational state. In a way, R shares this property with
standard cases of control predicates. For instance, all verbs with meanings like
motivate, urge, seduce, etc. select a direct object as the controller of a following
infinitive, which describes the ensuing action.
 Depictives and transparent adverbs 

Things look slightly different with other interpretations of R. Consider

again:
[42] b. He discovered sadly that the solution was incorrect.
c. λx ∃e (discover(e, x, p) & ∃s [sad(s, x) & s ◦ e & e cause s])

If the link between event and state consisted of pure causation and nothing else,
then it would not be semantically implausible that an adverb might be used to
describe the situation “John left Mary and this made Mary sad”:
(62) a. *Johni left Maryk sadlyk .
b. ∃e(leave(e, John, Mary) & ∃s[sad(s, Mary) & s ◦ e & e cause s])

If it is true, however, that the causal relation is always accompanied by other

meaning components, as hinted at above in the discussion at the end of Sec-
tion 4.3, this might lead to a solution. It should be possible to constrain the
orientation of transparent adverbs via some kind of requirement on the the-
matic role properties of the controller. For instance, these adverbs might have
to describe the way in which an agent is emotionally affected by its own ac-
tion, or the attitude of an agent towards its own action, although examples like
(42b) with the non-agentive verb discover show that agentivity is not the de-
cisive factor. I have to leave this point open for further research. Some sort of
stipulation will be inevitable, even if it only concerns the range of admissible
specifications of R.
In connection to this problem it should be noted that the orientation of
depictives is not entirely free, either. However, on closer inspection it seems
to be governed by a different kind of restriction. Since some significant facts
appear to be not widely known, let me dwell a little bit on this issue here and
in the following subsection.
Depictives can be controlled by subjects and direct objects, but usually not
by indirect objects or obliques12 – as can be seen from the following examples
given in Baker (1997: 90):
(63) a. John loaded the hayi onto the wagon greeni .
b. *John loaded the wagon with hayi greeni .
(64) a. I gave the meat i to Mary rawi .
b. *I gave the meat to Maryi hungryi .
c. *I gave Maryi the meat hungryi .

In a few cases, however, structures of the type (64c) do allow control of

a depictive. Pylkkänen (2002) cites the following example from a paper by
Maling (2001):
 Wilhelm Geuder

(65) The nurse gave the patient i his medication half-asleepi .

As Pylkkänen points out, examples of this type may consistently turn out to
be light verb constructions. Light verb constructions have the peculiar seman-
tic property that the verb only has a very elusive meaning, with an object NP
providing the content of the event description; in this way, the second object
argument patient, which appears to be a recipient in (65), would come close to
a theme argument on a purely semantic level of description.13
While orientation of adverbs to participants is effected via reference to the
event (see also Wyner 1998 who defends this position with respect to the adverb
type reluctantly), it is commonly assumed that the orientation of depictives is
regulated by syntactic mechanisms. However, I think, the case of the light verb
construction in (65) militates against the view that the conditions on possible
controllers for depictives are of a syntactic nature; they rather seem to arise on
an interpretational level, too14 (in the next subsection I will further substantiate
this position). If this is true, it bridges the apparent gulf between depictives
and adverbs.

. Control in passives

Depictives often cannot be predicated of implicit agents in a passive (Roberts

1987), as already mentioned at the beginning of this article:
[3] a. Hei left the room sadi .
b. ??The room was left sad.

Some more examples that illustrate the effect are:

(66) a. *The vegetables were bought hungry.
b. ?The living room was painted nude.

This seems to constitute a contrast to transparent adverbs, which easily allow

orientation towards an implicit agent in passives:15
(67) a. *The letter was forwarded to me angry.
b. ok The letter was angrily forwarded to me.

Alas, this picture is too simple. In contrast to the account given e.g. in Roberts
(1987), the behaviour of depictives in passives is not uniform. It is true that
subject depictives easily become unacceptable if their controller is removed
under passivisation. However, there are perfectly acceptable examples as well
(compare these with (66) above):
 Depictives and transparent adverbs 

(68) a. The shopping should never be done hungry.

b. The rites used to be performed nude.

As far as I can see, depictives can systematically occur with orientation towards
implicit arguments if the sentence is generic, habitual, or in some related mood.
(A similar effect has already been noted by Demonte 1988 for Spanish, where
it also holds for resultative adjectives, but it seems to have gone unnoticed
for English).
Given such examples, we are again led to the conclusion that it is not a
structural syntactic factor that decides the accessibility of an argument as a
controller for the depictive. The crucial distinction between the behaviour of
adverbs and depictives in passives is rather that, with depictives, the predication
of an implicit argument obeys certain restrictions which are not relevant to
adverbs. Intuitively, the source of the restrictions can best be described as some
kind of “salience.” This is a “functional” notion, not a semantic one. Vague as
it is, it seems to be the kind of notion that is general enough to tie together the
diverse effects that can be observed.
Let me offer some speculations to explicate this idea: The generic pas-
sive sentences that worked fine with depictives (cf. 68) make reference to an
“arbitrary” agent. Arbitrary arguments, along with overt arbitrary pronouns
like German man, are to be viewed as referring to the salient group in the
logophoric centre of a context (cf. Kratzer 2000). These cannot be explicitly in-
troduced into a discourse, but are given as salient entities by the context itself.
In passive sentences which describe a single episode, a clash between implicit
controllers and depictives occurs because the depictive in this case needs a fore-
grounded specific referent, and the passive serves to remove and background
it. However, in view of their special discourse status, the backgrounding dis-
course effect of the passive probably cannot arise with arbitrary arguments,
hence there would be no clash with the salience requirement of the depictive.
It seems possible to extend this functional perspective on the licencing of
depictives to cover the restrictions on the construction as well. Another re-
striction on depictives not mentioned so far is the fact that object depictives,
by and large, can only predicate of objects that undergo a change in the event,
whereas subject depictives can be used more freely. Rapoport (1999) points out
the following contrasts:
(69) a. Jonesi cut the bread drunki .
Jones cut the breadi hot i .
b. Jonesi phoned Smith sadi .
Jones phoned Smithi sadi .
 Wilhelm Geuder

Furthermore, not all verbs allow subject depictives. Guémann (1990) points
out the following examples:
(70) a. *The palacei swarmed with enemy troops protectedi by artillery.
b. *The guni shot Mary proppedi against a tree.
c. *The lakei receives water from the river half-fulli .
d. *The medicinei helped Mary undilutedi .

Guémann proposes to distinguish participants by their degree of “involve-

ment” in the situation. The basic idea, adapted from work by Hopper and
Thompson (1980), is that the most salient participants of an event are those
which define the begin and the endpoint of a force transmission schema. Many
authors have advocated an approach that tries to reduce these notions to struc-
tural notions. For instance, Rapoport (1999) proposes that the licencing of
depictives can in general be captured by the structural relation “subject”; to
this end, she distinguishes affected object arguments from inert ones, claim-
ing that they occupy different positions in an abstract syntax that reflects the
lexical decomposition of a predicate:
(71) a. activity b. accomplishment / causative
VP VP

NP V' NP V'
Jones Jones

V NP V VP
pushed Smith cut

NP V'
the bread
V (AP)
(cut) ()

With these structures, any surface subject is a structural subject (even in the
case of unaccusatives) and so is any surface object that is the argument of a
change-of-state predicate. Given that this model does not represent a syntac-
tic surface, it remains to be seen whether the information it expresses is truly
syntactic or whether it is lexical-semantic in nature. I don’t want to go into the
discussion of whether lexical decomposition in the syntax is feasible or empir-
ically superior to a lexical approach (Kiparsky 1997 expresses scepticism with
respect to the first, Rappaport Hovav and Levin 1998 with respect to the sec-
ond point). I would simply like to point out that a functional condition on
 Depictives and transparent adverbs 

depictive licencing is superior since it also encompasses the case of unaccept-

able subject depictives in (70) (the problem is that all those subjects would
have to be removed from a structural subject position in abstract decompo-
sitional syntax, but the lowest complement position is the only non-subject
position available). Furthermore, the passive data cannot be explained in terms
of syntactic structures alone, and finally, the functional condition ties in with
the fact that depictives are restricted to a clause-final position in which they
have to receive stress. This requirement is perfectly natural if depictives serve
to present a property of a salient participant, which must then probably itself
be a salient property.
To reiterate, these considerations are highly speculative, but they would
yield a fairly unified picture according to which both adverbial and depictive
predication are subject to restrictions that relate to their interpretation. The
difference is that the adverbs seem to have access to an argument via thematic
role information while depictives select the target of predication not accord-
ing to information from event concepts, but rather according to functional
conditions. This state of affairs corresponds to the difference in the semantic
representations, as for example in:
(72) She read the invitation perplexed(ly).
a. ∃e [read(e, x1 , the invitation) & ∃s (perplex(s, x2 )& s◦e
b. . . . & Rcause (e,s))]

It is only the adverb which, via the relation R, makes reference to thematic role
information. The underspecified representation of depictives (i.e. (72a) alone),
forces an interpretation procedure which instead makes use of functional and
pragmatic factors to constrain the output in a way that makes the utterance of
a depictive informative and relevant.

. Conclusion

I have started out from the simple view that the distinction between depictives
and adverbs expresses a distinction between predicates of individuals and pred-
icates of events. What turned out to be wrong about this view was mainly its
overly simplistic view of the adverbial meanings. By acknowledging a class of
transparent adverbs that denote states of individuals, too, and that are linked
to the event via a specific causal, motivational (or other) relation, the view has
emerged that the behaviour of depictives can be understood on the basis of
 Wilhelm Geuder

their contrast to these adverbial adjuncts. Depictives are also associated with
the event variable of the clause, but only in terms of temporal cooccurrence.
Due to their lack of any further connection to the event, and due to the fact
that they enter into a contrast with the more specific meanings of adverbs, they
end up with an interpretation as event-independent states (which is stronger
than what is strictly speaking indicated in their semantic repesentation).
The results of this investigation speak in favour of a view that does not
assume a very strict distinction between depictives and adverbs as seman-
tic categories. In terms of semantic types, I have posited only two variants:
adjuncts which are predicates of events, i.e. manner modifiers, and adjuncts
which denote states of their own. The state-denoting adjuncts can be endowed
with a variety of relations that connect it more tightly to the event than just
in terms of temporal cooccurrence. However, the morphological distinction
in English is not between manner-denoting and state-denoting adjuncts, but
rather between adjuncts that are closely related to the event (whether they
denote separate states or just manners) and adjuncts without any type of event-
dependence (depictives). It would be conceivable that in other languages, the
border is drawn elsewhere, and in this respect the present work ties in with
the results of the recent crosslinguistic study by Schultze-Berndt and Him-
melmann (2004). For languages like German with their notorious lack of an
adjective-adverb distinction, the result is that there is no need to impose the
categorisation of adjuncts found in English onto the German data where this
is not supported by morphological distinctions. Transparent adverbs and de-
pictives may in principle also be members of a single undifferentiated semantic
category.

Acknowledgements

This paper has profited from comments on an earlier version provided by two
anonymous reviewers and the editors. The work started as a follow-up on
a chapter of my dissertation (Geuder 2000); earlier versions of this material
have been presented at the University of Leipzig (June 2001) and the Wup-
pertal colloquium on adverbs (November 2001), and parts of the final version
at LAGB in Oxford (September 2003). I want to thank all these audiences for
their questions and comments. Moreover, I want to thank a number of col-
leagues for comments and advice they have given in personal communication
at various stages of this work (some of them long ago!), especially Miriam Butt,
Veronika Ehrich, Graham Katz, Claudia Nohl, Gillian Ramchand, Ben Shaer,
 Depictives and transparent adverbs 

Arnim v. Stechow, Hubert Truckenbrodt, and Matthew Whelpton. This paper

was written with support from the SFB 471 “Variation und Entwicklung im
Lexikon” funded by the Deutsche Forschungsgemeinschaft (DFG).

Notes

. Rapoport (1999), again in a note, treats barefoot as a depictive, without further argumen-
tation.
. The connection between events and tense is made via the temporal trace function τ (e),
which maps an event e onto its runtime. The relation between the runtime of an event and
the reference time of the sentence can vary: in the first clause of (14b) we are dealing with
a perfective aspect, hence the runtime of the event is fully included in the interval t. The
second clause features a state, which, as a rule, merely includes the reference time. So when
a state is linked to some interval, it is not necessarily the maximal phase of the state that is
contained in it, it may rather hold before and after t as well. To put it differently, the past
tense of a state does not create a perfective interpretation. Following Klein (1994), aspect
can be defined as the relation between the time of the event and the reference time; the tense
component is then a relation between the reference time and the utterance time t0 .
. Due to the homomorphism from events to times, the overlap relation can be stated di-
rectly for events, even though it is a “temporal” statement. Using the event variable avoids
the implausible distinction between two sorts of time variables (i.e. reference time and event
time).
. Since the holder of the state tired is implicit this time, I am showing only the VP with
lambda abstraction over both occurrences of the x-argument in order to bring out the fact
that control holds.
. Measure adverbials like for an hour don’t provide unambiguous evidence because their
application to a sum event in connection with TP connectedness as formulated in (20) ends
up truth-conditionally equivalent to modifying the verb alone.
. There is a great deal of variation among speakers with respect to the absolute extent to
which manner adverbs are tolerated in preverbal position, but one never observes prefer-
ences that are the reverse of the judgements in (29b) vs. (29d).
. A reviewer objects that the adverb sadly in this sentence should actually be considered
parenthetical. Such a construction seems indeed frequent, but for me the adverb doesn’t
sound as if it must be assigned a parenthetical status in a strict syntactic sense. (An internet
search shows that roughly half of the attested examples are indeed orthographically marked
as parenthetical, but the other half is not.) It is interesting, though, that the question arises at
all and that there is quite generally some amount of uncertainty among speakers about non-
manner adverbs in such positions. Something that might contribute to the impression of a
parenthetical is the fact that this adverb is a non-restrictive modifier in a position (sentence-
medially after the verb) where it is destressed. Transparent adverbs often resist being stressed.
 Wilhelm Geuder

. The provisional representation (33b) indicates that if transparent adverbs are interpreted
in scenarios that make them appear semantically similar to manner adverbs, this is due to a
“depictive component”, as it were, in their meaning. In the same way, we can expect that the
presence of depictives, too, may rub off on the manner of an event, but only on the level of
inferencing.
. There is also a manner reading available for reluctantly, but it is of course the pure attitude
meaning that is relevant here.
. http://www.geocities.com/ravenmockerv/book1/choose4.html
. Take for example Dowty’s (1991) theory of agents and themes in terms of a collection of
proto-role entailments; or the use of topological schemata in cognitive semantics.
. Guémann (1990: 297) contradicts this generalisation and claims that certain NPs in PPs
are able to control a depictive, e.g. in (i); however, the reviewers of the present paper judge
it to be unacceptable.

(i) John carved on the clayi dampi

I suspect that people who manage to find (i) acceptable do so by treating carve on the clay as
a complex predicate. Maybe this then aligns with a class of adjectival adjuncts that can live
without any syntactically visible controller, such as He bathed hot.
. Cf. Butt and Geuder (2001) for some illustration of this problem.
. Subject and object depictives differ in syntactic position (cf. e.g. Rothstein 2001: 125–
127). My claim that non-structural factors regulate control refers only to the question which
types of control constellations are in principle available; it is not to be denied that different
types of control may be implemented in different syntactic constellations.
. Some speakers even report a very neat contrast between adverbial and adjectival forms of
participles, which is interesting because of the lack of a palpable meaning difference between
the two forms:

(i) He read the invitation delighted / delightedly

(ii) The invitation was read *delighted / ok ... delightedly

However, there is enormous variation among speakers and others don’t get this contrast
at all. I might add that still others even reject the adverbial form delightedly as such (and
require with delight or similar expressions). . .

References

Baker, M. (1997). “Thematic roles and syntactic structure.” In L. Haegeman (Ed.), Elements
of Grammar. Handbook of Generative Syntax (pp. 73–137). Dordrecht: Kluwer.
Bowers, J. (1993). “The syntax of predication.” Linguistic Inquiry, 24, 591–656.
Butt, M., & Geuder, W. (2001). “On the (semi)lexical status of light verbs.” In N. Corver
& H. van Riemsdijk (Eds.), Semi-lexical Categories (pp. 323–370). Berlin: Mouton de
Gruyter.
 Depictives and transparent adverbs 

Déchaine, R.-M. (1993). “Predicates across categories.” PhD dissertation, University of

Massachusetts, Amherst.
Demonte, V. (1988). “Remarks on secondary predicates: c-command, extraction, and
reanalysis.” The Linguistic Review, 6, 1–39.
Dowty, D. (1979). Word Meaning and Montague Grammar. Dordrecht: Reidel.
Dowty, D. (1991). “Thematic Proto-Roles and argument selection.” Language, 67, 547–619.
Ernst, T. (2002). The Syntax of Adjuncts. Cambridge: Cambridge University Press.
Geuder, W. (2000). “Oriented adverbs. Issues in the lexical semantics of event adverbs.”
PhD dissertation, Universität Tübingen, [Available online:
http://w210.ub.uni-tuebingen.de/dbt/volltexte/2002/546].
Guémann, S. (1990). “Secondary predication in English and Spanish.” PhD dissertation,
University of California, Berkeley.
Hopper, P., & Thompson, S. (1980). “Transitivity in grammar and discourse.” Language, 56,
251–99.
Johnston, M. (1994). “The syntax and semantics of adverbial adjuncts.” PhD dissertation,
University of California, Santa Cruz.
Kiparsky, P. (1997). “Remarks on denominal verbs.” In A. Alsina et al. (Eds.), Complex
Predicates (pp. 473–499). Stanford: CSLI.
Klein, W. (1994). Time in Language. London: Routledge.
Kratzer, A. (1988/1995). “Stage-level and individual-level predicates.” In G. Carlson & F. J.
Pelletier (Eds.), The Generic Book (pp. 125–175). Chicago: University of Chicago Press,
[Manuscript from 1988].
Kratzer, A. (2000). “German impersonal pronouns and logophoricity.” Paper presented at
the conference Generic Pronouns and Logophoricity, São Paolo 2000, [Available online
via: http://semanticsarchive.net].
Krifka, M. (1992). “Thematic relations as links between nominal reference and temporal
constitution.” In I. Sag & A. Szabolcsi (Eds.), Lexical Matters (pp. 29–53). Stanford:
CSLI.
Larson, R., & Segal, G. (1995). Knowledge of Meaning: An Introduction to Semantic Theory.
Cambridge, MA: MIT Press.
Maling, J. (2001). “Dative: the heterogeneity of the mapping among morphological Case,
grammatical functions and thematic roles.” Lingua, 111 (4–7), 419–464.
McNally, L. (1994). “Adjunct predicates and the individual/stage distinction.” In E. Duncan
et al. (Eds.), The Proceedings of the Twelfth West Coast Conference on Formal Linguistics
(pp. 561–576). Stanford: CSLI.
Parsons, T. (1990). Events in the Semantics of English. Cambridge, MA: MIT Press.
Pesetsky, D. (1995). Zero Syntax. Cambridge, MA: MIT Press.
Pylkkänen, L. (2002). “Applicatives and depictive secondary predication.” Ms., MIT,
[Available online: http://web.mit.edu/liina/Public/Talks/Depictive-ling-lunch.pdf].
Rapoport, T. (1999). “Structure, aspect, and the predicate.” Language, 75, 653–677.
Rappaport Hovav, M., & Levin, B. (1998). “Building verb meanings.” In M. Butt &
W. Geuder (Eds.), The Projection of Arguments (pp. 97–134). Stanford: CSLI.
Roberts, I. (1987). The Representation of Implicit and Dethematized Subjects. Dordrecht:
Foris.
Roberts, I. (1988). “Predicative APs.” Lingustic Inquiry, 19, 703–710.
 Wilhelm Geuder

Rothstein, S. (2001). Predicates and their Subjects. Dordrecht: Kluwer.

Rothstein, S. (2003). “Secondary predication and aspectual structure.” In E. Lang,
C. Maienborn, & C. Fabricius-Hansen (Eds.), Modifying Adjuncts (pp. 553–590). Berlin:
Mouton de Gruyter.
Schultze-Berndt, E., & Himmelmann, N. (2004). “Depictive secondary predicates in cross-
linguistic perspective.” Linguistic Typology, 8, 59–131.
Williams, E. (1980). “Predication.” Linguistic Inquiry, 11, 203–238.
Williams, E. (1994). Thematic Structure in Syntax. Cambridge, MA: MIT Press.
Winkler, S. (1997). Focus and Secondary Predication. Berlin: Mouton de Gruyter.
Wunderlich, D. (1997). “Argument extension by lexical adjunction.” Journal of Semantics,
14, 95–142.
Wyner, A. (1998). “Subject-oriented adverbs are thematically dependent.” In S. Rothstein
(Ed.), Events and Grammar (pp. 333–348). Dordrecht: Kluwer.
 Degree Phrases versus Quantifier Phrases
in prenominal and preverbal positions
A hybrid explanation for some
distributional asymmetries*

Dagmar Haumann
University of Erfurt

This paper surveys the distribution and licensing of adjectival and adverbial
Degree Phrases, e.g. [DegP so clumsy / clumsily] and Quantifier Phrases, e.g.
[QP less gloomy / gloomily], in verbal and nominal projections. It will emerge
that DegPs and QPs, while equally admissible in right-/left-peripheral
positions within verbal and nominal projections, display unexpected
distributional asymmetries in preverbal and prenominal position (viz. he
{*so / less} clumsily lit a cigarette; a {*so / less} gloomy prognosis). The
conclusion will be that the distributional asymmetries under consideration
do not have a uniform source: the (non)admissibility of preverbal DegPs and
QPs is contingent on the (non)availability of a result clause, while the
(non)admissibility prenominal DegPs and QPs is reliant on the
(non)predicate status of the adjective involved.

. Introduction

Heterogeneity is a trademark of all traditional (sub)classes of adverbs, and the

functionally defined subclass of degree adverbs, e.g. abundantly, as, enough,
exactly, fairly, less, more, so, too, very, well, etc., is no exception. The lowest
common denominator for joint class membership of these elements lies in
their functioning as degree modifiers within adjectival and adverbial expres-
sions, i.e. degree adverbs operate on properties denoted by relative adjectives
and adverbs. And this is where similarities end for the group consisting of as,
enough, less, more, so, and too.
Starting out from Corver’s (1997) Split-DegP Hypothesis, according to
which the above-mentioned degree elements fall into two distinct syntactic
 Dagmar Haumann

categories (Deg(ree) for as, so, and too and Q(uantifier) for enough, less, and
more), this paper addresses the distributional (a)symmetries displayed by ad-
jectival and adverbial projections containing overt Deg and Q heads. Focusing
on the external structure of these constructions, I will show that headed DegPs
and QPs, as well as adjectival and adverbial projections lacking overt Deg and
Q heads, are distributionally equivalent only if they assume non-left-peripheral
positions, i.e. the postnominal and postverbal position. In positions other than
these, two types of asymmetries can be observed. First, headed DegPs and
QPs, as opposed to adjectival and adverbial projections lacking overt Deg and
Q heads, may assume a left-peripheral position within the extended nomi-
nal projection. Secondly, headed DegPs and QPs are not equally admissible
in prenominal and preverbal positions. It will be shown that the distributional
asymmetries displayed by headed DegPs and headed QPs do not have a uni-
form source: the (non)admissibility of headed DegPs and QPs in preverbal po-
sition is subject to the interplay between the (non)availability of result clauses
and the (non)interpretability of Deg and Q, whereas the (non)admissibility of
DegPs and QPs in prenominal position is intimately tied to the (non)predicate
status of the adjective involved.
In the following, I first present the empirical arguments that led to severing
as, enough, less, more, so, and too from the traditional class of degree adverbs
(Section 1.1), and then turn to briefly review Abney’s (1987) DegP Hypothesis
(cf. Section 1.2), according to which these elements are recategorized as func-
tional heads within the extended adjectival and adverbial projection. Section
1.3 focuses on Corver’s (1997) Split-DegP Hypothesis. Section 2 provides an
outline of the theoretical background against which the syntactic analysis of
headed DegPs and QPs is set and introduces the asymmetry puzzle posed by
headed DegPs and QPs in non-right-peripheral positions. In Section 3, the fo-
cus is on the various factors involved in the (non)admissibility of DegPs and
QPs in preverbal and prenominal position. Section 4 provides a conclusion.

. A class of their own

As has been argued by Bresnan (1973: 287ff.), Bowers (1975: 549ff.) and Corver
(1991: 36), as, enough, less, more, so, and too differ from run-of-the-mill degree
adverbs, e.g. abundantly, exactly, fairly, very, well, etc., in that they may – to-
gether with the adjective they operate on – precede the indefinite article, as
shown in (1):
 Degree Phrases versus Quantifier Phrases 

 
(1) a. as beneficial, happy enough 
too simple, less complicated an arrangement
 
more economic, so wonderful
b. *{ abundantly, fairly, very clear } an argument

Moreover, the elements under consideration differ from all other traditional
degree adverbs in that they resist though attraction (cf. Culicover (1980: 15ff.)
and Corver (1991: 36)):
 
(2) a. *as beneficial, ? happy enough 
too simple, less complicated though the arrangement was
 
more economic, so wonderful
b. { abundantly, fairly, very clear } though the argument was

A further systematic difference between as, enough, less, more, so, and too, on
the one hand, and abundantly, exactly, fairly, very, etc., on the other lies in their
(in)ability to license a result clause, as shown in (3) (cf. Bresnan (1973: 316ff.),
Bowers (1975: 530, 545), Abney (1987: 312ff.), and Corver (1991: 39ff.) for
discussion):1
(3) a. At the end of the day, a system is only {as, *fairly good} as the user wants
it to be. (CBT 1688)
b. The actor was {keen enough, *very keen} on the project to commit
himself to the project months ahead. (K54 996)
c. He was {so, *completely drunk} that his words came thickly and with
great difficulty. (B0U 2911)
d. I was {too, *extremely tired} to clean my teeth. (CEX 3057)
e. Suddenly the man seemed {more, less, *fairly attractive} than he had at
home. (Collins Cobuild)

. DegP

On the basis of the differences sketched, Abney (1987: 298ff.) proposes

severing as, enough, less, more, so, and too from the traditional class of degree
adverbs2 and recategorizing these elements as members of the functional
category Deg(ree), which figures in the extended projection of relative adjec-
tives and adverbs, as illustrated in (4):3
(4) a. [DegP [Deg {as, enough, less, more, so, too}] [AP [A tired]]]
b. [DegP [Deg {as, enough, less, more, so, too}] [AdvP [Adv slowly]]]
 Dagmar Haumann

Although Abney’s (1987) analysis properly accounts for the contrasts between
as, enough, less, more, so, and too, on the one hand, and all the other traditional
degree adverbs, on the other (cf. (1)–(3) above), it fails to account for the fact
that more, less, and enough have a much wider distribution than as, so, and too
(cf. Bresnan (1973: 280ff.) for discussion), as illustrated in (5)–(8) (respective
result clauses omitted):4
(5) a. There had not been {enough, *as} evidence. (K5D 4136)
b. This meant they could drink {more, *too}. (B06 231)
c. There are {enough, *so} of us. (EWC 1585)
(6) a. These aspects of language performance are {more, *too} under conscious
control. (CCV 722)
b. She wasn’t up {enough, *so} in foreign words. (AC3 193)
(7) a. Uniforum is still {more, *as } a Unix celebration. (CT3 429)
b. But I wasn’t stupid {enough, *so}. (CEE 344)
(8) a. Those who sleep {less, *so} do so by choice. (EVA 1458)
b. By now you will have stimulated the circulation {enough,*as}.
(B06 1502)

While enough, less, and more may occur in such environments as within an
extended nominal projection, as in (5), within an extended prepositional pro-
jection, as in (6), and as degree modifiers within extended nominal and verbal
projections, as in (7) and (8), as, so, and too are systematically barred from
occurring in these positions.
Moreover, extended adjectival and adverbial projections containing the
elements Abney (1987) categorizes as Deg are not equally admissible in
prenominal and preverbal positions, as shown in (9)–(10) (cf. Section 2 for a
detailed discussion). For (10a), one way of ensuring that the intended reading
of so subtly is being considered is to place stress on so:
(9) a. *A so gloomy prognosis that even John was impressed.
b. A less gloomy prognosis assumes that there is sufficient matter in the
universe... (Collins Cobuild)
(10) a. *She so subtly hinted at the possibility of murder.
b. She subtly enough hinted at the possibility of murder.

The data in (5) to (10) suggest that much, less, and enough, on the one hand,
and as, too, and so, on the other hand, are not members of one category,
despite the fact that these elements share a host of properties, as shown in
(1)–(3) above.
 Degree Phrases versus Quantifier Phrases 

. Split-DegP

Corver (1997) proposes modifying Abney’s (1987) DegP analysis to incorpo-

rate Bresnan’s (1973) contention that more, less, enough, etc. are categorially
distinct from as, so, too, etc. with the former belonging to the functional cat-
egory Q and the latter to the functional category Deg, both of which project
onto a phrasal level, as illustrated in (11) (cf. Corver (1997: 128)):
(11) [DegP [Deg {so, too, as}] [QP [Q [{more, less, enough}] [A(dv)P ]]]

Corver (1997) maintains that adjectival and adverbial expressions, as in (12),

are quantificational, of which more, less, and enough are genuine quantifiers,
whereas so, too, and as are parasitic on the presence of a genuine, albeit pho-
netically empty Q head within the extended adjectival and adverbial projection
(cf. below for discussion):
(12) a. You’re [QP [Q [A young]i [Q enough] [AP [A ti ]] to be dramatic about
everything. (FET 1420)
b. Even here he tends to play [QP [Q less] [AdvP softly]] than he should.
(ED6 2259)
c. She expressed herself [QP [Q more] [AdvP subtly.]] (H97 1414)
d. Half the people were [DegP [Deg as [QP [Q e] [AP ready]]] to drop out as
he was. (BOU 2238)
e. I had just lived through it [DegP [Deg so [QP [Q e] [AP vividly]]] that the
palms of my hands were sweating.] (BOU 2697)
f. But they’d have been [DegP [Deg too [QP [Q e] [AP sloshed]]] to notice.
(G1S 1297)

Evidence for both the distinct functional head status of the elements under
consideration and the structures in (11) and (12) derives from the well-known
fact that so-pronominalization may apply to either the entire extended adjectival
or adverbial projection, as in (13a), or a part of it, as in (13b) (data from Corver
(1997: 126)):
(13) a. John seems too tall to serve on a submarine, and Bill seems so too.
b. The police searched the big room carefully, but the small room [less so.]

As can be seen from the contrasts between the grammatical strings in (13) and
the ungrammatical strings in (14), the availability of so-pronominalization is
closely tied to the presence of more, less, and enough, i.e. the subset of degree
adverbs that make up category Q ((14a) adapted from Corver (1997: 127)):
 Dagmar Haumann

(14) a. *The weather was hot in Cairo – [DegP [Deg so] [AP so]], that we stayed
indoors all day.
b. *The vampire drank the red liquid hastily, probably [DegP [Deg too] [AP
so]] to realize that it was cranberry juice.

Corver (1997: 127) points out that the strings in (14) become acceptable when
the pleonastic quantifier much intervenes between the DegP headed by so and
too and [AP/ AdvP so], as shown in (15). Specifically, he argues that pleonastic
much is inserted into the head position of QP, which, as we have seen in (11)
above, is sandwiched between DegP and A(dv)P ((15a) adapted from Corver
(1997: 127)):
(15) a. The weather was hot in Cairo – [DegP [Deg so] [QP [Q much] [AP so]]]
that we stayed indoors all day.
b. The vampire drank the red liquid hastily, probably [DegP [Deg too][QP[Q
much] [AP so]]] to realize that it was cranberry juice.

Corver’s (1997: 128) account of pleonastic much in (15) as a Last Resort is

driven by the Principle of Full Interpretation, FI, which requires that every
element receive an appropriate interpretation (cf. Chomsky (1986a: 98ff.)).
In the present context, two considerations are relevant. First, the argument
structure of gradable adjectives contains a referential d(egree) argument which
ranges over degrees of properties (cf. Zwarts (1992: 48f., 137ff.)) and which
must be Θ-bound. Second, Deg and Q are operators that quantify over degrees
of properties, and as such they must bind a variable, namely the referential
argument, d, contained in the adjective’s argument structure. Corver (1997)
maintains that Θ-binding of d through Deg or Q is restricted to applying
under strictly local government. This condition is automatically met in (16),
where Q and A are in a local government configuration so that Θ-binding can
apply in situ, as indicated by coindexation of Q and d (adapted from Corver
(1997: 133)):5
(16) He will never be [QP [Q morei ] [AP [A clever<di> ]] than her.

Θ-binding of d amounts to fixing the degree of the property denoted by the

adjective on a scale (cf. Zwarts (1992: 142ff.) and Corver (1997: 132)). Thus,
for example, by binding the referential argument of the adjective intelligent, the
quantifier more in (16) restricts the denotation of the adjective “by picking out
one degree from the set of degrees of intelligence and compares this degree to
the degree of intelligence expressed or implied by the than-phrase” (cf. Corver
(1997: 132)).
 Degree Phrases versus Quantifier Phrases 

In contexts in which Q is phonetically empty, but Deg is not, A must raise to Q

in order to create a strictly local government configuration in which Θ-binding
can apply, as shown in (17) (adapted from Corver (1997: 132f)):
(17)

If A-to-Q raising did not take place, d would not be Θ-bound by Deg because –
under Relativized Minimality (cf. Rizzi (1990)) – [Q e] would be a closer po-
tential governor for A, and thus block Θ-binding of d by Deg, which, in turn,
would leave Deg vacuously quantifying and would thus induce a violation of
FI. Moreover, since [Q e] cannot bind the referential argument of tall, because
it lacks quantificational force, d would be left undischarged and so would give
rise to a violation of FI. Thus, A-to-Q raising serves to overcome a two-fold
violation of FI.
Since so in (15) is a phrasal pro-form replacing the entire AP, this ele-
ment – in accordance with the Uniformity Condition on Chains (cf. Chomsky
(1995: 253)) – is barred from raising to Q, as indicated in (18):
(18)

Corver (1997: 134ff.) suggests that much-insertion, as in (15) above, be viewed

as a Last Resort operation which prevents a two-fold violation of FI.
Under his analysis, much copies the referential argument d of the adjectival
pro-form so, and thus makes d available in the head position of QP, where it
can be accessed by the Θ-binder in Deg, as illustrated in (19) (adapted from
Corver (1997: 135)):
(19) [DegP too / so / as / howi [QP much<di> [AP so<di> ]]]

Corver’s (1997) Split-DegP Hypothesis, in connection with much-insertion,

neatly accounts for the contrasts in acceptability of Q and Deg heads in the
extended nominal, prepositional, and verbal projection in (5)–(8) above.6
While differences in distribution between [Q {enough, less, more}] and [Deg
{as, too, so}] in (5)–(8) above have been widely discussed in the relevant lite-
rature, distributional differences between adjectival and adverbial projections
containing [Q {enough, less, more}] and [Deg {as, too, so}], as in (9) and (10)
above, have largely gone unnoted. As we shall see in the following, this is mainly
due to the fact that – in the data that have been considered – headed DegPs
and QPs occur in right-peripheral positions, such as the predicative position,
 Dagmar Haumann

which are equally accessible to headed DegPs and QPs and also to extended ad-
jectival and adverbial projections that lack overt functional heads (henceforth
[QP [Q e] [A(dv)P ]]), as illustrated in (20):
(20) a. He was [DegP so drunk] that his words came thickly and with great
difficulty. (BOU 2911)
b. In fact, this decision was [QP less revolutionary] than it appeared.
(B0M 432)
c. She was [QP tired.]

Taking the syntax of adjectival and adverbial projections with and without
overt functional heads in right-peripheral positions, i.e. postnominal, postver-
bal, and predicative positions (cf. Section 2.1), as a point of departure, I turn
to the presentation of two types of asymmetries.
The first type of asymmetry to be discussed concerns the distribution
and licensing of adjectival projections in a left-peripheral position within the
extended nominal projection (cf. Section 2.2). As can be seen in (21), this
position is accessible only to headed DegPs and QPs:
(21) a. He found the bright lights of the big city [DegP too attractive] a prospect
to refuse. (B0L 940)
b. You’re [QP good enough] an actress to be invited to our Fenice theatre.
(H94 2000)
c. *He found the bright lights of the big city [QP attractive] a prospect to
refuse.

The asymmetry displayed by headed DegPs and QPs vs. adjectival projections
that lack an overt functional head will be shown to be intimately tied to quan-
tification: only extended adjectival and adverbial projections that contain an
overt functional head, i.e. Deg or Q, are quantificational and thus susceptible
to raising from a DP-internal predicative position to the left-peripheral scope
position within the extended nominal projection.
The second type of asymmetry is constituted by headed DegPs vs. headed
QPs and [QP [Q e][A(dv)P ]] in prenominal and preverbal positions, as shown in
(22) and (23) (cf. Section 2.3):
(22) a. *Many teachers do not have a [DegP so strong] grasp of the language that
they could teach in it effectively.
b. Many teachers do not have a [QP strong enough] grasp of the language
to teach in it effectively. (B12 691)
c. Many teachers do not have a [QP strong] grasp of the language.
 Degree Phrases versus Quantifier Phrases 

(23) a. *She [DegP so subtly] hinted at the possibility of murder.

b. She [QP subtly enough] hinted at the possibility of murder.
c. She [QP subtly] hinted at the possibility of murder.

The asymmetries in (22) and (23), of which the asymmetries between headed
DegPs and QPs are at heart of my analysis, will be shown to have a nonuniform
source (cf. Section 3). I will show that the nonadmissibility of headed DegPs
in preverbal position derives from the interpretive requirements of as, so, and
too, which cannot be met in the configurations under consideration (cf. Sec-
tion 3.1), while the nonadmissibility of headed DegPs in prenominal position
derives from the general inability of as, so, and too to figure in the extended
projection of nonpredicative adjectives (cf. Section 3.2).

. The distribution and licensing of adjectival and adverbial projections

containing headed DegPs and QPs

. Headed DegP and QP in right-peripheral positions

Headed DegPs and QPs are generally acceptable in right-peripheral positions,

i.e. in predicative position, as in (24), in postnominal position, as in (25), and
in postverbal position, as in (26):
(24) a. He was [DegP so drunk] that his words came thickly and with great
difficulty. (BOU 2911)
b. In fact, this decision was [QP less revolutionary] than it appeared.
(B0M 432)
(25) a. [...] covering subjects [DegP as diverse] as nuclear waste disposal and the
Irish fears of an accident at Hinkley. (AN9 1380)
b. [It] could be written on a piece of paper [QP small enough] to fit in a
walnut. (CBC 8209)
(26) a. The violins had been taking the tempo [DegP slightly too fast.] (B11 1398)
b. He was speaking [QP very much more quietly] than when he first came
in. (BOU 363)

In these positions, which will be discussed in turn, headed DegPs and QPs
are subject to the same licensing conditions as [QP [Q e][A(dv)P ]], i.e. extended
adjectival and adverbial projections lacking overt Deg and Q heads, as in (27):
(27) a. The service was [QP [Q e] [AP lousy.]]
b. I’ve never met a pathologist [QP [Q e] [AP cheerful about her work.]]
 Dagmar Haumann

c. He complained [QP [Q e] [AdvP vociferously]] about the meal.

.. Predicative position

The predicative position is assumed by the extended projections of predica-
tive adjectives, i.e. adjectives whose argument structure contains an external
argument which is structurally realized in the specifier position of AP, as
shown in (28):
(28) a. . . . [DegP [Deg so] . . . [AP [D he] [A’ [A drunk]]]]]
b. . . . [QP [Q less] [AP [DP this decision] [A’ [A revolutionary]]]]
c. . . . [QP [Q e] [AP [D she] [A’ [A tired]]]]]]

The external argument of the adjective is a phonetically realized DP which

raises overtly to the sentential subject position for Case checking purposes, as
shown in (29):
(29) a. [AgrP [D he]i . . .was. . . [DegP [Deg so]. . . [AP ti [A’ [A drunk]]]]]
b. [AgrP [D this decision]i. . .was. . . [QP [Q less] [AP t i [A’ [A revolution-
ary]]]]]
c. [AgrP [D she]i . . . was . . . [QP [Q e] [AP t i [A’ [A tired]]]]]

.. Postnominal position

Since adjectival projections in postnominal position pattern like predica-
tive and unlike attributive adjectival projections, I suggest analyzing them as
secondary predicates, i.e. reduced relatives, that take a PRO subject, as shown
in (30):7
(30) a. . . . [DegP [Deg too] . . . [AP [D PRO] [A’ [A fond] of their children]]]
b. . . . [QP [Q less] [AP [D PRO] [A’ [A qualified]]]]
c. . . . [QP [Q e] [AP [D PRO] [A’ [A responsible for this]]]]

Evidence for the predicative, rather than attributive nature of postnominal ad-
jectival projections derives from the fact that all and only adjectival projections
which also occur in post-copula position may occur postnominally. Adjectives
such as former, seeming, and sheer in (31) are strictly prenominal, i.e. they are
systematically barred from predicative positions (see also Cinque (1994: 94f.)):
(31) a. *Her husband (is) former.
b. *The paradox (is) seeming.
c. *Denial (is) sheer.
 Degree Phrases versus Quantifier Phrases 

Moreover, postnominal adjectival projections differ from their prenominal

counterparts in interpretation. While prenominal, i.e. attributive, adjectives
can receive an individual-level interpretation (i.e. they express inherent prop-
erties), complementless postnominal adjectives receive a stage-level interpreta-
tion (i.e. they express temporal or accidental properties), as can be read off the
paraphrases in (32) and (33) (adapted from Sadler & Arnold (1994: 192)):8
(32) We need to find a responsible person.
= person inherently responsible (i.e. trustworthy)
(33) We need to find the person responsible.
= person that is to blame

A further argument in favor of analyzing postnominal adjectival projections

as predicates rather than as attributes derives from the fact that both predica-
tive and postnominal adjectival projections in German, as in (34), lack overt
agreement morphology, whereas prenominal adjectives are overtly marked for
number and gender, as shown in (35) (case marking omitted):9
(34) a. eine Studentin (ist) ihrer Arbeit überdrüssig
a(sg, fem) student(sg, fem) is her(sg, fem) work fed-up-withØ
*überdrüssige
*fed-up-with(fem,sg)
‘a student fed up with her work’
b. eine Studentin (ist) stolz auf ihre Mutter
a(sg, fem) student(sg, fem) (is) proudØ on her(sg,fem) mother
*stolze
*proud(sg, fem)
‘a student proud of her mother’
(35) a. eine ihrer Arbeit überdrüssige Studentin
a(sg, fem) her(sg, fem) work fed-up-with(sg, fem) student(sg, fem)
*überdrüssig
*fed-up-withØ
‘a student fed up with her work’
b. eine auf ihre Mutter stolze Studentin
a(sg, fem) on her(sg, fem) mother proud(sg, fem) student(sg, fem)
*stolz
*proudØ
‘a student proud of her mother’

As for the accommodation of the adjectival projections in (30), I adopt Koster’s

(1999) parallel construal analysis for relative clauses which allows for expand-
 Dagmar Haumann

ing structures “to the right with asyndetic specifications” (cf. Koster (1999: 4)).
Structure expansion is achieved by merging into the structure the invisible
or transparent functional projection ConjP, whose specifier position hosts the
relativized constituent, i.e. the antecedent of the relative clause (as indicated
by coindexation). The relative clause, i.e. the secondary adjectival predicate, is
realized as the complement within ConjP, as shown in (36):
(36) ConjP

Conj'

DegP/QP

DP Conj Deg/Q AP

D A'

A ...

[ parents]i too [ PRO]i fond of their children

[ people]i less [ PRO]i qualified
[ the person]i responsiblej [ PRO]i tj for this

.. Postverbal position

The core case of postverbal adverbial projections headed by relative adverbs is
presented in (37):
(37) a. [He] had thrown back his head and was snoring [DegP so loudly] that all
the neighbors could hear him.
b. Even here he tends to play [QP less softly] than he should.

While Alexiadou (1997: 129ff.) takes extended projections of manner adverbs

to originate as complements within VP, as shown in (38), Cinque (1999: 19ff.)
and Laenzlinger (1996: 115f.) maintain that the projections of manner ad-
verbs are generated in a preverbal specifier position10 and that postverbal oc-
currences of extended projections headed by manner adverbs are “illusions”
created by overt verb movement across the specifier position hosting the afore-
mentioned projections, as shown in (39):
(38) a. He was . . . [VP [V snoring] [DegP [Deg so] [AdvP loudly.]]]
b. He tends to . . . [VP [V play] [QP [Q less] [AdvP softly.]]]
 Degree Phrases versus Quantifier Phrases 

c. He . . . [VP [V rose] [QP [Q e] [AdvP carefully.]]]

(39) a. He was [. . . [V snoring]i . . . [[DegP so loudly] . . . [[VP [V t i ]]]]
b. He tends to [. . . [V play]i . . . [[QP less softly] . . . [[VP [V t i ]]]]
c. He [. . . [V rose]i . . . [[QP [Q e] [AdvP carefully.]]]]

Crucially, under both types of analyses, the extended projections of manner ad-
verbs are formally licensed under specifier head agreement with an appropriate
functional head in the extended verbal projection, either at LF, as under Alex-
iadou’s (1997) analysis, or in situ, as under Cinque’s (1999) and Laenzlinger’s
(1996) analyses. For the purpose of this paper, I subscribe to Alexiadou’s (1997)
analysis since it provides a straightforward account of the data in (37) without
stipulating additional movement operations (cf. Section 2.3.2 for refinements).

. Raising from the right to the left periphery within extended nominal
projections – The Big Mess Construction11

In addition to assuming right-peripheral positions, as in (24) and (26) above,

the extended projections of predicative adjectives containing headed DegPs and
QPs may precede the indefinite article within predicative DPs, as shown in (40),
whereas [QP [Q e][AP ]], as shown in (41), is systematically barred from this
position:12
(40) a. They would also learn to respect him and trust him, if he is [DP [DegP as
good] a leader] as you say. (CM7 2206)
b. He found the bright lights of the big city [DP [DegP too attractive] a
prospect] to refuse. (B0L 940)
c. The knowledge of books is [DP [DegP so important] a requirement] [...]
that an organisation of staff based on bibliography seems self-evident.
(B07 1503)
d. You’re [DP [QP good enough] an actress] to be invited to our Fenice
theatre. (H94 2000)
e. It turned out to be [DP [QP less ridiculous] an idea] than we thought.
f. A pity Rahmi is [DP [QP so much more attractive] a man] than this cop.
(CM7 474)
(41) a. *They would also learn to respect him and trust him, if he is [DP [QP e
good] a leader.]
b. *He found the bright lights of the big city [DP [QP e attractive] a prospect.]
c. *The knowledge of books is [DP [QP e important] a requirement.]
d. *You’re [DP [QP e good] an actress.]
e. *It turned out to be [DP [QP e ridiculous] an idea.]
 Dagmar Haumann

f. *A pity Rahmi is [DP [QP so much e attractive] a man.]

I take the strings in (40) to be derived from underlying relative clause con-
structions as in (36) above: headed DegPs and QPs encapsulating an AP whose
specifier position hosts PRO undergo predicate raising within the extended
nominal projection, as schematically indicated in (42):13
(42) a. [??? [DegP too attractive]i [ConjP [DP [D a] [NP prospect]] [Conj’ [Conj ]
[DegP ti ]]]]
b. [??? [QP good enough]i [ConjP [DP [D an] [NP actress]] [Conj’ [Conj ] [QP
ti ]]]]

Following Zamparelli (2000), I assume that predicate raising, i.e. DegP and
QP raising across the indefinite article, targets a specifier position within the
extended nominal projection. Under Zamparelli’s (2000: 117ff.) analysis, DP is
split into at least two functional layers, as shown in (43):
(43) [SDP [SD {every, both, the}] [PDP [PD {few, many, a}] . . . [NP ]]]]

The topmost layer, i.e. Strong Determiner Phrase (SDP), hosts so-called strong
quantifiers, e.g. every, both, each, the, this, these, most, ’s, whereas the Predicate
Determiner Phrase (PDP) hosts weak quantifiers, e.g. numerals, few, many, a,
several.14
Zamparelli (2000: 99) proposes analyzing the big mess construction in
(40) as in (44), with the head of PDP taking an adjectival small clause as its
complement (adapted from Zamparelli (2000: 99, 146f.)):
(44) a. [SDP [SD ] [PDP [PD a] [DegP [NP prospect] [DegP too attractive]]]]]
b. [SDP [SD ] [PDP [PD an] [QP [NP actress] [QP good enough]]]]]

The strings in (40) are derived from the structures in (44) by optional predicate
raising to specSDP, as shown in (45) (cf. Zamparelli (2000: 144ff.) for a detailed
discussion):
(45) a. [SDP [DegP too attractive]i [SD’ [SD ] [PDP [PD a] [DegP [NP prospect]
[DegP t i ]]]]]]]]
b. [SDP [QP good enough]i [SD’ [SD ] [PDP [PD an] [QP [NP actress]
[QP t i ]]]]]]]]

According to Zamparelli (2000: 144ff), predicate raising to specSDP is triggered

by the fact that DegP and QP are quantifying expressions and as such must be
in a position from which they can take scope over both the entire nominal pro-
jection and the result clause. Crucially, Zamparelli (2000: 146) takes the ability
 Degree Phrases versus Quantifier Phrases 

of headed DegPs and QPs to license a result clause to be a prerequisite for rais-
ing to specSDP. Under this premise, the fact that [QP [Q e][AP ]] in (41) above
is systematically barred from occurring in specSDP comes as a natural con-
sequence, i.e. empty headed QPs are neither quantificational in the relevant
sense, nor do they license result clauses.
While Zamparelli’s (2000) proposal adequately captures the fact that post-
nominal adjectival projections are secondary predicates which – if they contain
an overt Deg or Q head – are susceptible to raising, the structural analysis
of adjectival predicates as small clauses with an overt subject is problematic.
There are two initially plausible implementations of the idea that adjectival
predicates are small clauses with an overt subject, but neither of them is sat-
isfactory. Under the assumption that the external argument of the adjective is
base-adjoined to the extended adjectival projection, as the structure in (44)
suggests, predicate raising cannot be accounted for, since the lower DegP/QP
has the status of a segment, and as such is precluded from partaking in syntac-
tic movement operations. An alternative would be to assume that the external
argument of the adjective originates in specAP. Here we would have to stipulate
that [NP prospect] and [NP actress] leave the extended adjectival projection and
raise to some specifier position within the functional architecture dominating
DegP/QP prior to predicate raising, as in (45).
At this point, I would like to come back to the analysis of postnominal
adjectival projections presented in Section 2.1.2. Under this analysis, post-
nominal adjectival projections are secondary predicates by virtue of specAP
hosting PRO, and by PRO taking the lexically overt DP as its antecedent. The
analysis of secondary predication presented in Section 2.1.2 can be easily rec-
onciled with Zamparelli’s (2000) Split-DP Analysis if we assume that ConjP is
merged into structure as the complement of the highest functional head within
the split DP, i.e. SDP, and that DegP/QP raises to specSDP to take quantifica-
tional scope over both the entire nominal projection and the result clause, as
shown in (46):
 Dagmar Haumann

(46) SDP

SD'

SD ConjP

PDP Conj'

PD NP Conj DegP/QP

Deg/Q AP

D A'

a. a prospect [Degtoo] PRO attractive

b. an actress [Q[Agood]j[Qenough]] PRO tj

This analysis allows us – without additional stipulations – to maintain that

all postnominal adjectival projections are predicative in nature, and that only
those adjectival projections containing a headed DegP/QP are susceptible to
raising to specSDP.

. Distributional asymmetries between QPs and DegPs

The data considered so far suggest that headed DegP and QP as extended ad-
jectival and adverbial projections are identical in distribution. However, if we
take into consideration non-right-peripheral positions, such as the prenominal
and preverbal position, this nice picture gets blurred.

.. Prenominal position

Prenominal adjectives occupy specifier positions of rigidly ordered agreement
related functional projections (AgrP) within the extended nominal projection,
as shown in (47a) for event nominals and in (47b) for object denoting nouns
(cf. Cinque (1994: 92ff.)).15 Crucially, prenominal APs are merged into the
specifier positions they occupy (i.e. they do not raise there from postnomi-
nal positions as argued by Kayne (1994: 100ff.) and Alexiadou (1997: 217ff.)),
and they are licensed under specifier-head agreement with an appropriate
functional head:
 Degree Phrases versus Quantifier Phrases 

(47) a. [PDP [PD a] [AgrP [AP sudden] [Agr’ [Agr ] . . . [AgrP [AP hideous] [Agr’ [Agr ]
. . . [NP invasion of her privacy]]]]]]]
b. [PDP [PD ] [AgrP [AP huge ] [Agr’ [Agr ] ... [AgrP [AP yellow ] [Agr’ [Agr ] ...
[NP teeth]]]]]]

As can be seen from the following contrasts, headed DegPs, as in (48), are gene-
rally barred from occurring in prenominal position, while headed QPs, as in
(49), are admissible:16
(48) a. *An [AgrP [DegP as solid] [Agr’ ... [NP door]]] gave way.
b. *I met a [AgrP [DegP so good ] [Agr’ ... [NP student.]]]
c. *She bought a [AgrP [DegP too big] [Agr’ ... [NP van.]]]
(49) a. Many teachers do not have a [AgrP [QP strong enough] [Agr’ ...[NP grasp
of the language] to teach in it effectively. (B12 691)
b. A [AgrP [QP less gloomy] [Agr’ ... [NP prognosis] assumes that there is
sufficient matter in the universe...
c. Costume plays a [AgrP [QP more important] [Agr’ ... [NP role for Ander-
son] than for us. (ED9 408)

.. Preverbal position

Extended adverbial projections that assume preverbal positions are of two
types: specifier-type adverbs and raised complement-type adverbs, e.g. man-
ner adverbs, as in (51) (cf. Section 2.1.3). While the former, like prenominal
adjectives in the extended nominal projection, are merged into the struc-
ture as specifiers of rigidly ordered agreement-related functional projections
(AgrP) within the extended verbal projection, as shown in (50), the latter, iff
noncomplex (cf. Alexiadou (1997)), raise to preverbal specifier positions, as
schematically indicated in (51):
(50) a. He [AgrP [AdvP often] [Agr’ ... [VP pretends to be stupid.]]]
b. He [AgrP [AdvP deliberately] [Agr’ ... [VP lit his cigarette clumsily.]]]
(51) a. He [AgrP [AdvP clumsily]i [Agr’ ... [VP opened the jar [AdvP t i ]]]]
b. He [AgrP [AdvP vociferously]i [Agr’ ... [VP denied the claim [AdvP t i ]]]]
(K35 723)

Complex complement-type adverbial projections, as in (52), are barred from

raising overtly (under the intended reading, the boldface strings are manner
adverbials):17
(52) a. *He [AgrP [DegP so clumsily]i [Agr’ . . . [VP opened the jar [AdvP t i ]]]]
b. *He [AgrP [QP less clumsily]i [Agr’ ... [VP opened the jar [AdvP t i ]]]
 Dagmar Haumann

c. *He [AgrP [QP very clumsily]i [Agr’ . . . [VP opened the jar [AdvP t i ]]]

According to Alexiadou (1997: 41, 139ff.), the unacceptability of strings like

(52) derives from complexity restrictions, i.e. only noncomplex and unstressed
complement-type adverbial projections as in (51) may raise overtly to a pre-
verbal specifier position (see also Chomsky (1995: 334) for discussion). How-
ever, headed DegP and QPs, as in (52a) and (52b), as well as QPs containing
specifier-type degree modifiers, as in (52c), are complex by definition, and thus
cannot raise overtly.
While neatly accounting for the contrasts in acceptability between (51) and
(52) under the manner reading of the adverbial projection, the complexity vs.
noncomplexity distinction cannot be employed as a diagnostic for the contrasts
between (53) and (54) since the extended adverbial projections occupy base-
generated specifier positions within the extended verbal projection, i.e. neither
the headed DegPs nor the headed QPs originate as complements within VP.
Headed DegPs as in (53) are systematically barred from occurring in preverbal
position, whereas headed QPs, as shown in (54), are generally admissible.18
Notice that, under the intended reading, the extended adverbial projections in
(53) and (54) are subject-oriented, and they are merged into the structure as
specifiers:
(53) a. *[They] will [AgrP [DegP as carefully] ] [Agr’ ... [VP test the completeness of
the text.]]]
b. *He [AgrP [DegP so clumsily] [Agr’ ... [VP lit a cigarette.]]]
c. *Then he [AgrP [DegP too cleverly] [Agr’ ... [VP started talking about love.]]]
(54) a. [He] will [AgrP [QP more carefully] [Agr’ ... [VP test the completeness of
the text.]]]
b. He [AgrP [QP less clumsily] [Agr’ ... [VP lit a cigarette.]
c. Then he [AgrP [QP subtly enough] [Agr’ ... [VP started talking gently
about love.]

In the following section, I shall focus on the factors involved in the unex-
pected asymmetry between headed DegPs and headed QPs in prenominal and
preverbal positions.

. Towards an explanation

The asymmetry puzzle posed by headed DegPs and headed QPs in prenominal
and preverbal positions will be shown not to have a uniform solution. While
 Degree Phrases versus Quantifier Phrases 

the nonadmissibility of headed DegP vs. the admissibility of headed QPs in

preverbal position, as in (53) and (54), is essentially due to the requirement of
that DegPs in extended adverbial projections be associated with an overt result
clause, which cannot be met in preverbal position (cf. Section 3.1), the nonad-
missibility of headed DegPs vs. the admissibility of headed QPs in prenominal
position, as in (48) and (49), is intimately tied to the predicative nature of the
adjectival nucleus heading DegP (cf. Section 3.2).

. The syntax of result clauses and the interpretation of DegP and QP

In this section, I will show that the nonadmissibility of headed DegPs in prever-
bal position ultimately derives from a conflict between the interpretive require-
ment of [Deg {as, so, too}] to be accompanied by a result clause and syntactic
reality, which precludes the licensing of result clauses by preverbal DegPs (cf.
Section 3.1.2).
Since my account of the nonadmissibility of headed DegPs in preverbal
position rests on the assumption that the interpretive requirements of DegPs
can only be met if they are associated with a result clause, a review of the syntax
and licensing of result clauses is in order.

.. The syntax and licensing of result clauses

As we have seen in Section 1, Deg and Q are restricted to occurring in the
extended projections of relative adjectives and adverbs and, in addition to that,
qualify as licensers of result clauses. To account for these facts, most authors
take these elements to be combined into one constituent, i.e. DegP or QP.19 In
this section, I will show that analyses of this type, e.g. Abney (1987) and White
(1997), are misguided because they are exclusively built on data as in (55),
where headed DegPs and QPs occur in right-peripheral positions and where
the result clause is right-adjacent to A(dv)P:
(55) a. He was [DegP so drunk] that his words came thickly and with great
difficulty. (BOU 2911)
b. He was speaking [QP very much more quietly] than when he first came
in. (BOU 363)

In addition to being oblivious to the asymmetry puzzle under consideration,

these analyses fail to account for the empirical fact that result clauses are not
tied to occurring right-adjacent to A(dv)P, as shown in (56):
 Dagmar Haumann

(56) a. Many teachers do not have a [QP strong enough] grasp of the language
to teach in it effectively. (B12 691)
b. Costume plays a [QP more important] role for Anderson than for us.
(ED9 408)

The discussion of Abney’s (1987) and White’s (1997) analyses in Section 3.1.1.1
is followed by a brief outline of Rijkhoek’s (1998) proposal according to which
result clauses are not contained within DegP and QP, but are licensed as second
conjuncts in asymmetric conjunction structures (cf. Section 3.1.1.2).

... Result clauses within DegP and QP. Abney (1987: 314ff.) proposes an-
alyzing both A(dv)P and result clauses as complements of Deg, as shown in
(57). According to Abney (1987: 315), this analysis allows us to “generate the
extent clause as sister to the degree word in its surface position” and to do away
with extraposition:20
(57) a. He was so drunk that his words came thickly. . . (BOU 2911)
b. [DegP [Deg so] [AP [A drunk]] [CP that his words came thickly]]

Besides being ternary-branching, the structure in (57) runs counter to the over-
all conception of functional heads, which entails that functional heads take
only one complement with which they form an extended projection and on
which they are (morphologically) dependent (cf. Abney (1987: 64f.)). More-
over, this analysis is problematic with respect to strings as in (58), where the
result clause is not adjacent to AP:
(58) a. A less gloomy prognosis than the one advanced by John assumes that
there is sufficient matter in the universe. (Collins Cobuild)
b. It turned out to be less ridiculous an idea than we thought.

To account for the strings in (58), we would either have to subscribe to an

extraposition analysis or to Abney’s (1987: 322ff.) ternary-branching analysis
in conjunction with his analysis of prenominal and prearticle adjectival pro-
jections, according to which the adjective takes the NP or DP it modifies as its
complement (cf. note 13), as shown in (59):21
 Degree Phrases versus Quantifier Phrases 

(59) a. DP

D QP

Q AP CP

A NP

a less gloomy prognosis than the one advanced by John

b.
QP

Q AP CP

A DP

less ridiculous an idea than we thought

In addition to allowing for functional heads to take two structural comple-

ments, Abney’s (1987: 322ff.) analysis of prenominal adjectives in (59) is prob-
lematic since it forces us to assume that the adjective displays some of the
properties of functional heads, such as f(unctional)-selectional properties and
feature parasitism on their respective structural complement. However, the fact
that adjectives allow for multiple selection, e.g. NP-complements as in (59) and
AP-complements as in (60), undermines the idea of f-selection (cf. Sadler &
Arnold (1994: 202f.):
(60) [DP [D a] [AP [A poisonous] [AP [A big] [AP [A black] [NP [N spider]]]]]]

The assumption that as, enough, less, more, so, and too select both A(dv)P
and a result clause is also at the heart of White’s (1997) proposal. How-
ever, his analysis differs crucially from Abney’s (1987) in that the functional
head Deg projects Larsonian shells. As illustrated in (61b), A(dv)P is gene-
rated in specDegP and CP as the complement of Deg. The derivation of the
string in (61a) involves overt raising and adjunction of Deg2 to Deg1 , as il-
lustrated in (61c):
 Dagmar Haumann

(61) DegP1

Deg1 DegP2

AP Deg'2

Deg2 CP

b. drunk so that his words came thickly

c. so drunk ti that his words came thickly

White’s (1997) analysis, like Abney’s (1987), is problematic with respect to the
strings in (58) above, where the result clause is not adjacent to AP, i.e. the
derivation of these strings would have to involve extraposition. In addition
to these problems, White’s (1997) analysis does not allow for the functional
head to Θ-bind the referential argument of the relative adjective. Moreover, as
White (1997: 326) points out himself, there is neither independent evidence for
the presence of the outer DegP shell, i.e. DegP1 , nor is there any obvious trigger
for Deg2 -to-Deg1 raising as in (61c).
Differences aside, the analyses sketched above fail to properly account for
the distribution and licensing of result clauses in contexts in which the ex-
tended adjectival or adverbial projection assumes a non-right-peripheral po-
sition, i.e. the prenominal position in (58a) and the pre-article position in
(58b). Moreover, these analyses do not account for the empirical fact that
headed DegPs and headed QPs are not equally admissible in positions other
than right-peripheral positions and those derivationally related to them (cf.
Section 2.3).
These problems primarily result from the premise that result clauses ori-
ginate within DegP and QP. What is called for is an analysis which allows us
to capture the two relations into which Deg and Q enter (i.e. the Θ-binding
relation with A and Adv and the relation with the result clause) without forcing
us to assume that result clauses are contained within the extended adjectival or
adverbial projection (cf. note 19 for previous proposals along these lines).

... Result clauses as second conjuncts. An analysis which, with slight modi-
fications,22 allows for both capturing the empirical facts and keeping syntactic
operations at a minimum is that presented by Rijkhoek (1998). Under her
analysis, result clauses are conceived of as second conjuncts in asymmetric con-
junction structures, as illustrated in (62) (adapted from Rijkhoek (1998: 123)):
 Degree Phrases versus Quantifier Phrases 

(62) ConjP

DegP Conj'

Deg AP Conj CP

so drunk that his words came thickly

The structure in (62) is reminiscent of that which Koster (1999) assumes for
relative clauses (cf. Section 2.1.2 above), i.e. ConjP is transparent “in the sense
that it behaves as if it were the constituent in its specifier position” (cf. Rijkhoek
(1998: 123)). In addition to allowing for the right-expansion of structure (cf.
Koster (1999: 4)), ConjP functions as a mediator between the first and the
second conjunct, i.e. between DegP or QP in specConjP and the result clause
complement, in the sense that the head of ConjP both enters into a specifier-
head agreement relation with the constituent containing DegP or QP and
selects the type of clause that the degree element would select (cf. Rijkhoek
(1998: 125)).23 The assumption that ConjP is transparent in the above-stated
sense not only allows for merging it into structure at various levels, as shown in
full in (63) for extended adjectival and adverbial projections in right-peripheral
positions, but also accounts for the nonadjacency of the extended adjecti-
val projection containing a headed QP and the result clause, as illustrated in
full in (64):
 Dagmar Haumann

(63) a. AgrP

D Agr'

Agr TP

T ConjP

DegP Conj'

Deg AP Conj CP

D A'

hei wasj tj so ti drunk that his words came out thickly

b.
T'

T VP

V ConjP

QP Conj'

Q AdvP Conj CP

Adv

he tends to play less softly than he should

 Degree Phrases versus Quantifier Phrases 

c.
VP

V ConjPresult clause

PP Conj'

P ConjPreduced relative Conj CP

DP Conj'

Conj QP

Q AP

D A

opposed to people less PRO qualified than doctors delivering

health care

(64) AgrP

ConjP Agr'

DP ConjP Agr ...

D AgrP Conj CP

QP Agr'

Q AP Agr N

a less gloomy prognosis than ... assumes that ...

As can be seen in (63) and (64), ConjP is distributionally equivalent to the

constituent contained in its specifier position, i.e. ConjP is licensed wherever
the constituent contained in its specifier position is formally licensed, e.g. in
post-copula position in (63a), as the complement of V in (63b), in specAgrP
in (64a), or in postnominal position in (63c) and (64b), i.e. within a reduced
 Dagmar Haumann

relative. Notice that in (63c) and (64b), the result clause is licensed prior to
DegP raising to specSDP.
Rijkhoek’s (1998) analysis is clearly to be preferred over analyses which take
result clauses as originating within DegP and QP, since it allows for capturing
the empirical facts without stipulating otherwise unmotivated and undesired
movement operations. However, there are some problems left unsolved by
Rijkhoek’s (1998) analysis as it stands. One problem is that it does not solve
the asymmetry puzzle posed by headed DegPs vs. headed QPs in prenominal
and preverbal positions, as illustrated in (65) and (66) (cf. Section 2.3):
(65) a. *Many teachers do not have a [AgrP [DegP so strong] [Agr’ . . . [NP grasp of
the language.]]]
b. Many teachers do not have a [AgrP [QP strong enough] [Agr’ . . . [NP grasp
of the language.]]]
(66) a. *He [AgrP [DegP too clumsily] [Agr’ ... [VP lit a cigarette.]]]
b. He [AgrP [QP less clumsily] [Agr’ ... [VP lit a cigarette.]

Moreover, under the assumption that ConjP can freely be merged into the
structure, the unacceptability of the strings in (67) comes as a surprise (subtly
and clumsily, under the intended reading, have subject-orientation):
(67) a. *Then he subtly enough started talking gently about love to be made out
a loony.
b. *He { less / more } clumsily lit a cigarette than his girlfriend.

Note that in the ungrammatical sentence-final occurrences of the result clauses

in (67), ConjP assumes a relatively high position within the extended verbal
projection, i.e. the constituent hosted in specConjP must correspond to the
entire matrix clause, as shown in (68) for (67a):
(68) *[Conj [AgrP [D He] [Agr’ ] [Agr ] . . . [AgrP [QP subtly enough] [Agr’ [Agr ]
[VP started talking gently about love]]] [Conj’ [Conj ] [CP to be made out a
loony.]]]]]

However, the antecedent of the result clause, i.e. QP, is too deeply embedded
within specConjP to qualify as the licenser of the result clause.24 The only struc-
tural configuration in which QP in specFP could license a result clause is that
in (69). However, under this scenario the result clause would be realized in a
preverbal position, and not in a sentence-final one, thus giving rise to a vio-
lation of what Haider (2000: 99) dubs the Edge Effect, i.e. preverbal adverbial
projections must be head-final:25
 Degree Phrases versus Quantifier Phrases 

(69) *[AgrP [D He] [Agr’ ] [Agr ] . . . [AgrP [Conj [QP subtly enough] [Conj’ [Conj ] [CP to
be made out a loony]]] [Agr’ [Agr ] [VP started talking gently about love.]]]]]

.. (Non)Interpretability of DegP and QP

Depth of embedding, as in (68) above, would also seem to account for the
unacceptability of the strings in (70), where DegP in preverbal position fails to
license a sentence-final result clause:
(70) a. *Then he as subtly started talking gently about love as John had done
before.
b. *He so clumsily lit a cigarette that he burnt his nose.
c. *The vampire too hastily drank the red liquid to realize that it was
cranberry juice.

However, depth of embedding cannot be employed as a criterion in account-

ing for the contrasts in acceptability between (71) and (72), where neither the
preverbal DegPs nor the preverbal QPs are accompanied by result clauses:
(71) a. *Then he [AgrP [DegP as subtly] [Agr’ ... [VP started talking gently about
love.]]]
b. *He [AgrP [DegP so clumsily] [Agr’ ... [VP lit a cigarette.]]]
c. *The vampire [AgrP [DegP too hastily] [Agr’ ... [VP drank the red liquid.]
(72) a. Then he [AgrP [QP subtly enough] [Agr’ ... [VP started talking gently
about love.]
b. He [AgrP [QP {less / more} clumsily] [Agr’ ... [VP lit a cigarette.]

The contrasts in acceptability, despite the structural identity of (71) and (72),
suggest that the interpretive requirements of [Deg {as, so, too}] are not met in
the configurations under consideration.
As we have seen in Section 1.3, relative adjectives and adverbs are associ-
ated with argument structures that contain a referential argument, d, which
must be Θ-bound by an appropriate functional head in order for the expres-
sion containing a relative adjective or adverbs not to induce a violation of FI,
i.e. Θ-binding of d by Deg or Q amounts to fixing the degree to which the
property or quality denoted by the relative adjective or adverbs obtains.
While the interpretation of extended adjectival or adverbial projections not
containing a headed DegP or QP either involves reference to scales, as is the
case with the measure expression six foot in (73a), or reference to a default or
standard average norm, as is the case with [QP [Q e][AP ]] in (73b),26 the inter-
pretation of extended projections of relative adjectives and adverbs containing
 Dagmar Haumann

a headed DegP or QP involves reference to a relative standard which is intro-

duced by [Deg {as, so, too}] and [Q {enough, less, more}]. The relative standard is
typically expressed by a result clause:
(73) a. She’s [QP [DP six foot] [AP tall]] and races sled dogs. (Collins Cobuild)
b. They are [QP [Q e] [AP old]] (for lemmings.)

That is to say that the elements under consideration are relational items which
relate the degree to which a given property or quality applies either to a degree
of another given property or quality, as is the case with [Deg as], [Q less] and
[Q more], or to the state of affairs denoted by the extent clause, as is the case
with [Q enough], [Deg so] and [Deg too].
Given these considerations, the contrasts in acceptability between (71) and
(72) suggest that [Deg {as, so, too}] and [Q {enough, less, more}] are contingent on
different interpretive requirements: while [Q {enough, less, more}] allow for in-
ferring the relative standards they introduce, the relative standards introduced
by [Deg {as, so, too}] cannot be inferred from the context.
However, a closer look at the data reveals that the generalization that DegP
cannot be interpreted without a result clause is too strong. In right-peripheral
positions, as in (74)–(76), as, so, and too can be construed without a result
clause, provided the property or quality denoted by the adverb is focalized:
(74) a. He opened the door very carefully, as if he was afraid it might fall off.
He slid out [DegP just as carefully], planting his feet firmly on the
ground. . . (Collins Cobuild)
b. Now that Sarah was back at work Anne was not visiting the Redmond
house [DegP as frequently], and she missed her visits there. (G16 1241)
(75) a. The car suddenly screeched to a shuddering halt and Nicole . . . was
thrown forward and cracked the side of her head painfully on the back
of the passenger seat. She was still struggling to sit up when she re-
alized why the driver had been forced to brake [DegP so violently.]
(Collins Cobuild)
b. ...but by using an auxiliary collecting tank I am able to increase the ca-
pacity by more than eight time, thereby avoiding having to empty the
vac [DegP so frequently.] (A0X 1745)
(76) a. He rebuked me because I’d apparently stepped in [DegP too quickly]
with the next question before he’d finished answering the last one.
(EVN 149)
b. All the drugs weakened the fine control of blood pressure, so that patients
were liable to faint if they stood up [DegP too suddenly]. (ARF 1181)
 Degree Phrases versus Quantifier Phrases 

Given that the occurrence of as, so, and too without a result clause is parasitic
on the focalized adverb and given that focalized adverbs are confined to the
postverbal position (cf. Alexiadou (1997: 142), Cinque (1999: 16)), the unac-
ceptability of the strings in (71) comes as a natural consequence. That is to
say that, since preverbal adverbs cannot be focalized, the relative standard in-
troduced by as, so, and too cannot be inferred from the context, but must be
identified by a result clause. However, in the configuration under considera-
tion, DegP is too deeply embedded to be able to license a sentence final result
clause (cf. (68)). Also, the interpretive problem cannot be salvaged by realizing
the result clause right-adjacent to the adverb, since the resulting configuration
would induce a violation of the Edge Effect (cf. (69) above).

. Deg-heads as predicate parasites

As we have seen in Section 2.3.1 above, headed DegPs are generally barred from
occurring in prenominal position,27 while headed QPs are perfectly acceptable
in this position. Interestingly, this asymmetry persists even if Q and Deg are in
construction with a DP-final result clause, as shown in (77) and (78):
(77) a. *[ConjP [DP An [AgrP [DegP as solid] [Agr’ ... [NP door]]]] [Conj’ [Conj ] [PP
as steel]]] gave way.
b. *I met [ConjP [DP a [AgrP [DegP so good ] [Agr’ ... [NP student]]]] [Conj’
[Conj ] [CP that she could easily pass the exam.]]]
c. *She bought [ConjP [DP a [AgrP [DegP too big] [Agr’ ... [NP van]]]] [Conj’
[Conj ] [CP to go under the bridge.]]]
(78) a. Many teachers do not have [ConjP [DP a [AgrP [QP strong enough] [Agr’ ...
[NP grasp of the language]]]] [Conj’ [Conj ] [CP to teach in it effectively.]]]
(B12 691)
b. [ConjP [DP A [AgrP [QP less gloomy] [Agr’ ... [NP prognosis]]]] [Conj’ [Conj ]
[CP than the one advanced by John]]] assumes that there is sufficient
matter in the universe... (Collins Cobuild)
c. Costume plays [ConjP [DP a [AgrP [QP more important] [Agr’ ... [NP role]
. . . [PP for Anderson]. . . ]] [Conj’ [Conj ] [PP than for us.]]] (ED9 408)

The asymmetry between (77) and (78) cannot be taken to derive from struc-
tural differences since DegP and QP occupy identical positions and therefore
should both be just as able or unable to license a DP-final result clause. At least
theoretically, the strings in (77) could license a DP-final result clause. The fact
that they do not improve with one clearly shows that the asymmetry between
(77) and (78) cannot be taken to result from the (non)inferrability of relative
 Dagmar Haumann

standards which are introduced by the elements heading DegP and QP (cf. (71)
and (72) above).
A closer look at the positions that DegP felicitously assumes reveals that
the asymmetry between (77) and (78) is related to the nonpredicative nature
of the adjectival heads of the extended projections under consideration. That is
to say, headed DegPs are licensed in all and only in predicative positions, as in
(79a) and (79b), or in positions derivationally related to predicative positions,
such as the prearticle position, as in (79c) (cf. Section 2.2):
(79) a. Covering [ConjP [ConjP [DP subjects]i [Conj’ [Conj ] [DegP [Deg as] [AP
[D PRO]i [A diverse]]]]] [Conj’ [Conj ] [PP as nuclear waste disposal and
the Irish fears of an accident at Hinkley.]]]
b. [AgrP [D He]i . . .was. . . [ConjP [DegP [Deg so]. . . [AP t i [A drunk]]]] [Conj’
[Conj ] [CP that his words came thickly and with great difficulty.]]]
c. [ConjP [SDP [DegP [Deg too] [AP [D PRO] [A attractive]]]i [SD’ [SD ] [ConjP
[PDP [PD a] [NP prospect] [Conj’ [Conj ] [DegP t i ]]]]] [Conj’ [Conj ] [CP to
refuse.]]]

The hypothesis that the admissibility of headed DegPs feeds on the predicative
status of the adjective allows us to account for the limited distributional range
of extended adjectival projections containing a headed DegP.28

. Conclusion

Corver’s (1997) Split-DegP Hypothesis, which focuses on the internal struc-

ture of extended adjectival and adverbial projections, paves the way for an
account of the (a)symmetries between headed DegPs and QPs in extended
adjectival and adverbial projections as investigated in this paper. Starting out
from the observation that DegP and QPs, together with [QP [Q e][A(dv)P ]], i.e.
extended projections that lack overt functional heads, are both equally admissi-
ble in right-peripheral positions and subject to identical licensing conditions, I
turned to discuss two cases of asymmetry. The first case derives from the quan-
tificational nature of both headed DegP and QP vs. the nonquantificational
nature of [QP [Q e][AP ]] and thus is not essential with respect to the catego-
rial distinction between Deg and Q. The second type of asymmetry, which –
since most syntactic analyses focus on right-peripheral occurrences of headed
DegP/QP – has largely gone uncommented on, concerns the nonadmissibility
of headed DegPs and the admissibility of headed QPs in preverbal and prenom-
inal position. I have argued that the nonadmissibility of headed DegPs within
 Degree Phrases versus Quantifier Phrases 

extended verbal projections ultimately derives from the interplay between the
interpretational necessity of being associated with a result clause and the fail-
ure of the DegP (due to depth of embedding) to license a VP-/sentence-final
result clause. Although headed QPs, like headed DegPs, are too deeply embed-
ded to license a result clause, they are perfectly acceptable in preverbal position
because the relative standards introduced by enough, less, and more can be in-
ferred. The failure of headed DegPs in prenominal position was argued not
to derive from a conflict between interpretational requirements and structural
reality, but to be intimately tied to the (non)predicative nature of the adjec-
tive heading the entire extended projection, i.e. headed DegPs are parasitic
on predicative adjectives, whereas headed QPs co-occur indiscriminately with
predicative and attributive adjectives. The latter type of asymmetry not only
explains the fairly limited distributional range of headed DegPs, as opposed
to the relative versatility of headed QPs, but also suggests that the presence of
DegP and QP within the extended adjectival projection is not only dependent
on the presence of a referential argument in the argument structure of the ad-
jective but also on the presence of an external argument, which is either realized
as a lexical DP or as PRO.

Notes

* I am indebted to the editors and two anonymous reviewers for detailed and helpful com-
ments on earlier versions of this paper. Thanks to John Gledhill, Britta Mondorf, and Bob
Weissberg for allowing me to pick their brains about adverbial virtues and vices.
. Unless indicated otherwise, the data are taken from the British National Corpus (BNC),
and modified accordingly for the ungrammatical examples. The source specifications are
those employed in the BNC.
. Since degree adverbs, e.g. abundantly, rather, and very, are irrelevant to the topic of this
paper, suffice it to say that these elements, together with so-called noun phrase measure
phrases, e.g. a tad and nine meters, are standardly taken to occupy the specifier position
within DegP (cf. Abney (1987: 305ff.), Corver (1991: 43ff.), and Schütze (2001)).
. Extended adjectival and adverbial projections containing how, that, and most, which Ab-
ney (1987) takes to be members of category Deg, as well as the extended projections of
adjectives and adverbs in the synthetic comparative and superlative, will be exempted from
our discussion.
. Thanks to Jennifer R. Austin for reminding me that (8a) is ungrammatical only under
the intended reading, not if so is interpreted as “in such and such a way”.
. According to Corver (1997: 124, Note 5, 132), DegP is not projected if the head of QP is
lexicalized, as in (i):
 Dagmar Haumann

(i) *This exercise is [DegP so [QP more [AP strenuous]]].

Given that both Deg and Q have quantificational force and that the adjective strenuous
heads the minimal complement of Q and that [Q more] binds the referential argument of
strenuous, [Deg so] would be quantifying vacuously, thus inducing a violation of FI.
. Notice that the ungrammatical strings in (5)–(8) improve if much (or many) is inserted
into the head position of QP, as shown in (i) and (ii):

(i) There had not been as much evidence as we had hoped for.

(ii) There are so many of us that we have to rent a bigger car.

However, as shown in (iii) and (iv), much-insertion does not amend the illicit strings in (9a)
and (10a) above, where the extended adjectival and adverbial projection containing an overt
Deg head occurs in prenominal and preverbal positions:

(iii) *She bought a too much big van to go under the bridge.

(iv) *He so much clumsily lit a cigarette that he burnt his nose.

Since both big in (iii) and clumsily in (iv) are heads and as such susceptible to raising to the
head of QP, much-insertion, for reasons of economy, does not apply (cf. Corver (1997: 136)).
We shall come back to the nonadmissibility of headed DegPs in prenominal and preverbal
positions in Section 3.
. See also Cinque (1994: 92ff.) and Laenzlinger (2000: 65ff.) concerning the assumption
that postnominal adjectival projections of the type illustrated in (30) are predicative.
. Cf. Bolinger (1967), Carlson (1977), Svenonius (1993: 3), Cinque (1994: 94f.).
. See also van Riemsdijk (1983) and Cinque (1994: 94, Note 14).
. While Cinque (1999: 19ff.) takes the projections of manner adverbs to be base-generated
in specVoiceP, Laenzlinger (1996: 116f.) locates them in the A’-specifier of VP.
. The term Big Mess Construction is due to Berman (1974).
. According to Bresnan (1973: 283ff., 299ff.) and Zamparelli (2000: 79), constructions of
the type illustrated in (40) are usually barred from argument positions, as shown in (i) and
(ii):

(i) *[DP {[DegP As good], [QP more attractive]} a man drank beer.]

(ii) *I saw [DP {[DegP too attractive], [QP less ridiculous]} a man.]

. Abney (1987: 323ff.) takes the strings in (40) as evidence for his claim that prenomi-
nal adjectives take as their complement the nominal expression they modify. Moreover, he
points out that, in some dialects (which he does not specify), of may be interpolated between
the adjective and DP, as shown in (i) & (ii) (adapted from Abney (1987: 324)):

(i) [DegP [Deg too] [AP [A big ] [PP [P of ] [DP [D a] [NP [N house]

(ii) [DegP [Deg as] [AP [A nice ] [PP [P of ] [DP [D a] [NP [N man]
 Degree Phrases versus Quantifier Phrases 

. Zamparelli’s (2000: 250ff.) proposal to analyze strong and weak quantifiers as distinct
functional heads allows for a principled account of the differences between these two types
of quantifiers, i.e. their (non)admissibility in existential sentences (i) and (ii) and their
(in)ability to function as the subject of individual level predicates (iii) and (iv) (data gleaned
from Milsark (1977: 4, 8, 10ff.)):

(i) There were {*every, *both, *each, several, two} people cycling along the creek.

(ii) There is {*the, *this, a, one} wolf at the door.

(iii) {Every, each, the, etc.} crocodile is a reptile. / {Both, these}animals are crocodiles.

(iv) ?? {Several, two, etc.} crocodiles are reptiles. / ?? {A, one} crocodile is a reptile.

Not only does this analysis account for the strong-weak dichotomy in (i) to (iv), but also for
the fact that weak quantifiers may be embedded under strong quantifiers, as shown in (v)
and (vi) (cf. Bowers (1975: 541f.) for discussion):

(v) And the tally of crown successes over its opponents far outnumbers its few failures.
(AE4 410)

(vi) And I had a strong need to pit myself against her many suitors, and win. (AEA 1180)

Moreover, Zamparelli’s (2000) analysis properly accounts for the differences between the
two readings of weak quantifiers, i.e. the weak or adjectival / cardinal reading in (vii) and
the strong or proportional / partitive reading in (viii) (cf. Milsark (1977: 16ff.)):

(vii) In this house, there are {two, few, many} furry pets.

(viii) In this house, {two, few, many} pets are cats.

. There are as many AgrPs as there are attributive adjectives. Cinque (1994: 95) is vague
about the categorical nature of the agreement related functional projections, which he labels
XP.
. In general, all types of QP are admissible in prenominal position. The trivial case of
prenominal adjective placement is that seen in (22c) and (23c), where the head of QP is
phonetically empty, i.e. the extended adjectival projection has the format [QP [Q e] [AP ]].
. Notice that (52b) and (52c) are fully acceptable if QP has subject-orientation, in which
case QP would have been merged into the structure as a specifier-type adverbial, rather than
a complement-type one.
. In addition to QPs of the type [QP [Q e] [AdvP ]], as in (54), QPs containing specifier-type
degree modifiers are admitted in the preverbal position:

(i) She [AgrP [QP [AdvP very] [Q’ [Q ] [AdvP cautiously]] [Agr’ . . . [VP avoided her boss on
Friday afternoon.]]]

. Cf. Bresnan (1973: 331ff.) and Bowers (1975: 530ff.) for extraposition analyses of extent
clauses, Corver (1990: 47ff.) for a binary branching structure in which the extent clause func-
tions as a right branch constituent (adjunct or specifier) within the projection of Deg/Q, and
Jackendoff (1977: 207ff.) for an analysis of extent clauses as complements within AP/AdvP.
 Dagmar Haumann

Alternative accounts involve the base-generation of extent clauses in sentence-final position,

e.g. Andrews (1975: 167), Rouveret (1978: 175ff.), Guéron (1980: 648), as well as Rijkhoek’s
(1998) asymmetric conjunction analyses, which is discussed in Section 3.1.1.20.
. See also Rothstein (1991). However, Rothstein (1991: 102ff., 108f.) differs from Abney
(1987) in that she takes degree words to be minor functional heads (MFHs) which are cate-
gorially parasitic on the phrases they take as their complement, i.e. the AP complement of
the MFH so in (i) continues to project its categorial features:

(i) [AP so [AP drunk] [CP that his words came thickly]]

. Abney (1987: 328) argues that the fact that prenominal adjectives do not take internal
arguments, i.e. complements to which they assign a Θ-role, comes as a natural consequence
under his analysis: if a transitive adjective, e.g. cheerful in (i) takes the NP it modifies as its
complement, it cannot take its usual complement, e.g. about her work in (ii):

(i) The pathologist was [DP [D a] [AP [A cheerful] [NP woman]]]

(ii) *The pathologist was [DP [D a] [AP [A cheerful] [PP about her work] [NP woman]]]

A brief glance at German reveals that Abney’s (1987) analysis of prenominal adjectives is
plain wrong since prenominal adjectives, such as überdrüssig ‘fed up’ and stolz ‘proud of ’
in (i) and (ii), take their usual complements, e.g. [DP ihrer Arbeit] ‘her work’ and [DP ihre
Mutter] ‘her mother’ (see also van Riemsdijk (1983: 224ff.)):

(i) eine ihrer Arbeit überdrüssige Studentin

a(fem, sg) her(fem, sg) work fed-up-with(fem, sg) student(fem, sg)
‘a student fed up with her work’

(ii) eine auf ihre Mutter stolze Studentin

a(fem, sg) on her(fem, sg) mother proud(fem, sg) student(fem, sg)
‘a student proud of her mother’

. Rijkhoek’s (1998) analysis is restricted to sentence-final occurrences of result clauses

licensed by so and more within the extended nominal projection, but can easily be extended
to all Degs and Qs in both the extended verbal and nominal projection.
. According to Rijkhoek (1998: 126f., 130ff.), DegP or QP contained within the con-
stituent occupying specConjP must raise and adjoin covertly to the maximal constituent
in which it originates in order for the head of ConjP to be endowed with the selectional
properties of Deg or Q.
. In her discussion of result clause licensing, Rijkhoek (1998: 151ff.) presents the ungram-
matical strings in (i) and (ii):

(i) *...plots [that so many conspirators know about] have been hatched [that the government
has lost all credibility]

(ii) *[[that so many conspirators know about plots] is shocking news] [that the government
has lost all credibility]
 Degree Phrases versus Quantifier Phrases 

She suggests that the unacceptability of these strings be accounted for in terms of Subjacency
(cf. Chomsky (1986b: 34ff.)), i.e. the result clause may be separated from its antecedent by
at most one CP boundary, provided that CP is argumental.
. Cf. Williams’ (1982: 160) Head Final Filter on prenominal APs. See also Emonds
(1976: 19) and van Riemsdijk (1983: 228ff.).
. See also Higginbotham (1985: 563ff.), Hamann (1991: 660), and Neeeleman, van de
Koot & Doetjes (2004: 27ff.) for discussion.
. Combinations of too and nonpredicative adjective are marginally acceptable if too can
be re-interpreted as a “non-relative” degree expression (see also Zamparelli (2000: 144)):

(i) Truman had replaced the too-liberal Henry Wallace. (Collins Cobuild)

(ii) ...live with them without a too-crushing sense of guilt. (Collins Cobuild)

. Interestingly, Fischer (2000: 157ff.) makes a similar point with respect to the (non)admis-
sibility of degree modifiers, e.g. þearle ‘vigorously’ in (i), in Old English adjectival pro-
jections ((i) is taken from The Brooklyn-Geneva-Amsterdam-Helsinki Parsed Corpus of Old
English):

(i) Heo wæs soÞlice [[Adv Þearle] [AP wlitig.]] (APOLLO,36.48.7)

she was truly vigorously radiant
On the basis of an empirical investigation of Old English adjective placement within nom-
inal projections, Fischer (2000) argues that only strong adjectives like wlitig ‘radiant’ in (i),
which she takes to be predicative in nature, take degree modifiers.

References

Abney, S. (1987). The English Noun Phrase in Its Sentential Aspects. PhD, MIT, Cambridge,
MA.
Alexiadou, A. (1997). Adverb Placement: A Case Study in Antisymmetric Syntax. Amsterdam
and Philadelphia: John Benjamins.
Andrews, A. (1975). “Studies in the syntax of relative and comparative clauses.” PhD, MIT,
Cambridge, MA.
Berman, A. (1974). Adjectives and adjective complement constructions in English. PhD,
Harvard University, Cambridge, MA.
Bolinger, D. (1967). “Adjectives in English: Attribution and predication.” Lingua, 18, 1–34.
Bowers, J. (1975). “Adjectives and Adverbs in English.” Foundations of Language, 13, 529–
562.
Bresnan, J. (1973). “Syntax of the comparative clause constructions in English.” Linguistic
Inquiry, 4, 275–343.
Carlson, G. N. (1977). “A unified analysis of the English bare plural.” Linguistics and
Philosophy, 8, 413–457.
Chomsky, N. (1986a). Knowledge of Language. Its Nature, Origin, and Use. New York:
Praeger.
 Dagmar Haumann

Chomsky, N. (1986b). Barriers. Cambridge, MA: MIT Press.

Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press.
Cinque, G. (1994). “On the evidence for partial N-movement in the Romance DP.” In
G. Cinque, J. Koster, J.-Y. Pollock, L. Rizzi, & R. Zanuttini (Eds.), Paths Towards
Universal Grammar: Studies in Honor of Richard S. Kayne (pp. 85–110). Washington,
D.C.: Georgetown University Press.
Cinque, G. (1999). Adverbs and Functional Heads. A Cross-Linguistic Perspective. Oxford.
Oxford University Press.
Corver, N. (1990). The Syntax of Left Branch Extractions. PhD, Katholieke Universiteit
Brabant.
Corver, N. (1991). “Evidence for DegP.” In T. Sherer (Ed.), Proceedings of the 21st Meeting
of the North East Linguistic Society (pp. 33–47). Amherst, MA: GLSA University of
Massachusetts.
Corver, N. (1997). “Much-support as a Last Resort.” Linguistic Inquiry, 28, 119–164.
Culicover, P. (1980). “Though-attraction.” Social Sciences Research Reports 80. University of
California, Irvine.
Emonds, J. (1976). A Transformational Approach to English Syntax. New York: Academic
Press.
Fischer, O. (2000). “The positions of the adjective in Old English.” In R. Bermúdez-Otero,
D. Denison, R. M. Hogg, & C. B. McCully (Eds.), Generative Theory and Corpus Studies.
A Dialogue from 10 ICEHL (pp. 153–181). Berlin: Mouton de Gruyter.
Guéron, J. (1980). “On the syntax and semantics of PP extraposition.” Linguistic Inquiry, 11,
637–678.
Haider, H. (2000). “Adverb placement – convergence of structure and licensing.” Theoretical
Linguistics, 26, 95–134.
Hamann, C. (1991). “Adjectives.” In A. von Stechow & D. Wunderlich (Eds.), Semantics: An
International Handbook of Contemporary Research (pp. 657–673). Berlin: de Gruyter.
Higginbotham, J. (1985). “On semantics.” Linguistic Inquiry, 16, 547–593.
Jackendoff, R. (1977). X’-Syntax: A Study of Phrase Structure. Cambridge, MA: MIT Press.
Kayne, R. (1994). The Antisymmetry of Syntax. Cambridge, MA: MIT Press.
Klima, E. S. (1964). “Negation in English.” In J. A. Fodor & J. J. Katz (Eds.), The Structure of
Language (pp. 246–323). Englewood Cliffs, NJ: Prentice Hall.
Koster, J. (1999). “Empty objects in Dutch.” Ms., University of Groningen.
Laenzlinger, C. (1996). “Adverb syntax and phrase structure.” In A.-M. Di Sciullo
(Ed.), Configurations. Essays on Structure and Interpretation (pp. 99–127). Somerville:
Cascadilla Press.
Laenzlinger, C. (2000). “French adjective ordering: perspectives on DP-internal movement
types.” Generative Grammar in Geneva, 1, 55–104.
Milsark, G. (1977). “Towards an explanation of certain peculiarities in the existential
construction in English.” Linguistic Analysis, 3, 1–30.
Neeleman, A., van de Koot, H., & J. Doetjes (2004). “Degree expressions.” Linguistic Review,
21, 1–66.
Rijkhoek, P. (1998). On Degree Phrases & Result Clauses [Groningen Dissertations in
Linguistics 27]. Groningen: Rijksuniversiteit.
Rizzi, L. (1990). Relativized Minimality. Cambridge, MA: MIT Press.
 Degree Phrases versus Quantifier Phrases 

Rizzi, L. (1997). “The fine structure of the left periphery.” In L. Haegeman, L. (Ed.), Elements
of Grammar (pp. 281–337). Dordrecht: Kluwer.
Rothstein, S. (1991). “Heads, projections, and category determination.” In K. Leffel, K. & D.
Bouchard (Eds.), Views on Phrase Structure (pp. 97–112). Dordrecht: Kluwer.
Rouveret, A. (1978). “Result clauses and conditions on rules.” In S. J. Keyser (Ed.), Recent
Transformational Studies in European Languages (pp. 159–187). Cambridge, MA: MIT
Press.
Sadler, L., & Arnold, D. J. (1994). “Prenominal adjectives and the phrasal / lexical distinc-
tion.” Journal of Linguistics, 30, 187–226.
Schütze, C. T. (2001). “Semantically empty heads as last resorts.” In N. Corver & H. van
Riemsdijk (Eds.), Semi-Lexical Categories (pp. 127–187). Berlin: Mouton de Gruyter.
Svenonius, P. (1993). “The structural location of the attributive adjective.” Ms., University
of California at Santa Cruz.
van Riemsdijk, H. (1983). “The case of German adjectives.” In Heny, F. & B. Richards (Eds.),
Linguistic Categories: Auxiliaries and Related Puzzles, Vol. I (pp. 223–252). Dordrecht:
Reidel.
White, J. R. (1997). “Result clauses and the structure of degree phrases.” UCL Working Papers
in Linguistics, 9, 315–333.
Williams, E. (1982). “Another argument that passive is transformational.” Linguistic Inquiry,
13, 160–163.
Zamparelli, R. (2000). “Layers in the Determiner Phrase.” Ms., University of Rochester.
Zwarts, J. (1992). X’-Syntax X’-Semantics. On the Interpretation of Functional and Lexical
Heads. Utrecht: OTS Dissertation Series.

Other Sources:

British National Corpus (BNC), http://sara.natcorp.ox.ac.uk/lookup.html.

Collins Cobuild, http://titania.cobuild.collins.co.uk.
The Brooklyn-Geneva-Amsterdam-Helsinki Parsed Corpus of Old English. ©2000.
Pintzuk, S., Haeberli, E., van Kemenade, A., Koopman, W. and Beths. F.
 A feature-based theory of adverb syntax*

Christopher Laenzlinger
University of Geneva

This paper is a comparative study of adverb distribution in French and

German. On the basis of a feature-based theory of adverb licensing (Cinque
1999 contra Ernst’s 2002 scope-based theory) in which adverbs are merged as
unique specifiers of semantically related functional projections, the various
positions of (co-occurring) adverbs in the Vorfeld, Mittelfeld and Nachfeld of
both languages are analysed in terms of (i) the same set of clause-internal
structures (a fixed subject position with floating object and verb positions)
and (ii) the same set of transformations (subject and object movement, and
remnant/pied-piping VP movement). From a parametric point of view, it is
shown that French and German differ in (i) the range of positions which is
targeted by verb projection raising (VP-movement) and object scrambling
(high in German, low in French), and (ii) short verb movement (to
Subj(=Agrs ) in French and to C in German main clauses).

. Introduction

This paper is a comparative study of adverb distribution in Romance (mainly

French) and Germanic (mainly German and English) within the feature-based
theory of adverb syntax (Alexiadou 1997; Cinque 1999; Laenzlinger 1996, 1998,
2000a). In this framework adverbs merge as specifiers of clause-internal func-
tional projections. Alternatively, Ernst (2002) proposes a scope-based theory
of adverb distribution. He argues against the feature-based theory by giving
pieces of evidence against the uniqueness of adverb positions and the left-to-
right linearity of adverb co-occurrences. The two issues are illustrated in the
following examples:
(1) a. (Amicalement,) Jean (, amicalement,) a (amicalement)
friendlily Jean friendlily has friendlily
 Christopher Laenzlinger

(amicalement) salué (amicalement) le professeur

friendlily greeted friendlily the professor
(amicalement.)
(friendlily.)
‘Jean friendlily greeted the professor.’
b. (*Déjà), Jean (*déjà,) a (déjà) salué (?? déjà) le professeur
Already Jean already has already greeted already the professor
(*déjà.)
(already)
‘Jean already greeted the professor.’
(2) a. Jean l’a refait rapidement souvent maintenant.
Jean it-has redone quickly often now
‘Jean now often quickly did it again.’
b. Jean l’a maintenant souvent rapidement refait.
Jean it-has now often quickly redone
‘Jean now often quickly did it again.’
c. Jean l’a souvent gentiment fait de nouveau.
Jean it-has often kindly done again
‘John often did it kindly again.’
d. Jean l’a (de nouveau) souvent (de nouveau) gentiment fait.
Jean it-has again often again kindly done
‘John often did it kindly again.’

One can observe that the manner adverb amicalement in (1a) can occupy var-
ious positions within the clause, whereas the aspectual adverb déjà in (1b) is
confined to a single position. What differences between the two adverbs can
account for the contrast in (1a–b)? The adverb amicalement differs from the
adverb déjà in two ways. First, déjà is considered a “light” adverb in comparison
with the “heavy” form amicalement. Second, the adverb déjà is quantificational
(quantification over time), while the adverb amicalement is qualificational (cir-
cumstantial).
In (2a–b) one can note that the postverbal sequence of adverbs is the mir-
ror image of the preverbal sequence. This is unexpected under the view that
adverbs are merged as left specifiers of specific projections and that the order
[V AdvP] results from verb (projection) movement. Finally, in (2c–d) we have
a clear case of scope ambiguity. The repetitive adverb de nouveau in postverbal
position in (2c) has scope over the manner adverb, but may or not have scope
over the frequency adverb. The scope relations are expressed linearly in the sur-
face order possibilities in (2d). Such issues will be discussed in this paper with
respect to the feature-based theory of adverb syntax.
 A feature-based theory of adverb syntax 

The scope of the paper is to compare the distribution of adverbs within

the three domains of the clause in French and German. It will be assumed that
these two languages are subject to the same set of transformations. Verb move-
ment will be analyzed as an instance of remnant/pied-piping VP-movement
to the specifier of an inflectional/auxiliary projection (InflP/AuxP). Head-
movement will be restricted to very local domains containing the same type
of categorical projections. In addition to VP-raising, French and German share
the property of argument scrambling of the subject to Spec-SubjP and of the
objects to specifiers of ObjP.1 More precisely, all arguments must escape from
the VP-shell, for they have uninterpretable formal features that need to be
checked outside VP before Spell-Out.
The aim of this paper is to show that the feature-based theory of adverb
licensing is able to handle the problem of adverb order variation. The various
positions of (some) adverbs will be explained by (i) distinct positions for ad-
verb merger and (ii) different locations for verb (projection) movement and
object scrambling. As regards the second point, I will postulate for French and
German the same core of clause structure (principles), which contains a base
of functional projections, morpho-functional projections like InflP/AuxP, and
derived argument positions (SubjP, ObjP). More specifically, the clause struc-
ture contains recursive chunks of sub-structures, composed of the sequence
SubjP>InflP>ObjP potentially available between every adverb-related func-
tional projection. French and German clauses minimally differ in (their param-
eter setting for) (i) the set of InflP and ObjP which can be activated within the
middle field (henceforth Mittelfeld)2 and (ii) the (non) application of head-
movement to Subj or C. These parametric differences account for the varia-
tions in the distribution of adverbs within the clause in French and German.
The paper is organized as follows. After the introduction in Section 1, Sec-
tion 2 sets forth the theoretical background by introducing mechanisms for
licensing adverbs, phrase structure merging, and clause structure composi-
tion. In Section 3, the distribution of adverbs in the Mittelfeld is discussed
for French (Section 3.1) and German (Section 3.2). Section 3.1.1 introduces
the main structural properties of French, such as SVO order, generalized verb
projection raising, and short scrambling of arguments. Section 3.1.2 presents
a discussion of the range of A’-positions (for adverbs) and A-positions (for
arguments) in the Mittelfeld. Section 3.1.3 deals with the set of transforma-
tions in the Mittelfeld affecting the linear placement of adverbs. In Section
3.2.1 the main properties of the German Mittelfeld are discussed in relation
to basic word order (OV vs. VO), VP-raising, and high/low scrambling. In Sec-
tion 3.2.2 the positions of adverbs and scrambled arguments are identified on
 Christopher Laenzlinger

the basis of their respective ordering. The main transformations of the German
Mittelfeld are presented, such as verb projection raising, argument movement,
and adverb movement. Section 4 deals with the distribution of adverbs in the
Nachfeld, mainly in French and English, and includes a discussion of adverb in-
terference in the VP-shell structure. In Section 4.1 the order [V Adv Compl] is
analyzed with respect to different classes of adverbs and different types of verb
complements. Section 4.2 is concerned with sentence-final adverbs, focussing
on sequences of postverbal adverbs and mirror image ordering (triggered by
a “snowballing” sort of movement). Section 5 deals with the occurrence of
adverbs in the Vorfeld, mainly in French and German. In Section 5.1, the se-
quences [Adv Subj Aux/V] and [Subj Adv Aux/V] are analyzed in Romance
and English from a comparative point of view. Section 5.2 deals with adverb
preposing in German in relation to the phenomenon of remnant and pied-
piping VP-topicalization in German. Finally, Section 6 contains the conclusion.

. Theoretical background

There exist several theories of adverb distribution in generative grammar. For

the most part, they address the question of the syntax-LF/PF interface, which
is certainly relevant to adverb distribution. Apart from purely phonological,
semantic, or stylistic approaches, there are three main trends in the syntax-
semantics analysis of adverbs within the framework of the Principles & Pa-
rameters theory: (i) the theory of predication (Roberts 1985; Rochette 1990);
(ii) the scope-based theory (Haider 2000; Ernst 2002; Frey 2000);3 (iii) the
feature-based theory (Travis 1988; Laenzlinger 1996, 1998, 2000a; Alexiadou
1997; Cinque 1999). Arguments against the theory of predication are pro-
vided in Laenzlinger (1998: 69–72). This paper focuses on the feature-based
theory of adverb licensing in which a one-to-one relation between the posi-
tion (specifier) and the interpretation of adverbs is established. By comparison,
the scope-based theory specifies possible positions for adverbs with respect
to lexical selection and independent semantic rules. Put simply, the feature-
based theory stresses the role of syntax in the semantics of adverbs, while the
scope-based theory stresses the role of semantics in the syntax of adverbs.
The feature-based theory relies mainly on Checking Theory (Chomsky
1995). Thus, adverbs must occur in a checking configuration, essentially in a
Spec-head relation with the appropriate head (Laenzlinger 1996, 1998; Alex-
iadou 1997; Cinque 1999) for full adverbs and in a head-head relation for
clitic adverbs. The features to be checked on adverbs are not formal (unin-
 A feature-based theory of adverb syntax 

terpretable) features, but substantive (interpretable) ones. They are checked at

the root, by means of pure Merge, similarly to thematic or selectional features.
Laenzlinger (1998: 83–88) introduces a distinction between quantificational
and qualificational adverbs. There it was argued that subtly different licens-
ing mechanisms are at play for the two kinds of adverbs. While qualificational
adverbs adhere to the standard picture described above, quantificational ones
stand for operators in specifier positions, as shown by the relativized minimal-
ity effects in combien extraction (Obenauer 1983; Rizzi 1990) in (3):
(3) *Combien as-tu probablement / souvent / presque / à peine /
How.many has.you probably / often / almost / hardly /
beaucoup lu [t] de livres de Chomsky?
a lot read t of books of Chomsky?
‘How many books of Chomsky did you probably / often / almost / hardly
read / a lot?’

In addition, the quantificational adverbs, like neg/wh/foc-operators, are “frozen”

in their licensing position once they have satisfied the Adv-Criterion, their spe-
cial licensing requirement. They cannot be displaced from their base position
to a fronted position, as illustrated in (4).4
(4) a. Tu as beaucoup / à peine / presque dormi.
you have a lot / hardly / almost slept
‘You almost / hardly slept / a lot.’
b. *Beaucoup / à peine / presque, tu as dormi.
a lot / hardly / almost you have slept
The possibility of having probablement ‘probably’ and souvent ‘often’ in
sentence-initial position in (5b) does not result from movement of the adverb.
As will be discussed in Section 5.1.2, these adverbs are in their root positions,
where they are licensed.
(5) a. Tu as probablement / souvent dormi.
You have probably / often slept
‘You probably / often slept.’
b. Probablement / souvent, tu as dormi.
‘Probably / often you slept.’

Work by Jackendoff (1972), McConnell-Ginet (1982), Cinque (1999), and

Ernst (2002) gives rise to a relatively complete typology of adverbs. The
classes of adverbs relevant to this paper are given in (6), taken from Cinque
(1999: 106). Each class is exemplified by an adverb in English, French, and Ger-
 Christopher Laenzlinger

man. Since adverbs are unique specifiers of semantically related projections,

the clause structure is assigned a large number of functional projections, as
represented in (6). There is a potential slot in the clause structure for each
class of adverb following the hierarchy of functional projections given by UG.
In the absence of adverbs (or lexicalized heads) the functional projections are
assigned a default value.5
(6) [Frankly/ franchement/ offen gestanden Moodspeech act >
[unfortunately/ malheureusement/ unglücklicherweise Moodevaluative >
[apparently/ apparemment/ anscheinend Moodevidential >
[probably/ probablement/ wahrscheinlich Modepistemic >
[once/ autrefois/ einmal Tpast >
[then/ ensuite/ dann Tfuture >
[maybe/ peut-être/ vielleicht Mod(ir)realis >
[necessarily/ nécessairement/ notwendigerweise Modnecessity >
[possibly/ (possiblement)/ möglich Modpossibility >
[deliberately/ intentionnellement/ absichtlich Modvolitional >
[inevitably/ inévitablement/ zwangsläufig Modobligation >
[cleverly/ intelligemment/ geschickt Modability/permission >
[usually/ habituellement/ gewöhnlich Asphabitual >
[again/ de nouveau/ wieder Asprepetitive (I) >
[often/ souvent/ oft Aspfrequentative >
[quickly/ rapidement/ schnell Aspcelerative (I) >
[already/ déjà/ schon Tanterior >
[no longer/ plus/ nicht mehr Aspperfect >
[still/ encore/ noch Aspcontinuative >
[always/ toujours/ immer Aspperfective >
[just/ juste/ gerade Aspretrospective >
[soon/ bientôt/ bald Aspproximative >
[briefly/ brièvement/ kurz Aspdurative >
[typically/ typiquement/ typisch Aspgeneric/progressive >
[almost/ presque/ beinahe Aspprospective >
[completely/ complètement/ völlig AspSgCompletive (I) >
[all/ tout/ alles AspPlCompletive >
[well/ bien/ gut Voice >
[fast/ vite/ schnell Aspcelerative (II) >
[completely/ complètement/ ganz AspSgCompletive (II) >
[again/ de nouveau/ wieder Asprepetitive (II) >
[often/ souvent/ oft Aspfrequentative
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
 A feature-based theory of adverb syntax 

Adverbs in English, French, and German are full maximal lexical categories
(Adv → AdvP), which can be modified (very often, rather slowly). A few of
them can take a PP complement (independently of, similarly to). Some adverbs
in French are weak forms (bien ‘well’, mal ‘badly’, see Abeillé & Godard 2000)
whose distribution is very constrained (e.g. J’ai bien dormi ‘I slept well’ vs.
??
J’ai dormi bien/*Bien, j’ai dormi). They seem to be subject to an incorporation
requirement, which distinguishes them from heavy adverbial forms.
Within the feature-based theory, each class of adverb in (6) is confined to a
single position, which is identified as the specifier position of a corresponding
semantically related functional projection. Such an approach readily fits in with
the LCA (Kayne 1994) conception of phrase structures, as advocated by Alexi-
adou (1997) and Cinque (1999). Specifiers are unique left-branching adjoined
phrases. In Kayne’s system, multiple adjunction is banned, as well as right at-
tachment of a specifier. The only possible configuration is [Spec X◦ Compl].
Adverbs are adjoined specifiers attached to the left. Linearly, they precede the
head with which they are associated.6
Kayne/Cinque’s model of phrase and clause structure requires a certain
number of transformations (movement) to account for the right placement of
adverbs with respect to the head that licenses them. For instance, verb/auxiliary
movement affects the linear placement of adverbs (Pollock 1989; Belletti 1990;
Cinque 1999), as illustrated in (7) for French.
(7) a. Jean mange probablement [V t] une pomme.
Jean eats probably [V t] an apple
‘Jean is probably eating an apple.’
b. Jean a mangé probablement [V t] une pomme.
Jean has eaten probably [V t] an apple
c. Jean a probablement [Aux t] mangé une pomme.
Jean has probably [Aux t] eaten an apple
‘Jean probably ate an apple.’

Ernst (2002: 92–95) points out two issues that arise from the LCA approach to
adverb placement. The first one is related to the identification of the various
landing sites for movement; the second one is related to the various (morpho-
logical or other) triggers of such movement. The present paper discusses these
issues and proposes answers to the questions of landing site and trigger.7
The clause can be divided in three domains, as indicated in (8). The Mit-
telfeld goes from the complementizer position (non-included) to the VP-shell
structure (non-included). The pre-field (henceforth Vorfeld) corresponds to
the Comp-domain, which is assumed to be a rich complex structure follow-
 Christopher Laenzlinger

ing Rizzi (1997). Finally, the post-field (henceforth Nachfeld) is equated with a
Larsonian VP-shell structure (see Chomsky 1995).

(8) ForceP > TopP > FocP > TopP > FinP  (Vorfeld)
> MoodP > ModP
 > NegP > TP > AspP1 > AspP2 (Mittelfeld)
> vP > VP (Nachfeld)

The analysis of adverbs in the French Mittelfeld undertaken in the next section
will focus on their distribution with respect to verbs and auxiliaries.

. Adverbs in the Mittelfeld

. French

.. General structural properties

French is an SVO language displaying verb movement (V-to-I movement fol-
lowing Pollock’s (1989) analysis). In accordance with the LCA, specifiers are
on the left of their head, while complements are on the right. The subject is
moved from Spec-vP to its surface position, which is identified by Chomsky
(1995) as Spec-TP. Following the clause structure in (6) and (8), the specifier
of TP is an adverbial position. I propose that a position different from Spec-TP
serve as the landing site for subject movement. Let us call this position Spec-
SubjP, corresponding to the previous Spec-Agrs P. In terms of Checking Theory,
the EPP-features to be checked by the subject do not occur on T, but on Subj.
Object complements merge with the VP-shell structure, and as will be argued
later, can undergo short movement out of VP. Negation (i.e. NegP) stands for
the borderline between sentential adverbs and VP-adverbs. Sentential adverbs
correspond to adverbs pertaining to mood, modality, and aspect (high), while
VP-adverbs include adverbs of (low) aspect and manner.
Most importantly, verb movement is analyzed here differently from Pol-
lock’s (1989) V-to-I movement (head movement). In the spirit of Koopman
& Szabolcsi (2000) (see also Mahajan 2000), I propose that V-to-I move-
ment be reduced to an instance of remnant/pied-piping VP-movement. Head-
movement remains a very constrained operation, restricted to projections of
the same categorial layer, that is, v to V (VP-layer), Aux to I (IP-layer), Fin
to Force (CP-layer). The analysis according to which verb movement to the
IP-domain is an instance of XP-movement raises the issues of the displaced cat-
egory, the landing site, and the motivation for this movement. I assume that the
displaced constituent should be identified as VP/vP (pied-piping verb move-
 A feature-based theory of adverb syntax 

ment), from where the objects can be extracted (remnant VP). The target of
this movement is related to its trigger. Movement results from the selectional
relation between inflection (possibly lexicalized by an auxiliary) and VP. VP
raises to Spec-InflP or AuxP to have its selectional features checked.8
Let us consider the analysis of the French sentence in (9).
(9) Jean n’a probablement pas récemment souvent donné une
Jean neg-has probably not recently often given an
pomme à Marie.
apple to Marie
‘Jean probably did not recently often give an apple to Marie.’

The indirect object merges as Compl of the participial verb (V), while the di-
rect object merges as its Spec. As will be argued later, they raise to object-related
projections, namely Spec-ObjP. The subject is root-merged as Spec-vP and
raises to Spec-SubjP to get its EPP-features checked. The auxiliary is merged
as Aux and raises to Subj to check the agreement features of the subject. V
raises to v and the remnant vP moves to Spec-AuxP. The adverb probablement
merges as the specifier of ModP, the adverb pas as the specifier of NegP, the
adverb récemment as the specifier of TP and the adverb souvent as the specifier
of AspP. The complete derivation is reproduced in (10).
 Christopher Laenzlinger

(10) CP

SubjP

DP Subj+

Subj ModP

AdvP NegP

AdvP TP

AdvP AspP

AdvP AuxP

vP Aux+

Aux ObjP

DP ObjP

PP vP

SUBJ v+

v VP

DP V+
Jean n’a
probablement
pas V PP
récemment
souvent donné une pomme á Marie

The above analysis has important shortcomings on the placement of adverbs

with respect to the verb and the object.9 Take, for instance, a simple tensed
clause like (11):
 A feature-based theory of adverb syntax 

(11) Jean ne donne probablement pas souvent une pomme

John neg give probably not often an apple
à Marie
to Mary
‘John probably does not often give an apple to Mary’

The adverbs occur between the verb and its objects. In Pollock’s (1989) analysis
this configuration results from V-to-I movement. Within the present frame-
work it results from VP-movement to Spec-InflP, as illustrated in (12). It is
an instance of remnant VP-movement, since the objects have first been ex-
tracted from VP. The direct object DP and the indirect object PP both raise to
object-related projections (i.e. Spec-ObjP).10 I also assume that the verb (V) in
Spec-InflP raises to Subj as an instance of head-movement, as indicated in (12).
 Christopher Laenzlinger

(12) CP

SubjP

DP Subj+

Subj ModP

AdvP TP

AdvP AspP

AdvP InflP

vP ObjP

v DP ObjP

PP vP

SUBJ v+

v VP
Jean
donne DP V+
probablement
déja
souvent V PP
une pomme
á Marie

As argued by Pollock (1989), verb movement applies to French, but not to En-
glish. In the English sentences in (13a–b) one can observe that the lexical verb
occurs in a position following the adverbs, not only in compound tense, but
also in simple tense.
(13) a. John has probably not often kindly given an apple to Mary.
b. John probably already often kindly gives an apple to Mary.

Following the VP-movement analysis, I assume that verb projection raising

operates not only in French, but also in English, as shown in (14), the only dif-
ferences being that in English the simple tensed verb does not raise to Subj and,
 A feature-based theory of adverb syntax 

as we will see later, the direct object must remain in a position right-adjacent
to the verb.
(14) CP

SubjP

DP Subj+

Subj ModP

AdvP NegP

AdvP AspP

AdvP MannP

AdvP AuxP

vP Aux+

Aux ObjP

DP ObjP

PP vP

SUBJ v+

v VP

John has DP V+
probably not often
kindly given
John probably often kindly gives V PP
an apple to Mary
an apple to Mary

In the next section I will give further evidence in favor of (10/12) and (14) based
on the various positions of adverbs with respect to the verb, the auxiliary, and
the objects.
 Christopher Laenzlinger

.. The French Mittelfeld

The French Mittelfeld displays the adverb hierarchy presented in (6). Let us ex-
amine the following adverb classes : MoodPspeech-act : franchement ‘frankly’ >
ModPepistemic : probablement ‘probably’ > ModPvolitional : volontiers ‘willingly’ >
NegPnegation : pas ‘not’ > TPanterior : récemment ‘recently’ > AspPfrequentative : sou-
vent ‘often’ > AspPproximative : à peine ‘hardly’ >
MannPverb-oriented : péniblement ‘painstakingly’. Cinque’s hierarchy of adverb-
related functional projections makes the correct prediction for the left-to-right
order of adverbs in the French examples in (15). All the adverbs occur in the
structural space between the auxiliary and the participial verb (the so-called
Mittelfeld).11
(15) a. Jean a franchement probablement spontanément souvent
Jean has frankly probably spontaneously often
mangé une pomme.
eaten an apple
‘Frankly Jean probably spontaneously often ate an apple.’
b. Jean n’a pas déjà souvent à peine mangé une pomme.
Jean has not already often hardly eaten an apple
‘Jean did not already often hardly eat an apple.’
c. Jean n’a franchement probablement pas récemment souvent
Jean has frankly probably not recently often
refait son travail.
redone his work
‘Frankly Jean probably did not recently often do his work again.’
d. Jean a spontanément souvent péniblement refait son travail
Jean has spontaneously often painstakingly redone his work
‘Jean spontaneously often painstakingly did his job again.’

The hierarchy exemplified in (15) relies on the clause structure in (16).

 A feature-based theory of adverb syntax 

(16) CP

SubjP

DP Subj+

Subj ModPspeech-act

AdvP ModPepistemic

AdvP ModPvolitional

AdvP NegP

AdvP TP

AdvP AspPfrequentative

AdvP AspPproximative
Jean a
franchement
probablement AdvP MannP
spontanément
pas AdvP AuxP
récemment
souvent
à peine VP...
péniblement mangé une pomme

Ernst (2002) points out some cases that apparently violate Cinque’s hierarchy
in (6) and (16). They involve functional adverbs displaying what Ernst calls a
“loose hierarchy”. A particular example is the frequency adverb souvent, which
seems to have a wide range of positions within the clause structure. For in-
stance, it can occur below negation, as predicted by (16), but also above it. The
two sentences in (17) are equally grammatical with the expected interpretation.
(17) a. Jean n’a pas souvent pleuré aux enterrements.
Jean has not often cried at.the burials
‘Jean did not often cry at the burials.’
b. Jean n’a souvent pas pleuré aux enterrements.
Jean has often not cried at.the burials
‘Jean often did not cry at the burials.’
 Christopher Laenzlinger

As suggested by Cinque (1999: Chapter 4), the frequency adverb targets two
possible positions: one is event-related (‘there is a frequent event which is John
not crying at the burials’) and the other is process-related (‘there is a non-
frequent event which is John crying at the burial’). The former position is
situated higher than NegP (but lower than ModP) and the latter is realized
as in (16). The structure of (17b) containing a modal adverb is given in (18).
(18) [SubjP Jean n’a [ModP probablement [AspP souvent [NegP pas [AuxP [VP pleuré
aux enterrements.]]]]]]

Ernst (2002: 357–366) points out that adverbs of the same functional class
(say AspPhigh ), such as déjà ‘already’ and de nouveau ‘again’, may vary in their
respective position, as shown in (19) for French.
(19) a. Jean a déjà de nouveau mangé une pomme.
Jean has already again eaten an apple
‘Jean already ate an apple.’
b. Jean a de nouveau déjà mangé une pomme.
Jean has again already eaten an apple
‘Jean again already ate an apple.’

Ernst takes facts like (19) as evidence against Cinque’s tight-fit approach to
adverb hierarchy. However, similarly to what is proposed for souvent, the repet-
itive adverb de nouveau can be merged in two different positions according to
(6): one is below TPanterior hosting déjà and the other is higher than TPanterior .
Hence, the order variation observed in (19) is expected.
In short, the so-called loose hierarchy only occurs when a given adverb is
assigned two positions among the adverb-related functional projections of the
clause. A well-known case is the ambiguous adverb intelligemment ‘cleverly’,
which has a manner reading in (20a) and a (subject-oriented) factive/event-
related reading in (20b).
(20) a. Jean n’a pas intelligemment répondu à la question.
Jean has not cleverly answered to the question
‘Jean did not cleverly answer the question.’
b. Jean n’a intelligemment pas répondu à la question.
Jean has cleverly not answered to the question
‘Jean cleverly did not answer the question.’

In (20b) the adverb occupies Spec-ModPability above NegP and entails the read-
ing ‘John is intelligent in not answering the question’. In (20a) the adverb
merges as Spec-MannP below NegP and is assigned the reading ‘John did
 A feature-based theory of adverb syntax 

not answer the question in an intelligent way’. The two distinct positions of
intelligemment are represented in (21).
(21) [SubjP Jean [ModP intelligemment [NegP pas [MannP intelligemment [AuxP [VP
répondu à la question.]]]]]]

To sum up, French displays the following structural and derivational prop-
erties: (i) short V-movement of auxiliaries and simple verb from Aux/Infl to
Subj; (ii) vP/VP-raising to Spec-AuxP/InflP; (iii) argument raising to Spec-
SubjP (subject) and to Spec-ObjP (complements) and (iv) ambiguous merger
of adverbs like intelligemment, de nouveau, souvent, etc.
In the next section, I will discuss the distribution of adverbs and argu-
ments in the German Mittelfeld. I will show from a comparative point of view
that the structural properties established for French also hold for German.
The latter differs from the former in that it displays high scrambling and verb
movement to C.

. German

.. General structural properties

In accordance with Kayne’s (1994) LCA, I follow the hypothesis that German
is basically an SVO language. Thus, German is assigned the same structural
representation of functional projections as French, as in (16) (see also (8)).
Despite basic SVO order in German, the verb can surface in a sentence-final
position in some contexts, as for instance in the embedded clause in (22a).
(22) a. weil der Mann das Buch gelesen hat
because the man the book read has
‘because the man read the book’
 Christopher Laenzlinger

b.
CP

C° SubjP

DP ...ObjP

...AuxP

VP Aux+

Aux ... VP

V DP

weil der Mann das Buch gelesen hat

The tensed auxiliary occupies a sentence-final position, and is immediately pre-

ceded by the participial verb. The direct object occurs on the left of the two
verbal forms. On the basis of the clause structure in (22b), the surface order is
obtained after VP-movement to Spec-AuxP and object raising. VP-movement
applies here in the same way as in French. The auxiliary remains on the right of
the participle, because it does not raise to Subj. One of the parametric differ-
ences between German and French lies in Aux-to-Subj movement. As for object
raising, it must be effected to Spec-ObjP. In simple tense, as in (23), VP-raising
to InflP operates vacuously, since there is no lexical material in Infl.
(23) [CP weil [SubjP der Mann [ObjP das Buch [InflP [VP las ]]]]]

German also displays object scrambling past the subject (Frank, Lee, & Ram-
bow 1991; Grewendorf & Sternefeld 1990; Haider 1993; Moltmann 1991), as in
(24a). Laenzlinger (1998: 275–278) proposes that the object in a high scram-
bling position targets the specifier position of a topic phrase, which is not
related to the CP-system, but to the IP-system.12 Laenzlinger calls this phe-
nomenon “discourse topicalization” (cf. “aboutness-topic”, Frey 2000). The
structure for (24a) is given in (24b).
(24) a. weil dieses Buch der Mann gelesen hat
because this book the man read has
‘because the man read this book’
b. [CP weil [TopP das Buch [SubjP der Mann [InflP [VP gelesen ] hat]] ] ]
 A feature-based theory of adverb syntax 

Consider now the occurrence of adverbs in the Mittelfeld. As proposed for

French, the German clause structure contains the set of functional projec-
tions in (16) on the basis of (8). Given the derived DP-positions in (22b) and
(24b), the German Mittelfeld is assigned the structure in (25b) for the sen-
tence in (25a) involving a modal adverb, negation, an aspectual adverb, and a
manner adverb.
(25) a. weil (diese Sonate) der Mann wahrscheinlich (diese Sonate)
because this sonata the man probably this sonata
nicht oft gut gespielt hat
not often well played has
‘because the man probably not often played the sonata well’
 Christopher Laenzlinger

b.
CP

C° TopP

SubjP

ModP

ObjP

NegP

AspP

MannP

AuxP

VP Aux+

Aux ‚P

DP VP

V DP
wahrscheinlich
weil der Mann nicht oft gut gespielt hat
diese Sonate diese Sonate

As shown in (26a), (tensed) clausal object complements remain in their

postverbal position. This configuration derives from remnant VP-movement:
the object complement is first extracted from the verbal projection before
the vP raises to Spec-AuxP/InflP. The landing site of this object movement
is a functional projection below AuxP, namely an object position (ObjP)
reserved for hosting sentential complements.13 The relevant derivation is
represented in (26b).
(26) a. weil der Mann gesagt hat, daß er kommen wird
because the man said has that he come will
 A feature-based theory of adverb syntax 

‘because the man said that he would come’

b. [CP weil [SubjP der Mann [AuxP [vP gesagt]i hat [ObjP [CP dass er kommen
wird]j [vP t i [CP t j ]]]]]]

The next section is devoted to a more detailed discussion of the interaction of

adverb positioning with object scrambling in the Mittelfeld.

.. Positions for adverbs and scrambled arguments in the Mittelfeld

Cinque’s (1999) adverb hierarchy holds for German too, as shown in (27).
These sentences are based on the hierarchy in (8), partially repeated in (6).
(6) MoodPspeech-act offen gestanden ‘frankly’
< ModPepistemic wahrscheinlich ‘probably’
< ModPvolitional freiwillig ‘willingly’
< NegPnegation nicht ‘not’
< TPanterior schon ‘already’
< AspPfrequentative oft ‘often’
< AspPproximative kaum ‘hardly’
< MannPverb-oriented schwerfällig ‘painstakingly’
(27) a. weil Hans offen gestanden wahrscheinlich freiwillig einen
because Hans frankly probably willingly an
Apfel oft gegessen hat
apple often eaten has
‘because frankly Hans probably spontaneously often ate an apple’
b. weil Hans einen Apfel nicht schon kaum gegessen hat
because Hans an apple not already hardly eaten has
‘because Hans did not already hardly eat an apple’
c. weil Hans offen gestanden wahrscheinlich seine Arbeit nicht
because Hans frankly probably his work not
schon beendet hat
already finished has
‘because frankly Hans probably did not already finish his work’
d. weil Hans freiwillig seine Arbeit oft schwerfällig
because Hans spontaneously his work often painstakingly
beendet hat
finished has
‘because Hans spontaneously often finished his work painstakingly’

The direct object complement, which has a specific (non-existential) in-

terpretation (Diesing 1992; de Hoop 1993), is located above the nega-
tion and the VP-adverbs, but below the modal adverbs. Thus, the projec-
 Christopher Laenzlinger

tion ObjP is situated between ModPs and NegP, as shown in (28). Each
adverb merges as the specifier of its corresponding functional projection,
namely MoodPspeech-act , ModPepistemic, ModPvolitional, NegP, TP, AspPfrequentative ,
AspPproximative , and MannP.
(28) CP

C° SubjP

DP ModPspeech-act

AdvP ModPepistemic

AdvP ModPvolitional

AdvP ObjP

DP NegP

AdvP TP

AdvP AspPfrequentative

AdvP AspPproximative
weil Hans
offen gestanden
wahrscheinlich AdvP MannP
freiwilling
einen Apfel
seine Arbeit AdvP AuxP
nicht
schon VP Aux+
oft
kaum
schwerfällig Aux trace
gegessen hat
beendet

Most interestingly, one can observe that the structure and the derivation of
compound tensed sentences differ minimally in French and German, namely
in: (i) the positioning of ObjP and (ii) movement to Subj.
Like French, German displays cases of a “loose” adverb hierarchy with am-
biguous items like oft or wieder. For instance, these adverbs can occur either
above the negation or below it, as illustrated in (29).
 A feature-based theory of adverb syntax 

(29) a. Hans hat (oft) nicht (oft) während der Beerdigung geweint.
Hans has often not often during the burials cried
‘Hans (often) did not (often) cry at the burials.’
b. Hans hat (wieder) seine Arbeit nicht (wieder) gemacht.
Hans has again his work not again done
‘Hans (again) did not do his work (again.)’

The adverbs in parenthesis in (29) can occur in two distinct structural posi-
tions. The aspectual adverbs can merge as the Spec of an aspectual projection
(AspP) which is either higher or lower than NegP.14
As regards sentential adverbs, they must occur above the negation, al-
though they can be placed in a position lower than the specific direct object, as
illustrated in (30a–c).
(30) a. weil Hans die Sonate wahrscheinlich / offen gestanden (nicht)
because Hans the sonata probably / frankly (not)
gespielt hat
played has
‘because Hans frankly/probably (did not) play the sonata’
b. weil Hans seine Arbeit freiwillig (nicht) gemacht hat
because Hans his work voluntarily (not) done has
‘because Hans voluntarily (did not do) did his work’

The configuration [Objspecific Advsentential Neg] derives from direct object rais-
ing to a high scrambling position. This position is situated at the top of
the highest adverb-related functional projection, immediately below SubjP, as
represented in (31).
(31) weil [SubjP Subject [ObjP Direct Object [MoodP offen gestanden [ModP wahr-
scheinlich [ [NegP nicht [AuxP [VP V ] Aux ]]]]]]]

Additional ObjP projections must be postulated in the Mittelfeld, among the

adverb-related functional projections, to host the direct object in the various
scrambled positions in (32).
(32) weil Hans (die Sonate) offen gestanden (die Sonate) wahrscheinlich
because Hans the sonata frankly the sonata probably
(die Sonate) oft (die Sonate) nicht gespielt hat
the sonata often the sonata not played has
‘because frankly Hans probably often did not play the sonata’

At least four distinct ObjP projections must be established in the Mittelfeld for
(32), as represented in (33).15
 Christopher Laenzlinger

(33) weil [SubjP Hans [ObjP die Sonate [MoodP offen gestanden[ObjP die Sonate
[ModP wahrscheinlich [ObjP die Sonate [AspP oft [ObjP die Sonate [NegP nicht
[AuxP gespielt hat ]]]]]]]]]]

Consider now the placement of adverbs in the pre-subject domain of the Mit-
telfeld, as illustrated in (34)
(34) a. weil offen gestanden / wahrscheinlich / sorgfältigerweise Hans
because frankly/ probably carefully Hans
die Sonate gespielt hat
the sonata played has
‘because frankly / probably / carefully Hans played the sonata’
b. weil oft / *nicht / schwerfällig Hans die Sonate gespielt
because often / *not / painstakingly Hans the sonata played
hat
has
‘because often / *not / painstakingly Hans played the sonata’

One can observe that adverbs (except negation) can be scrambled in a pre-
subject position, which is the same position as the one occupied by the scram-
bled object in (25a). Thus, adverbs can move from their position of interpreta-
tion to a topic position internal to the Mittelfeld, as represented in (35).
(35) [CP weil [TopP offen gestandeni [SubjP Hans [MoodP t i [ObjP die Sonate nicht
gespielt hat]]]]]
[TopP wahrscheinlichi [ModP t i ]]
[TopP sorgfältigerweisei [ModP t i ]]
[TopP oft i [AspP t i ]]
[TopP schwerfällig i [MannP t i ]]

In double object constructions the combinatory possibilities between argu-

ments and adjuncts increase exponentially. It is often pointed out in the lit-
erature (see Haeberli 1993 and references cited therein) that the subject and
the objects can scramble in any order, as schematized in (36).16
(36) dass {der Mann / diesen Brief / meinem Onkel} {diesen
that {the-nom man/ this-acc letter / my-dat uncle} {this
Brief / meinem Onkel / der Mann} {meinem Onkel / der Mann /
letter / my uncle / the man} {my uncle / the man /
diesen Brief } gesendet hat
this letter} sent has
‘that the man sent this letter to my uncle’
 A feature-based theory of adverb syntax 

A modal adverb like wahrscheinlich ‘probably’ or a temporal adverb like schon

‘already’ can occupy any position among the arguments in the Mittelfed in
(36), as indicated in (37).
(37) dass (wahrscheinlich / schon) {der Mann / diesen Brief /
that (probably / already) {the-nom man / this-acc letter /
meinem Onkel} (wahrscheinlich / schon) {diesen Brief / meinem
my-dat uncle} (probably / already) {this letter / my
Onkel / der Mann} (wahrscheinlich / schon) {meinem Onkel / der
uncle / the man} (probably / already) {my uncle / the
Mann / diesen Brief } (wahrscheinlich / schon) gesendet hat
man / this letter} (probably / already) sent has
‘that (probably) the man (already) sent this letter to my uncle’

The adverbs and complements that precede the subject merge as Specs of re-
cursive topic projections occurring between CP and SubjP, as represented in
(38b) for the example in (38a).
(38) a. weil wahrscheinlich / schon diesen Brief der Mann meinem Onkel
gesendet hat
b. [CP weil [TopP wahrscheinlich / schon [TopP diesen Brief [SubjP der Mann
[ObjP meinem Onkel . . . ]]]]]

As regards scrambling below the subject (i.e. SubjP) I assume that the Mit-
telfeld contains a pair of unordered object positions (i.e. ObjP) among every
adverb-related functional projection, as shown in (39).
 Christopher Laenzlinger

(39) CP

C SubjP

DP ObjP

weil Hans DP ObjP

í
DP ModP

diesen Brief/meinem Onkel AdvP ObjP

wahrscheinlich DP ObjP

í
DP AspP

diesen Brief/meinem Onkel AdvP ObjP

schon DP ObjP

í
DP AuxP

diesen Brief/meinem Onkel VP ...

To sum up, the German Mittelfeld is composed of rigid adverb-related func-

tional projections, a fixed subject position, recursive topic projections between
the complementizer and the subject position, and unordered object positions
between the functional projections. A series of transformations applies to the
German Mittelfeld. The verb moves as a remnant VP to Spec-AuxP/InflP, as in
French. The object arguments are extracted from VP and raise to object po-
sitions, lower than negation for non-specific DPs, and higher than NegP for
specific DPs. Numerous object positions in the Mittelfeld serve as landing sites
for argument scrambling. Finally, adverbs as well as object arguments can move
to a topic domain between the complementizer and the subject position.
French and German share important structural properties, such as a strict
hierarchy of functional projections, a fixed subject position, an operation of
verb projection raising, and the phenomenon of object scrambling.
The next section is devoted to a discussion of adverb distribution in the
Nachfeld, i.e. the domain corresponding to the VP-shell structures. The anal-
 A feature-based theory of adverb syntax 

ysis will focus on French and English, since adverbs are not allowed in the
Nachfeld in German.

. Adverbs in the Nachfeld: The case of French (and English)

. The order [verb adverb complement]

Following Larson’s (1988) VP-shell structure, a verb’s arguments all root-merge

with the VP-shell (Chomsky 1995). The subject merges as Spec-vP, the direct
object as Spec-VP and the indirect object as Compl-V. The verbal head (V)
raises to v following Chomsky (1995: 315). Contrary to Ernst (2002) I assume
that no adverb merges with the VP-shell structure, which delimits the verb’s
thematic domain. Adverbs are confined to a functional domain going from vP
(not included) to SubjP (not included). MannP is the functional projection
which is the nearest to vP (for manner adverbs). As proposed in Section 3.1,
the verb raises as vP to AuxP/InflP, after the arguments have all been extracted
from the VP-shell structure. The subject raises to Spec-SubjP, while the objects
move to Spec-ObjP, as represented in (40).17
(40) [SubjPDP[Subj+Subj[AuxP/InflPVP[Aux+/Infl+Aux/Infl[ObjPDP[ObjPPP

[íP t[í+V[VP t[V+V t]]]]]]]]]]

í

This analysis in terms of remnant/pied-piping VP-movement has the advan-

tage over the analysis in terms of V-movement of accounting for the distribu-
tion of adverbs in the Nachfeld. Consider first the intervention of the adverb
between the participial verb and its complement(s), as illustrated in (41).
(41) a. Jean a donné gentiment une pomme (à Marie.)
Jean has given kindly an apple (to Marie)
‘Jean kindly gave an apple to Mary.’
b. Jean a présenté récemment son amie à ses parents.
Jean has introduced recently his friend to his parents
‘Jean recently introduced his (girl)friend to his parents.’

As already pointed out in Section 3.1, these adverbs can occur in a preverbal
position, as in (42). The configuration [Aux Adv V Obj] is obtained after VP-
movement to Spec-AuxP situated below TP and MannP.
 Christopher Laenzlinger

(42) [CP [SubjP Jean [Subj+ a [TP récemment [MannP gentiment [AuxP [VP donné]
[ObjP une pomme [ObjP à Marie [VP ]]]]]]]]]

The order [Aux V Adv Obj] in (41) results from remnant VP-movement to a
position higher than that of the adverbs, as in (43).
(43) [CP [SubjP Jean [Subj+ a [AuxP [VP donné] [TP récemment [MannP gentiment
[ObjP une pomme [ObjP à Marie [VP ]]]]]]]]]

The adverbs récemment and gentiment can also be placed between the two
objects, as exemplified in (44).
(44) a. Jean a donné une pomme gentiment à Marie.
Jean has given an apple kindly to Marie
‘Jean gave an apple kindly to Mary.’
b. Jean a présenté son amie récemment à ses parents.
Jean has introduced his friend recently to his parents
‘Jean recently introduced his (girl)friend to his parents.’

The configuration [Aux V Objdirect Adv Objindirect ] in (44) arises from move-
ment of the direct object to an object position situated higher than the func-
tional projections hosting the adverbs. This is represented in (45).18,19
(45) [CP [SubjP Jean [Subj+ a [AuxP [VP donné [ObjP une pomme [TP récemment
[MannP gentiment [ObjP à Marie [VP ]]]]]]]]]

From a comparative point of view, English differs from French in having a

so-called adjacency constraint on Case assignment (Stowell 1981). As shown
in (46a), an adverb may not intervene between the verb and the direct object
DP in English, as opposed to the corresponding French sentence in (41). In
contrast, the adverb may intervene between the verb and the complement when
the latter is a prepositional phrase, as in (46b).
(46) a. John has given *kindly / *recently an apple (to Mary.)
b. John has spoken kindly / recently to Mary.

In Section 3.1, I assumed that VP and object complements also move to Spec-
InflP/AuxP and Spec-ObjP, respectively, in English. The only condition on the
direct object is that it occur in the ObjP immediately dominated by InflP/AuxP
(the so-called adjacency constraint between V and nominal direct object in the
sense of Stowell’s (1981) Case resistance). In such a configuration there is no
position for an adverb to intervene between the verb and the direct object. As
for the indirect object PP, it can occupy the Spec-ObjP position distant from
 A feature-based theory of adverb syntax 

InflP/AuxP. The contrast between a direct object ObjP and an indirect object
ObjP is represented in (47).
(47) [AuxP [vP given [VP [DP an apple]]] [TP/MannP recently / kindly [ObjP [PP to
Mary] [vP t ]]]]

There are also differences between French and Italian as to the classes of ad-
verbs that can occur near the participle within the Mittelfeld. Consider the
following paradigm.
(48) a. Jean a probablement / souvent / gentiment résolu tes
Jean has probably / often / kindly solved your
problèmes.
problems
b. Jean a résolu ? probablement 20 / souvent / gentiment tes
Jean has solved probably / often / kindly your
problèmes.
problems
‘Jean probably/often/kindly solved your problems.’
c. Gianni ha ok probabilmente / *spesso / ok gentilmente risolto i tuoi prob-
lemi.
d. Gianni ha risolto *probabilmente / ok spesso / ok gentilmente i tuoi prob-
lemi.
(cf. Gianni ha ok probabilmente risolto ok spesso / ok gentilmente i tuoi
problemi.)

The order [Aux Adv V] in (48a), which also holds for English (e.g. John has
probably often kindly solved your problems), is derived by VP-raising to the
Spec-AuxP situated below the adverb-related functional projections. In (48b)
VP-raising applies to a domain higher than these functional projections.21 In
Italian the possibilities of VP-raising vary according to the adverb-related func-
tional projection involved. As shown in (48d), VP-raising cannot apply to a do-
main higher than ModP, while it can apply to domains higher than AspPfrequency
and MannP. Following (48c) the VP can raise to a position below ModP and
MannP, but not below AspPfrequency . The derivation possibilities for VP-raising
in Italian are represented in (49).
(49) . . . [SubjP [*AuxP [ModP probabilmente [AuxP [VP risolto] [AspP spesso [*AuxP
[MannP gentilmente [AuxP [VP risolto. . . ]

Thus, the parametric difference between French and Italian lies in the pos-
sibility of VP-raising to AuxP in specific domains in the Mittelfeld. In French
 Christopher Laenzlinger

these domains are identified as a functional space between every adverb-related

functional projection, whereas in Italian AuxP is confined to a domain between
ModP and AspP, and between MannP and VP.22
To end this section, let us examine the case of mirror image order of ad-
verbs, as in (50b), in comparison with the left-to-right linear order in (50a)
derived from Cinque’s hierarchy.
(50) a. Jean a souvent gentiment donné une pomme à Marie.
Jean has often kindly given an apple to Marie
‘Jean often kindly gave an apple to Mary.’
b. Jean a gentiment donné souvent une pomme à Marie.
Jean has kindly given often an apple to Marie
The adverbs in (50a) merge as Spec of their corresponding functional projec-
tion, while the verb raises as a remnant VP to AuxP above the adverb-related
functional projections. The mirror image sequential order in (50b) derives
from a kind of “snowballing” pied-piping XP-movement. It is not the VP cat-
egory that raises to AuxP above AspP, but the MannP category, as represented
in (51). Note that before MannP raising the objects have been extracted from
VP to “scrambling” object positions occurring between MannP and AspP.
(51) [SubjP Jean [Subj+ [Subj a] [AuxP [MannP gentiment donné] [AspP souvent [ObjP
une pomme [ObjP à Marie [MannP t] ]]]]]]

This kind of pied-piping XP movement will be further discussed in the next

section, which deals with the distribution of adverbs in sentence-final position,
i.e. on the right of the complements.

. Sentence-final adverbs

The examples in (52) involve a temporal adverb, an aspectual adverb, and a

manner adverb in a sentence-final position.23
(52) a. Jean a lu un livre récemment.
Jean has read a book recently
‘Jean read a book recently.’
b. Jean a lu la Bible souvent.
Jean has read the Bible often
‘Jean read the Bible often.’
c. Jean a embrassé Marie tendrement.
Jean has kissed Marie tenderly
‘Jean kissed Marie tenderly.’
 A feature-based theory of adverb syntax 

The adverbs in (52) surface in a sentence-final position after VP-raising to

Spec-AuxP and object raising to Spec-ObjP in a domain higher than the rele-
vant adverb-related functional projections. This is represented in (53).24
(53) [SubjP Jean a [AuxP [VP lu] [ObjP la Bible [TP récemment [AspP souvent [MannP
tendrement. [VP ]]]]]]]

Interestingly, the verb and its object cannot raise past functional projections
hosting low aspectual and light adverbs (cf. *Jean a lu la Bible beaucoup/bien
à Marie ‘Jean read the Bible a lot/well to Marie’). Thus, these adverbs are not
expected in sentence-final position.25
As exemplified in (54), more than one adverb can occur on the right of the
complement(s), either in a linear order respecting Cinque’s hierarchy (54a–b)
or in a mirror image order (54c–d).
(54) a. Jean a embrassé Marie souvent tendrement.
Jean has kissed Marie often tenderly
‘Jean kissed Marie often tenderly.’
b. Jean a mangé une pomme parfois entièrement.
Jean has eaten an apple sometimes entirely
‘Jean sometimes ate an apple entirely.’
c. Jean a mangé une pomme entièrement récemment.
Jean has eaten an apple entirely recently
‘Jean recently ate an apple entirely.’
d. Jean lit la Bible attentivement maintenant.
Jean reads the Bible carefully now
‘Jean is carefully reading the Bible now.’

The linear order of adverbs (54a–b) results from VP-raising to AuxP and object
movement to ObjP past the two adverb-related projections (AspP and MannP),
as represented in (55).
(55) a. [SubjP Jean a [AuxP [VP embrassé] [ObjP Marie [AspP souvent [MannP ten-
drement [VP . . . ]]]]]]
b. [SubjP Jean a [AuxP [VP mangé] [ObjP une pomme [AspP parfois [AspP
entièrement [VP . . . ]]]]]]

The mirror image order in (54c–d) derives from pied-piping (i.e. snowballing)
movement of the projection containing the right-hand adverb, the verb, and
its object. The derivation for (54c), for instance, is given in (56). The projec-
tion AuxP containing the raised VP (mangé), the object (une pomme), and the
adverb entièrement moves to a functional position immediately above TP.26
 Christopher Laenzlinger

(56) ...SubjP

Jean Subj+

a FP27

TP

récemment AuxP

mangé ObjP

une pomme MannP

entièrement VP

V DP

Similarly, such pied-piping movement applies to (54d), where the complex

sub-structure (i.e. InflP) containing the verb (lit), the raised object (la Bible),
and the manner adverb (attentivement) moves past the temporal adverb
(maintenant).28
The next section is devoted to the distribution of adverbs in the Vorfeld.

. Adverbs in the Vorfeld

. The order [Adv Subj V] vs. [Subj Adv V] in Romance and English

Adverb preposing is relatively frequent in French and English. The adverbs in

(57a–b) occupy a fronted topic position above the subject position.29 Within
Rizzi’s (1997) split CP structure, they merge as Spec of TopP (labeled ModP in
Rizzi 2002).30
(57) a. [TopP Probably / amicably [SubjP John will greet the professor]]
b. [TopP Probablement / amicalement, [SubjP Jean saluera le prof.]]

In French these “high” adverbs may not intervene between the subject and
the tensed verb/auxiliary, unless they are parentheticals, as shown in (58a–d).
 A feature-based theory of adverb syntax 

In contrast, these adverbs may occur between the subject and the tensed
verb/auxiliary in English, as illustrated in (58e–f).
(58) a. *Jean probablement / amicalement saluera le professeur.
Jean probably amicably will-greet the professor
b. *Jean probablement / amicalement a salué le professeur.
Jean probably amicably has greeted the professor
c. Jean, probablement / amicalement, saluera le professeur.
‘Jean probably / amicably will greet the professor.’
d. Jean, probablement / amicalement, a salué le professeur.
‘Jean probably / amicably greeted the professor.’
e. John probably / amicably greeted the professor.
f. John probably / amicably will greet the professor.

The difference between (58a) and (58e) lies in the application of VP-raising
to InflP in French (i.e. adjacency between VP in Spec-InflP and SubjP), but
not in English. Hence, any clause-internal adverb can intervene between the
subject and the verb in English (e.g. John kindly left). As regards (58b), an ad-
verb cannot intervene between the subject in Spec-SubjP and the verb, because
the latter raises to Subj (cf. [SubjP Jean [Subj a] [ModP probablement dormi]] ‘Jean
has probably slept’). As for (58f), one is led to assume that the modal may, but
need not, raise to Subj : [SubjP John [Subj will] [ModP probably leave]] vs. [SubjP John
[Subj ] [ModP probably will leave]]. Similar facts are observed in Italian, Spanish,
and Portuguese, as illustrated in (59a), (59b), and (59c) respectively.
(59) a. Gianni probabilmente / spesso telefonerà alle 5.
‘Gianni probably / often will phone at 5.’
Gianni probabilmente / spesso ha telefonato alle 5.
‘Gianni probably / often has phoned at 5.’
b. Juan probablemente / a menudo llamará a las cinco.
Juan probablemente / a menudo ha llamado a las cinco.
c. O João provavelmente / frequentemente telefonara as 5 da tarde.
O João provavelmente / frequentemente teria telefonado as 5 da tarde.

In Romance, it is assumed that the tensed verb/auxiliary raises to a position

adjacent to the subject, say Subj. Therefore, the intervention of an adverb be-
tween the subject and the raised verb is not expected. As far as the Italian
sentences in (59a–b) are concerned, Belletti (1990) shows that the subject is
topicalized, since it is incompatible with a negative quantifier: *Nessuno prob-
abilmente/spesso telefonera. Thus, the subject Gianni in (59a) is topicalized as
Spec-TopP in the Comp domain, which leaves enough space for the insertion of
 Christopher Laenzlinger

an intervening adverb, either in Spec-ModP/AspP or Spec-TopP.31 In (58f) and

(59b–c) the subject is not topicalized, since a negative quantifier is possible in
these constructions, as for instance in English: nobody probably/often will phone
at 5. The subject occurs in its canonical position, identified as Spec-SubjP, and
the adverbs occupy the specifier position of their semantically related func-
tional projection. The configuration [Subj Adv V/Aux] is obtained after verb
projection raising to a position lower than the adverb-related functional pro-
jections, as represented in (60).
(60) SubjP

DP ModP/AspP

AdvP Mod+/Asp+

Mod/Asp AuxP

VP Aux+

Aux ...

John probably will phone

often
Juan probablemente ha llamado
a menudo
O João provavelmente teria telefonado
frequentemente

The next section deals with the very restricted distribution of adverbs in
the German Vorfeld with respect to the V2 constraint and remnant VP-
topicalization.

. Adverbs in the German Vorfeld

In German, adverbs can be preposed in the CP domain, either as a topic or as

a focus, with the tensed verb in second position (the V2 constraint). This is
illustrated in (61).
(61) a. [CP=TopP Oft [C hat] [SubjP Hans dieses Buch gelesen]].
often has Hans this book read
‘Often Hans read this book.’
 A feature-based theory of adverb syntax 

b. [CP=FocP GESTERN [C hat] [SubjP Hans dieses Buch gelesen]].

yesterday has Hans this book read
‘Yesterday Hans read this book.’

Most interestingly, adverbs can also occur in the Vorfeld by means of a transfor-
mation called “Remnant VP-topicalization”, which brings along the participial
verb and possibly some constituent(s) of the Mittelfeld to a fronted position, as
in (62). In the Principles & Parameters framework, VP-topicalization is treated
as an instance of remnant and/or pied-piping VP-movement to Spec-CP (see
den Besten & Webelhuth 1990; Haider 1993 among others) with the tensed
auxiliary in C (V2).
(62) a. [CP Das Buch nicht gelesen [C hat] [SubjP Hans gestern.]]
the book not read has Hans yesterday
‘Read the book Hans did not yesterday.’(?)
b. [CP Gelesen [C hat] [SubjP Hans das Buch gestern.]]
read has Hans the book yesterday
‘Read the book Hans did yesterday.’(?)

According to (63a–b) VP-adverbs and event-related adverbs can be fronted

with the participial verb and the scrambled direct object, whereas sentential ad-
verbs, as in (63c), cannot be involved in pied-piping VP-topicalization. Adverbs
can be used as clues for identifying the categorical status of the FP projection
involved in (63).
(63) a. [CP [FP Die Sonate gut gespielt] [C hat] Hans.]
The sonata well played has Hans
‘Play the sonata well Hans did.’
b. [CP [FP Schnell / gern / oft / gestern / kürzlich die Sonate
Quickly / willingly / often / yesterday / recently the sonata
gespielt] [C hat] Hans.]
played has Hans
‘Quickly / willingly / often / yesterday / recently play the sonata Hans
did.’
c. *[CP [FP Leider / glücklicherweise / vielleicht die Sonate gespielt]
Unfortunately / fortunately maybe the sonata played
[C hat] Hans.]
has Hans
The borderline for VP-topicalization can be identified as the functional pro-
jection immediately below ModP. Thus, any clause-internal projection below
ModP can be fronted. For instance, in (62b) the projection that is topicalized
 Christopher Laenzlinger

contains only the participle. It can be identified as the VP (or the AuxP) from
which the object has been extracted. In (62a) the topicalized projection, which
contains the scrambled object and negation, can be equated with the ObjP
immediately above NegP. As for (63a-b), the displaced FP constituent can be
identified as a functional projection below ModP containing the participle, the
direct object, and the relevant adverb. Any clause-internal projection below
ModP can be potentially topicalized in Spec-CP, as represented in (64).32
(64)

VP-topicalization can involve an adverb while leaving another one behind, as

in (65a–b). In these examples the adverb hierarchy is preserved since the top-
icalized projection contains an adverb whose position is lower than that of
the adverb left behind. Thus, scope can be correctly computed at the level
of the trace, that is, in the Mittelfeld.33 In contrast, the examples in (65c–d)
are ungrammatical because VP-topicalization cannot take a functional domain
containing a high adverb while leaving a low adverb behind. As shown by the
location of the trace in (65c–d) such a derivation is impossible.34
(65) a. Schnell die Sonate gespielt hat Hans oft [t].
Quickly the sonata played has Hans often
‘Quickly play the sonata Hans did often.’
b. Sorgfältig die Sonate gespielt hat Hans nicht [t].
Carefully the sonata played has Hans not
‘Carefully play the sonata Hans did not.’
c. *Oft die Sonate gespielt hat Hans [t] schnell.
Often the sonata played has Hans quickly
d. *Nicht die Sonate gespielt hat Hans [t] sorgfältig.
Not the sonata played has Hans carefully
To sum up, the Vorfeld in German can host adverbs in first position (Spec-CP)
and the verb in second position (C). The fronted adverbs are either topical-
ized or focalized. Adverbs can also occur in the Vorfeld by means of remnant
and pied-piping VP-topicalization. Formally, any clause-internal functional
projection below ModP can be topicalized in Spec-CP.
 A feature-based theory of adverb syntax 

. Conclusion

The question of adverb syntax has been tackled from a comparative point of
view by analyzing the distribution of adverbs in French and German within
the three domains of the clause: the Vorfeld, the Mittelfeld, and the Nachfeld.
Adverbs are regarded as unique specifiers of semantically related func-
tional projections (basically MoodP < ModP < TP < AspP). The hierarchy of
clause-internal functional projections dictates the hierarchy of co-occurrences
of adverbs. In addition to adverb-related functional projections, the Mittelfeld
structure contains argument-related projections (ObjP) and lexico-functional
projections (AuxP, InflP). The VP-shell structure stands for the Nachfeld, while
the split-CP layer constitutes the Vorfeld.
I have shown that the clause structure in French and German contains
the same set of functional projections. The adverbs and the subject have a
fixed position within the clause structure. Both German and French display
verb projection (VP) raising to AuxP/InflP and argument scrambling. Pre-
sumably, there are recursive chunks of structures (i.e. SubjP>InflP>ObjP)
among adverb-related functional projections made available by UG. German
and French differ in (i) the range of positions for AuxP/InflP and ObjP (high
scrambling for German, i.e. ObjP must dominate AuxP/InflP, and low scram-
bling for French, i.e. AuxP/InflP must dominate ObjP) and (ii) Infl/Aux move-
ment to Subj (in French) and to C (V2 in German). Scrambling movement
targets different landing sites in French and German, which accounts for the
variations of positioning observed in the two languages among co-occurring
adverbs and derived arguments.
When adverbs occur in the Nachfeld in French, their distribution interacts
with that of complements. The various positions of adverbs with respect to
objects are derived from transformations like object extraction from VP and
pied-piping VP-raising. Co-occurring adverbs in the Nachfeld either respect
Cinque’s hierarchy of functional projections or display the mirror image order
of this hierarchy. The former case results from cyclic object-movement and
VP-raising, whereas the latter case derives from “snowballing” pied-piping VP-
raising (involving MannP, AspP, etc.). Thus, high adverbs can surface on the
right of low adverbs, since their respective scope properties can be recovered
by reconstruction. Adverbs do not occur in the Nachfeld in German because
(non-CP) arguments are obliged to scramble past AuxP/InflP.
Finally, the distribution of adverbs in the Vorfeld has been analyzed by
means of topicalization in the left periphery of the clause. Adverbs in the
French Vorfeld are structurally prominent (as Spec-TopP). No adverb is al-
 Christopher Laenzlinger

lowed to intervene between the subject and the tensed verb/auxiliary, since the
former occurs in Spec-SubjP, while the latter raises to Subj. In English, Spanish,
and Portuguese, the tensed verb/auxiliary does not need to move up to Subj,
hence making adverb intervention possible. In Italian, the sequence [Subj Adv
V] is analyzed as an instance of subject topicalization. In German, adverbs can
occur in the German Vorfeld as topics in Spec-TopP. They can also surface
in the German Vorfeld by means of remnant/pied-piping VP-topicalization,
which consists of raising any clause-internal functional projection lower than
ModP to Spec-CP.

Notes

* I would like to thank Thierry Etchegoyhen, Sandra Schwab, and Stéphanie Dürrleman
for comments on the first draft of this paper. This work is part of my thesis d’habilitation
à diriger des recherches defended on the 13th of December 2002 at the University of Lille 3
(France).
. SubjP and ObjP are projections in which the arguments can have their uninterpretable
formal features (say Case, phi-features) checked. I assume that such a checking process is
independent of verb movement.
. InflP must dominate ObjP in French, whereas ObjP must dominate InflP in German.
This may represent another parameter.
. See Nilsen (2002) for an alternative semantically based theory.
. As noted by Wilder (1999: 697), a quantificational adverb (e.g. much) can be defrozen
after having met its licensing condition, the Adv-Criterion, if it is accompanied by a wh-
feature. This is shown in the contrast between (i) and (ii).

(i) *[Much]+Q he has worked.

(ii) [How [much] +Q ] +wh has he worked?

Such a “defrozen” property results from the fact that the Q-features are superseded by the
wh-features applying to the whole constituent.
. In the spirit of Giorgi & Pianesi (1997), Laenzlinger (2000a) proposes that the clause be
assigned a minimal semantically legitimate structure, namely (i).

(i) Mood >> Modal >> Tense >> Aspecthigh >> Aspectlow >> v >> V

The order of projections inside MoodPs, ModPs, and AspPs is established by the geometry
of feature matrices associated with each major functional head. The set of features associ-
ated with the heads in (ii) is specified in the Numeration as to the hierarchical organization
among them. In the case of co-occurrences of several Mood, Modal, and Aspect phrases, as
in the co-occurrences of adverbs, the clause structure is organized into a hierarchy according
to the feature geometry in (ii).
 A feature-based theory of adverb syntax 

(ii) MoodP >> ModP >> AspP (high) >> AspP (low)
F1=speech-act F1=epistemic F1=habitual F1=perfect(I)
> F2=evaluative > F2=irrealis > F2=repetitive > F2=continuative
> F3=evidential > F3=necessity > F3=frequent. > F3=perfect(II)
> F4=possibility > F4=celerative > F4=retrospective
> F5=volitional > > F5=proximative
> F6=durative
> F7=generic/
> progressive
> F8=prospective
> F9=completive
> etc.
I assume that the clause structure proposed by Cinque (1999) is potentially available in
cases of co-occurrences of adverbs, although the minimal functional structure required by
the semantics of full propositions (i.e. Event > Process > Activity/State) is the one given in
(7).
. Chomsky (1995) proposes the system of Bare Phrase Structure as an alternative way of
deriving phrase structures from non-primitive entities. Chomsky’s proposal runs counter to
those of Kayne’s antisymmetry system. Multiple specifiers are allowed (by successive set-
Merge), and adjunction is available (by pair-Merge). According to this model of phrase
structure, adverbs could be either specifiers or adjoined phrases. Adjunction has been tenta-
tively formalized following the Principles & Parameters model of X’-theory (see May 1985;
Chomsky 1986). For instance, the structure preserving constraint forces XP to adjoin to
maximal projections and X◦ to heads. Adjunction to the X’-level is not permitted, contrary
to what Ernst (2002) proposes in his scope-based theory of adverbs. In this paper, however,
I follow the feature-based theory of adverbs, without free adjunction, observing the LCA.
. PP adverbials will not be treated here. Cinque (2002) argues that like adverbs DP/PP-
adverbials are licensed in Spec-positions, but the latter surface in positions different from
those of true adverbs because they must also be licensed for Case along the lines of Kayne’s
(2000) analysis of prepositional phrases. Note that some adverbial PPs like à peine, sans
doute, au mieux, etc. can be classified as adverbs given their distributional properties. True
PP adverbials tend to be parentheticals in the position between the auxiliary and the partici-
ple in French, as illustrated in the contrast between (i) and (ii). See also Note 17:

(i) Jean a sans doute / à peine mangé la pomme

Jean has doubtlessly / hardly eaten the apple

(ii) Jean a, à mon avis / dans une certaine mesure, mangé la pomme
Jean has, in my opinion / to some extent eaten the apple
. Selection can be understood here either as morphological (Verb+Inflection) or func-
tional (Aux+Verb). Note that selection can be established in two configurations: (i) in a
head-complement relation realized under a local Agree (Chomsky 2000) relation (previ-
ously head-government) and lexically specified (Aux/Infl, [__, VPpp ]) (ii) in a spec-head
relation triggered by overt selectional feature checking.
 Christopher Laenzlinger

. The computational cost of remnant VP-movement in comparison with V-movement

is balanced by the wide range of data that can be treated in Romance and Germanic by
independent object movement, notably with respect to adverb placement (see Section 3.1.2).
. See Note 1. PP-objects are subject to a particular Case checking process, which applies to
a domain external to VP following Kayne (2002). For purposes of simplification, the object
positions related to direct objects (DP) and indirect objects (PP) are not distinguished here
(namely ObjPs).
. There is a limit on the number of co-occurring adverbs in the Mittelfeld for reasons
that do not concern internal (syntactic, semantic) factors, but external factors (human
processing, prosody).
. There exist alternative analyses, such as an adjunction analysis (Müller & Sternefeld
1993; Haider 1993) or an analysis in terms of A-position (Moltmann 1991; Haeberli 1993).
Laenzlinger (1998) proposes that this topic position be a non-operator A’-position.
. The object position hosting sentential complements is clearly below MannP, as shown
in (i) for English and (ii) for French.

(i) He said (ok kindly) that he would be late (*kindly)

(ii) Il a dit (ok gentiment) qu’il serait en retard (*gentiment)

The sequence generating (i) and (ii) must be the following:

(iii) [InflP/AuxP VP [MannP gently / gentiment [ObjP CP [vP ]]]]

The fact that CP-complements must evacuate the VP-shell implies that they too have unin-
terpretable formal features to get checked outside VP. One can observe that these features
are checked in a low Object position. I leave this question for future research.
. The negation (i.e. NegP) functions as the borderline between sentential adverbs
(proposition- and event-related adverbs) and VP-adverbs (process- and activity-related
adverbs). It also serves to mark the domains of specific and non-specific interpretation
(Diesing 1992; de Hoop 1993). Thus, the indefinite complement in (i) below NegP and
low AspP (oft) must have an existential reading. As represented in (ii), the non-specific DP
occupies an object position lower than NegP and AspPlow .

(i) weil Hans nicht oft Sonaten gespielt hat

because Hans not often Sonatas played has
‘because Hans did not often play Sonatas’

(ii) weil [SubjP Hans [NegP nicht [AspP oft [ObjP Sonaten [AuxP [VP gespielt hat ]]]]]]

. The adverb sorgfältigerweise must have a subject-oriented sentential reading (hence it is
root-merged as Spec-ModPability ).
. There are conditions to be posed on this kind of scrambling related to specificity
(Diesing 1992; de Hoop 1993), focus, and pragmatic contexts (Abraham 1986). Note that
some German speakers are reticent about free argument scrambling in the sense that they
accept a limited number of XP in front of the subject.
 A feature-based theory of adverb syntax 

. In this paper I do not discuss the distribution of DP and PP adverbials, for I consider
that for the most part they belong to a distributional class different from that of true adverbs
(Adv). (See also Note 7). As proposed by Larson (1985) and Haider (2000), they may be
licensed in an (extended) VP-shell structure.
. High aspectual adverbs behave in the Nachfeld in the same way as temporal and manner
adverbs, as in (i). In contrast, negation and measure (low aspectual) adverbs are prohibited
in that domain, as shown in (ii)–(iii). For mood and modal adverbs, the only acceptable
position in the Nachfeld is between the verb and the first complement, as in (iv). Finally,
“light” adverbs like the manner bien ‘well’ may not occur in the Nachfeld, as shown in (v).

(i) Jean a donné (souvent / de nouveau) une pomme (souvent / de nouveau) à

Jean has given often / again an apple often / again to
Marie.
Marie

(ii) Jean n’a donné (*pas) une pomme (*pas) à Marie.

Jean neg-has given not an apple not to Marie

(iii) Jean a lu (*beaucoup) la Bible (*beaucoup) à ses enfants.

Jean has read a lot the Bible a lot to his children

(iv) Jean a donné (évidemment / probablement) une pomme (*évidemment /

Jean has given evidently / probably an apple evidently /
probablement) à Maire.
probably to Marie

(v) Jean a lu (*bien) un livre (*bien) à Marie.

Jean has read well a book well to Marie
The placement of frequency adverbs in the Nachfeld can be analyzed in the same way as (45)
and (47), namely: the adverb-related functional projection (AspP) can be dominated by an
AuxP/InflP projection hosting the VP and by an ObjP projection hosting the object. As for
mood and modal adverbs, their corresponding functional projection is dominated only by a
projection hosting VP, not by an ObjP. Finally, NegP and AspPmeasure are dominated by none
of these projections. Thus, the complete derivational possibilities are represented in (vi).
. The two complements can permute in the Nachfeld, especially if the DO is indefinite (see
Belletti & Shlonsky 1995 for the focus condition on the order IO<DO). This is illustrated in
(i). The possibility of DO-IO/IP-DO permutation is due to the unordered character of ObjP
projections, as in German. This is represented in (ii).

(i) Jean a donné gentiment à Marie une pomme.

Jean has given kindly to Marie an apple
‘Jean kindly gave Mary an apple.’

(ii) [SubjP Jean [Subj a [AuxP [VP donné] [ManP gentiment [ObjP une pomme / à Marie [ObjP à
Marie / une pomme. . . ]]]]]]
 Christopher Laenzlinger

As expected, an adverb can intervene between the two inverted complements, as in (iii). This
means that the Obj positions within the Mittelfeld can host either the direct object or the
indirect object, as represented in (iv).

(iii) Jean a donné à Marie gentiment / récemment une pomme.

Jean has given to Marie kindly / recently an apple
(iv) [SubjP Jean [Subj a [AuxP [VP donné] [ObjP à Marie / une pomme [ManP gentiment [ObjP une
pomme / à Marie. . . ]]]]]]

. There seems to be some difference of grammaticality judgment among French speakers
as to the possibility for a modal adverb to occur between the participial verb and its com-
plement. As far as I am concerned, a mood adverb like certainement ‘certainly’ may occupy
such a position: e.g. il a donné certainement une pomme à Marie ‘he certainly gave an apple
to Marie’.
. So far the sequence [(Aux) V Adv Compl] has been analyzed as an instance of rem-
nant pied-piping VP-movement, which must be distinguished from true cases of rightward
extraposition, as illustrated in (i)–(iv).

(i) Jean a donné hier / ce matin un livre à Marie.

Jean has given yesterday / this morning a book to Marie
(ii) John read recently / yesterday a very nice book that deals with linguistics.

(iii) John talked yesterday / this morning to Mary.

(iv) Jean parlait hier / ce matin à Marie.

The sentence in (i) differs from (41a) in that the pronoun en cannot be extracted from the
direct object DP, as shown by the following contrast: *Jean en a donné hier un [e] ‘Jean gave
one of it yesterday’ vs. ok Jean en a donné récemment/gentiment un [e] ‘Jean recently/kindly
gave one of it’. This contrast can be explained by the different position of the direct object
DP in the two structures. In (41a) the direct object DP occurs in a leftward canonical Obj
position within the Mittelfeld, a kind of “properly governed” position in GB terms. The
extraposition in (i) is presumably not a syntactic leftward transformation, but a rightward
PF-transformation (see Ernst 2002: 226ff.), and no syntactic extraction can apply in the PF-
component. Similarly, the well-known case of Heavy NP-shift in (ii) is a post-Spell-Out
phenomenon, which is not subject to the syntactic adjacency constraint invoked for (46a).
The intervention of the strong adverbial (of DP category) in (iii–iv) also results from right-
ward PF-extraposition. The extraposed PP stands for a barrier to extraction, as shown by the
following case of preposition stranding in English: *Who did you talk yesterday/this morning
to [e]? vs. ok Who did you talk recently/kindly to [e]?
. This parametric difference between French and Italian does not seem to result from
morphological properties of the verb. Rather, it looks like an idiosyncratic property of AuxP.
. Mood and modal adverbs cannot occur in a sentence-final position, unless they are par-
entheticals, as in (i)–(ii), while low aspectual and light adverbs cannot occupy this position,
not even in their parenthetical use, as shown in (iii)–(iv).
 A feature-based theory of adverb syntax 

(i) Jean n’a pas lu la Bible, heureusement.

Jean neg-has not read the Bible fortunately
‘Jean did not read the Bible, fortunately.’

(ii) Jean a lu la Bible, probablement.

Jean has read the Bible probably
‘Jean read the Bible, probably.’

(iii) Jean a lu la Bible *(,)beaucoup.

Jean has read the Bible a.lot
‘Jean read the Bible a lot.’

(iv) Jean a embrassé Marie *(,)bien.

Jean has kissed Marie well
‘Jean read the Bible well.’

. Alternatively, the verb and the direct object can move together as a VP to AuxP. How-
ever, I have adopted an analysis of verb projection raising that applies independently of
object raising in order to deal with the cases of adverb intervention between the verb and
the raised object(s) (e.g. Jean a embrassé souvent Marie tendrement ‘Jean (has) often kissed
Marie tenderly’).
. As far as the light adverb bien is concerned, it cannot occur in the Nachfeld, because it
has to incorporate to the linearly adjacent verb after Spell-Out (see Section 4.1). As regards
the measure adverb beaucoup, its very restricted distribution remains mysterious (it cannot
be considered a light adverb).
. Note that this kind of snowballing pied-piping movement is limited to one-step move-
ment in French and cannot imply modal/mood adverbs, as shown in (i) and (ii).

(i) ?? Jean a lu la Bible attentivement souvent récemment.

Jean has read the Bible carefully often recently

(ii) *Jean a lu la Bible attentivement (souvent) spontanément.

Jean has read the Bible carefully often spontaneously
Laenzlinger (2000b) argues for a similar restriction on the sequence of three postnominal
adjectives realized in a mirror image order and puts forward an explanation in terms of
derivational complexity.
. This functional projection can be identified as another InflP/AuxP available in the upper
chunk of sub-structure above the temporal adverb related projection.
. This type of analysis can account for the scope ambiguity displayed by the sentence-
final adverbs in (i) below. Provided that scope can be computed at any step of the derivation
(by any member of the chain following the Scope Principle in Ernst (2002: 317)) under
c-command, the adverb intentionnellement has scope over deux fois at the root (Cinque’s
hierarchy), while after pied-piping “snowballing” movement the adverb deux fois has scope
over intentionnellement. By way of contrast, the adverbs in (ii) do not display scope am-
biguity. The scope relation intentionnellement > deux fois derives from Cinque’s hierarchy,
namely ModPvolitional > AspPmeasure , which is preserved during the derivation.
 Christopher Laenzlinger

(i) Il a frappé à la porte deux-fois intentionnellement.

he has knocked on the door twice deliberately
‘He knocked on the door twice deliberately.’
[deliberately [twice]] or [twice [deliberately]] (ambiguous)

(ii) Il a frappé à la porte intentionnellement deux-fois.

He has knocked on the door deliberately twice
‘He knocked on the door twice deliberately.’ (non ambiguous)

. This kind of topicalization differs from the topic-comment relation observed by Rizzi
(1997) for clitic left dislocation (see also Cinque 1990). Rizzi (2002) refers to it as a case
of “structural prominence”. I still assume that the fronted adverbs in (57) occupy a topic
position labeled TopP.
. As already mentioned in Section 3.2, adverb topicalization cannot involve quantifica-
tional adverbs, i.e. true operators, as shown by the ungrammaticality of (i)–(ii). Once they
have satisfied the licensing mechanism of the Adv-Criterion, these operators are “frozen” in
their checking position.

(i) *Beaucoup, Jean a lu la Bible.

Much Jean has read the Bible

(ii) *Presque, Jean a mangé sa soupe.

Almost Jean has eaten his soup

(iii) *Bien, il a mangé sa soupe.

Well he has eaten his soup

(iv) *Entièrement, Jean a lu ce livre.

Entirely Jean has read this book
Light adverbs like bien in (iii) cannot be topicalized, since they must be linearly adjacent to
the verb in order to incorporate with it after Spell-Out. The problem in (iv) lies in the pred-
ication requirement on the adverb: entièrement must be in a local c-command relation with
the object (i.e. the entire book). The adverb in (iv) is not in such a local configuration with the
direct object (cf. ok Entièrement, ce livre a été lu ‘Entirely this book has been read’). Likewise,
the adverbs in (i)–(iv) cannot be focalized (i.e. moved to FocP), as shown in the cleft/focused
sentences in (v)–(viii), for the same reasons as they cannot be topicalized: quantificational
adverbs are frozen in their clause-internal position, light adverbs require an adjacency con-
figuration with the verb, and object-oriented adverbs need to be in a local relation with the
object of the verb.

(v) *C’est [FocP beaucoup qu’ il a lu la Bible.]

It.is much that.he has read the Bible
*[FocP BEAUCOUP, il a lu la Bible.]
MUCH he has read the Bible
 A feature-based theory of adverb syntax 

(vi) *C’est [FocP presque que Jean a mangé sa soupe.]

It.is almost.that Jean has eaten his soup
*[FocP PRESQUE, Jean a mangé sa soupe.]
ALMOST Jean has eaten his soup

(vii) *C’est [FocP bien qu’ il a mangé sa soupe.]

It.is well that.he has eaten his soup
*[FocP BIEN, il a mangé sa soupe.]
WELL he has eaten his soup

(viii) *C’est [FocP entièrement qu’il a lu ce livre.

It.is entirely that.he has read this book
*[FocP ENTIEREMENT, il a lu ce livre]
ENTIRELY he has read this book
. If the adverb occurs in Spec-ModP/AspP, the structure assigned to (59a) is given in (i).
If the adverb occurs in Spec-TopP, the resulting structure is the one in (ii).

(i) [TopP Giannii [SubjP t i [ModP/AspP probabilmente / spesso [telefonerà / ha telephonato alle
5.]]]]

(ii) [TopP Giannii [TopP probabilmente / spesso [SubjP t i [telefonerà / ha telephonato alle 5.]]]]

The structure in (i) is related to the sentence Gianni telefonerà probabilmente/spesso alle
5, while the structure in (ii) is related to the construction Probabilmente/spesso Gianni
telefonerà alle 5.
. As expected from (64), VP-topicalization can involve solely the adverbs or solely the
direct object, as shown in (i)–(iv).

(i) [CP [FP Gut / schnell / gern / gestern / kürzlich / oft gespielt] [C hat]
well / quickly / willingly / yesterday / recently / often played has
Hans die Sonate.]
Hans the sonata

(ii) *[CP [FP Leider / glücklicherweise / vielleicht gespielt] [C hat] Hans die Sonate.]
unfortunately / happily / maybe played has Hans the sonata

(iii) [CP [FP Die Sonate gespielt] [C hat] Hans gut / schnell / gern / gestern /
the sonata played has Hans well / quickly / willingly / yesterday /
kürzlich.]
recently

(iv) [CP [FP Die Sonate gespielt] [C hat] Hans leider / glücklicherweise /
the sonata played has Hans unfortunately / happily /
vielleicht.]
maybe
 Christopher Laenzlinger

(ii) is ungrammatical, since no projection above TP can raise to Spec-CP. In short, any ObjP
and adverb-related functional projection below ModP in (64) is allowed to topicalize in
Spec-CP. In double object constructions either complements can be fronted with the par-
ticiple, as in (v), or only one of the two objects, as in (vi)–(vii). As regards the adverbs, only
those licensed below ModP can be raised to CP. The analysis of (v)–(vii) is based on the as-
sumption that any ObjP (see structure (39)) and adverb-related functional projection below
ModP can be topicalized in Spec-CP.
ok
(v) *Leider/ gestern / ok oft / ok gern dieses Buch dem Jason
Unfortunately / yesterday / often / willingly this book the-dat Jason
geschenkt hat sie.
offered has she

(vi) *Leider / ok gestern / ok oft / ok gern dieses Buch geschenkt hat sie dem Jason.

(vii) *Leider / ok gestern / ok oft / ok gern dem Jason geschenkt hat sie dieses Buch.

. This analysis fits in well with Chomsky’s (1995) Copy Theory of reconstruction phe-
nomena.
. From a comparative point of view VP-topicalization in German differs from VP-
fronting in English in that in the latter event-related adverbs and negation cannot be fronted
with the verb in the latter language. As pointed out by Jackendoff (1972), only VP-adverbs
can be fronted with the verb in sentence-initial position. This is shown by the contrast
between (i) and (ii). Thus, only functional projections below NegP can be moved to a
sentence-initial position in English, whereas VP-topicalization in German can involve any
functional projection below ModP.

(i) Painstakingly read the book he has.

(ii) *Maybe / not / recently read the book he has.

References

Abeillé, A., & Godard, D. (2000). “French word order and lexical weight.” In R. Borsley (Ed.),
The Nature and Function of Syntactic Categories [Syntax and Semantics 32] (pp. 325–
358). New York: Academic Press.
Abraham, W. (1986). “Word order in the middle field of the German clause.” In W. Abraham
& S. de Meys (Eds.), Topic, Focus and Configurationality (pp. 15–38). Amsterdam and
Philadelphia: John Benjamins.
Alexiadou, A. (1997). Adverb Placement: a Case Study in Antisymmetric Syntax. Amsterdam
and Philadelphia: John Benjamins.
Belletti, A. (1990). Generalized Verb Movement: Aspects of Verb Syntax. Turin: Rosenberg &
Sellier.
Belletti, A., & Shlonsky, U. (1995). “The order of verbal complements.” Natural Language
and Linguistic Theory, 13, 489–526.
 A feature-based theory of adverb syntax 

den Besten, H., & Webelhuth, G. (1990). “Stranding.” In G. Grewendorf & W. Sternefeld
(Eds.), Scrambling and Barriers (pp. 72–92). Amsterdam and Philadelphia: John
Benjamins.
Chomsky, N. (1986). Barriers. Cambridge, MA: MIT Press.
Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press.
Chomsky, N. (2000). “Minimalist inquiries: the framework”. In R. Martin, D. Michaels,
& J. Uriagereka (Eds.), Step by Step: Essays on Minimalist Syntax in Honor of Howard
Lasnik (pp. 89–155). Cambridge, MA: MIT Press.
Cinque, G. (1990). Types of Ā Dependencies. Cambridge, MA: MIT Press.
Cinque, G. (1999). Adverbs and Functional Heads: a Cross-Linguistic Perspective. Oxford:
Oxford University Press.
Cinque, G. (2004). “Issues in adverbial syntax”. In A. Alexiadou (Ed.), Taking up the gauntlet
– Adverbs across frameworks [Lingua, Volume 114, Issue 6] (pp. 683–710).
Diesing, M. (1992). Indefinites. Cambridge, MA: MIT Press.
Ernst, T. (2002). The Syntax of Adjuncts. Cambridge: Cambridge University Press.
Frank, R., Lee, Y.-S., & Rambow, O. (1991).“Scrambling as non-operator movement and the
special status of subjects.” LCJL3 Proceedings (pp. 135–154). Leiden.
Frey, W. (2000). “Syntactic requirements on adverbs.” In C. Fabricius-Hansen, E. Langand,
& C. Maienborn (Eds.), Approaching the Grammar of Adjuncts (ZASPIL 17) (pp. 107–
134). Berlin: ZAS.
Giorgi, A., & Pianesi, F. (1997). Tense and Aspect: From Semantics to Morphosyntax. New
York: Oxford University Press.
Grewendorf, G., & Sternefeld, W. (1990). Scrambling and Barriers. Amsterdam and
Philadelphia: John Benjamins.
Haeberli, E. (1993). “Scrambling and feature checking.” Geneva Generative Papers, 1 (2),
26–47.
Haider, H. (1993). Deutsche Syntax – Generativ. Tübingen: Gunter Narr.
Haider, H. (2000). “Adverb placement – Convergence of structure and licensing.” Theoretical
Linguistics, 26, 95–134.
de Hoop, H. (1993). “Case configuration and NP interpretation.” PhD, University of
Groningen.
Jackendoff, R. (1972). Semantic Interpretation in Generative Grammar. Cambridge, MA:
MIT Press.
Kayne, R. (1994). The Antisymmetry of Syntax. Cambridge, MA: MIT Press.
Kayne, R. (2002). “On Some Prepositions That Look DP-internal: English of and French de.”
Catalan Journal of Linguistics, 1, 71–115.
Koopman, H., & Szabolcsi, A. (2000). Verbal Complexes [Current Studies in Linguistics 34].
Cambridge, MA: MIT Press.
Laenzlinger, C. (1996). “Adverb syntax and phrase structure.” In A.-M. di Sciullo (Ed.),
Configurations: Essays on Structure and Interpretation (pp. 99–127). Somerville, MA:
Cascadilla Press.
Laenzlinger, C. (1998). Comparative Studies in Word Order Variation: Adverbs, Pronouns
and Clause Structure in Romance and Germanic. Amsterdam and Philadelphia: John
Benjamins.
 Christopher Laenzlinger

Laenzlinger, C. (2000a). “More on adverb syntax.” In A. Alexiadou & P. Svenonius (Eds.),

Linguistics in Potsdam, Vol. 6 (pp. 103–132). Potsdam: University of Potsdam.
Laenzlinger, C. (2000b). “French adjective ordering: Perspectives on DP-internal movement
types”. Generative Grammar @ Geneva, 1, 55–104.
Larson, R. K. (1985). “Bare NP adverbs.” Linguistic Inquiry, 16, 595–621.
Larson, R. K. (1988). “On the double object construction.” Linguistic Inquiry, 19, 335–392.
Mahajan, H. (2000). “Two types of remnant movement”. Talk given in the Workshop on
Antisymmetry Theory, Cortona.
May, R. (1985). Logical Form. Cambridge, MA: MIT Press.
McConnell-Ginet, S. (1982). “Adverbs and logical form: A linguistically realistic theory”.
Language, 58, 144–184.
Moltmann, F. (1991). “Scrambling in German and the specificity effect.” Ms., MIT,
Cambridge, MA.
Müller, G. & W. Sternefeld (1993). “Improper movement and unambiguous binding.”
Linguistic Inquiry, 24, 461–507.
Nilsen, Ø. (2004). “Domains for Adverbs”. In Alexiadou (Ed.), Taking up the Gauntlet –
Adverbs across Frameworks [Ligua, Volume 114, Issue 6] (pp. 809–847).
Obenauer, H. (1983). “On the identification of empty categories.” The Linguistic Review, 4,
153–202.
Pollock, J.-Y. (1989). “Verb movement, UG and the structure of IP.” Linguistic Inquiry, 20,
365–425.
Rizzi, L. (1990). Relativized Minimality. Cambridge, MA: MIT Press.
Rizzi, L. (1997). “The fine structure of the left periphery.” In L. Haegeman (Ed.), Elements
of Grammar (pp. 281–337). Dordrecht: Kluwer.
Rizzi, L. (2002). “Locality and left periphery.” In A. Belletti (Ed.), Structures and Beyond. The
Cartography of Syntactic Structures, Vol. 3. Oxford: Oxford University Press.
Roberts, I. (1985). The Representation of Implicit and Dethematized Subjects. Dordrecht:
Foris.
Rochette, A. (1990). “The selectional properties of adverbs.” In M. Ziolkowski, M. Noske,
& K. Deaton (Eds.), Papers from the 26th Regional Meeting of the Chicago Linguistics
Society, Vol. 1 (pp. 425–445). Chicago: Chicago Linguistics Society.
Stowell, T. (1981). “Origins of phrase structure.” PhD, MIT, Cambridge, MA.
Travis, L. (1988). “The syntax of adverbs.” McGill Working Papers in Linguistics: Special Issue
on Comparative Germanic Syntax, 280–310.
Wilder, C. (1999). “Review of Christopher Laenzlinger, Comparative Studies in Word Order
Variation: John Benjamins: Amsterdam 1998.” Studies in Language, 23, 694–719.
 Where syntax and semantics meet
Adverbial positions in the German middle field*

Karin Pittner
University of Bochum

This paper argues that adverbials fall into several classes according to their
base position in the German middle field: sentence adverbials, frame
adverbials, event-related adverbials, event-internal adverbials, and
process-related adverbials. While there may be semantic preferences for a
certain order within these classes, this order is not syntactically determined.
The findings suggest that these classes of adverbials relate to different kinds of
semantic entities, which are accessible only in certain structural
configurations. Thus, the base positions are due to the interplay of syntax and
semantics reflecting the mapping mechanism between them. Finally, it is
argued that these adverbial classes appear in reverse order in the English
postverbal position and German extraposition field because in these
environments, the scope of adverbials is calculated from right to left.

. Introduction

How adjunct positions can be explained is a highly contentious issue. There

is hardly any view imaginable that has not been proposed. On one extreme is
the view that adverbs and, more generally, adjuncts can be placed practically
anywhere and that they are base generated wherever they appear. The other
extreme has recently been proposed by several authors working within a mini-
malistic framework, e.g. by Cinque (1999), who argues for a universal hierarchy
of functional projections which contains a special specifier position for every
type of adjunct.
For German, which exhibits great freedom of word order, it has been
controversial whether there are syntactically determined positions for adver-
bials at all. Some authors have maintained that the placement of adverbials is
completely free (e.g. Haider 1996; Hetland 1992 for sentence adverbials).
 Karin Pittner

In Frey and Pittner (1998) and Pittner (1999) we have endeavoured to

demonstrate that the tests which are used to identify the base positions of verbal
arguments can be fruitfully applied to determine the base positions of adverbial
adjuncts as well (cf. also Maienborn 1996, 2001 on locatives).
Moreover, ambiguous adverbs are especially interesting in this respect. The
ambiguity often disappears in certain positions, which shows that particular
syntactic positions of adverbials lend themselves to particular interpretations.1
Ambiguous adverbs will also be drawn upon as evidence for the close connec-
tion between syntactic position and interpretation of adverbials in this paper.
If there are base positions for adverbials, the question arises what deter-
mines them. In this paper, the basic assumption is that the semantic entity to
which an adverbial constituent applies determines its positions. That is, the
syntactic position of an adverbial is determined by its semantic relations to
the rest of the sentence, which is reflected in its c-command relations to other
adverbials and the verbal arguments.
I will argue that adverbials fall into different classes as far as their base posi-
tions are concerned and that these different base positions reflect the semantic
objects the adverbials are related to. There is a differentiation to be made
between proposition-related adverbials (sentence adverbials), event-related ad-
verbials, event-internal adverbials, and process-oriented adverbials.2 The paper
is organized as follows: first, some properties of the German middle field will
be presented. In the third section, tests for base positions will be discussed.
In Section 4, I will discuss the base positions of several types of adverbials. A
comparison with other approaches is made in 5. Section 6 deals with the mir-
ror image of adverbial positions in final position in English and in the German
extraposition field. A short summary of the results of this paper is given in 7.

. Some properties of the German middle field

The middle field is the space between the parts of the finite verb and infi-
nite verbs constituting the verbal bracket in main clauses with verb-second or
verb-first position, or alternatively between the complementizer and the verbal
elements in subordinate clauses with verb-end position. Since German is an
OV-type language, the base position of the finite verb is to the right of all verbal
arguments in the right part of the bracket. Verb-first and verb-second position
are derived by movement and are possible only if there is no complementizer
or other subordinating element.
 Adverbial positions in the German middle field 

Major constituents can be permuted quite freely in the middle field of the
German sentence:
(1) a. weil Hans seiner Freundin (IO) ein Bild (DO) gegeben
because John his friend (IO) a picture (DO) given
hat
has
‘because John gave his (girl)friend a picture’
b. weil seiner Freundin Hans ein Bild gegeben hat
because his friend (IO) John a picture (DO) given has
c. weil ein Bild seiner Freundin Hans gegeben hat
because a picture (DO) his friend (IO) John given has
The possible word orders are not all equally normal or neutral, since some
require special contexts. Within generative grammar, it has been established by
now as the view of the majority that verbal arguments have a normal or base
order and orders which are derived by scrambling.
The base order of verbal arguments is not the same for all verbs, however.
While many three-place verbs take subject > indirect object > direct object as
the base order, there are verbs with a base order DO > IO, and for some verbs,
the subject is not the first argument in the base order. This first argument will
be called the highest argument.3 It has been convincingly argued that the base
order of arguments is determined by the argument structure of the verb, which
is present on the level of semantics and projected into the syntax (e.g. Haider
1993: 101ff.).
It is controversial whether adverbials, which for the most part are free
adjuncts, have base positions and, if there are base positions for them, what
determines these base positions. Based on earlier work done with W. Frey, I
will argue in this paper that not only verbal arguments, but also adverbials
have base positions which can be identified by applying a number of tests.

. Tests for base positions

. Focus projection

One reason to assume that there are base orders and derived orders is that some
orders allow for wide focus or focus projection, as it is called in the German
literature, whereas others do not. According to Höhle (1982), focus projection
 Karin Pittner

is possible if the accent is placed on the constituent adjacent to the verb and if
normal word order obtains.
A test for maximal focus projection is whether a sentence can occur as an
answer to a question like what happened? This is possible only with (2a). Both
(2b) and (2c) could only be answers to questions in which some of the material
is mentioned.
(2) a. Gestern hat Hans seiner Freundin (IO) ein Bild (DO)
yesterday has John his girlfriend (IO) a picture (DO)
gegeben.
given
‘Yesterday John gave his girlfriend a picture.’
b. Gestern hat seiner Freundin Hans ein Bild gegeben.
yesterday has his friend John a picture given
c. Gestern hat ein Bild seiner Freundin Hans gegeben.
yesterday has a picture his friend John given

. Theme-rheme condition

Lenerz (1977) observed that some orders of two constituents allow either con-
stituent to be either theme or rheme, whereas other orders are not possible if
the first constituent is rheme and the second theme. Lenerz employs questions
as a test for theme and rheme. The theme is already mentioned in the question
and the rheme corresponds to the question word:
(3) Wem hast du das Geld gegeben?
Who did you the money give
‘Who did you give the money to?’
a. Ich habe dem KasSIERer das Geld gegeben.
I have the cashier the money given
‘I gave the money to the cashier.’
b. Ich habe das Geld dem KasSIERer gegeben.
I have the money the cashier given
(4) Was hast du dem Kassierer gegeben?
What have you the cashier given
‘What did you give to the cashier?’
a. Ich habe dem Kassierer das GELD gegeben.
I have the cashier the money given
b. *Ich habe das GELD dem Kassierer gegeben.4
?

I have the money the cashier given

 Adverbial positions in the German middle field 

This test demonstrates that the order IO > DO is unmarked, hence the base
order. They also show that focussed constituents may not scramble.

. Complex frontings

Some constituents can be topicalized together with the verb, but this is not
possible for all kinds of elements. Only constituents which have a base posi-
tion adjacent to the final verb can be part of a complex fronting. The reduced
grammaticality of sentences like (b) is arguably due to a trace in the fronted
string that is ungoverned because its antecedent remains in the middle field
(cf. Haider 1993).
(5) a. Das Buch gegeben habe ich dem Kind.
the book given have I the child
‘I gave the book to the child.’
b. ?? Dem Kind gegeben habe ich das Buch.
the child given have I the book

. Quantifier scope

Frey (1993) observes that ambiguities of quantifier scope occur in derived

orders, but not in base orders.
(6) a. WEIL fast jeder Deutsche mindestens einen Goethe-Roman
because nearly every German at.least one Goethe-novel
kennt 5 (∀∃)
knows
‘because nearly every German knows at least one novel by Goethe’
b. WEIL mindestens einen Goethe-Roman fast jeder Deutsche
because at.least one Goethe-novel nearly every German
kennt (∃∀, ∀∃)
knows
The only possible reading for (6a) is one in which the universal quantifier has
scope over the existential quantifier: nearly every German knows at least one
novel by Goethe. (6b) is ambiguous. Besides the reading of (6a), it also has a
reading in which there is one novel by Goethe which every German knows.
Frey’s explanation for this is that scope is determined by c-command relations,
which are ambiguous in (6b) since the surface position of the existential quan-
tifier c-commands the universal quantifier, but in the base order it is the other
 Karin Pittner

way round so that there are two readings. Cf. Frey’s (1993) scope principle in a
simplified version:
(7) Scope principle: A quantified expression α can have scope over a quanti-
fied expression β if the head of the local chain of α c-commands the base
position of β.

The trace which is due to the scrambling of the object in (6b) adds a scope
option which leads to the ambiguity described.

. Existentially interpreted w-pronouns

It has been observed that existentially interpreted w-pronouns cannot be

scrambled. This makes them useful as a diagnostic for base positions.
(8) a. weil eine Studentin wen beleidigt hat
because a student someone insulted has
‘because a student insulted someone’
b. ?? weil wen eine Studentin beleidigt hat
because someone a student insulted has
(8) provides evidence for a base position of the object to the right of the subject.
Principle-C effects can also be used as diagnostics for determining base
positions. For reasons of space, I will not discuss them here and refer the reader
to the discussions of this test in Frey and Pittner (1998) and Maienborn (2001).
In the next section, I will show how these tests can be applied in order to
determine the base positions of several types of adverbials.

. Adverbial positions in the middle field

. Directional adverbials

It has been observed by a number of authors that directional adverbials in their

base position are adjacent to the verbal elements in such a way that neither
sentence negation nor other elements can occur between them. Not only di-
rectional adverbials, but also predicative elements belong to this special verbal
domain which will be called the “verbal complex”.
(9) a. Sie ist nicht nach München gegangen.
she is not to Munich gone
‘She did not go to Munich.’
 Adverbial positions in the German middle field 

b. *Sie ist nach München nicht gegangen.

she is to Munich not gone
c. Er ist nicht Präsident geworden.
he is not president become
‘He did not become president.’
d. *Er ist Präsident nicht geworden.
he is president not become
Frey (1993: 37) points out that the verbal complex is the smallest verbal projec-
tion that can be topicalized, i.e. be put into the position before the finite verb
(the “prefield”). The following shows that directional adverbials are part of this
projection, since the verb alone cannot be topicalized.
(10) a. *Gelegt HAT er die Bücher in das Regal.6
put has he the books into the shelves
‘He has put the books into the shelves.’
b. In das Regal gelegt HAT er die Bücher
into the shelves put has he the books
It is noteworthy that resultative predicates exhibit parallels to directional ad-
verbials with respect to the two kinds of verb complex phenomena discussed
above, as (11) shows. Two additional parallels are that both resultative predi-
cates and directional adverbials can create additional argument places and be
focus exponents.7
(11) a. *Er hat leer den Teller gegessen.
he has empty the plate eaten
‘He ate the plate clean.’
b. Er hat den Teller nicht leer gegessen.
he has the plate not empty eaten
‘He didn’t eat the plate clean.’
c. *Gegessen hat er den Teller leer.
eaten has he the plate empty
d. Leer gegessen hat er den Teller.
empty eaten has he the plate
A last observation with regards to the base order of directional adverbials can
be made for those cases when both a goal and a path are specified, the latter by a
starting point and points passed on the way. The normal order is starting point
> points on the way > goal, since we get focus projection in (a), but not in (b):
 Karin Pittner

(12) a. Wir sind von München über Nürnberg nach BerLIN

we are from Munich via Nuremberg to Berlin
gefahren.
driven
‘We went from Munich via Nuremberg to Berlin.’
b. Wir sind über Nürnberg von München nach BerLIN
We are via Nuremberg from Munich to Berlin
gefahren.
driven
In this case, the order of adverbials is iconically motivated.

. Manner adverbials

Manner adverbials behave like directional adverbials in some respects. Sentence

negation may not occur to the right of them. We have argued that manner
adverbials are adjoined to the verbal complex (Frey & Pittner 1998: 501ff.; cf.
Frey 2003). This is not uncontroversial, however: Eckardt (2003), for instance,
argues that manner adverbials are adjoined to a higher projection, containing
the verb and its object.
The tests in Section 3 support the assumption of a base position of man-
ner adverbials adjacent to the verbal complex. An indefinite w-pronoun as
object occurs to the left of the manner adverbial. Since the w-pronoun can-
not scramble, this shows that the manner adverbial is base generated to the
right of it.
(13) weil Maria heute was (nicht) sorgfältig gelesen hat
because Mary today something (not) carefully read has
‘because today Mary has (not) read something carefully’

Quantifier scope points in the same direction. We have scope ambiguity with
respect to the quantifier of the manner adverbial in (14b), but not in (14a), (cf.
Frey 1993; Frey & Pittner 1998: 502). While in (a) there is one woman which is
courted in nearly every way, (b) has an additional reading where for every way
of courting there is (at least) one woman. This shows that the manner adverbial
in (b) has been scrambled over the object, whereas in (a) all constituents are in
their base position:
 Adverbial positions in the German middle field 

(14) a. WEIL er mindestens eine Frau auf fast jede Weise

because he at.least one woman in nearly every way
umwarb (∃∀)
courted
‘because he courted at least one woman in nearly every way’
b. WEIL er auf fast jede Weise mindestens eine Frau
because he in nearly every way at.least one woman
umwarb (∀∃, ∃∀)
courted
Complex fronting supports this view as well:
(15) a. Mit bebender Stimme angesprochen hat Hans die elegante
with trembling voice addressed has John the elegant
Frau.
woman
‘With a trembling voice John addressed the elegant woman.’
b. ?? *Die elegante Frau angesprochen hat Hans mit bebender
the elegant woman addressed has John with trembling
Stimme.
voice
It has to be noted, however, that indefinite objects very often occur after man-
ner adverbials:
(16) weil sie schüchtern einen Prinzen geküsst hat
because she shyly a prince kissed has
‘because she kissed a prince shyly’

Eckardt sees this as evidence for a base position of manner adjuncts higher
than the object. In our view, data like (16) are due to the “integration” of the
object into the predicate in the sense of Jacobs (1993).8 Jacobs assumes that
a sister constituent may be either informationally autonomous or integrated
into its head. If it is integrated, this means that it is not processed separately,
but only as part of the head. Among the examples Jacobs gives are the subjects
of thetic sentences and certain kinds of objects. An integrated object is not
conceptualized as a separate entity, but merely as a part of a process. As Jacobs
argues, integration of an object is possible if the object exhibits proto-patient
characteristics as defined by Dowty (1991). Focus on an integrated object can
be wide focus.
 Karin Pittner

(17) a. Sie hat ein BUCH gelesen. (wide focus possible)

She has a book read
‘She read a book.’
b. Sie hat einen KolLEGen verachtet. (only narrow focus)
she has a colleague despised
‘She despised a colleague.’

The patient object in (17a) can be integrated whereas this is not possible for the
stimulus object in (17b) which does not exhibit proto-patient characteristics.
It has also been observed that distributive quantification prevents integration
(cf. Jacobs 1993: 80f.):
(18) a. Sie hat jedes HEMD gebügelt. (only narrow focus)
she has each shirt ironed
‘She ironed each shirt.’
b. Sie hat alle HEMDen gebügelt. (wide focus possible)
she has all shirts ironed
‘She ironed all the shirts.’

It is crucial for our explanation that manner adjuncts can occur only in front
of integrated objects. The following sentences with non-integrable objects are
not acceptable.
(19) a. *Ich habe abgrundtief den Mann verachtet.
I have deeply the man despised
‘I despised the man deeply.’
b. ?? Er hat sorgfältig jedes Hemd gebügelt.
he has carefully each shirt ironed
‘He ironed each shirt carefully.’

We assume that an integrated object extends the verbal complex. Since any
element intervening between an integrated object and the verb would pre-
vent integration, manner adverbials are adjoined to the left of the object in
these cases.
Additional support for the assumption of a base position of manner ad-
verbials adjacent to the verbal complex comes from the interpretation of ad-
verbials which are ambiguous between a process (manner) reading and an
event-related reading like langsam ‘slowly’ and schnell ‘quickly’.
(20) Er ging schnell.
he went quickly
‘He went quickly/he quickly went.’
 Adverbial positions in the German middle field 

On its manner reading, the adverb refers to the speed of the movements of
legs which constitute the activity of walking. On its event reading, the ad-
verb refers to the time span between some reference point and his leaving.
There is a clear preference among native speakers of German for the manner
interpretation of the verb-adjacent adverbial and for the event-related inter-
pretation of ambiguous adverbials to the left of the object. When the context
requires a process interpretation, the order object > manner adverb in (21a) is
preferred.
(21) Damit niemand etwas hört, hat der Einbrecher
in.order.that nobody something hears has the burglar
‘So that nobody hears anything, the burglar has
a. die Tür langsam geöffnet.
the door slowly opened
opened the door slowly.’
b. langsam die Tür geöffnet.
slowly the door opened
If the context requires an event reading, the order adverbial > object in (22a) is
preferred.
(22) Das Zimmer ist schon gut gelüftet –
the room is already well aired
‘The room is already well aired –’
a. du könntest langsam das Fenster zumachen.9
you could slowly the window close
‘you could close the window now.’
b. du könntest das Fenster langsam zumachen.
you could the window slowly close
‘you could close the window now.’

The strong tendency of native speakers of German to interpret ambiguous ad-

verbials to the left of objects as event-related adverbials cannot be explained
if this is the base position for manner adverbials.10 It strongly suggests a base
position for manner adverbials adjacent to the verbal complex.
If an adverbial which is ambiguous between a manner reading and an
event-related reading occurs to the left of the highest argument, only the event
reading is available, as in (23).
(23) weil langsam wer das Essen kochen könnte
because slowly somebody the meal cook could
 Karin Pittner

In this section, it was argued that the base position of manner adverbials is
adjacent to the verbal complex. The following condition holds:
(24) Process-oriented adverbials minimally c-command the verbal complex.

. Instrumentals and comitatives

There is some evidence that the base position of instrumentals and comitatives
is to the left of the objects. We have focus projection only in (b):
(25) a. Er hat eine Dose mit einem MESSER geöffnet. (narrow focus)
he has a can with a knife opened
‘He opened a can with a knife.’
b. Er hat mit einem Messer eine DOSE geöffnet. (wide focus)
he has with a knife a can opened
Complex frontings provide further support for this view:
(26) a. ? Mit dem Messer geöffnet hat sie die Dose.
with the knife opened has she the can
‘With the knife she opened the can.’
b. Die Dose geöffnet hat sie mit dem Messer.
the can opened has she with the knife
These tests suggest a base position of the instrumental to the left of the object.11
Focus projection shows that the position of instrumentals is lower than that
of subjects, since we get wide focus only in (27a).
(27) a. weil Hans mit einem Hammer das FENster einschlägt
because John with a hammer the window breaks
‘because John breaks the window with a hammer’
b. weil mit einem Hammer Hans das FENster einschlägt
because with a hammer John the window breaks
The tests show the same results for comitatives if they are related to subjects.
Comitatives are closely related to one of the verbal arguments, usually the
subject. A (rough) paraphrase of a sentence with a comitative is a sentence
where the comitative is coordinated with the argument it relates to. There are
cases, however, where the comitative relates to the object. Interestingly, the
object-related comitatives take their base position to the right of the object,
cf. complex frontings:
 Adverbial positions in the German middle field 

(28) a. Sie hat das Fleisch mit der Suppe in den Topf getan.
she has the meat with the soup into the pot done
‘She put the meat with the soup into the pot.’
b. Mit der Suppe in den Topf getan hat sie das Fleisch.
with the soup into the pot done has she the meat
??
c. Das Fleisch in den Topf getan hat sie mit der Suppe.
the meat into the pot done has she with the soup
It can be concluded that the comitative has to be c-commanded by the argu-
ment it is related to (cf. Frey & Pittner 1998: 506; Pittner 1999: 165).
As for relative order between instrumental and comitative, the theme-
rheme condition, focus projection, and complex frontings hint at a slight pref-
erence for comitative > instrumental, but no scope ambiguities arise (cf. Frey
& Pittner 1998: 406ff.; Pittner 1999: 165ff.). We have argued that the lack of
scope ambiguities between subject-oriented comitatives and instrumentals is
due to the fact that these adverbials belong to the same class of adverbials and
therefore can be base-generated freely with respect to each other.
Nevertheless, there is a semantically motivated preference for comitative >
instrumental, which is not due to syntactically differentiated base positions. In
Pittner (1999), I have argued that this corresponds to the greater inherence of
instruments to certain actions: while any action can be carried out by nearly
any person, the instruments for carrying out certain actions are usually re-
stricted (such as i.e. cutlery or hands for essen ‘eat’, a hammer or rock for einen
Nagel einschlagen ‘to drive in a nail’, various tools and adhesives for ein Fahrrad
flicken ‘to mend a bicycle’ etc.). Instruments are more inherent to the actions
denoted by the verb (often together with the object) than comitatives, which is
reflected in a slight preference to place them after subject-oriented comitatives.

. Mental attitude

Mental attitude adverbials denote the attitude of one of the participants in an

event, usually the subject. Interestingly, the ambiguous adverb gerne ‘willingly’,
‘frequently’ has no mental attitude reading if it occurs to the left of the subject
as in (a):
(29) a. weil hier gerne jemand arbeitet (only frequency reading)
because here often somebody works
‘because often somebody works here’
 Karin Pittner

b. weil hier jemand gerne

because here somebody willingly / often
arbeitet (mental attitude reading)
works
‘because somebody works here often / willingly’

It can be concluded that mental attitude adverbials have to be c-commanded

by the argument they are related to, which is usually the subject. With verbs
whose highest argument is not the subject, it is the highest argument which
c-commands the mental attitude adverb: 12
(30) weil wem versehentlich was rausgerutscht ist
because someone inadvertently something slipped.out is
‘because someone inadvertently said something’

If mental attitude adverbials are related to the subject they have their base
position to the left of the object, cf. complex frontings:
(31) a. Den Fehler gemacht hat Peter absichtlich.
the mistake made has Peter intentionally
‘Peter made the mistake intentionally.’
b. ? Absichtlich gemacht hat Peter den Fehler.
intentionally made has Peter the mistake
Mental attitude adverbials share with instrumentals and comitatives the prop-
erty of being related to one of the arguments, which led us to classify them as
event-internal adverbials (cf. Frey & Pittner 1998: 508ff.) for which the follow-
ing condition holds:
(32) Event-internal adverbials are minimally c-commanded by the argument
they are related to.

As was shown in this and the preceding section, this condition determines
the base position of subject-oriented comitatives and instrumentals between
subject and object and for object-oriented comitatives below the object. The
base position of mental attitude adverbials is below the argument to which this
attitude is attributed, which is usually the subject.

. Locatives

Locative and temporal adverbials are considered by many authors to be similar

and therefore are often treated together (e.g. Laenzlinger 1998; Haider 2000).
According to the tests, however, they exhibit different base positions and can
 Adverbial positions in the German middle field 

be assigned to different adverbial classes on the basis of their distribution. I will

argue that canonical locatives are another instance of event-internal adverbials
whereas temporal adverbials are event-related adverbials.
Locatives can relate to different kinds of semantic entities which is reflected
in different base positions. First I will deal with canonical locatives which are
commonly assumed to localize events.
These locatives take their base position higher than the object. This is borne
out by focus projection which is possible in (33a), but not in (33b):
(33) Was hat Hans gemacht?
what has John done
‘What did John do?’
a. Hans hat im Kino MARIA geküsst.
John has in.the cinema Mary kissed
‘John has kissed Mary in the cinema.’
b. *Hans hat Maria im KINO geküsst.
John has Mary in.the cinema kissed
This view is supported by scope data (cf. Frey 1993: 194):
(34) a. WEIL Paul zumindest einen fast überall traf
because Paul at.least one nearly everywhere met
‘because Paul met at least one nearly everywhere’
b. WEIL Paul fast überall zumindest einen traf
because Paul nearly everywhere at.least one met
Only (a) is ambiguous with respect to the scope of the quantified locative due
to scrambling of the object over the locative.
The theme-rheme test points in the same direction:
(35) Wen hat Anne in der Kantine beleidigt?
Whom has Ann in the canteen insulted?
‘Who did Ann insult in the canteen?’
a. Anne hat in der Kantine den KOCH beleidigt.
Ann has in the canteen the cook insulted
‘Ann insulted the cook in the canteen.’
b. ? Anne hat den KOCH in der Kantine beleidigt.
Ann has the cook in the canteen insulted
According to the theme-rheme test, the locative has a base position below the
subject of transitive verbs, which is corroborated by w-pronouns:
 Karin Pittner

(36) Wo hat gestern jemand den Koch beleidigt?

where has yesterday somebody the cook insulted
‘Where did someone insult the cook yesterday?’
a. ?? Gestern hat in der KanTIne jemand den Koch beleidigt.
yesterday has in the canteen somebody the cook insulted
b. Gestern hat jemand in der KanTIne den Koch beleidigt.
yesterday has somebody in the canteen the cook insulted
(37) a. Dies wird hoffentlich jemand wo bemerken.
this will hopefully somebody somewhere notice
‘This will be hopefully noticed somewhere by someone.’
b. ?? Dies wird hoffentlich wo jemand bemerken.
this will hopefully somewhere somebody notice
Thus we have the following order of base positions with a transitive verb:
NPNom > AdvLoc > NPAcc .
Taking into account verbs whose highest argument is not the subject, it
becomes clear that the base position of locatives is below the highest argument,
as focus projection and the position of w-pronouns show:
(38) weil dem Mann auf der Brücke der HUND entlief
because the man (dat) on the bridge the dog ran away
‘because the man lost his dog on the bridge’
(39) a. weil ihm in einem Park wer entlaufen ist
because him (dat) in a park somebody run.away is
b. ?? weil ihm wer in einem Park entlaufen ist
because him somebody in a park run.away is
c. dieses Tier ist einem Wärter wo entlaufen
this animal is a guard somewhere run.away
‘This animal ran away from the guard somewhere’
d. ?? dieses Tier ist wo einem Wärter entlaufen
this animal is somewhere a guard run.away
With these two-place sein-verbs (‘ergative verbs’) we have this order of base
positions: NPDat > AdvLoc > NPNom .
We have seen that canonical locatives are related to the highest argu-
ment. This means that these locatives do not relate to events directly, but
rather to the most prominent element in them which is represented by the
highest argument.
 Adverbial positions in the German middle field 

There are locatives, however, which take their base position below an object
of a transitive verb which is not the highest argument. These either locate the
object or are a process specification.
Object-locating locatives occur as adjuncts with verbs like see or as com-
plements to verbs with the basic meaning ‘to put something somewhere’ like
verstauen ‘to put away, to stow’, abstellen ‘to put down, to park’, unterbringen
‘to store, to accommodate’, verstecken ‘to hide’, where the place is specified by a
locative rather than by a directional adverbial.
(40) a. Er hat das Auto in der GaRAge abgestellt.
he has the car in the garage parked
‘He parked the car in the garage.’
b. Er hat das Geschenk im SCHRANK versteckt.
he has the present in.the cupboard hid
‘He hid the present in the cupboard.’

We get focus projection in (40) which indicates that this is the base order.
A canonical locative may be combined with an object-related one, with the
base position of the first higher than the object, and of the latter lower than the
subject. In the following sentences we get focus projection:
(41) a. Ich habe auf dem Balkon eine Palme im HOF gesehen.
I have on the balcony a palm in.the backyard seen
‘On the balcony, I saw a palm in the backyard.’
b. Sie hat in dem Terminal ein Flugzeug über den WOLken
She has in the terminal an airplane over the clouds
gesehen.
seen
‘In the terminal, she saw an airplane over the clouds.’

Maienborn (2001) points out that there are locatives which do not refer to a
situation as a whole but only to part of it. They can receive different interpreta-
tions, all of which are closely related to the action denoted by the verb and are
very similar in this respect to manner adjuncts which apply to processes. They
occur in the same base position as manner adjuncts, minimally c-commanding
the verbal complex.
(42) a. Sie haben das Hühnchen in der PFANne gebraten.
they have the chicken in the pan fried
‘They fried the chicken in the pan.’
b. Sie haben die Gangster auf dem MoTORrad verfolgt.
they have the gangsters on a motorbike pursued
‘They pursued the gangsters on a motorbike.’
 Karin Pittner

We have focus projection in (42) which shows that the base position of these
locatives is adjacent to the verbal complex. Moreover, like manner adverbials,
this kind of locative can occur only after sentence negation.
A third class of locatives are so-called frame setting locatives, which are
dealt with in Section 4.8.

. Temporals

Focus projection shows that temporal adverbials take their base position higher
than that of objects and higher than canonical locatives:
(43) a. Hans hat am Montag in seiner Wohnung den FUSSboden
John has on Monday in his apartment the floor
geschrubbt. (wide focus)
cleaned
‘On Monday, John cleaned the floor in his apartment.’
b. Hans hat den Fussboden am MONtag geschrubbt. (narrow focus)
John has the floor on Monday cleaned
Scope data support this view. (44b) is ambiguous with regard to quantifier
scope, indicating that the locative adverbial has scrambled over the temporal
adverbial:
(44) a. weil Hans an jedem Tag in mindestens einem Lokal
because John on every day in at.least one bar
einkehrte (∀∃)
stopped.off
‘because every day John stopped off at in least one bar’
b. weil Hans in mindestens einem Lokal an jedem Tag
because John in at.least one bar at every day
einkehrte (∃∀, ∀∃)
stopped.off
There is some evidence that the base position of temporals is higher than that
of the subject, cf. the position of the indefinite w-pronoun.
(45) a. weil morgen wer den Boden putzen sollte
because tomorrow somebody the floor clean should
‘because tomorrow somebody should clean the floor’
??
b. weil wer morgen den Boden putzen sollte
because somebody tomorrow the floor clean should13
 Adverbial positions in the German middle field 

The scope test points in the same direction:

(46) a. WEIL an fast jedem Tag mindestens einer nichts tat (∀∃)
because on almost every day at.least one nothing did
‘because almost every day at least one person did nothing at all’
b. WEIL mindestens einer an fast jedem Tag nichts
because at.least one on almost every day nothing
tat (∀∃, ∃∀)
did
I conclude that temporal adverbials c-command the base position of the high-
est argument. This reflects the fact that they are event-related, i.e. situate events
in time.14
Now the question arises why locatives, which are often considered to be
very similar to temporals, should behave differently. Here I can only offer some
speculations. Locatives are not necessarily related to events, but may localize
only certain parts of it (cf. Wunderlich & Herweg 1991). Locatives which ap-
pear to localize whole events are related to the most prominent figure in it, the
highest argument. Thus it is plausible that they are event-internal. Temporals,
on the other hand, characterize whole events.15
The following condition holds for event-related adverbials:
(47) Event related adverbials c-command the base position of all verbal argu-
ments as well as the base positions of event- internal adverbials.

. Causals

Causals are meant here in a broad sense and comprise all adverbials which
express causal or conditional relations. Intuitively, it makes sense to say that
these are event-related.
With respect to temporals, scope data suggest that they are not ordered in
syntax. Since no scope ambiguity occurs in (48), both orders of the adverbials
can be considered to be base generated.
(48) a. WEIL wegen mindestens einem Artikel an fast
because on.account.of at.least one article on almost
jedem Abend Streit herrschte (∃∀)
every evening quarrel was
‘because there was a quarrel on account of at least one article almost
every evening’
 Karin Pittner

b. WEIL an mindestens einem Abend wegen fast

because on at.least one evening on.account.of almost
jedem Artikel Streit herrschte (∃∀)
every article quarrel was
Other tests can be applied with the same result. I take this as evidence that
causals belong to the same category as temporals, namely event-related adver-
bials which c-command the base positions of all verbal arguments.

. Sentence adverbials

The position of the indefinite pronoun shows that sentence adverbials have
their base positions higher than all verbal arguments.
(49) a. weil wahrscheinlich wer schläft
because probably somebody sleeps
‘because somebody is probably sleeping’
b. *weil wer wahrscheinlich schläft
because somebody probably sleeps
Sentence adverbials are sensitive to topic-comment structure. Only topics may
occur to the left of sentence adverbials (cf. Frey & Pittner 1998: 517; Pittner
1999: 175).16 This becomes evident with elements that cannot be a topic, since
they have no referent. According to Lambrecht (1994: 154ff.), topichood pre-
supposes referentiality.
(50) a. ?? weil keiner wahrscheinlich Zeit hat
because nobody probably time has
‘because probably nobody has time’
b. weil wahrscheinlich keiner Zeit hat
because probably nobody time has
The semantic entity that sentence adverbials relate to are propositions. These
propositions contain completely specified events including their temporal
specification. This is reflected in a c-command relation over the finite verb.
This condition is always fulfilled in the middle field, since German is of the
OV-type and all constituents in the middle field c-command the finite verb or
its trace. Therefore, this condition can be violated only in complex frontings as
in (a) and (b).
(51) a. *Glücklicherweise viel getanzt wird in diesem Club.
fortunately much danced is in this club
 Adverbial positions in the German middle field 

‘In this club there is fortunately much dancing that goes on.’
b. *Vermutlich getanzt wird heute Abend.
probably danced is today evening
‘This evening, there will probably be dancing.’
c. Es wurde juristisch betrachtet vermutlich falsch entschieden.
it was legally seen probably wrongly decided
‘From a legal point of view, it was probably a wrong decision.’
d. *[Vermutlich falsch entschieden] wurde juristisch betrachtet.
probably wrongly decided was legally seen
There are ordering restrictions between different kinds of sentence adverbials.
I take these to reflect the scope relations among these classes of sentence adver-
bials, where, as a first approximation, a > b > c > d holds:
(52) a. evaluative adverbs (glücklicherweise ‘fortunately’, leider ‘unfortu-
nately’, . . .)
b. evidentials (offensichtlich ‘apparently’, klarerweise ‘clearly’, . . .)
c. epistemic adverbs (wahrscheinlich ‘probably’, sicherlich ‘sup-
posedly’, . . .)
d. subject-oriented adverbs (klugerweise ‘cleverly’, arroganterweise ‘arro-
gantly’, ...)

Consider the following sentences:

(53) a. *dass Petra wahrscheinlich leider verreist ist
that Petra probably unfortunately away is
‘that Petra probably unfortunately is away’
b. dass Petra leider wahrscheinlich verreist ist
that Petra unfortunately probably away is
‘that unfortunately Petra is probably away’
c. *Paul ist vermutlich laut CNN gestern abgereist.
Paul is supposedly according.to CNN yesterday left
‘According to CNN, Paul has supposedly left yesterday.’
d. Paul ist laut CNN vermutlich gestern abgereist.
Paul is according.to CNN supposedly yesterday left
(53a) shows that an evaluative adverbial cannot occur within c-command of an
epistemic adverbial. While it makes sense to give an evaluation of a proposition
which has a certain degree of probability, it seems to make no sense to judge
the probability of an evaluated proposition. Likewise, (53c) is semantically de-
viant, since it makes no sense to name the source for the proposition which has
evaluated its content as true, and then to make an epistemic judgement about
 Karin Pittner

the proposition. The only possible interpretation is that the epistemic adverbial
is used in (53c) to focus on the source alone.
We think that semantic scope relations are sufficient to account for the
ordering restrictions among different kinds of sentence adverbials. There is
no evidence for different syntactic base positions for these classes of sentence
adverbials. The ungrammaticality of (53a, c) can even be considered as coun-
terevidence for any claim of different base positions for sentence adverbials. If
the only thing that matters for ordering within this group is surface position,
the observed ungrammaticality follows. On the other hand, if (53a) could be
derived from (53b) by scrambling, then the trace in the base order should give
the scope option that makes sense, namely the one with scope of the epistemic
over the evaluative sentence adverbial. However, if there are no different base
positions for different kinds of sentence adverbials, scope can only be calculated
with recourse to the surface order, namely from left to right.
All the sentence adverbials considered so far occur before the so-called
subject-oriented sentence adverbials.
(54) a. Peter hat leider arroganterweise / *arroganterweise
Peter has unfortunately arrogantly / *arrogantly
leider nicht gegrüßt.
unfortunately not greeted
‘Unfortunately Peter arrogantly did not say hello.’
b. Peter hat wahrscheinlich arroganterweise / *arroganterweise
Peter has probably arrogantly / *arrogantly
wahrscheinlich nicht gegrüßt.
probably not greeted
‘Peter probably did not say hello arrogantly.’

The class of subject-oriented sentence adverbials stands out among the others
because of its members special relation to the subject. These subject-oriented
sentence adverbials give a judgement of the speaker about the subject refer-
ent with regard to its participation in the event denoted in the sentence. This
special relation to the subject is reflected in the c-command relations. Subject-
oriented adverbials have to occur within the c-command domain of the subject
in the surface structure.
(55) a. weil Hans arroganterweise nicht grüßte
because Hans arrogantly not greeted
‘because Hans arrogantly did not say hello’
b. ?? weil arroganterweise Hans nicht grüßte
because arrogantly Hans not greeted
 Adverbial positions in the German middle field 

Ambiguous adverbs can serve as a diagnostic. An adverb like dummerweise can

either be evaluative (‘unfortunately’) or subject-oriented (‘stupidly’). Accord-
ing to my intuition, the second reading is only possible if there is an agent
which c-commands the surface position of the adverb:
(56) a. weil dummerweise alle antworteten
because unfortunately all replied
‘because unfortunately all replied’
b. weil alle dummerweise antworteten
because all stupidly replied
Subject-oriented adverbials, like other sentence adverbials, have to c-command
the finite verb, which becomes visible in complex frontings. This differentiates
them from mental attitude adverbials.
(57) a. ?? Klugerweise / dummerweise den Brief beantwortet hat Peter.
intelligently / stupidly the letter answered has Peter
‘Intelligently / stupidly, Peter answered the letter.’
b. Freiwillig / gerne beim Putzen geholfen hat Peter gestern.
voluntarily / willingly with cleaning helped has Peter yesterday
‘Peter voluntarily / willingly helped with the cleaning yesterday.’

To sum up: the strict ordering restrictions among different classes of sentence
adverbials corresponds to the fact that they are not ordered by the syntax and
ordering restrictions are only semantically determined.
The following conditions hold for the base position of sentence adverbials:
(58) Sentence adverbials (proposition-related adverbials):
i. The base position of sentence adverbials c-commands the base posi-
tion of all other kinds of adverbials (and hence the base positions of
all verbal arguments.)
ii. The base position of sentence adverbials c-commands the base posi-
tion of the finite verb.

A note is necessary here on the proposition-related use of locative and tem-

poral adverbials, the so-called frame adverbials. In this function, locative and
temporal adverbials apply to whole propositions and restrict their validity to
certain times or places.17 In this case, they serve as topics (Chinese-style topics
in the sense of Chafe18 ) and appear to the left of all verbal arguments as well as
to the left of all types of adverbials.
 Karin Pittner

Often, the reference of other elements in the sentence is restricted by this

type of adverbial, such as the reference of viele Leute ‘many people’ to America.
This type of adverbial can be combined with canonical locatives and temporals:
(59) a. In Amerika essen viele Leute in Fastfood- Restaurants.
In America eat many people in Fast.food restaurants
‘In America, many people eat in fast food restaurants.’
b. Im Mittelalter tranken die Mönche während der Fastenzeit
in.the Middle.Ages drank the monks during the Lent
Bier.
beer
‘In the Middle Ages, the monks drank beer during Lent.’

Different frame adverbials may be combined. It appears that there is an un-

marked order epistemic frame19 > temporal frame > locative frame.
(60) weil in seinem Traum im Mittelalter in Europa alle
because in his dream in.the Middle.Ages in Europe all
Frauen schön waren
women beautiful were
‘because in his dream in the Middle Ages in Europe all women were
beautiful’

Using scope data, Maienborn (1996: 115f.) shows that the position of locative
frames is above the base position of canonical temporal adverbials. In Frey and
Pittner (1998), we came to the conclusion that frame adverbials have their base
position higher than sentence adverbials. The position to the left of sentence
adverbials, however, may be an effect which is due to the fact that frame adver-
bials usually are topics. As Frey (2003: 168) points out, there are non-referring
frame adverbials which cannot be topics. These have to occur to the right of
sentence adverbials:
(61) a. weil wahrscheinlich in keinem Land alle Beamten korrupt
because probably in no country all officers corrupt
sind
are
‘because probably in no country all officers are corrupt’
b. ?? weil in keinem Land wahrscheinlich alle Beamten korrupt
because in no country probably all officers corrupt
sind
are
 Adverbial positions in the German middle field 

This suggests that frame adverbials take base positions below those of sentence
adverbials and usually have to occur to the left of them only because they are
topics. This makes sense because sentence adverbials are sensitive to sentence
mood (i.e. they cannot occur freely with any sentence type), whereas frame
adverbials are not. Sentence mood, whether introduced by an illocutionary
operator or represented in a syntactic phrase (as suggested by Rizzi 1997), is
usually considered to be left-peripheral.

. A look at other approaches

Classes of adverbials, which are primarily semantically defined, have repercus-

sions on the syntax. As has been indicated in the beginning, there are very
different conceptions as to how syntax and semantics are related in this respect.
We hope to have provided sufficient evidence that the assumption of a
completely free positioning of adverbial adjuncts can be refuted. A somewhat
less extreme view close to completely free positioning is advocated by Haider
(2000). He maintains that the syntax only provides potential slots for adverbial
adjuncts which can be filled in by adjuncts without restrictions as long as their
relative hierarchy is observed. This hierarchy is seen to be a reflex of a hierarchy
of semantic types:
(62) proposition > event > process/state

A very similar view is held by Ernst (1998), whose hierarchy is a bit more elab-
orated and includes speech act > fact > proposition > event > specified event.
Both Haider and Ernst assume that the positioning of adjuncts is grammatical
as long as the hierarchy is observed. They maintain that there are no restrictions
on adjunct positions relative to the arguments in the sentence.
Our findings have shown, however, that there are regularities of adverbial
positioning which hold in relation to arguments. This was demonstrated with
instrumentals, comitatives, mental attitude adverbials, and locatives (cf. 4.3–
4.5). The position of adjuncts reflects their semantic relations to the sentence
in intricate ways.
Another difference between the approach presented here and the approach
by Ernst and Haider is that our approach allows for scrambling of adverbials,
whereas Haider and Ernst have to assume that adverbials are base-generated
wherever they appear. If adverbials belonging to different classes are permuted,
scope ambiguities can be observed which are evidence for scrambling. The am-
 Karin Pittner

biguity of (63b) indicates that the manner adverbial has scrambled over the
temporal adverbial:
(63) a. WEIL Hans an fast jedem Tag auf mindestens eine Art
because John on nearly every day in at.least one way
seinen Chef verärgert (∀∃)
his boss worries
‘because John worries his boss in at least one way nearly every day’
b. WEIL Hans auf mindestens eine Art an fast jedem Tag
because John in at.least one way on nearly every day
seinen Chef verärgert (∃∀, ∀∃)
his boss worries
No such effect occurs if adverbials belonging to the same class are permuted, as
for instance instrumentals and comitatives. These facts support our assump-
tion that instrumentals and comitatives have the same base position, whereas
temporal and manner adverbials belong to different classes as far as their base
position is concerned. It is unclear how differences like this could be captured
by the semantic hierarchy approach.
Recently, it has been maintained within a minimalistic framework that ad-
juncts occur in the specifier positions of functional projections (Cinque 1999;
cf. Alexiadou 1997; Laenzlinger 1998 for similar proposals). It is assumed that
there is a suitable functional projection for every semantic type of adjunct
imaginable, which results in an enormous amount of functional projections. As
far as the syntax-semantics mapping is concerned, the semantics is completely
represented in the syntax which determines the order of adjuncts.
This is not the place to go into a detailed critique of this approach, since this
has been done elsewhere (e.g. Frey & Pittner 1998; Haider 2000). I would like
to point out only briefly that the ordering restrictions among adverbials exist
independently of syntactic projections. The same ordering restrictions can be
observed for modifiers within a nominal phrase.
(64) a. the probable quick end of the war
b. *the quick probable end of the war

This is naturally explained if there is a semantically determined hierarchy be-

tween operators, but under the assumption of syntactically fixed positioning,
it leads to an enormous amount of functional projections. More importantly,
however, the same ordering restrictions can be observed among arguments (cf.
Shaer 1998):
 Adverbial positions in the German middle field 

(65) a. It is lucky for us that it is probable that it is easy for John.

b. *It is probable that it is lucky that it is easy for John.
c. The probability of the quickness of the end of the war.
d. *The quickness of the probability of the end of the war.

In (a)–(d) the ordering restrictions cannot be due to functional projec-

tions, which once again shows that the observed ordering restrictions ex-
ist independently of functional projections. The ordering restrictions found
for nominal modifiers (64) and arguments (65), which correspond to the
scope relations between adverbials, are thus strong evidence for a semantically
motivated account.
If these ordering restrictions exist independently of syntactic configura-
tions, it is not clear why they have to be duplicated by means of functional
projections in syntax, leading to an enormous inflation of functional projec-
tions, and moreover to a semantification of syntactic configurations. It has
been claimed that functional projections provide a more restrictive way of cap-
turing adverbial positions. But if it is taken into account that certain adverbials
can appear in several places without any movement of other elements involved,
then there are two possibilities: either a respective functional projection to host
these adverbials has to be assumed in all these places or these facts are cap-
tured by means of traditional adjunction. But if traditional adjunction cannot
be done away with, then an approach which explains all adverbial positions by
adjunction is more economical.
If this is correct, then it is not necessary to explain the ordering restric-
tions by hierarchically ordered functional projections. We claim that functional
projections are not needed in order to capture the regularities of adverbial
positioning.

. Mirror images: Adverbials in final position in English and in the

German extraposition field

For the German middle field, it holds that elements have scope over elements
to their right.
However, scope is not calculated from left to right in all languages. There
are environments where the order is reversed and scope is calculated from right
to left. This is the case in English in postverbal position and in the extraposition
field in German.
 Karin Pittner

Let us first look at English. Quirk/Greenbaum (1973: 241) note that the
normal order of adverbials in postverbal position is instrumental/manner >
locative > temporal adverbial:
(66) He was working with his shears in the garden the whole morning.

Also, it can be observed that in postverbal position in English the normal order
is locative > frame adverbial:
(67) Many people eat in fast food restaurants in America.

These examples must suffice here to show that in English the normal order of
adverbials in postverbal position is reversed in comparison with the German
middle field.20 In addition, there is considerable evidence that the base order
of adverbials is reversed in the German extraposition field. This is especially
evident with adverbial clauses. In Pittner (1999), I have shown that the base
positions in the middle field apply to adverbial clauses as well. In the extrapo-
sition field, however, the order of base positions is reversed. Let us take as an
example an adverbial clause which is ambiguous between a process-related and
an event-related reading: in (a) the event of telling someone that he is a spy will
come about quickly whereas in (b) the process of telling is going on quickly:
(68) a. Hans wird Maria erzählen, dass Peter ein Spion ist, so schnell
Hans will Mary tell that Peter a spy is as fast
er kann. (event)
he can.
‘Hans will tell Mary that Peter is a spy as fast as he can.’
b. Hans wird Maria erzählen, so schnell er kann, dass Peter ein
Hans will Mary tell as fast he can that Peter a
Spion ist. (process)
spy is.
One central question that arises is how the reversed order can be accounted for.
One apparently straightforward explanation would involve the assumption of
right adjunction. For one thing, the scopal relations between adverbials would
be as expected: adverbials to the right have scope over adverbials to their left.
(69) a. He helped his friend willingly frequently.
(frequently has scope over willingly)
b. He attended classes at at least one university nearly every week.
(nearly every week has scope over at least one university)
 Adverbial positions in the German middle field 

Although the order of adverbials and the scopal relations between them are as
expected, objections have been raised against right adjunction both for empir-
ical and conceptual reasons. On conceptual grounds, it has been argued that
syntactic structures in general can only be right-branching, which rules out
right adjunction as a possibility (e.g. Kayne 1994; Haider 1993).
Larson (1988) used binding data and other indicators for c-command-
relations like negative polarity in order to establish that postverbal adverbials
are c-commanded by the object and therefore cannot be right-adjoined:
(70) a. John visited every colleague1 on his1 birthday.
b. Eve insulted no-one for any reason.

Larson’s approach, however, cannot capture the scope data exemplified in

(69). Thus, we are faced with a dilemma: scope and the order of adver-
bials suggest c-command from right to left which would be expected under
a right-adjunction analysis, whereas binding data and negative polarity suggest
c-command from left to right, hence right-branching structures.
So if one does not want to give up the well-founded idea that binding pre-
supposes a c-command relation, the solution to this puzzle has to be looked for
elsewhere.
In order to capture the order of adverbials Cinque has made an innovative
suggestion. According to him, adverbials are base-generated in the order which
corresponds to the scopal relations exemplified in (69). He assumes a new
movement called intraposition which moves constituents around an adverbial.
By this movement, first (b) and subsequently (c) is derived from (a):
(71) a. He has [frequently [willingly [helped his friend.]]]
b. He has [frequently [[helped his friend]1 willingly t1 ]]]
c. He has [[[helped his friend]1 [frequently t1 ]2 willingly t2 ]

Since this is an innovation, it is not clear at all which restrictions an operation

like intraposition is subject to. But at any rate, this kind of movement can-
not capture binding data correctly and therefore is no solution for the puzzle
outlined above.
There are two different ways out of the dilemma. One solution is to give
up the connection between binding and c-command. This is the way Ernst
(2002) is going. Another way has been pointed out in Frey and Pittner (1999),
where it is suggested that the adverbials in final position cannot be interpreted
semantically.21 Their interpretation becomes possible by way of “modification
markers” which link the adverbials to their respective preverbal place, i.e. the
constituent modified by them.22 The order of adverbials is explained by a con-
 Karin Pittner

dition of non-crossing lines of the modification markers. Thus, the order and
scopal relations of adverbials as well as the binding facts could be captured, the
first by the modification markers, the latter by the c-command relations.
Whatever the best solution to the problems outlined in this section will
turn out to be, the parallels between the English postverbal position and the
German extraposition field give rise to the expectation that the explanation for
the order of adverbials in both these environments will be the same.

. Summary

By applying the tests for identifying base positions to adjuncts, we established

that there are five classes of adverbial adjuncts as far as their base positions
are concerned: (I) sentence adverbials, (II) frame adverbials, (III) event-related
adverbials, (IV) event-internal adverbials and (V) process-related adverbials.
Within these classes, there may be semantic preferences for a certain order, but
this order is not syntactically determined. The base position of these classes
and their c-command-relations reflect their semantic relations to the rest of
the sentence.
(72) Base positions of adverbials:
i. sentence adverbials: c-command the finite verb and the base positions
of all adverbial adjuncts and verbal arguments
ii. frame adverbials: c-command the base positions of event-related ad-
verbials
iii. event-related adverbials: c-command the base position of the high-
est argument and the base positions of event-internal adverbials (e.g.
time, cause)
iv. event-internal adverbials: they are minimally c-commanded by the ar-
gument they are related to, i.e. no other argument can intervene (e.g.
instrumental, comitative, mental attitude adverbials)
v. process adverbials: c-command minimally the verb or “verbal com-
plex” (manner)

Our findings are evidence for a close connection between syntactic base posi-
tion and semantic interpretation of adverbials. They also suggest that adverbial
modifiers do not uniformly relate to an event variable, but that they relate
to very different kinds of semantic entities, e.g. processes (as parts of events),
partial events, events, propositions, and speech acts.
 Adverbial positions in the German middle field 

As far as the syntax-semantics interface is concerned, the following overall

picture emerges. Syntax itself imposes no restrictions on adverbial adjunction,
adverbials can be adjoined freely. The semantic entities that adverbials apply
to, however, are accessible only in certain structural configurations, resulting
in base positions for the various types of adverbials. Thus, the base positions
which were advocated in this paper are due to the interplay of syntax and se-
mantics, and reflect the mapping mechanism between syntactic and semantic
structures.
Since the ordering restrictions for adverbials are semantically based, we
expect that the conditions on base positions hold not only for German, but
for other languages as well. Obvious differences in adverbial placement are
due to different sentence structures (cf. Frey & Pittner 1999; Frey 2003 for a
comparative study). This promises to be a worthwhile topic for further study.

Notes

* This paper is based on earlier work done with Werner Frey. I would like to thank the
editors as well as two anonymous reviewers for their helpful comments.
. This has been very intensively studied for wieder (e.g. von Stechow 1996; Fabricius-
Hansen 2001; Pittner 2003 and the references given there).
. Because of their parenthetical nature, there is arguably no base position for speech act
oriented adverbials (cf. Pittner 1999 for a discussion of speech act adverbials).
. Often it is called the external argument, but for German it can be argued that this argu-
ment is part of the verbal projection and therefore not “external” to it. (Haider 1993: 142ff.)
. Capitals indicate accents.
. In these and similar following examples the accent is placed on the complementizer or
on the verb, since accented arguments may lead to additional scope options. The elements
fast ‘nearly’ and mindestens ‘at least’ ensure a quantificational reading of the quantifiers.
. It is important that the judgements apply to sentences with an accent on the finite verb
which indicates verum-focus, i.e. focus on the truth of the proposition (usually occurring in
contexts where the truth is under debate).
. It can be argued that resultatives and directional adverbials are part of the verbal complex
due to a process of integration as described by Jacobs (1993). According to him, constituents
may be either informationally autonomous with regard to their head or integrated into it.
Jacobs notes that integrated elements may contribute to the valency of their head and are
focus exponents. The role integration plays for the formation of the verbal complex has yet
to be explored, cf. also Section 4.2.
. For a more detailed discussion of this the reader is referred to Frey and Pittner (1998: 498–
501).
 Karin Pittner

. Any element intervening between the adverbial and the verb prevents a manner interpre-
tation of the adverb, cf. the following example with a modal particle:

(i) Du könntest das Fenster langsam mal zumachen.

you could the window slowly PARTICLE close
‘You could really close the window now’

This supports the view that manner adverbials minimally c-command the verbal complex.
. Besides their position, intonation has also a disambiguating effect because event-related
adverbials usually remain unaccented. Their accentuation would require very special con-
texts, for instance phonetic correction.
. Note that the PPs in applicative constructions, which look very similar to instrumentals,
are not adjuncts but rather arguments of their verbs and have their base position to the right
of the object:

(i) Er hat die Wand mit FARbe beschmiert. (wide focus)

he has the wall with paint covered
‘He covered the wall with paint.’

(ii) Er hat mit Farbe die WAND beschmiert. (narrow focus)

he has with paint the wall covered
. If mental attitude adverbials occur above all arguments the attitude might be attributed
to, then it is attributed to some other person which is present in the context.

(i) weil alle Schüler absichtlich fehlten

because all students intentionally were.absent

(ii) weil absichtlich alle Schüler fehlten

In (i) it is intentional of all students that they are not present, whereas in (ii) the intention
has to be attributed to somebody else.
. Not all speakers find (b) to be clearly worse than (a). But it is certainly worse than a
corresponding example with an instrumental instead of a temporal adverbial. This shows
that temporal adverbials have base positions higher than instrumentals.

(i) weil wer mit dem Besen den Boden kehren sollte
because somebody with the broom the floor sweep should
‘because somebody should sweep the floor with the broom’

. Temporal adverbials relating to the reference time behave like frame-setting adverbials,
cf. 4.8.
. A note is necessary here on frequency adverbials. For frequency adverbials it can be ar-
gued that they can take their base positions in different places (cf. Frey & Pittner 1998: 515f.;
Frey 2003), which corresponds to their property of taking different parts of events into their
scope.
 Adverbial positions in the German middle field 

. Frey (2004) provides ample evidence for a structural position reserved for topics which
is located to the left of sentence adverbs. As a consequence of this, adverbials can appear in
front of sentence adverbs if they are topics.

(i) Peter wird auf diese Weise wahrscheinlich seine Reisen finanzieren.
Peter will in this way probably his trips pay.for.
‘In this way, Peter will probably pay for his trips.’

. Cf. Parsons (1990: 209) and Maienborn (1996: 168ff.) and (2001) for a discussion of the
characteristics of frame adverbials.
. Cf. the definition of topic by Chafe (1976: 50): “What the topic appears to do is to limit
the applicability of the main predication to a certain restricted domain [...] the topic sets a
spatial, temporal, individual framework within which the main predication holds.”
. Cf. Maienborn (2001) on epistemic frames, cf. Parson’s (1990) distinction between real
and unreal frames.
. In this respect it is interesting to note that in postverbal position in English, not all the
observed orderings are reversed. One exception is the relative order of different directional
adverbials, which I explained as iconically motivated:

(i) They carried the man from the third floor over the staircase to the ground floor.

(ii) ?They carried the man to the ground floor over the staircase from the third floor.

This supports the view that there is iconic motivation for this order and therefore it cannot
be reversed.
. The idea goes back to Chomsky (1995), who remarks that phrases on the right end
of the clause “might be supported by empty heads below the main verb”. Haider (2000)
assumes that these empty heads are mere structural licensers which cannot be interpreted
semantically.
. The idea resembles the concept of scope markers developed by Williams (1986).

References

Alexiadou, A. (1997). Adverb Placement: A Case Study in Antisymmetric Syntax. Amsterdam

and Philadelphia: Benjamins.
Chafe, W. L. (1976). “Givenness, contrastiveness, definiteness, subject, topics and point of
view.” In C. N. Li (Ed.), Subject and Topic (pp. 25–56). London: Academic Press.
Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press.
Cinque, G. (1999). Adverbs and Functional Heads. A Cross-Linguistic Perspective. Oxford:
Oxford University Press.
Dowty, D. (1991). “Thematic proto-roles and argument selection”. Language, 67 (3), 547–
619.
 Karin Pittner

Eckardt, R. (2003). “Manner adverbs and information structure. Evidence from the
adverbial modification of verbs of creation.” In E. Lang, C. Maienborn, & C. Fabricius-
Hansen, (Eds.), Modifying Adjuncts (pp. 261–305). Berlin: Mouton de Gruyter.
Ernst, T. (1998). “The scopal basis of adverb licensing.” In P. Tamanji & K. Kusumoto
(Eds.), Proceedings of the 28th Meeting of the North East Linguistic Society (pp. 127–142).
Amherst, MA: GLSA University of Massachusetts.
Ernst, T. (2002). The Syntax of Adjuncts. Cambridge: Cambridge University Press.
Fabricius-Hansen, C. (2001). “‘Wi(e)der’ and ‘Again(st)’”. In C. Fery & W. Sternefeld (Eds.),
Audiatur vox sapientiae: a Festschrift for Arnim von Stechow (pp. 101–130). Berlin:
Akademie-Verlag.
Frey, W. (1993). Syntaktische Bedingungen für die semantische Interpretation. Über Bindung,
implizite Argumente und Skopus. [studia grammatica 35]. Berlin: Akademie-Verlag.
Frey, W. (2003). “Syntactic conditions on adjunct classes.” In E. Lang, C. Maienborn, &
C. Fabricius-Hansen (Eds.), Modifying Adjuncts (pp. 163–209). Berlin: de Gruyter.
Frey, W. (2004). “A medial position for topics in German.” Linguistische Berichte, 198, 153–
190.
Frey, W., & Pittner, K. (1998). “Zur Positionierung von Adverbialen im deutschen
Mittelfeld”. Linguistische Berichte, 176, 489–534.
Frey, W., & Pittner, K. (1999). “Adverbialpositionen im deutsch-englischen Vergleich.” In M.
Doherty (Ed.), Sprachspezifische Aspekte der Informationsverteilung. [studia grammatica
47] (pp. 14–40). Berlin: Akademie-Verlag.
Haider, H. (1993). Deutsche Syntax – Generativ. Tübingen: Narr.
Haider, H. (1996). “Wenn die Semantik arbeitet, – und die Syntax sie gewähren läßt”. In
G. Harras (Ed.), Wenn die Semantik arbeitet. Klaus Baumgärtner zum 65. Geburtstag
(pp. 7–27). Tübingen: Niemeyer.
Haider, H. (2000). “Adverb placement – Convergence of structure and licensing.” Theoretical
Linguistics, 26, 95–134.
Hetland, J. (1992). Satzadverbien im Fokus. Tübingen: Narr.
Höhle, T. N. (1982). “Explikation für ‘normale Betonung’ und ‘normale Wortstellung”’. In
W. Abraham (Ed.), Satzglieder im Deutschen (pp. 75–153). Tübingen: Narr.
Jacobs, J. (1993). “Integration.” In M. Reis (Ed.), Wortstellung und Informationsstruktur
(pp. 63–116). Tübingen: Niemeyer.
Kayne, R. (1994). The Antisymmetry of Syntax. Cambridge, MA: MIT Press.
Laenzlinger, C. (1998). Comparative Studies in Word Order Variation: Adverbs, Pronouns and
Clause Structure in Romance and Germanic. Amsterdam and Philadelphia: Benjamins.
Lambrecht, K. (1994). Information Structure and Sentence Form. Topic, Focus and the Mental
Representation of Discourse Referents. Cambridge: Cambridge University Press.
Larson, R. (1988). “On the double object construction.” Linguistic Inquiry, 19, 335–391.
Lenerz, J. (1977). Zur Abfolge nominaler Satzglieder im Deutschen [Studien zur deutschen
Grammatik 5]. Tübingen: Narr.
Maienborn, C. (1996). Situation und Lokation. Die Bedeutung lokaler Adjunkte von
Verbalprojektionen. Tübingen: Stauffenburg.
Maienborn, C. (2001). “On the position and interpretation of locative modifiers.” Natural
Language Semantics, 9 (2), 191–240.
 Adverbial positions in the German middle field 

Parsons, T. (1990). Events in the Semantics of English. A Study in Subatomic Semantics.

Cambridge, MA: MIT Press.
Pittner, K. (1999). Adverbiale im Deutschen. Untersuchungen zu ihrer Stellung und Interpre-
tation. Tübingen: Stauffenburg.
Pittner, K. (2003). “Process, eventuality and wieder/again.” In E. Lang, C. Maienborn, &
C. Fabricius-Hansen (Eds.), Modifying Adjuncts (pp. 365–391). Berlin: Mouton de
Gruyter.
Quirk, R. and Greenbaum, S. (1973). A University Grammar of English. London: Longman.
Rizzi, L. (1997). “The fine structure of the left periphery.” In L. Haegeman (Ed.), Elements
of Grammar. Handbook of Generative Syntax (pp. 281–337). Dordrecht: Kluwer.
Shaer, B. (1998). “Adverbials, functional structure and restrictiveness.” In P. Tamanji &
K. Kusumoto (Eds.), Proceedings of the 28th Annual Meeting of the North East Linguistics
Society, Vol. 1 (pp. 391–407). Amherst, MA: GLSA University of Massachusetts.
von Stechow, A. (1996). “The different readings of wieder ‘again’: A structural account.”
Journal of Semantics, 13, 87–138.
Williams, E. (1986). “A reassignment of the functions of LF.” Linguistic Inquiry, 19, 135–146.
Wiltschko, M. (1994). “Extraposition in German.” Wiener Linguistische Gazette, 48–50, 1–30.
Wunderlich, D., & Herweg, M. (1991). “Lokale und Direktionale.” In A. von Stechow
& D. Wunderlich (Eds.), Handbuch Semantik. [Handbücher zur Sprach- und Kom-
munikationswissenschaft] (pp. 758–785). Berlin: de Gruyter.
 Left/right contrasts among English
temporal adverbials*

Benjamin Shaer
Centre for General Linguistics, Typology and Universals Research (ZAS),
Berlin

In this paper, I investigate the claim that interpretative asymmetries observed

between “left-peripheral” (LP), (i.e., sentence-initial) and “right-peripheral”
(RP) (i.e., sentence-final) temporal adverbials in English have a structural
source, an essential connection existing between certain interpretations and
higher or lower positions. I reconcile this claim with the problematic
availability of “lower” interpretations for LP adverbials, taken to occupy a
high position, by claiming that LP adverbials have an unspecified rather than
high position. While this somewhat overstates the interpretations generally
available to LP adverbials, I remedy this by showing how the unavailability of
certain interpretations for them is attributable to their function (which RP
adverbials lack) as “links”, which relate information in the current utterance
to that of the prior context.

Introduction

It has long been recognized that temporal adverbials on the left periphery of the
English sentence make subtly but significantly different contributions to the
sentence than those on its right periphery. It has also been recognized that the
most plausible source of this asymmetry is a structural one – in other words,
that there is “systematic correlation between the position and the meaning” of
temporal adverbials and that left- and right-peripheral positions underwrite
these differences in meaning (de Swart 1999: 338).
This claim has been at the heart of the analysis of various left/right
contrasts reported in the literature. One such contrast is that between left-
peripheral (LP) and right-peripheral (RP) temporal adverbials containing
quantificational expressions and involves the possibility of wide or narrow
scope interpretations for these expressions. Another contrast, between LP and
 Benjamin Shaer

RP duration adverbials occurring with the present perfect, involves the avail-
ability of “existential” or “up-to-now” interpretations – respectively, those that
relate to some interval in the set that the adverbial describes and to the in-
terval that extends to the time of speech. These and similar contrasts have
alternatively been described in terms of the “position definiteness” of the ad-
verbial – that is, its ability to identify a particular interval on the time axis. A
final contrast involves the ability of LP and RP adverbials to identify the sen-
tence’s “reference time”, the time that a sentence is “about”, or its “event time”,
the time that the situation described by the sentence holds or occurs.
Despite obvious differences between these left/right contrasts, what has
been taken to underlie them all, according to most analyses, is their source
in a structural asymmetry: namely, the high position of LP adverbials and low
position of their RP counterparts. The basic strategy of most analyses, then,
has been to identify ambiguities in sentences with LP and RP temporal ad-
verbials and to associate these different interpretations with either the higher
or the lower position. Accordingly, LP positions are associated with “wide
scope”, “up-to-now”, “position-definite”, and “reference time” interpretations;
and RP positions with “narrow scope”, “existential”, “non-position-definite”,
and “event time” interpretations. The claim, then, is that “lower” interpre-
tations are unavailable in the higher position but not vice versa, given, for
example, the possibility of movement in the former but not the latter case.
Yet, there are real difficulties with this view of left/right contrasts, as certain
studies have noted. One is that the contrasts reported seem more like strong
tendencies than anything structurally determined, with “lower” interpretations
being available to LP temporal adverbials in at least certain cases. Another is
that most of the ambiguities attributed to a structural asymmetry between LP
and RP adverbials turn out to be analyzable in other terms.
At the same time, any effort to account for left/right contrasts without in-
voking a structural difference between LP and RP adverbials will founder on
data like those in (1), to which it is difficult not to attribute a structural source:
(1) a. On Sunday morning, Julia even / only goes to church.
b. Julia even / only goes to church on Sunday morning. (based on de Swart
1999: 344, ex. 17)

So the question seems to be just what sort of structural difference is involved

here. The problem is that this structural difference must somehow under-
write the different contributions of LP and RP adverbials to the rest of the
sentence without committing us to a truth-conditional difference between sen-
 Left/right contrasts among English temporal adverbials 

tences with LP and RP temporal adverbials, which these sentences do not

always evince.
The solution to this problem, I shall be claiming, lies in the possibility of
a structural difference between LP and RP temporal adverbials that does not
rest on the former being in a higher position than the latter. While this sounds
contradictory, the contradiction disappears once we understand this possibil-
ity to involve a syntactically unspecified position for LP elements. What this
means is that there is no grammatically determined “higher” interpretation for
LP temporal adverbials after all, since the semantic relation of these adverbials
to their host sentences would be similarly unspecified, leaving more than one
interpretation available to them.
On this account, LP temporal adverbials should be compatible with a
greater range of interpretations than their RP counterparts, whose syntactic
position is fixed. Since this is the opposite of what is generally claimed, the ap-
proach just sketched faces the significant task of explaining this discrepancy.
Although this task ultimately requires a more detailed account of RP temporal
adverbials than I am able to offer here, what I shall be doing is identifying
certain relevant properties of these adverbials and thus the differences that
they display with respect to their LP counterparts. I shall be making two ba-
sic claims. One is that the reported ambiguity of RP temporal adverbials in
sentences with perfect forms has a structural source, but one associated neither
with IP and VP positions nor with any general structural ambiguity in the right
periphery. This leads to the other, which is that the interpretations available to
RP temporal adverbials beyond the structural ambiguity just noted derive from
their underspecified semantic relation to the situations designated by the VP.
In other words, while both left- and right-peripheral adverbials are associated
with semantic indeterminacy, this has two quite different sources: a basically
syntactic source in the case of the former, and a basically semantic source in
the case of the latter.
Consistent with these two sources of semantic indeterminacy are, I shall
argue, two distinct means by which this indeterminacy is resolved. One, which
applies to RP and LP adverbials alike, involves compatibility requirements on
pairings of temporal adverbials and VPs, as related to their respective temporal
properties. The other is specific to LP adverbials and pertains to a discourse
function that distinguishes them from their RP counterparts. The function in
question is that of a “link”, an element that relates “the information presented
in the current utterance” to “information evoked in or [inferable] from the
prior context” (Birner & Ward 1998: 20–22).
 Benjamin Shaer

Seen as a whole, the analysis of temporal adverbial contrasts that I shall be

presenting in this study is very much a “modular” one, on which syntactic po-
sition is only one factor behind the interpretative patterns associated with LP
and RP temporal adverbials. Admittedly, this analysis remains informal and
highly programmatic, and awaits more detailed investigation of the syntac-
tic, semantic, and discourse dimensions of left/right contrasts among temporal
adverbials. I nevertheless hope to show that this view of these contrasts is a
compelling one, which does warrant further study.
The rest of this paper is organized as follows. In Section 1, I review vari-
ous left/right contrasts among temporal adverbials reported in the literature,
showing that the contrasts claimed do not reflect general patterns and giving
a more adequate characterization of these patterns. In Section 2, I argue that
the patterns that emerge among LP temporal adverbials have a plausible analy-
sis in terms of the syntactically unspecified position of LP temporal adverbials
in English. I then address the question of why interpretations of LP temporal
adverbials seem more rather than less restricted than their RP counterparts,
given the syntactically unspecified position of the former. My response to this
question has two parts. The first, which I offer in Section 3, pertains to restric-
tions on the interpretations of temporal adverbials that arise from interactions
between lexical properties of adverbials and VPs. The second, which I offer in
Section 4, pertains to the status of LP temporal adverbials as “links” and the
discourse principles to which they must adhere. Finally, in Section 5, I offer
some concluding remarks.

. Some left/right contrasts

Let us proceed to the left/right adverbial contrasts noted above. As also noted,
the standard claim is that LP adverbials have “higher” interpretations and
RP adverbials both “higher” and “lower” ones, consistent with a structurally
higher position for the former adverbials. While this picture of adverbial posi-
tion and interpretation is a very compelling one, it becomes considerably less
so when subjected to closer scrutiny.

. Scope asymmetries

We might begin with contrasts that have, not surprisingly, attracted the most
attention of semanticists: those involving scope interactions between adverbials
and expressions in the host sentence. We shall first consider data discussed by
 Left/right contrasts among English temporal adverbials 

de Swart (1999). These include sentences like those in (2), which contain quan-
tifiers in the adverbial and the host sentence;1 those in which the host contains a
negative element, as in (3) to (4); and those in which the host contains a focus-
sensitive particle like only or even, as in (5) (based on (1)). De Swart claims that
the LP adverbial in each case has wide scope with respect to the quantifier or
operator, whereas its RP counterpart may have either wide or narrow scope.
(2) a. On a beautiful Sunday in spring, every student on campus went hiking.
b. Every student on campus went hiking on a beautiful Sunday in spring.
(3) a. At three o’clock, the bomb didn’t explode.
b. The bomb didn’t explode at three o’clock.
(4) a. When Bill left the house, he didn’t turn off the stove.
b. Bill didn’t turn off the stove when he left the house.
(5) a. On Sunday morning, Julia only goes to church.
b. Julia only goes to church on Sunday morning.
(de Swart 1999: 344, ex. 15; 339–40, ex. 6, 10; 344, ex. 17)

Thus, according to de Swart, (2a) describes a situation in which every student

went hiking on the same Sunday in spring, whereas (2b) describes either this
situation or one in which every student went hiking on some beautiful spring
Sunday, but not necessarily the same one. Similarly, (3a) has only an inter-
pretation whereby “what happened at three o’clock was that the bomb didn’t
explode”, with the time adverbial outside the scope of negation. The sentence
in (3b) has this interpretation, but also one according to which the bomb did
explode, but not at three o’clock, with the temporal adverbial in the scope of
negation (de Swart 1999: 339). The sentences in (4) display the same pattern:
(4a) has a single “wide scope” interpretation, according to which Bill’s failure
to turn off the stove is temporally located at the time of his leaving the house;
while (4b) has both this and the “narrow scope” interpretation, according to
which “Bill did turn off the stove, but not when he left the house” (de Swart
1999: 340). Finally, the sentences in (5) display analogous effects, in that (5a)
permits only an interpretation whereby the temporal adverbial is outside the
scope of only and the assertion of the sentence is restricted to “the contextually
relevant things Julia could do on Sunday morning”; whereas (5b) permits both
this interpretation and one whereby the temporal adverbial is in the scope of
only and the sentence asserts that “the only time Julia goes to church is Sunday
morning” (de Swart 1999: 345).
While the conclusions that de Swart draws about left/right contrasts on the
basis of sentences like those in (2) to (5) are plausible ones, closer inspection of
 Benjamin Shaer

these and related data suggests that her conclusions are not correct. Consider
sentences involving two quantificational expressions such as those in (6):
(6) a. Every Sunday, a certain group of students goes hiking in the foothills.
b. A certain group of students goes hiking in the foothills every Sunday.
(based on de Swart 1999: 360, Note 4)

Contrary to de Swart’s claim that a sentence like that in (6a) has only an inter-
pretation on which the universal quantifier takes wide scope, this sentence can
easily receive the corresponding “narrow scope” interpretation. The relevant
interpretation of this sentence becomes more salient when it is embedded in a
discourse like the following one:
(7) Every Sunday, a certain group of students goes hiking in the foothills. They
always wear Norwegian sweaters, and so are known by their classmates as
the “Norwegians”.

Such counterexamples can easily be multiplied, indicating that the one given
here cannot be dismissed as somehow exceptional. For example, alongside
sentences like (2a), in which a “wide scope” interpretation of the temporal
adverbial is the only one available, are the structurally similar (a) sentences
below, in which the LP temporal adverbials are as easily construed as having
either wide or narrow scope as their RP counterparts in the (b) sentences:
(8) a. One Sunday during the semester, every student did volunteer work.
b. Every student did volunteer work one Sunday during the semester.
(9) a. At some point in their lives, each twin will want to leave home.
b. Each twin will want to leave home at some point in their lives.

That is, in both (a) and (b) sentences in (8), the “one Sunday during the
semester” can be the same Sunday for every student or can be different Sun-
days. And in both (a) and (b) sentences in (9), the “point in their lives” can be
the same one or different ones for the twins.
De Swart’s claim that LP temporal adverbials have only wide scope with
respect to negation is likewise open to doubt, although here the effects are ad-
mittedly not as clear-cut. Consider again the sentence given in (3a), now placed
in a discourse context favoring a “narrow scope” interpretation:
(10) At three o’clock, the bomb didn’t explode. This happened only at four o’clock.

This discourse seems perfectly acceptable, and clearly involves an interpreta-

tion for the first sentence whereby an explosion did occur but not at three
o’clock. While this interpretation is not a very salient one, it is neither inaccessi-
 Left/right contrasts among English temporal adverbials 

ble nor available only through contrastive stress on the adverbial. This suggests
that the unavailability of “narrow scope” interpretations of sentences like that
in (3a) is not an absolute one, rigidly determined by the grammar.
There is one pattern of “wide scope” interpretations, however, that does
seem robust. This is the one associated with focus-sensitive particles, as illus-
trated in (5) and in the following pair of sentences:
(11) a. After midnight, Julia even drinks vodka.
b. Julia even drinks vodka after midnight. (de Swart 1999: 341, ex. 8)

As de Swart rightly observes, LP temporal adverbials behave as elements out-

side the scope of even and only, so that (11a) indicates that “the unlikely thing
happening is that Julia drinks vodka”, whereas (11b) can also indicate that “it
is Julia’s drinking of the vodka after midnight that is unexpected” (de Swart
1999: 339–340). A straightforward explanation of this pattern is that focus-
sensitive particles must c-command the constituents that they place into focus
and do so in the case of RP but not LP adverbials. The lack of a c-command
relation between the particle and the LP temporal adverbial in (11a) would
generally be assumed to indicate that this adverbial is in a high position in the
tree, as de Swart’s analysis suggests.2 Yet, the scope data in (6) to (10) cast some
doubt on this possibility, since they cannot be handled in these terms. What
I shall be arguing in Section 2 is that the patterns in (11) and (6) to (10) can
indeed be traced to a common structural source, although one rather different
from that suggested by the standard view of LP and RP adverbial positions.
The problems that we have seen for de Swart’s claim that LP temporal
adverbials consistently have wide scope suggest that one key correlate of this
“wide scope” property – that of providing a “reference time” for the sentence –
is likewise open to doubt. We shall be taking up this point in Section 1.4.
To summarize the findings of this section: examination of temporal ad-
verbial scope data, as discussed by de Swart (1999), has offered little reason
to adopt her conclusion that the grammar rigidly determines a “wide scope”
interpretation for LP temporal adverbials while leaving both “wide” and “nar-
row” scope possibilities open to their RP counterparts. Although there is ad-
mittedly a strong tendency for the former to exhibit the properties in question,
the association of these properties with the LP position appears to be just that –
a strong tendency – rather than a matter of grammatical specification.
 Benjamin Shaer

. “Up-to-now” and “existential” interpretations with the present perfect

The same conclusion emerges for another kind of left/right scope contrast
among temporal adverbials. This contrast, which is related to the interaction of
duration adverbials with the present perfect, takes the following form. Dura-
tion adverbials in RP position are claimed to give rise to an ambiguity between
“up-to-now” and “existential” interpretations, which (as noted above) desig-
nate the interval that extends to the time of speech and some interval in the set
that the adverbial identifies, respectively. In contrast, the same adverbials in LP
position have only the former interpretation. The contrast in the availability
of these two interpretations has been taken to support the view that LP tem-
poral adverbials are structurally higher than their RP counterparts. Just like
de Swart’s left/right contrasts, however, this one also turns out to be at odds
with available data, which indicate that “existential” interpretations are avail-
able with at least some LP temporal adverbials. In addition, evidence based
on an ambiguity test discussed by Gillon (2004) suggests that the difference
between “up-to-now” and “existential” interpretations does not even reflect a
true ambiguity, structural or other.
The original case for a left/right contrast in the availability of “existential”
and “up-to-now” interpretations was made by Dowty (1979: 343), who based
his claim on sentences like those in (12):
(12) a. For four years, John has lived in Boston.
b. John has lived in Boston for four years.

According to Dowty, the (b) sentence can have the “up-to-now” interpretation,
according to which John still lives in Boston; or the “existential” interpreta-
tion, according to which he has spent some four-year period there (which, in
fact, subsumes the first interpretation, a point to which we shall be return-
ing). Crucially, the difference between these is a matter not of “conversational
principles”, but of “a true syntactic ambiguity”. Dowty’s evidence for this is
that only the “up-to-now” interpretation is available for sentences like that in
(12a), where the temporal adverbial is in LP position (Dowty 1979: 343). This
strongly suggests that the two interpretations of sentences like that in (12b) are
associated with distinct syntactic configurations; and that the single “up-to-
now” interpretation of LP temporal adverbials corresponds to a configuration
in which the adverbial has wide scope.
The interpretative contrasts that Dowty finds in present perfect sentences
appear to be paralleled in past-tensed sentences like those in (13), suggesting
the same structural ambiguity:
 Left/right contrasts among English temporal adverbials 

(13) a. In five minutes, John was solving the puzzle.

b. John was solving the puzzle in five minutes.
(Dowty 1979: 346, ex. 16–17)

Here, the RP temporal adverbial in (13b) identifies either an interval before

the start of puzzle-solving or one within which the puzzle-solving occurs,
whereas the LP temporal adverbial in (13a) identifies only the former (Dowty
1979: 346).
Yet, examination of a broader range of examples reveals “existential” in-
terpretations for LP temporal adverbials after all. Examples of LP adverbials
naturally construed in this way are given in (14):
(14) a. For two hours, he has been asleep, and for ten hours, he has watched TV.
(Abusch & Rooth 1990: 12, ex. 34)
b. For two weeks, John has been a milkman. For three weeks, he has been a
paperboy. Now he wants a desk job.
c. For at least three weeks in my life, I have tasted true happiness.

On the basis of such data, Abusch and Rooth (1990: 12) conclude that the
relation between LP temporal adverbials and the “up-to-now” interpretation
of sentences with present perfect forms is “more a strong tendency than an
absolute correlation”, and thus that “a strictly grammatical explanation for
Dowty’s observation is inappropriate.” In other words, the left/right contrasts
that Dowty has observed do not obviously reflect a structural ambiguity.
In fact, there is good evidence that “existential” and “up-to-now” interpre-
tations do not even reflect a true ambiguity, but only different conditions under
which present perfect sentences with duration adverbials may be true. This ev-
idence can be assembled on the basis of the detailed discussion of ambiguity
provided in Gillon (2004). As Gillon suggests, the ambiguity of an expression
lies not so much in its ability to be judged true with respect to distinct states
of affairs, but in its ability to be judged both true and false for a given state of
affairs, where such judgments can be traced to the assignment of more than
one phrase marker to the sentence.3 This view of ambiguity is highlighted in a
familiar example like (15):
(15) Bill saw a man with a telescope.

This sentence is generally taken to be associated with two different phrase

markers: one in which (on traditional assumptions) the PP [with a telescope]
modifies the N [man], the other in which it modifies the V [see a man]. As
such, it can be judged both true and false given a state of affairs in which there
 Benjamin Shaer

is a boy, Bill, who has a telescope, and a man, Fred, who does not; and Bill uses
his telescope to see Fred. It can likewise be judged both true and false given a
state of affairs in which Bill has no telescope, Fred does, and Bill sees Fred with
his naked eyes (Gillon 2004: 158–160).
Gillon (2004: 178–180) concludes from his review of ambiguity tests in the
literature that the “alternate truth value” test just described has the broadest
application and relies on the fewest and least controversial assumptions. This
recommends such a test for closer inspection of the purported ambiguity of
sentences like (12b). Since this test’s action depends on the invoking of states
of affairs according to which truly ambiguous sentences are judged both true
and false, a particularly effective version of the test is one that supplies a con-
tinuation that contradicts one hypothesized interpretation (Brendan Gillon,
personal communication). This naturally leads to alternating judgments, as-
suming the relevant state of affairs: that the sentence is a contradiction, on the
one hand, given the effect of the continuation on one interpretation; and that
the sentence is true, on the other, given the availability of the other interpreta-
tion. The workings of this test can be seen in (16), where it is applied, relative
to the two states of affairs described above, to the two interpretations hypoth-
esized for the sentence in (14) and provides support for these interpretations:
(16) Bill saw a man with a telescope. . .
a. although he himself was using binoculars.
(Given a state of affairs in which Fred has a telescope and Bill has used
binoculars rather than a telescope to see him:)
True if [with a telescope] modifies [man].
False (contradiction!) if [with a telescope] modifies [see a man].
b. although the man himself had nothing with him but a hat.
(Given a state of affairs in which Bill has used a telescope to see Fred
and Fred has a hat but no telescope:)
True if [with a telescope] modifies [see a man].
False (contradiction!) if [with a telescope] modifies [man].

If we now apply this test to the two interpretations hypothesized for RP tem-
poral adverbials and the single interpretation hypothesized for their LP coun-
terparts, as described above, we see that the test produces rather different re-
sults. More specifically, both LP and RP adverbials seem to be compatible with
continuations that respectively force “existential” and “up-to-now” interpreta-
tions; and the perception of contradiction that emerged with the structurally
ambiguous sentence in (16) fails to materialize here:
 Left/right contrasts among English temporal adverbials 

(17) a. John has lived in Boston for four years, just not the past four years.
b. For four years, John has lived in Boston, just not the past four years.
Both (a) and (b) true if John lived in Boston for some four-year period
that does not overlap with the four-year period extending to the time
of speech.
No perception of contradiction for either (a) or (b).
(18) a. John has lived in Boston for four years – namely, the past four years.
b. For four years, John has lived in Boston – namely, the past four years.
Both (a) and (b) true if John has lived in Boston for the past four
years.
No perception of contradiction for either (a) or (b).

One way to interpret these results, given that the “up-to-now” interpretation
is only a specific case of the “existential” interpretation, is to take the former
interpretation to reflect simply one salient state of affairs consistent with the
meaning of present perfect sentences containing duration adverbials, this state
of affairs defeasibly inferred by the hearer on the basis of information available
in the discourse.4 An interval abutting the time of speech is, of course, a highly
salient one, given the discourse salience of the time of speech.
We can conclude this section much as we concluded the previous one:
by observing that the left/right contrasts among duration adverbials that we
have examined here give little indication of conforming to a rigidly determined
grammatical pattern.

. “Position definiteness” asymmetries

The conclusion from the previous section extends to the left/right contrasts
identified by Hitzeman (e.g. 1997), which are related to a temporal adverbial’s
“position-definiteness”. This property, as described by Klein (1992, 1994), per-
tains to the ability of a temporal expression to identify “a fixed position on the
time axis” (Hitzeman 1997: 88). Accordingly, temporal adverbials interpreted
as “position-definite” (PD) identify such a position and those interpreted as
“non-position-definite” (∼PD) do not. Hitzeman’s claim about the PD/∼PD
contrast – which subsumes the contrast between “up-to-now” and “existential”
interpretations described above – closely parallels the claims that we have al-
ready examined. Her claim is that PD interpretations are associated with both
LP and RP occurrences of temporal adverbials, whereas ∼PD interpretations
are associated only with RP occurrences. The sole exception to this generaliza-
tion is LP temporal adverbials in generic sentences, which have only the latter
 Benjamin Shaer

interpretation (Hitzeman 1997: 96, 98–99). This PD/∼PD pattern is illustrated

in (19) and (20):
(19) a. Martha will be in her office at noon.
a . At noon, Martha will be in her office.
b. The rain will begin on Saturday.
b . On Saturday, the rain will begin.
c. John will live in Brooklyn in the summer.
c . In the summer, John will live in Brooklyn.
(based on Hitzeman 1997: 89, ex. 7)
(20) In one hour Americans consume five million gallons of fuel.
(Hitzeman 1997: 90, ex. 10)

These patterns, on Hitzeman’s analysis, fall out of the “tree-splitting” operation

described by Diesing (1992), which divides the syntactic tree into the restrictor
and the nuclear scope – corresponding, respectively, to the IP and VP areas of
the tree – and determines how elements in the sentence will be bound. If an
operator is present, elements base-generated in or moved to the IP area are in
its restrictor and thus bound by it; whereas those in the VP area are bound by
an existential closure operation. If no operator is present, elements in the IP
area are interpreted as specific and those in the VP area as non-specific. Given
the assumption that LP temporal adverbials are base-generated in the IP area
and that RP temporal adverbials are base-generated in the VP area but may
raise to the IP area at LF, the “tree-splitting” operation produces the following
results. LP temporal adverbials have the PD, or “specific”, interpretation illus-
trated in (19a , b , c ) when no operator is present, and the ∼PD interpretation
illustrated in (20) when an operator – in this case, the generic operator – binds
them in its restrictor clause. In contrast, RP temporal adverbials have both the
PD and the ∼PD, or “non-specific”, interpretations illustrated in (19a, b, c),
given both “operator binding” and “existential closure” possibilities for them
(Hitzeman 1997: 92–97).
While the data in (19) to (20) do suggest that the left/right contrasts
adduced by Hitzeman describe a rather well-behaved “higher/lower” pattern
among temporal adverbials, closer inspection once again reveals something
else. In this case, it is that the PD/∼PD contrast does not correlate neatly with
the IP/VP contrast after all, even with Hitzeman’s supplementary claims about
generics and the possibility of LF movement for temporal adverbials in the VP.
In fact, Hitzeman’s analysis of generic sentences does not extend in any case
to sentences like those in (21), which pertain to an interval determined with
 Left/right contrasts among English temporal adverbials 

respect to the moment of speech. In such sentences, LP temporal adverbials

can indeed receive PD interpretations:
(21) a. In a few years, bears in this park will probably get most of their calories
from garbage.
b. In one hour from now, the average American will be glued to the televi-
sion set.
c. In the old days, real men ate meat.

Hitzeman would thus have to supplement her claims further to incorporate

these sentences into her analysis – proposing, for example, that the generic
operator somehow does not bind the adverbial in these cases.
However, it is not this pattern but the more basic one of PD/∼PD con-
trasts that seems truly problematic for Hitzeman’s analysis. This is because
counterexamples to her claim that LP temporal adverbials do not have ∼PD
interpretations can easily be constructed even from sentences that appear to
offer impeccable support for it. Consider the sentence in (22a). Although it
contains a LP temporal adverbial that is naturally interpreted only as PD, this
PD interpretation is clearly cancelled by the continuation of (22a) contained in
(22b), where the two hours mentioned can no longer have a fixed position on
the time axis:
(22) a. For two hours, John has been unconscious.
b. For two hours, John has been unconscious, but not even John knows
exactly which two hours during the past twenty-four these were.

What is significant here is not only that a ∼PD interpretation is available to a

LP temporal adverbial, but also that it arises precisely when the PD interpre-
tation is canceled. Of course, what has done the canceling is just the sort of
continuation that we used to probe the supposed ambiguity of duration ad-
verbials. Here as there, the continuation did not give rise to any perception of
contradiction – as we would expect if the PD/∼PD contrast reflected a true
ambiguity – and thus leads inescapably to the conclusion that the ambiguity is
a spurious one.
This conclusion is strengthened by closer inspection of the sentences in
(19) and Hitzeman’s other evidence for the PD/∼PD ambiguity claim. Here we
find that the continuations of our “alternate truth value” test likewise do not
give rise to contradictions. To assess this evidence, however, it is first necessary
to recognize the biasing effects of many of Hitzeman’s examples, which favor
PD or ∼PD interpretations independently of the contributions of LP or RP
 Benjamin Shaer

temporal adverbials. Only once these are eliminated can we witness a truer
picture of the PD/∼PD contrast.
The biasing effects in question include those traceable to expressions such
as the temporal adverbial once, which forces a ∼PD interpretation of the RP
temporal adverbial in the winter in (23a). They also include those traceable
to the description of particular situations, such as the raining description in
(23b), which strongly favor a PD interpretation simply because of the unlike-
lihood of a forecast of rain for some day not contained in the current week
(cf. 23b ):
(23) a. Greg took the job at Kodak. He fixed copy machines in the winter once
(which is why they offered him the job).
b. The rain will begin on Saturday.
(Hitzeman 1997: 91, ex. 13b; 89, ex. 7b)
b . The program will air on Saturday.

Of course, a reasonable response to these remarks is that such lexical choices

serve simply to make one or another grammatically determined interpretation
more salient, consistent with Hitzeman’s claims. In this case, though, it is these
choices themselves that are more likely responsible for the interpretations in
question, since the relevant differences between LP and RP temporal adver-
bials largely fade away when these biasing effects are neutralized, as shown in
(24) and (25):
(24) Greg took the job at Kodak.
a. He fixed copy machines in the winter.
b. In the winter, he fixed copy machines.
(25) Greg took the job at Kodak.
a. He fixed copy machines in the winter once.
b. Once, in the winter, he fixed copy machines.

Once we eliminate these biasing effects, we can see that the supposed ambigu-
ity of RP temporal adverbials, as revealed by our “alternate truth value” test,
is chimerical. This is clear from the fact that the continuations in the exam-
ples below simply cancel PD or ∼PD interpretations rather than giving rise to
contradictions:
(26) a. Greg fixed copy machines in the winter – but not this past winter, since
he went traveling then.
b. Martha will be in her office at noon – but unfortunately not today at
noon, because she has a lunch engagement.
 Left/right contrasts among English temporal adverbials 

c. The program will air on Saturday, but not this coming Saturday, when
it will be preempted by election coverage.
d. John will live in Brooklyn in the summer, but not this coming summer,
since he’ll still be at college.
(27) a. Greg fixed copy machines in the winter – but not just any old winter:
this past winter.
b. Martha must be in her office at noon – but not any old noon: today at
noon!
c. The program will air on Saturday – but not any old Saturday: this
coming Saturday.
d. John will live in Brooklyn in the summer – but not any old summer: this
coming summer.

For at least some speakers, the effects illustrated in the above examples are par-
alleled in sentences with LP temporal adverbials, once more contradicting the
claim that these adverbials can have only PD interpretations:
(28) a. In the winter, Greg fixed copy machines – but not this past winter, since
he went traveling then.
b. At noon, Martha will be in her office – but unfortunately not today at
noon, because she has a lunch engagement.
c. On Saturday, the program will air, but not this coming Saturday, when
it will be preempted by election coverage.
d. In the summer, John will live in Brooklyn, but not this coming summer,
since he’ll still be at college.
(29) a. In the winter, Greg fixed copy machines – but not just any old winter:
this past winter.
b. At noon, Martha must be in her office – but not any old noon: today at
noon!
c. On Saturday, the program will air – but not any old Saturday: this
coming Saturday.
d. In the summer, John will live in Brooklyn – but not any old summer:
this coming summer.

It is important to note, however, that other speakers do perceive the sentences

in (28) to be contradictory; and that even those speakers who find these sen-
tences acceptable do not consider them interchangeable with their RP counter-
parts, which they still find more acceptable. This suggests that the position of a
temporal adverbial in a sentence not only influences the sentence’s acceptabil-
ity,5 it also makes a significant contribution to the sentence’s interpretation –
 Benjamin Shaer

even if this contribution is not quite the one that Hitzeman suggests. We shall
be examining this contribution further in Section 4.
What we have seen in this section, then, is that another interpretative con-
trast among temporal adverbials, that between their “position-definiteness”
and “non-position-definiteness”, does not correlate neatly with LP and RP po-
sitions after all. However, since the PD/∼PD contrast has shown itself to be
of considerable heuristic value in our investigation of temporal adverbials, we
shall continue to use it in the discussion below.

. “Reference time” and “event time” asymmetries

We come to a final left/right contrast involving elements that have played a key
role in the linguistic analysis of temporality: namely, those related to “reference
time” (R) and “event time” (E), two of the constituents of tense meanings ac-
cording to Reichenbach’s (1947) analysis. What we shall see once again is that
the tight connection between adverbial positions and interpretations claimed
by some authors is not borne out by the data, which are better described in
terms only of tendencies for LP adverbials to receive R interpretations and
RP adverbials to receive E or R interpretations. More surprisingly, however,
the R/E distinction itself will turn out to be less useful than is commonly as-
sumed. This is because “E interpretations” and “R interpretations” will emerge
more as cover terms for a range of interpretations than as descriptors of unique,
well-defined ones.
Before we proceed, it might be worth reviewing the basic features of
Reichenbach’s analysis of tenses, highlighting those that will be most relevant
to us here.6 According to this analysis, the meanings of tenses are defined in
terms of two relations. One is between R, the time for which the speaker makes
a claim (see, e.g., Klein 1992: 535) or more generally that a sentence is “about”;
and the “speech time” (S), the time at which a speaker utters a sentence. The
other is between R and E, the time at which a situation – a state or event –
is located.
While S and E are fairly straightforward notions, R is, in fact, rather more
complicated. In particular, as many researchers have noted (see, e.g., Kamp &
Reyle 1993: 593–594), reference times have both a “sentence-internal” and a
“sentence-external” aspect. The former involves the relation of reference times
to event times, and is often described in terms of their providing a particular
“temporal perspective” on an event. This can be brought out by the contrast
between English sentences with past and present perfect forms, as illustrated in
(30), which are a key motivation for Reichenbach’s analysis:
 Left/right contrasts among English temporal adverbials 

(30) a. I saw John.

b. I have seen John. (based on Reichenbach 1947: 290)

As Reichenbach (1947: 289) describes it, the past event of seeing John is viewed
in (30a) “from a reference point situated also in the past”, but in (30b) “from
a point of reference which coincides with the point of speech”. It is the present
reference time associated with present perfect forms, as in the latter sentence,
that is behind the ability of sentences containing these forms to “affect us with
the immediacy of a direct report”.
The “sentence-external” aspect of reference times involves the relation of
the R of one sentence to that of the previous or following sentence, and the role
of event- or state-designating sentences in controlling the current position of
the R in a discourse. That is, if the situation described by a sentence is an event,
then it typically “follow[s] the contextually given reference point”, whereas if it
is “a state, then it typically includes the reference point”. However, when tempo-
ral adverbials are present, these are assumed to “override, as it were, the effect
of the antecedent context”, triggering an updating of the reference time (Kamp
& Reyle 1993: 527, 529).
Both “sentence-internal” and “sentence-external” aspects of R have figured
in analyses of temporal adverbials in the literature. Those related to the former
have developed Reichenbach’s (1947) original claim that temporal adverbials
serve to identify R (e.g., Declerck 1991; McGilvray 1991) or, departing from this
claim, have taken these adverbials to identify either E or R (e.g., Hitzeman 1997;
Hornstein 1990; Klein 1994; Musan 2002). In each case, the temporal adverbial
can be seen to identify the relevant time directly or indirectly. That is, it may
identify the interval at which or within which a situation is located or a point
that represents the upper or lower bound of the interval at which the situation
is located, all of these possibilities representing what might, broadly speaking,
be called “E interpretations”. Similarly, it may identify the interval or the upper
or lower bound of the interval for which the speaker makes a claim (see, e.g.,
Klein 1994: 162–165, 187–199), all of these representing “R interpretations”.
Of greatest interest to us here is the link that researchers like Hitzeman
(1997) have sought to establish between R interpretations and IP positions,
on the one hand, and E interpretations and VP positions, on the other – such
claims, of course, closely resembling those discussed in previous sections. The
link in question can be seen to follow from the assumption that the value of R
in a sentence is determined on the basis of tense information encoded in the
head of IP, while the value of E is determined on the basis of information en-
coded by the verb and its arguments in the VP. Complementing these claims are
 Benjamin Shaer

ones related to the “sentence-external” aspect of reference times. These include

de Swart’s (1999) claim, which figures in her analysis of left/right contrasts re-
viewed in Section 1.1, that temporal adverbials in LP position always provide
the R for a sentence, while those in RP position may, but need not, do so.
The temporal adverbial data most frequently discussed in connection with
the “sentence-internal” properties of R are those involving sentences with per-
fect forms and “positional” adverbials, as illustrated in (31):
(31) Andy had attacked Mal on Tuesday. (Andrews 1982: 315, ex. 9)

Sentences like this one provide the clearest indication of an R/E distinction
in the interpretations of temporal adverbials, which emerges in the sentence’s
entailing either “that the attack took place on Tuesday” or “that it took place
before Tuesday” (Andrews 1982: 315). In terms of E and R, these two different
interpretations involve the temporal adverbial’s locating on Tuesday either E,
the time at which the attack occurred, or R, the time for which the speaker is
making a claim. A point worth noting here, and one which we shall be taking
up again, is that the latter interpretation is considerably less salient than the for-
mer, although it is readily available for sentences with LP temporal adverbials,
such as that in (32):
(32) On Tuesday, Andy had attacked Mal.

Despite the lack of salience of the R interpretation in sentences like (31), there is
nevertheless good evidence that these sentences display a structural ambiguity
consistent with E and R interpretations. This involves the occurrence of the
adverbial in either lower or higher VPs, as in the labeled bracketings below:
(33) a. Andy [VP had [VP [VP attacked Mal] on Tuesday]]
b. Andy [VP [VP had [VP attacked Mal]] on Tuesday]
(Andrews 1982: 315, ex. 10)

The evidence adduced for this ambiguity takes the form of wh-clefting, VP
preposing, and though movement data, as shown in (34), where the (a), (b),
and (c) sentences signal a lower position for the adverbial, and the (a ), (b ),
and (c ) sentences signal a higher position (Andrews 1982: 315–316):
(34) a. What Andy had done on Tuesday is attack(ed) Mal.
a . What Andy had done is attack(ed) Mal on Tuesday.
b. We thought that Andy had attacked Mal on Tuesday, and attacked Mal
on Tuesday he had.
 Left/right contrasts among English temporal adverbials 

b . *We thought that Andy had attacked Mal (on Tuesday), and attacked Mal
he had on Tuesday.
c. Attacked Mal on Tuesday though he had, Andy still voted with the gov-
ernment.

c . *Attacked Mal though he had on Tuesday, Andy still voted with the gov-
ernment. (Andrews 1982: 315–316, ex. 11–13)

This structural ambiguity is also consistent with the results of the “alternate
truth value” test employed in previous sections. For sentences like the fol-
lowing one, the test gives rise to both contradictory and non-contradictory
interpretations, indicating that the sentence is ambiguous:
(35) John had done it all on his thirtieth birthday – although he had actually done
nothing on that particular day.
(Given a state of affairs in which John did it all in the past interval begin-
ning at his birth and abutting his 30th birthday:)
True if temporal adverbial is in higher VP.
False (contradiction!) if temporal adverbial is in lower VP.

Given the above evidence, the claim that sentences like (31) exhibit a struc-
tural ambiguity is thus a compelling one. However, there turns out to be little
support for the claim that sentences containing RP temporal adverbials are am-
biguous between R and E interpretations more generally, or that these E and R
interpretations are respectively related to higher and lower positions in the tree.
Similarly, there is little support for the claim that LP temporal adverbials receive
only R interpretations, whether this claim is related to the “sentence-internal”
or “sentence-external” aspect of R.
A closer look at the data of RP adverbials reveals that R interpretations and
thus R/E ambiguities are largely restricted to sentences with perfect verb forms.
For example, the sentences in (36), which all have simple verb forms, have no
obvious R interpretations for either LP or RP adverbials:
(36) a. Andy attacked Mal on Tuesday.
a . On Tuesday, Andy attacked Mal.
b. John did it all on his thirtieth birthday.
b . On his thirtieth birthday, John did it all.
c. He was in Beijing for two weeks.
c . For two weeks, he was in Beijing.

One might, however, point to temporal adverbials in sentences like those in

(37), which do seem to have the desired R interpretations:
 Benjamin Shaer

(37) a. John should leave by five.

b. John was tired at five.

In (37a), the VP is describing a punctual event, so the temporal adverbial by five

must be describing R rather than E, given that it is specifying the time by which
John should leave and not the time of John’s leaving itself. In (37b), the VP is
describing a state, so the temporal adverbial at five must also be describing R
rather than E, since it is specifying the time at which John is (revealed to be)
tired and not the time of John’s being tired.
While the interpretations of these two RP adverbials as just described do
sound like R interpretations (although whether they are best thought of as such
is another matter, which we shall be taking up below), there is still no R/E
ambiguity here: these interpretations appear to be the only ones available to
them. So it remains doubtful that the availability of R interpretations in such
cases has anything to do with structural ambiguity.
The same conclusion emerges from an examination of LP temporal ad-
verbials in “scheduling state” descriptions (McGilvray 1991: 40–43) like those
in (38), where the R interpretations predicted for such adverbials fail to ma-
terialize:
(38) This is John’s schedule. On Monday, he plays a match in London and meets
the press. On Tuesday, he plays a match in Sheffield. . .

At first sight, the temporal adverbials in these descriptions do appear to intro-

duce new reference times, consistent with the paraphrases ‘speaking of Mon-
day, that’s when John plays a match in London; speaking of Tuesday. . .’. Yet,
closer inspection indicates that the times that such descriptions are “about” –
that is, at which the speaker’s claim is evaluated as true or false – are cru-
cially not the times at which the scheduled events occur, but rather those
at which these events are (claimed to be) scheduled. In the case of present-
tensed scheduling state descriptions like those in (38), these times coincide
with the time of speech. This point can be made clearer by the emended ver-
sion of (38) given in (39a) (based on examples from McGilvray 1991: 41) and
paraphrased in (39b):
(39) a. This is John’s schedule. On Monday, he plays / *wins a match in London
and meets the press. On Tuesday, he plays / *loses a match in Sheffield. . .
b. . . . On Monday, John is scheduled to play a match in London. On Tues-
day, he is scheduled to lose a match in Sheffield.
 Left/right contrasts among English temporal adverbials 

Here, the unacceptability of wins a match and loses a match (barring the match
being fixed, in which case they are acceptable) derives from the observation just
given: namely, that these present-tensed scheduling state descriptions, though
locating events in the future, assert that they are determined at the time of
speech – which, of course, is contrary to our knowledge about winning and
losing. The paraphrases in (39a) also suggest what is wrong with our earlier
paraphrase of (38). This is that we are not “speaking of Monday” in giving a
scheduling state description of an event located on Monday, but asserting that
as of some reference time prior to Monday – namely, the time of speech –
this event is scheduled to occur on Monday.7 What these scheduling state
cases indicate, then, is that LP temporal adverbials need not have R interpre-
tations – and thus need not, pace de Swart (1999: 340), serve a reference time
updating function.
What they also indicate, once again, is the unlikelihood that temporal ad-
verbials are subject to any general structurally determined R/E ambiguity or
that R and E interpretations are respectively related to IP and VP positions.
It is worth emphasizing, then, that the clear cases of ambiguity that we have
seen, among RP adverbials like those in (35), turned out to be most plausibly
treated in terms of two positions in the VP, and as such provided no evidence
for the involvement of IP elements in the determination of a “higher” inter-
pretation. Of course, we could follow Hitzeman (1997) and take the “lower”
interpretation of these adverbials to be associated with a VP position in which
they are base-generated and the “higher” interpretation to be associated with
an IP position to which they are moved at LF. But such an appeal to LF move-
ment would have no obvious motivation beyond the “higher” interpretation
that it is meant to account for in the first place. Moreover, it leaves us with
a real predicament as regards LP adverbials, since, as we have already ob-
served, these may have E interpretations also – in accordance with the claims of
Klein (1994: 164, 189) and others and contrary to that of Hitzeman. Thus, the
claimed connection between R interpretations and IP positions leaves unan-
swered the difficult question of how RP temporal adverbials would actually
receive these interpretations.
As it happens, the “higher” interpretations that we observed in (35) and
other examples do not obviously depend on such a connection to IP posi-
tions, and can be described in terms of the properties of the VP alone. The
properties in question are related to the function of VP-internal adverbials,
which “‘restrict’ the range of. . . situations (partially) designated by the VP”
(McConnell-Ginet 1982: 159); the event and state descriptions contributed by
the VP; and what Moens and Steedman (1988) have called the “event nucleus”,
 Benjamin Shaer

which represents the hearer’s knowledge of events and in terms of which the
hearer interprets event descriptions.
The “event nucleus”, an event structure composed of the event itself, a
“preparation”, and a “consequent”, can help us to explain the interpretations
available to VP-internal temporal adverbials if we see these adverbials as able
to target larger and smaller constituents of this structure. One explanation for
the ability of these (and other) adverbials to do so is that the grammar does
not specify any particular constituent of an event nucleus to which VP-internal
adverbials must be related, the relevant constituent being determined on the
basis of world knowledge and contextual information (Maienborn 2001: 198).
A hearer’s knowledge base can thus license a range of values for this con-
stituent, which may, for example, be some proper part of the event related to
one of its participants, either of the event’s end-points, or the event’s prepara-
tion or consequent. These possibilities are illustrated in (40), (41), and (42),
respectively:
(40) a. The cook prepared the chicken in a Marihuana sauce.
b. The bank robbers fled on bicycles.
c. Paul is standing on his head. (Maienborn 2001: 195, ex. 9)
(41) a. Mary ran at five.
b. John should leave by Monday.
(42) a. John solved the puzzle in five minutes.
b. Can we open the window for five minutes? (Klein 1994: 194, ex. 21)

What is also clear from these examples is the key role played by the lexical
properties of temporal adverbials in triggering the inferences that lead to the
interpretations observed here, consistent with our observations about these
properties as related to the sentences in (37).
Given this analysis of VP-internal modification, we can also point to a basic
difference between sentences with perfect verb forms, as in (31), and those with
simple verb forms, as in (40) to (42). This is that “consequent” interpretations
are associated with the former through the explicit grammatical encoding of
the consequent phase of the event nucleus, achieved by means of the auxiliary
form have; whereas such interpretations are associated with the latter only by
virtue of the lexical properties of the adverbial, world knowledge, and context.
Similar remarks apply to sentences with progressive forms, like that in (43),
which can be seen as explicitly encoding the preparatory phase of the event:
(43) John was winning for five minutes.
 Left/right contrasts among English temporal adverbials 

On the above analysis, then, the interpretations generally available to RP tem-

poral adverbials emerge as varieties of E interpretations. Although this seems
to represent a significant divergence from analyses that distinguish E and R in-
terpretations of these adverbials, the divergence turns out to be a narrow one,
given that on Reichenbachian assumptions E is always related temporally to R.
This means that any E interpretation of a temporal adverbial can be under-
stood to involve an E that coincides with, precedes, or follows R, while by the
same token being analytically distinct from R. This takes us some way toward
explaining why it is difficult to perceive any significant ambiguity in sentences
like (31), repeated below:
(44) Andy had attacked Mal on Tuesday.

The idea, then, is that the difference between locating on Tuesday the event of
attacking Mal and the consequent state of having done so is a small one.8
Although the foregoing considerations help to explain why RP adverbials
with “higher” interpretations do not obviously occupy IP positions, we must
still address the question of whether there is indeed a distinct R interpretation
of temporal adverbials associated with a high syntactic position. Now, it is cer-
tainly true that LP temporal adverbials like those below display a clear “fram-
ing” effect, which can be understood to involve “set[ting] a context within
which the rest of the sentence is to be interpreted” (Parsons 1990: 209).
(45) a. On a beautiful Sunday in spring, every student on campus went hiking.
(= 2a)
b. At three o’clock, the bomb didn’t explode. (= 3a)
c. In five minutes, John was solving the puzzle. (= 13a)

It is also true, as we have seen in Section 1.3, that LP temporal adverbials com-
monly display (what seem to be closely related) “position-definiteness” effects,
as in the following examples:
(46) a. At noon, Martha will be in her office.
b. On Saturday, the rain will begin. (= 19a , b )

In each case, these effects are quite plausibly described as related to R, since
they appear to be implicated in the updating of a discourse or the identifica-
tion of some time that the discourse is “about”. Yet, the temporal adverbials in
many of these examples, including (45a) and (46a, b), could just as easily be
described as serving to identify E. This would make these effects orthogonal to
the R/E distinction and the claim that they reflect a distinct R interpretation at
best moot. Moreover, the same framing and position-definiteness effects also
 Benjamin Shaer

emerge with RP temporal adverbials – even if the effects are admittedly not as
strong in these cases. This casts further doubt on the existence of a distinct R
interpretation associated with a high position in the tree:
(47) a. Every student on campus went hiking on a beautiful Sunday in spring.
(= 2b)
b. The bomb didn’t explode at three o’clock. (= 3b)
c. John was solving the puzzle in five minutes. (= 13b)
(48) a. Martha will be in her office at noon.
b. The rain will begin on Saturday. (= 19a, b)

Similar comments apply even to sentences such as those in (49), in which

LP temporal adverbials clearly indicate the time about which the speaker is
making a claim:
(49) a. At five, John was tired.
b. By five, John had had enough.

Although the identification of such a “claiming time” is arguably the hallmark

of an R interpretation, the likelihood that this represents a distinct structurally
determined interpretation is rather small. This is not only because our previous
observations revealed that the lexical properties of the temporal adverbial and
the VP played a key role in effecting such interpretations, but also because such
R interpretations are likewise available for RP adverbials, as shown in (50):9
(50) a. John was tired at five.
b. John had had enough by five.

These considerations leave us little reason to recognize a distinct structurally

determined R interpretation for temporal adverbials – although identifying the
interpretations of the adverbials in (49) and (50) as “R interpretations” will cer-
tainly remain useful for expository purposes. Of course, this conclusion does
not mean that the strong tendency of LP temporal adverbials to display the
framing, position-definiteness, and R effects just described no longer calls for
an explanation. However, it does suggest that the explanation may well have
little to do with LP adverbials occupying a high position in the tree. This is the
idea that we shall be exploring in the following section, where we shall take a
closer look at the syntax of LP temporal adverbials and see what insights this
can offer us into how they are interpreted.
 Left/right contrasts among English temporal adverbials 

. The syntax of LP temporal adverbials

Let us take stock. In our investigation of LP and RP temporal adverbials so

far, we saw that the former adverbials had a strong tendency to receive what
we have called “higher” interpretations: namely, “wide scope”, “up-to-now”,
“position-definite”, or “reference time”. In contrast, the latter adverbials were
equally likely to receive these interpretations or what we have called “lower”
ones: namely, “narrow scope”, “existential”, “non-position-definite”, or “event
time”. This left/right contrast lent support to the possibility that LP adverbials
were base-generated in high left-peripheral positions, whereas RP adverbials
were base-generated in the VP but could move to higher positions at LF. Such
a possibility also found support in the clear c-command effects that we found
in the interpretations of focus-sensitive particles like even and only, which sug-
gested that LP but not RP adverbials were outside the scope of these particles.
Yet we have also found inconsistency in the behavior of LP adverbials, in the
form of numerous counterexamples to the claim that they do not have “lower”
interpretations. We have thus had to recognize that these left/right contrasts
reflected only tendencies.
Of course, generative frameworks offer a ready means to eliminate this in-
consistency: namely, to associate the “higher” and “lower” interpretations of
LP temporal adverbials with base-generation and movement possibilities, re-
spectively. Since the surface position of these adverbials is a standard target of
movement operations, the movement possibility for capturing their “lower”
interpretations seems quite reasonable.
There turns out, however, to be substantial evidence against a movement
analysis of these interpretations, which derives from binding, island, and VP
ellipsis data.10 The binding data take the form of LP temporal adverbials that
do not give rise to Principle C violations when they contain R-expressions
coindexed with subject NPs, as illustrated in (51):
(51) a. On John’si fiftieth birthday, hei had done it all.
b. For three whole years of John’si life, hei had built boxes.

If the LP adverbials in these sentences were, on their “lower” interpreta-

tions, moved from (and then reconstructed to) lower positions, then the R-
expressions in them should be A-bound by the subjects of these sentences, thus
creating Principle C violations. Yet these sentences seem perfectly acceptable,
suggesting that the adverbials have not moved after all.
Although the sentences in (51) thus clearly display “anti-reconstruction”
effects, the controversial status of Principle C in current grammatical theory
 Benjamin Shaer

(see, e.g., Pollard & Sag 1994: 247–248; Lasnik 2003: 125–138) means that this
form of evidence must be supplemented by others. As already noted, other
forms of evidence are, in fact, available. One involves the ability of LP temporal
adverbials to acceptably violate standard movement constraints, such as the
wh-island and complex NP constraints, as demonstrated below:
(52) a. At five o’clock in the morning, I wonder who finally stumbled home
drunk.
b. *At what time in the morning do you wonder who finally stumbled home
drunk?
(53) a. At five o’clock in the morning, I very much doubt John’s claim that he is
still perfectly coherent.
b. *At what time in the morning do you very much doubt John’s claim that
he is still perfectly coherent?

Another form of evidence involves the behavior of LP temporal adverbials with

respect to verb phrase ellipsis. What we find here – at least for some speak-
ers – is that LP adverbials pattern with their “parenthetical” and “afterthought”
counterparts, and in contrast to RP adverbials, as elements independent of the
antecedent verb phrase, much like the parenthetical expressions described by
McCawley (1982). The key point here is that this effect emerges even when
these adverbials clearly have “lower” interpretations, as they do in (54), which
would generally be associated with verb-phrase-internal base positions. It is
this hypothesized independence of the LP, parenthetical, and afterthought ad-
verbials alike that makes it possible for some speakers to construe these with
the first conjunct only, so that the continuations in (54a, b, c) do not give rise
to contradictions, whereas that in (54c ) does:
(54) a. At five, John finally signed the form, and Mary did too – though she, in
fact, a bit later.
b. John finally, at five, signed the form, and Mary did too – though she, in
fact, a bit later.
c. John finally signed the form, at five, and Mary did too – though she, in
fact, a bit later.
c . John finally signed the form at five, and Mary did too – though she, in
fact, a bit later.

Admittedly, many speakers do perceive (54a) to be contradictory – not sur-

prisingly, since it is natural to understand the two conjuncts of a sentence as
describing parallel situations (see, e.g., Asher 1993: 284–292) and thus to take
the LP adverbial to apply to both of them. However, the fact that some speak-
 Left/right contrasts among English temporal adverbials 

ers do not perceive this sentence to be contradictory, even given the strong
tendency to assign it a “parallel” interpretation, does offer further support for
the analysis of LP temporal adverbials as independent of the verb phrase and a
fortiori not moved out of it.
Yet, if this conclusion is correct and “lower” interpretations of LP temporal
adverbials are not associated with movement, how are we to capture with a sin-
gle syntactic structure the two interpretations that we have observed for them?
One intriguing possibility – which is often suggested for discourse adverbials,
appositive relative clauses, and various kinds of parenthetical expressions (see,
e.g., Espinal 1991: 726–727), and which has more recently been explored for LP
manner and instrumental adverbials (Shaer 2003: 246–249, 2004, to appear) –
is that such adverbials have the status of “orphans” (Haegeman 1991), elements
that are independent of their host sentences in the syntax. What such a status
means is that these elements occupy positions that have no hierarchical rela-
tion to the sentences with which they are associated, and thus no c-command
relation to them. Schematically, an orphan would have the kind of structure
indicated in the labeled bracketing below, where the orphan PP [at five] is dis-
tinct from its host sentence in much the same way as one sentence would be
distinct from a sentence adjacent to it:
(55) [PP At five] [IP John got drunk.]

An orphan status for the LP temporal adverbials described in this study is con-
sistent with the binding, island, and VP ellipsis data reviewed above, as well
as with the observation that these adverbials are typically set off intonation-
ally from the rest of the sentence, the idea being that speakers can signal the
presence of a distinct syntactic unit by this prosodic means.
Although the evidence just given is only indirect, more direct evidence for
the orphan status of these adverbials does exist, which is related to the behavior
of focus-sensitive particles like only. This evidence, which complements the
pattern given by de Swart (1999: 344), as described in Section 1.1, emerges in
the examples below:
(56) a. *Only at five, John ever got drunk.
b. Only at five did John ever get drunk.
c. ?Only at five John ever got drunk!
d. (cf. On Sunday morning, Julia only goes to church.)

The scope behaviour of LP temporal adverbials like the one in (56d) (repeated
from 5a), as observed by de Swart, indicated that these adverbials were not c-
commanded by particles like only. What the sentence in (56a) suggests is that
 Benjamin Shaer

these LP adverbials, in turn, do not c-command negative polarity items in the

sentence. This claim is further supported by the (more) acceptable sentences in
(56a, b), in which the temporal adverbials are both plausibly analyzed as moved
(in the latter case, as an instance of focus movement).
The orphan analysis of LP temporal adverbials proposed in this section
thus appears consistent with a range of data – in particular, those that indi-
cate that LP adverbials with “lower” interpretations fail to display standard
signs of movement. This analysis also puts us in a good position to explain
why these adverbials do not behave in any consistent fashion as elements occu-
pying either higher or lower positions in the tree: namely, that they are not in
either position, but simply independent of their host sentences in the syntax.
Taking these adverbials to have no syntactically specified relation to the sen-
tence thus gives us a direct way to capture their ability to have both “higher”
and “lower” interpretations without recourse to claims of structural ambigu-
ity, which the data reviewed above show to be empirically untenable. Of course,
what remains for us to explain is why LP temporal adverbials nevertheless have
such a strong tendency to be construed as wide scope, position-definite, R-
indicating, or framing expressions, as we have seen throughout this study. We
shall be addressing this question in the following sections.

. Temporal adverbials and Aktionsart

If we consider the many examples of LP temporal adverbials that have figured

in our investigation, it is clear that they have not exhibited the kind of un-
constrained interpretative freedom that the orphan analysis just given seems
to predict. So the question that naturally arises for such an analysis is why
such freedom, though granted to these adverbials in principle, is not obviously
exercised by them in practice.
In fact, this question has a straightforward answer, consistent with the
orphan idea that LP adverbials have no grammatically determined function
in a sentence. This is that the absence of any syntactic mechanism to relate
these adverbials to the expressions that they modify means that the process of
interpreting a sentence associated with an orphan adverbial must involve lin-
guistic and other information that goes surrogate for the “missing” syntactic
information.
The information that would figure most prominently in this process is ar-
guably that related to the lexical properties of the orphan adverbials themselves,
since any acceptable resolution of their syntactic indeterminacy must be con-
 Left/right contrasts among English temporal adverbials 

sistent with these properties. Moreover, since different VPs describe situations
that occur at or hold for different periods of time (see, e.g., Vendler 1967: 97–
121), it is clear that the assignment of a coherent interpretation to a sentence
with a temporal adverbial requires that the temporal properties of the adverbial
and the VP be compatible with each other. However, what counts as a compat-
ible pairing of adverbials and VPs turns out to be a more complicated matter
than a simple comparison of the respective temporal properties of these ele-
ments might suggest, given the availability of coercion processes which may
“rescue” otherwise illicit adverbial-VP pairs. What also emerges is that such
lexical compatibilities are no less relevant to RP than to LP temporal adver-
bials, since the former must likewise supply temporal information consistent
with that supplied by the VP.
In what follows, we shall be examining the patterns of acceptability dis-
played by various combinations of LP and RP temporal adverbials and VPs.
For expository purposes, we shall be making use of the classifications of tem-
poral adverbials given in Parsons (1990: 207) and Klein (1994: 149) and distin-
guishing two classes of “positional” and two classes of “durational” adverbials:
respectively, “point-time” adverbials such as at noon and “interval-time” ad-
verbials such as between 2:00 and 3:00, and in- and for-adverbials such as in
three hours and for three hours.11 As for VPs, we shall be adopting Vendler’s
(1967: 97–121) quadripartite division of them into “activities”, “accomplish-
ments”, “achievements”, and “states”.
The combinations of temporal adverbials and VPs that we shall be consid-
ering are those given in (57) to (60) below, where point-time, interval-time,
for-, and in-adverbials are respectively paired with achievements, accomplish-
ments, activities, and states:
(57) point-time adverbials:
a. (At five,) John had solved the puzzle (at five).
b. (At five,) John had written the letter (at five).
c. ?(At five,) John had pushed carts (at five).
d. (At five,) John had been drunk (at five).
(58) interval-time adverbials:
a. (Between 2:00 and 3:00,) John had solved the puzzle (between 2:00 and
3:00).
b. (Between 2:00 and 3:00,) John had written the letter (between 2:00 and
3:00).
c. (Between 2:00 and 3:00,) John had pushed carts (between 2:00 and
3:00).
 Benjamin Shaer

d. (Between 2:00 and 3:00,) John had been drunk (between 2:00 and 3:00).
(59) for-adverbials:
a. *?(For 3 hours,) John had solved the puzzle (for 3 hours).
b. *?(For 3 hours,) John had written the letter (for 3 hours).
c. (For 3 hours,) John had pushed carts (for 3 hours).
d. (For 3 hours,) John had been drunk (for 3 hours).
(60) in-adverbials:
a. (In 3 hours,) John had solved the puzzle (in 3 hours).
b. (In 3 hours,) John had written the letter (in 3 hours).
c. *?(In 3 hours,) John had pushed carts (in 3 hours).
d. (In 3 hours,) John had been drunk (in 3 hours).

Inspection of these temporal adverbial-VP pairs indicates that most of them

are fully acceptable, and that even those that are not become at least somewhat
improved given appropriate contextual support – a matter to which we shall
be returning below. What is more relevant here, though, are the different in-
terpretations on which these sentences are acceptable and the relation of their
acceptability to the respective temporal properties of the adverbial and the VP
that co-occur in them.
The table below summarizes the interpretations available to these adverbial-
VP pairs. Some clarification is in order regarding the notation and labels used
in indicating interpretations. An available but not particularly salient interpre-
tation for a given pair is indicated in parentheses. As for the interpretations

Table 1. Interactions between temporal adverbial and Aktionsart class

Temporal adverbial class

positional adverbials durational adverbials
point-time interval for in
Aktionsart class
achievement E E (*? repetitive E) E
consequent E (?? consequent E) (consequent E)
accomplishment consequent E E (*? repetitive E) E
(?? consequent E) (consequent E)
activity (? consequent E) E E (*? E)
(? preparatory E) (??consequent E) (*? consequent E)
state R R/E E R
consequent E (consequent E)†
(?? E)
† Available only for predicates like be dead, where it is the only natural interpretation for them
 Left/right contrasts among English temporal adverbials 

themselves, what are labeled “E” interpretations identify the interval at which
or within which the achievement, activity, accomplishment, or state situation is
located, with the labels “preparatory E” and “consequent E” respectively iden-
tifying the preparatory and consequent phases associated with achievements
and accomplishments. In addition, the label “repetitive E” indicates repeated
performance of the action described by the VP. Since (as noted in Section 1.4)
at least some R interpretations can be assimilated to consequent phase inter-
pretations, we shall reserve the label “R interpretation” for cases that cannot
obviously be assimilated in this way and more clearly involve specifications
of the interval (or the upper or lower bound of the interval) for which the
speaker makes a claim. Arguably all of these cases have the kind of “epistemic”
flavor highlighted by the sentence in (61), which can be captured in terms
of paraphrases such as ‘it seemed that’, ‘it became evident that’, and ‘it was
reported that’ (see, e.g., Declerck 1991: 364, Note 88; Shaer 1996: Chapter 2,
Section 1.3.3.1):
(61) At two o’clock / Between two and three / For one hour, John had solved the
puzzle, but at three o’clock / between three and four / for several hours after
that, his solution attracted only doubt.

Notwithstanding the assimilation of certain R interpretations to “consequent

E” interpretations, as noted above, what we still observe is that most of the tem-
poral adverbial-VP pairings indicated in the table give rise to some variety of E
interpretation, with R interpretations emerging in only a few instances. More-
over, inspection of the relevant examples in (57) to (60) reveals no discernible
difference in this respect between LP and RP occurrences of the adverbials in
question. This seems consistent with the conclusion reached in Section 1.4 that
R interpretations are not structurally determined.
Turning to the specific patterns associated with the four categories of tem-
poral adverbials, we can see that these all have rather straightforward explana-
tions. We might begin with point-time adverbials, as given in the first column
of the table. Since these adverbials serve to identify single moments in time,
they have a basic E interpretation with VPs describing achievements, which
“occur at a single moment” (Vendler 1967: 103). Moreover, because the onset
of the consequent state is adjacent to the culminating moment at which the
achievement occurs, these adverbials can also be understood as identifying the
onset of this state. Since none of the other three classes of VPs describe punctual
situations, point-time adverbials have no basic E interpretations when they oc-
cur with them. Thus, the interpretation that we find with accomplishment VPs
is the consequent state interpretation just described for achievements, which
 Benjamin Shaer

likewise identifies the onset of this state. VPs that describe activities and states –
which are both “homogeneous” situations in the sense that they have a constant
character throughout a given interval and thus have no culmination – have only
marginal E interpretations with point-time adverbials, since neither situation
type has any particularly salient moments that such adverbials can identify. The
interpretations that arise in the former case involve the identification of the on-
set of the consequent phase or – even less salient – the preparatory phase of a
temporally bounded activity. The interpretations that arise in the latter case in-
volve the – generally rather improbable – identification of the single moment
at which the state holds. This makes an R interpretation, involving the identi-
fication of the R at which some state becomes manifest, and a consequent state
interpretation, identifying the outcome of a temporally bounded state, more
salient possibilities.
The pattern for interval-time adverbials given in the second column closely
resembles that for point-time adverbials given in the first, both classes being
members of the larger positional class. The difference between these patterns
thus reduces to the difference between one class identifying single moments
and the other identifying extended periods of time. Since interval-time adver-
bials identify extended periods, they can readily identify intervals at which or
within which the situations described by all four classes of VPs hold or occur.
This means that we find basic E interpretations for them with each VP class.
In addition, because interval-time adverbials do not identify single moments,
they are not naturally interpreted as identifying the onsets of preparatory or
consequent phases of an event nucleus, as their point-time counterparts are.
This means that the consequent phase interpretations of achievement, accom-
plishment, and activity VPs that were acceptable with point-time adverbials
are at best marginal here. However, interval-time adverbials arguably have ac-
ceptable E as well as R interpretations with state VPs, the former involving the
interval at which the state actually holds, and the latter involving some interval
at which a state becomes manifest.
Finally, the pairings of for- and in-adverbials with VPs, as given in the third
and fourth columns, give rise to patterns that likewise follow from well-known
temporal properties of these adverbials. In particular, for-adverbials have ac-
ceptable E interpretations with state and activity VPs, indicating the duration
of the “homogeneous” situations that these VPs describe (with the exception
noted in the table for VPs like be dead, which describe “one-sided” states;
see Klein (1994: 188)). However, they have no analogous interpretations with
achievement and accomplishment VPs, since such interpretations are incom-
patible with the “culminating” character of the corresponding situations. Con-
 Left/right contrasts among English temporal adverbials 

versely, in-adverbials have acceptable E interpretations with achievement and

accomplishment VPs, which indicate the time within which the correspond-
ing events reached their culmination and some change thereby occurred. They
also have acceptable consequent state interpretations with these VPs, reflect-
ing the possibility that they can identify the onset of such states. However, they
have no E interpretations with states and activities, since the latter involve no
culmination and thus no change (see, e.g., Dowty 1979: 336). Of the VPs that
have no particularly salient E interpretations either with in-adverbials or with
for-adverbials, only the pairing of in-adverbials with state VPs results in a truly
acceptable interpretation – namely, an R interpretation identifying the interval
within which the state becomes manifest. In contrast, the other pairings result
in only remotely acceptable interpretations. These are E and consequent state
interpretations for in-adverbials with activity VPs, where the activity is under-
stood as temporally bounded (as it is with similarly marginal interpretations
with interval-time adverbials); and repetitive interpretations for for-adverbials
with achievement and accomplishment VPs.
The investigation of VP-temporal adverbial pairs just reported has admit-
tedly been brief; and a more exhaustive one, which takes into account not only
the interpretations of specific VP-adverbial pairs, inevitably glossed over here,
but also the contribution of progressive forms, is certainly required to verify
its results. Yet, this investigation has arguably fulfilled one objective: namely,
to show that the interpretations of temporal adverbials, whether LP or RP, are
constrained by lexical compatibility requirements that determine their ability
to combine with particular VPs. What this means is that any “wild” interpre-
tations permitted by an orphan analysis of LP temporal adverbials can still be
ruled out on the basis of such independent lexical factors.
The results reported also offer strong support for the idea that both LP
and RP adverbial interpretations are best described in “modular” terms, as
involving a hearer’s use of syntactic, lexical, and conceptual information in de-
termining compatibilities between temporal adverbials and VPs and thus licit
interpretations for these pairings. That the determination of these compati-
bilities is not a mechanical matter, but rather a process driven by the hearer’s
attempt to establish a coherent discourse is, in fact, strongly suggested by the
availability of interpretations that have the effect of “rescuing” otherwise in-
compatible pairings and by the varying degrees of acceptability of these inter-
pretations. Such “rescued” pairings are, in fact, commonly described in terms
of processes of “event coercion” (see, e.g., Moens & Steedman 1988), whereby
a hearer resolves a conflict between the lexical properties of the modifier and
those of the VP by inferring a situation of the appropriate kind, derived from
 Benjamin Shaer

the description given in the original linguistic input. Attention to such pro-
cesses, and more generally to the hearer’s task of inferring how a temporal
adverbial is related to a VP – which arguably plays a role not only in “rescu-
ing” certain VP-adverbial pairs that may, strictly speaking, be ungrammatical
(see, e.g., Maienborn 2003: Section 3), but also in licensing less salient but still
acceptable interpretations of other pairs – thus represents an important aspect
of future research.
A final issue to be addressed here, which has been a key one throughout
this study, is that of the distribution of “higher” and “lower” interpretations of
LP and RP temporal adverbials. Although we have seen that the two interpreta-
tions are available with both occurrences of these adverbials, it is still true that
R and position-definite interpretations, which may be thought of as serving
to relate the time identified by the adverbial to a time introduced in previous
discourse, are far more salient with LP than with RP occurrences. This might
be seen as consistent with the orphan analysis, given its claim that LP adver-
bials are more weakly integrated into the sentence than their RP counterparts,
and with the further possibility that weaker integration into the sentence is off-
set by a correspondingly greater integration into previous discourse. We shall
be exploring this suggestion in the next section, where we investigate certain
discourse properties of LP temporal adverbials.

. Toward a discourse account of LP temporal adverbials

In the previous section, we observed that LP temporal adverbials, though by

hypothesis independent of their host sentences in the syntax and thus in prin-
ciple unconstrained in their interpretation, are in practice constrained by inde-
pendent requirements that their temporal properties be compatible with those
of the VPs with which they occur. Here, we shall see how discourse princi-
ples governing the distribution of these adverbials impose further constraints
on their distribution and interpretation, to which their RP counterparts are
not subject.
To determine the nature of these discourse constraints, it is necessary to
consider what LP temporal adverbials contribute to a discourse and what pat-
terns they exhibit that their RP counterparts do not. We can do so by asking
such questions about LP temporal adverbials as (i) why they are often marked
or even unacceptable discourse-initially, representing the special case of tem-
poral adverbial modification while their RP counterparts represent the general
case; (ii) why position-definite interpretations are often preferred for them, and
 Left/right contrasts among English temporal adverbials 

why these often seem more accessible to them than to their RP counterparts,
as suggested by the sentences in (62) (repeated from 12); and (iii) why certain
interpretations, such as the consequent state interpretation exemplified in (63),
appear to be inaccessible to them:
(62) a. For four years, John has lived in Boston.
b. John has lived in Boston for four years.
(63) a. He has been dead for two weeks.
b. *For two weeks, he has been dead. (Klein 1994: 188–89, ex. 10)

One way to approach these questions is within the context of the analysis
of other preposing constructions, such as topicalization, inversion, and left
dislocation, as illustrated in (64a, b, c), respectively:
(64) a. A: You know this album?
B: This song I know. (Birner & Ward 1998: 44, ex. 38a)
b. Behind a cluster of microphones was Mike Ditka, holding yet another
press conference. (Birner & Ward 1993, in Birner 1998: 311, ex. 3a)
c. Gallstones, you have them out and they’re out. (Roth 1969, in Birner &
Ward 1998: 93, ex. 112)

One difficulty for this undertaking is that LP temporal adverbials have proper-
ties significantly different from those of the more familiar preposed elements
given in (64). In particular, they are not lexically governed, and are thus less
constrained in their distribution than the latter elements. This difference has
led Birner and Ward (1998: 31–32), for example, to conclude that the licens-
ing conditions for these two classes of preposed elements are distinct. Yet, LP
temporal adverbials still seem to have enough in common with these other pre-
posed elements to warrant an investigation of them using tools developed for
such elements.
To do so, we can make use of the analysis of preposing constructions that
Birner and Ward (1998) develop, which turns out to be quite well suited to
this task. According to this analysis, what is crucial for these constructions is
the requirement that “the preposed constituent represent information that is
old in some sense” (Birner & Ward 1998: 23), the relevant sense being ‘old
with respect to the discourse’. Crucially, though, the preposed element does
not represent information that has already been “explicitly evoked in the prior
discourse”, but instead stands “in some salient and relevant relationship to ele-
ments that have been evoked” (Ward, Birner, & Huddleston 2002: 1368). Birner
and Ward (1998) cash out this specific notion of “discourse-old information”
in terms of the more basic notions of “link”, “anchor”, and “trigger”, which they
 Benjamin Shaer

define as follows. A “link” is “the linguistic material representing information

which stands in a contextually licensed poset [partially ordered set] relation
with information evoked in or [inferable] from the prior context”. It is the pre-
posed constituent that serves as this “link” to the “inferred poset”, and as such
“as a point of connection between the information presented in the current
utterance and the prior context” (Birner & Ward 1998: 20). Birner and Ward
(1998: 21–22) refer to the poset in question as the “anchor” and the “linguistic
or situational material which licenses the inference to the anchor” as the “trig-
ger”. Thus, in a discourse like the one in (65), the first sentence constitutes the
prior context; the “link” is the preposed constituent on a/the counter; the “an-
chor” is {elements of a house} (which, as Birner and Ward note, is “ordered by
the relation part-of ”); and the “triggers for the inference to the anchor” are the
expressions the kitchen and the counter (Birner & Ward 1998: 22).
(65) I walked into the kitchen. On a / the counter was a large book. (Birner &
Ward 1998: 20, ex. 15a)

A key factor, then, in the acceptable use of these preposing constructions is

the salience of the “linking relation”, the relation between the “link” and the
“anchor” (Birner & Ward 1998: 21, 45).
It is worth emphasizing that such preposing constructions establish a
tighter connection with the previous discourse than their counterparts with
canonical word order, but cannot, generally speaking, be considered more ac-
ceptable than them. Accordingly, the counterpart of (65) given below is also
perfectly acceptable:
(66) I walked into the kitchen. There was a large book on a/the counter.

We might gain greater insight into the differences between preposed elements
and their counterparts in canonical positions by considering an analogous dif-
ference between pronouns and names, as described by Stevenson (2002: 184).
Stevenson notes that the use of a pronoun “promotes coherence because find-
ing the referent of a pronoun involves relating it to other aspects of the text”,
whereas “a name may contain sufficient information to identify its referent
without reference to other aspects of the text”. Thus, while the sequence of sen-
tences in (67a), with coreferential names in each sentence, is no less acceptable
than its counterpart in (67b), with a name in the first sentence and a corefer-
ential pronoun in the second, the latter sequence, by virtue of the anaphoric
properties of the pronoun in its second sentence, does impose greater coher-
ence requirements on its felicitous use than the former sequence:
 Left/right contrasts among English temporal adverbials 

(67) a. Johni is the best person for the job. Johni is the one I want on my staff.
b. Johni is the best person for the job. Hei is the one I want on my staff.

Similar remarks apply to the sequences in (65) and (66): the former, by virtue
of the anaphoric properties of the preposed element, also imposes greater
coherence requirements on its felicitous use than the latter.
If we now try to extend these insights about preposed elements to LP tem-
poral adverbials, we find that these adverbials do establish some “point of
connection” between the sentence with which they occur and the previous
discourse, just as other preposed elements do. However, we also find that an
analysis in terms of “links”, “anchors”, and “triggers”, as sketched above, is dif-
ficult to apply directly to the case of LP temporal adverbials. This is simply
because these notions as defined for other preposing constructions all involve
sets of entities, whereas LP temporal adverbials denote sets of intervals, and
have no obvious discourse relation to other sets of intervals. Yet, these adver-
bials do establish a “link”, at least in some intuitive sense, between an “anchor”,
or inferred poset of entities, and the previous discourse; and certain expres-
sions do serve as “triggers” for the inference to this poset. The task, then, is
to determine how these discourse notions can be reconciled with the specific
properties of LP temporal adverbials.
Our attempt to do so might start with the examination of the scheduling
state description repeated from (38), in which two LP temporal adverbials are
clearly related to prior context:
(68) This is John’s schedule. On Monday, he plays a match in London and meets
the press. On Tuesday, he plays a match in Sheffield. . .

In this discourse, much like that in (65), a relation is established between the
prior context, as represented by the first sentence, and a poset – namely, {sched-
uled events in John’s week} – with the expressions John’s schedule and Monday
serving as triggers for this poset. Unlike the discourse in (65), however, the
poset inferred here is a set of situations rather than entities; and the preposed
elements do not denote members of this poset, but rather temporal locations
of the situations that constitute its members. In fact, the nature of this poset
gives us a plausible explanation for the unusual E interpretations that we have
already observed for these scheduling state descriptions. This is that the poset
consists of scheduled events, so it is the times of these events, rather than times
about which the speaker is making claims, that establishes a more salient con-
nection between the current utterance and prior context. This example, then,
suggests a relevant way of understanding “link” here: as a linguistic expression
 Benjamin Shaer

that identifies not a member of the “anchor” directly, but rather the interval in
which a member of the “anchor” is temporally located.
Examination of additional examples suggests that LP temporal adverbials
can be related to these “situation posets” in various ways and thus be assigned
different interpretations, depending both on the prior context and on their
own lexical properties. Consider discourses like the following ones (based on
12a and 14b, respectively):
(69) a. John has been feeling old and bored. For four years, he has lived in
Boston. Now it’s time to move on.
b. John has had an interesting couple of months. For two weeks, John has
been a milkman. For three weeks, he has been a paperboy. Now he wants
a desk job.

The posets associated with these discourses might be {situations that are the
causes of John’s feelings} and {interesting events over the past two months
of John’s life}, respectively. Likely triggers for these are the expressions feel-
ing old and bored and interesting couple of months. Although the for-adverbials
in (69a) and (69b) simply identify the duration of the situations described by
the VPs in their host sentences, we can trace their respective “up-to-now” and
“existential” interpretations to the discourses in which they are embedded. In
particular, the prior context in (69a) of John’s feeling old and bored makes the
description of a situation that has not changed and thus an “up-to-now” inter-
pretation the more likely one. In contrast, the prior context in (69b) of John’s
having had an interesting couple of months – and the discrepancy between the
duration of this state and that of the states described subsequently in the dis-
course – makes an “existential” interpretation of the temporal adverbials in this
discourse the more likely one.
Lexical and contextual factors also seem to be behind the availability of
“wide scope” and “narrow scope” interpretations of LP adverbials in the fol-
lowing discourses:
(70) a. Students’ lives at Résumé University were not just about studying and
partying. For example, one Sunday during the semester, every student
did volunteer work. (based on 8a)
b. (What can I say about that delightful time, so many years ago? Only
this.) On a beautiful Sunday in spring, every student on campus went
hiking. (based on 2a)

The poset in the first discourse might be {student activities not involving study-
ing or partying}, triggered by not just about studying and partying. What is
 Left/right contrasts among English temporal adverbials 

relevant here is that the members of this poset are activities engaged in either
habitually or episodically and either individually or collectively, which suggests
that the poset should be consistent with either “wide scope” or “narrow scope”
interpretations of the LP temporal adverbial. With no further means to resolve
this indeterminacy, both interpretations become available in the discourse.
The poset in the second discourse might be the singleton set {that delight-
ful situation in the past}, triggered by that delightful time, which biases the
temporal adverbial significantly in favor of a “wide scope” interpretation. Yet,
such an interpretation is also made salient by the adverbial itself, which not
only makes the Sunday in question a specific one, but also seems to reflect con-
ventional scene-setting language. This is arguably what is behind the acceptable
occurrence of this adverbial at the beginning of the discourse (as indicated by
the parentheses around the prior context in this example). That is, the richness
of its lexical content may be sufficient for the hearer to infer an appropriate
poset – say, {situations that typically hold or occur on a sunny spring Sunday} –
on the basis of a null context.
A final demonstration of the significance of prior context in licensing LP
temporal adverbials can be made for the apparently unacceptable sentence
given in (63b). If we return to this sentence and its acceptable counterpart in
(63a), we can see that the particular temporal properties of its VP, be dead,
force a consequent state interpretation for the for-adverbial that occurs with
it (as we already noted about this VP in Section 3). Given the acceptable RP
occurrence of this adverbial and the reasons that we have seen to doubt strictly
grammatical accounts of these left/right contrasts, we might instead try to ac-
count for the unacceptability of (63b) in non-grammatical terms. We can do
so, following Klein (1994: 189), by first asking the following question: in what
prior context would the duration of the consequent state of dying be relevant?
The discourse given in (71) offers one possibility:
(71) It’s hard to know which is more horrible, the assassination itself or the way
that it has been ignored by the government and the media. For two weeks,
he has been dead, but the government has made absolutely no statement and
not a single newspaper has bothered to report the murder.

Here, we might say that a poset is triggered by the assassination and ignored by
the government and the media: namely, {the consequent state of his murder}.
The most salient interpretation of the temporal adverbial that serves as a “link”
to such a poset is clearly an “up-to-now” interpretation, identifying an interval
extending to the time of speech. This, then, is consistent with the observed
interpretation of (71).
 Benjamin Shaer

Although the analysis of this and the other examples given above has hardly
been exhaustive, it has nevertheless shown that attention to the discourse prop-
erties of LP temporal adverbials can help us to reconcile an orphan analysis of
these adverbials with the interpretations available to them, and to capture one
obvious way – namely, their tighter connection to the previous discourse – in
which they differ from their RP counterparts.

. Conclusion

In this study, we have examined various contrasts associated with the inter-
pretations of LP and RP temporal adverbials in English. These contrasts have
included ones between “wide scope” and “narrow scope”, “up-to-now” and
“existential”, “position-definite” and “non-position-definite”, and “reference
time” and “event time” interpretations, all of which we have placed under the
rubric of contrasts between “higher” and “lower” interpretations, respectively.
What we have seen is that these contrasts do not correlate in any neat
way with higher and lower syntactic positions. In particular, the claim that
LP adverbials have only “higher” interpretations whereas RP adverbials have
“higher” or “lower” interpretations has encountered numerous counterexam-
ples among all of the contrasts that we have examined. In addition, the con-
trasts between “up-to-now” and “existential” interpretations and “position-
definite” and “non-position-definite” interpretations have turned out not to
reflect true ambiguities, given evidence based on the diagnostics in Gillon
(2004). While we did find true ambiguities with both “wide scope” and “nar-
row scope” and “reference time” and “event time” interpretations, neither con-
trast provided any support for a general structural determination of “higher”
and “lower” interpretations.
In seeking to capture the failure of LP temporal adverbials to have con-
sistently “higher” interpretations, we investigated the possibility that such ad-
verbials were “orphans” – elements syntactically independent of their host
sentences – and found considerable evidence for this possibility. Although
the orphan analysis itself imposed no constraints on the interpretations of LP
temporal adverbials, we did discover two independent constraints on their in-
terpretations. One, which applied equally to RP temporal adverbials, involved a
compatibility requirement on temporal adverbials and the VPs with which they
occurred. The other, which was specific to LP temporal adverbials, involved a
requirement that they be related to a poset of situations inferable from prior
 Left/right contrasts among English temporal adverbials 

context, thus capturing the tighter connection that they display to previous
discourse than their RP counterparts.
Although the analysis of LP and RP temporal adverbials presented here –
in particular, that related to their respective interpretations and the nature of
the restrictions on them – has remained tentative and informal, the “modu-
lar” picture of left/right contrasts that it reflects nevertheless emerges as both
plausible and worthy of further investigation.

Notes

* This paper reports the results of research conducted under the auspices of the P9 project,
“Positions and Interpretations for Sentence Topichood”, at the Zentrum für Allgemeine
Sprachwissenschaft, Berlin. It grows out of work reported in Shaer (2003, 2004, to appear)
and presented in preliminary form at the Universities of Leipzig and Amsterdam. I wish to
thank the audiences of these presentations, in particular Evan Mellander and Martin Schäfer,
and Werner Frey, Brendan Gillon, Nikolas Gisborne, Ed Keenan, and Claudia Maienborn for
helpful discussion of this material. Special thanks go to the editors of this volume and to two
anonymous reviewers for their copious remarks on an earlier version of this paper. These I
have incorporated into the present version in so many places that specific reference to them
would put me even further over the page limit.
. Of course, the indefinite descriptions involved in this pattern have also received analyses
(e.g. Kamp & Reyle 1993: 74–85) on which they are not inherently quantificational. The
remarks in the text appear to apply equally to indefinite descriptions on such an analysis.
. Note that de Swart is not explicit about the syntactic position of LP temporal adverbials,
and claims only that their RP counterparts are generated in sentence-final positions, from
which they modify the VP. However, her claim that the difference between LP and RP ad-
verbials “is rooted in the syntax” (de Swart 1999: 340) strongly suggests that the former are
generated in a high position.
. The limiting case being that of lexical ambiguity, where the difference between phrase
markers consists only of “a single node and its label” (Gillon 2004: 162).
. As such, these resemble standard cases of conversational implicature, as exemplified in
(i), the hallmark of which is their ability to be cancelled without contradiction (Brendan
Gillon, personal communication):

(i) Mary has a cat – in fact, she has two.

Implicature cancelled: Mary has only one cat.

. The influence of adverbial position on acceptability can also be seen in the following
examples from Hitzeman (1997: 89, ex. 6c–d):

(i) a. *For an hour Martha will be in her office one day next week.
b. *One day next week for an hour Martha will be in her office.
 Benjamin Shaer

However, this pattern seems to have more to do with licit combinations of temporal ad-
verbials than with the unavailability of ∼PD interpretations for LP occurrences of them,
as Hitzeman claims, given the acceptability of alternate orderings of the same LP temporal
adverbials, as indicated in (ii):

(ii) a. One day for an hour next week, Martha will be in her office.
b. For an hour one day next week, Martha will be in her office.

. Inevitably, the discussion here must omit both significant developments and significant
points of disagreement regarding the characterization of “reference time”. For discussion of
these issues, see, e.g., Declerck (1991: 224–232, 250–252); Kamp & Reyle (1993: 523, 593–
595); Klein (1992: 533–537).
. For a more detailed defense of this claim, see Shaer (1996: Chapter 2, Section 1.2.5.).
. This echoes Ryle’s (1954: 102) remark (quoted in Vendler 1967: 103) that “I can say ‘I
have seen it’ as soon as I can say ‘I see it.’”
. Of course, one could still claim that such RP adverbials are adjoined to IP rather than VP,
as suggested by Hitzeman (1993: 96) and others. However, independent evidence for such a
claim seems hard to come by; see Andrews (1982) for further discussion.
. This closely parallels evidence against a movement analysis of certain LP manner and
instrumental adverbials in English, as reported in Shaer (2003: 246–249, 2004, to appear).
. Both Parsons and Klein identify a class of “frequency” adverbials, such as every day and
often. I shall be omitting this class from my discussion, leaving consideration of them for
future research.

References

Abusch, D., & Rooth, M. (1990). “Temporal adverbials and the English perfect.” In J. Carter
et al. (Eds.), Proceedings of the 20th Meeting of the North East Linguistic Society (pp. 1–
15). Amherst, MA: GLSA, University of Massachusetts.
Andrews, A. (1982). “A note on the constituent structure of adverbials and auxiliaries.”
Linguistic Inquiry, 13, 313–317.
Asher, N. (1993). Reference to Abstract Objects in Discourse. Dordrecht: Kluwer.
Birner, B. (1998). “Recency effects in English inversion.” In M. Ward, A. Joshi, & E. Prince
(Eds.), Centering Theory in Discourse (pp. 309–323). Oxford: Clarendon Press.
Birner, B., & Ward. G. (1993). “There-sentences and inversion as distinct constructions:
A functional account.” In J. Guenter, B. Kaiser, & C. Zoll (Eds.), Proceedings of the
19th Annual Meeting of the Berkeley Linguistics Society (pp. 27–39). Berkeley: Berkeley
Linguistics Society.
Birner, B., & Ward, G. (1998). Information Status and Noncanonical Word Order in English.
Amsterdam and Philadelphia: Benjamins.
Declerck, R. (1991). Tense in English: Its Structure and Use in Discourse. London and New
York: Routledge.
Diesing, M. (1992). Indefinites. Cambridge, MA: MIT Press.
 Left/right contrasts among English temporal adverbials 

Dowty, D. R. (1979). Word Meaning and Montague Grammar: The Semantics of Verbs and
Times in Generative Semantics and in Montague’s PTQ. Dordrecht: Reidel.
Espinal, M. T. (1991). “The representation of disjunct constituents.” Language, 67, 726–762.
Gillon, B. (2004). “Ambiguity, indeterminacy, deixis and vagueness: Evidence and theory.”
In S. Davis & B. Gillon (Eds.), Semantics: A Reader (pp. 157–190). Oxford: Oxford
University Press.
Haegeman, L. (1991). “Parenthetical adverbials: The radical orphan approach.” In S. Chiba
et al. (Eds.), Aspects of Modern English Linguistics: Papers Presented to Masatomo Ukaji
on His 60th Birthday (pp. 232–254). Tokyo: Kaitakushi.
Hitzeman, J. (1997). “Semantic partition and the ambiguity of sentences containing
temporal adverbials.” Natural Language Semantics, 5, 87–100.
Hornstein, N. (1990). As Time Goes By: Tense and Universal Grammar. Cambridge, MA: MIT
Press.
Kamp, H., & Reyle, U. (1993). From Discourse to Logic. Dordrecht, Boston and London:
Kluwer.
Klein, W. (1992). “The present perfect puzzle.” Language, 68, 525–552.
Klein, W. (1994). Time in Language. London: Routledge.
Lasnik, H. (2003). Minimalist Investigations in Linguistic Theory. London and New York:
Routledge.
Maienborn, C. (2001). “On the position and interpretation of locative modifiers.” Natural
Language Semantics, 9, 191–240.
Maienborn, C. (2003). “On Davidsonian and Kimian states.” Ms., Berlin: Humboldt-
Universität.
McCawley, J. (1982). “Parentheticals and discontinuous constituent structure.” Linguistic
Inquiry, 13, 91–106.
McConnell-Ginet, S. (1982). “Adverbs and logical form: A linguistically realistic theory.”
Language, 58, 144–184.
Moens, M., & Steedman, M. (1988). “Temporal ontology and temporal reference”. Journal
of Computational Linguistics, 14, 15–28.
McGilvray, J. A. (1991). Tense, Reference, and Worldmaking. Montréal and Kingston: McGill-
Queen’s.
Musan, R. (2002). The German Perfect: Its Semantic Composition and Its Interactions with
Temporal Adverbials. Dordrecht, Boston and London: Kluwer.
Parsons, T. (1990). Events in the Semantics of English: A Study in Subatomic Semantics.
Cambridge, MA: MIT Press.
Pollard, C., & Sag, I. (1994). Head-driven Phrase Structure Grammar. Chicago and London:
University of Chicago Press.
Reichenbach, H. (1947). Elements of Symbolic Logic. New York: Macmillan.
Roth, P. (1969). Portnoy’s Complaint. New York: Random House.
Ryle, G. (1954). Dilemmas. Cambridge: Cambridge University Press.
Shaer, B. (1996). Making Sense of Tense: Tense, Time Reference, and Linking Theory. PhD,
Montreal: McGill.
Shaer, B. (2003). “‘Manner’ adverbs and the association theory: Some problems and
solutions.” In E. Lang, C. Maienborn, & C. Fabricius-Hansen (Eds.), Modifying Adjuncts
(pp. 211–259). Berlin: Mouton de Gruyter.
 Benjamin Shaer

Shaer, B. (2004). “An ‘orphan’ analysis of long and short adjunct movement in English.” In
G. Garding & M. Tsujimura (Eds.), Proceedings of the 22nd West Coast Conference on
Formal Linguistics (pp. 450–463). Somerville, MA: Cascadilla Press.
Shaer, B. (to appear). “On the syntactic status of certain fronted adverbials in English.”
In P. Nowak & C. Yoquelet (Eds.), Proceedings of the Berkeley Linguistics Society 29.
Berkeley, CA: Berkeley Linguistics Society.
Stevenson, R. (2002). “The role of salience in the production of referring expressions.” In
K. van Deemter & R. Kibble (Eds.), Information Sharing: Reference and Presupposition
in Language Generation and Interpretation (pp. 167–192). Stanford: CSLI.
de Swart, H. (1999). “Position and meaning: Time adverbials in context.” In P. Bosch &
R. van der Sandt (Eds.), Focus: Linguistic, Cognitive and Computational Perspectives
(pp. 336–361). Cambridge: Cambridge University Press.
Vendler, Z. (1967). Linguistics in Philosophy. Ithaca: Cornell University Press.
Ward, G., Birner, B., & Huddleston, R. (2002). “Information packaging.” In R. Huddleston &
G. K. Pullum (Eds.), The Cambridge Grammar of the English Language (pp. 1363–1447).
Cambridge: Cambridge University Press.
 Author index

A E
Abney, S. 168–171, 186–188, 198, Eckardt, R. 22, 260f.
200 Engels, E. 15f, 28f., 38
Abusch, D. 297 Ernst, T. 10, 14f., 24f., 30, 51, 74–78,
Adger, D. 12, 18, 22–24, 34 88, 95f., 105, 107, 115, 119f., 122,
Alexiadou, A. 7, 12, 178f., 182, 184, 136, 142f., 145–147, 152f., 205,
211, 278 211, 219f., 231, 243, 247, 277, 281
Andrews, A. 306f.
Arnold, D. J. 187
F
Fischer, O. 201
B Fodor, J. 75–77
Baker, C. L. 69f., 74f. Frey, W. 16, 254f., 257–260, 266f.,
Belletti, A. 237 276, 281–285
Birner, B. 291, 323f. Fukui, N. 64
Borer, H. 47, 56
Bowers, J. 135, 168f.
Bresnan, J. 36, 168–171, 198 G
Geuder, W. 20f., 35, 38, 132
Gillon, B. 296–298
C
Greenbaum, S. 280
Chomsky, N. 46, 54f., 59, 85, 212,
231, 243 Grimshaw, J. 81, 97, 99
Cinque, G. 10–12, 51f., 124f., 178f., Guémann, S. 140, 160, 164
198f., 211, 219f., 225, 234f., 241,
243, 278, 281
H
Corver, N. 37, 167–173, 196f.
Haegeman, L. 315
Costa, J. 15, 34, 49
Haider, H. 15, 24, 192, 245, 255,
257, 277f., 285
D Hale, K. 63
Déchaine, R.-M. 135, 148 Harris, Z. 45f.
Declerck, R. 305, 319 Haumann, D. 7f., 36–38
Demonte, V. 159 Hitzeman, J. 31, 299–304, 309, 329f.
Diesing, M. 300 Höhle, T. N. 255f.
Dowty, D. R. 150, 296 Hopper, P. 160
 Author index

Huddleston, R. 323f. Ortmann, A. 42

J P
Jackendoff, R. S. 116, 250 Parsons, T. 311, 317
Jacobs, J. 22, 261f., 283 Pesetsky, D. 55f., 154
Johnson, K. 54, 82, 98 Pinto, M. 59f.
Pittner, K. 15–17, 22, 24, 30, 254,
265f., 280
K
Pollock, J.-Y. 211f., 215f.
Kamp, H. 127f., 305
Potsdam, E. 76f.
Kayne, R. S. 11f., 50f., 182, 211, 221,
Pylkkänen, L. 157f.
243f.
Keyser, S. J. 63
Kim, J.-B. 75–77 Q
Klein, W. 299, 304f., 309, 317, 327 Quirk, R. 280
Koster, J. 177f., 189
Kratzer, A. 134, 159
Krifka, M. 139 R
Rapoport, T. 159f.
Reyle, U. 127f., 305f.
L Rijkhoek, P. 188f., 192, 200
Laenzlinger, C. 12–14, 24, 29f., 41, Rizzi, L. 173, 209, 236, 248
178f., 209, 222, 242 Roberts, I. 133, 135f., 158
Lambrecht, K. 272 Rooth, M. 297
Larson, R. K. 8f., 15, 30, 231, 281 Rosengren, I. 41
Lasnik, H. 54 Rothstein, S. 138–141, 164, 200
Lenerz, J. 256
Lobeck, A. 76f., 96
S
Sadler, L. 187
M
Sag, I. 75–77
Maienborn, C. 23, 113f., 126f., 269,
Shaer, B. 31f., 38, 278f.
276, 310, 322
Sportiche, D. 7f.
McConnell-Ginet, S. 8, 309
Steedman, M. 309f., 321
McGilvray, J. A. 308
Steinitz, R. 1
McNally, L. 138
Stevenson, R. 324
Merchant, J. 98
Stroik, T. 9, 15
Moens, M. 309f., 321
Swart, H. de 289, 293–296, 306, 309,
Müller, G. 81
315, 329

N
T
Nilsen, Ø. 52f.
Thompson, S. 160
Torrego, E. 46, 55f.
O Travis, L. 6, 53
Obenauer, H. 209 Tsoulas, A. 12, 18, 22–24, 34
 Author index 

V Wilder, C. 242
Vendler, Z. 317 Wyner, A. 158
Verkuyl, H. 56

Z
W Zagona, K. 76f.
Ward, G. 291, 323f. Zamparelli, R. 180f., 198f., 201
White, J. R. 185–188 Zwarts, J. 172
 Subject index

Bold print of page numbers indicates that the subject is treated throughout the chapter
starting with that page number.

A directional adverb(ial)
accomplishment 150, 160, 317–321 domain adverb(ial)
achievement 317–321 epistemic adverb(ial)
activity 160, 317–321 evaluative adverb(ial)
adjectival projection; adverb within ∼ event-internal adverb(ial)
35f., 38, 167ff. event-related adverb(ial)
adjective (see prenominal position, evidential adverb(ial)
post-copula position, for-adverbial (durational
postnominal position, postverbal adverbial)
position, adverb) frame adverb(ial)
adjective order 35 frequency adverb(ial)
adjunct approach (to the syntax of in-adverbial (time-span
adverbs) 10, 14–17, 30, 67ff., adverbial)
103ff., 253ff. instrumental adverb(ial)
adjunction (see left adjunction, right intentional adverb(ial)
adjunction) interval-time adverbial
Adv-Criterion 209, 242, 248 locative adverb(ial)
adverb (see also light adverb); manner adverb(ial)
relationship of ∼ to adjective means-domain adverb(ial)
6f., 34–36, 39, 42, 62, 131ff. mental attitude adverb(ial)
adverb(ial) (see also adverb(ial) modal adverb(ial)
classes, ambiguous point-time adverb(ial)
adverb(ial), obligatory predicational adverb(ial)
adverbial, PP adverbial, VP process-related/oriented
adverb(ial)); adverb(ial)
use of the term ∼ 2f., 33–36 proposition-related adverb(ial)
adverb(ial) classes 8f., 12, 16–20, 38, pure domain adverb(ial)
253ff.; see also quantificational adverb(ial)
agent-oriented adverb(ial) repetitive adverb(ial)
aspectual adverb(ial) sentence adverb(ial)
causal adverb(ial) speaker-oriented adverb(ial)
circumstantial adverb(ial) speech act adverb(ial)
comitative adverb(ial) subject-oriented adverb(ial)
degree adverb temporal adverb(ial)
 Subject index

transparent adverb(ial) order of ∼ wrt adverbials 4f.,

adverbial clause 34, 280 (see also 67ff., 104, 123, 146,
when-clause) 212–222, 231–239, 243, 246
adverb(ial) movement 13, 99, 123 auxiliary contraction 70, 74, 83, 96
adverbial NP 133 auxiliary movement 74–76, 79, 97,
adverbial projection; 122, 211, 242
AuxP (Auxiliary Phrase) 13, 78, 94,
adverb as head within ∼ 7
97f., 207, 213, 221, 231–235, 241,
adverb as modifier within ∼
245, 247
35, 119f., 167ff.
adverb morphology 35, 42, 127,
131f., 153f. B
affected object 22f., 45f., 63, 160 base position of adverb(ial)s 8,
afterthought 314 11–17, 22, 24, 30–32, 34, 45ff.,
agent; 67ff., 122f., 205ff., 253ff., 309,
∼ and manner adverb(ial), 313f.
23f., 45–47, 53f., 63 basic order (see base position)
∼ and subject-oriented Big Mess Construction 37, 179–182
adverb(ial) 275 binding (data indicating adverb
position) 30, 32, 52, 258, 281f.,
∼ and transparent adverb(ial)
300, 313, 315
142f., 156–159
agent-oriented adverb(ial) 116, 123
AgrP (Agreement Phrase) 50f., C
121f., 183f., 191 case feature (case checking) 59, 63
aktionsart (see lexical aspect) Case resistance 232
ambiguity test 298f., 307 causal adverb(ial) 16f., 271f.
ambiguous adverb(ial) 16–19, 22, causation 149–152, 157, 161
31, 131ff., 220f., 226f., 254, 260ff., c-command and adverb order 15f.,
289ff. (see also scope ambiguity) 22, 30, 32, 58, 60, 62, 109, 119f,
antisymmetric syntax 11, 51f., 211, 253ff., 295, 313, 315f.
243 Checking Theory 13, 208f., 212
AP (Adjective Phrase) (see adjectival circumstantial adverb(ial) 45ff.,
projection) 146f., 206
classes of adverbials (see adverb(ial)
applicative construction 284
classes)
Asp-associate relation 57–60
clefting 93, 306
aspect (see lexical aspect)
combien-extraction 209
aspectual adverb(ial) 223f., 227, comitative adverb(ial) 264f., 277f.,
234f., 246f. 282
AspP (Aspect Phrase) 12, 23f., 54, common ground 112–114, 125
57, 117f., 121, 213f., 225–228, comparison class 117
234f., 238, 241–245 complement;
atelicity 47, 56–60, 138f. adverb(ial) as ∼ 7–9, 12, 27, 40,
auxiliary; 269
∼ and adverbial projection 11 ∼ and little v 51–53
 Subject index 

∼ of an adverb(ial) 6f., 33, EPP feature 13, 23, 41, 55–63, 213
155f., 211 evaluative adverb(ial) 116, 273–275
order of verbal ∼ relative to event argument 9, 19f., 126, 103ff.,
adverbial 45ff., 205ff., 131ff., 282
253ff. event-dependence of a state
relative and small clauses as ∼ 149–153, 155f.
178, 180, 186–189 event-internal adverb(ial) 16, 23,
complex fronting 17, 22, 257, 261, 119, 125, 254, 266f., 271, 282
264–266, 272, 275 event-related adverb(ial) 16, 137,
ConjP (Conjunction Phrase) 178, 220, 239, 244, 250, 253ff.
181, 189, 191f., 200 event structure 310
contracted auxiliary (see auxiliary event time; relation of adverbial to ∼
contraction) 21, 163, 290, 304–313, 328 (see
control 157–161, 164 also reference time)
conversational implicature 153, 329 evidential adverb(ial) 16, 28f.,
Copy and Deletion 85 68–75, 85–96, 99, 273
existential reading of present perfect
D 31, 290, 296–299, 313, 326, 328
DegP (Degree Phrase) 7f., 37f.,
167ff.
DegP Hypothesis 168–170 (see also F
Split-DegP Hypothesis) F-closure 98
degree adverb 7f., 33, 36, 105, 167ff. feature-checking 12, 24, 41, 45ff.,
degree argument 38, 172f., 188, 193, 121, 124f., 205ff.
197f. FocP (Focus Phrase) 212, 240, 248f.
depictive 20f., 35, 131ff. focus; adverb(ial) in 13f., 119,
directional adverb(ial) 258–260, 194–196
283, 285 focus movement 7, 316
discourse and the interpretation of focus projection (data indicating
adverbials 1f., 20, 30–32, 39, adverb position) 17, 22, 24,
159, 289ff. (see also theme-rheme 136, 255f., 259, 261–270, 284
structure, topic-comment focus-sensitive particle 295, 313, 315
structure) for-adverbial 163, 317–321, 326
distributive quantification 262 frame adverb(ial) 16, 19, 23, 104,
domain adverb(ial) 5, 17, 24f., 103ff. 275–277, 280, 282
durational adverb(ial) (see frequency adverb(ial) 16, 28f., 69,
for-adverbial) 71, 73–75, 78, 80, 85–88, 92–95,
206, 219f., 245, 284
E fronting 13f., 29, 67ff. (see also
edge effect 27, 33, 192, 195 complex fronting, left-periphery,
e-givenness 98 topicalization, VP fronting)
Empty Category Principle 76f. functional projection;
epistemic adverb(ial) 16, 28f., 68f., adverb(ial)s as specifiers in ∼
71, 73–75, 78, 88–95, 109, 126, 10–13, 45ff., 121–124,
273f., 276 205ff., 278f.
 Subject index

adverb as heads of ∼ 7, 33, IP / InflP (Inflection Phrase) 13,

167ff. 67ff., 135f., 205ff., 305, 309, 311
hierarchy of ∼ 205ff.

L
G Larsonian shell 9, 11–13, 41, 52,
gap construction 15, 27–29, 67ff. 187f., 212
Generalized Government Last Resort 172f.
Transparency Corollary 96 left adjunction 127
goal: left periphery;
∼ and directionals 9, 259 ∼ of the NP and VP 37f.,
∼ of a probe 23, 54–56, 59–61 167ff., 241
∼ of the sentence 27f., 30–32,
H 241, 289ff. (see also complex
H-associate relation 55, 59, 63 fronting, fronting,
head movement 7, 13, 96, 127, 212, topicalization, parenthetical)
215 left-right contrast 289ff.
hierarchy; lexical aspect 23f., 32, 47f., 45ff.,
∼ of OT-constraints 68, 77, 88, 139, 312–317
93f. light adverb 27, 206, 211, 235,
∼ of semantic types 77, 277f. 245–248
hierarchical order of adverbials locative adverb(ial) 8f., 12, 16,
3–6, 18, 36, 205ff., 253ff. 18–20, 23–25, 34, 45ff., 113,
thematic ∼ 8f. 126f., 266–271, 275–277, 280
locative inversion (see inversion)

I
iconic order of adverb(ial)s 260, 285 M
implicature (see conversational manner adverb(ial) 8f., 12, 14, 18,
implicature) 20–24, 27f., 34f., 40, 42, 45ff.,
in-adverbial 317–321 105, 116–120, 125–127, 131ff.,
individual level 134, 142, 147f., 177 178f., 183f., 206, 220, 223, 231,
(see also stage level) 234–236, 245, 260–264, 269f.,
instrumental adverb(ial) 264–266, 278, 280, 282, 284, 330
277f., 280–284, 315, 330 MannP (Manner Phrase) 217–221,
integration 261f., 283 225f., 231–235, 241, 244
intentional adverb(ial) 145–147, 152 means-domain adverb(ial) 105, 110,
interpretable feature 12, 24, 41, 56f., 125
60, 209 mental attitude adverb(ial) 142, 147,
interval-time adverbial 317–321 265f., 275, 277, 282, 284
intraposition (movement) 281 middle field (see Mittelfeld)
inversion Minimalist framework 12, 23, 40,
locative ∼ 60 45ff., 205ff., 278f.
subject-auxiliary ∼ 28, 67f., mirror order 26, 29f., 41f., 206,
71–76, 81–83, 88–91, 94, 99 234f., 241, 247, 279–282
subject-verb ∼ 47, 58–60 Mittelfeld 12, 16f., 205ff., 253ff.
 Subject index 

modal adverb(ial) 121, 220, 223, PDP (Predicate Determiner Phrase)

225f., 229, 245–247 180
modal auxiliary (and adverb order) phi-feature (φ-feature) 58–61, 242
80f., 92f., 96, 117, 237 pied-piping 205ff. (see also snowball
modification marker 30, 281f. movement)
ModP / ModalP (Modal Phrase) point-time adverb(ial) 317–320
30f., 117f., 121f., 212–214, 220f., position (in)definiteness 31,
226, 233f., 239–243, 249f. 299–304, 313, 322
MoodP (Mood Phrase) 218, 225f., post-copula position (of DegP/QP)
241f. 176, 191
much-insertion 172f., 198
postnominal position (of DegP/QP)
37, 168, 174–177, 181f., 191
N postverbal position;
Nachfeld 12, 212, 230, 231–236, 241, ∼ of adverbial 5, 14f., 21,
245, 247 (see also right periphery) 27–30, 34, 41f., 49, 95f., 123,
negation, adverb order wrt ∼ 79, 81, 136, 143, 195, 206, 279–282,
96f., 109, 136, 219, 223–228, 230, 285 (see also right periphery)
244f., 250, 258, 260, 270, 293f. ∼ of DegP/QP 37, 168, 174f.,
negative polarity (data indicating 178f.
adverb position) 30, 281, 316 PP adverbial 34, 50, 243, 245
NegP (Negation Phrase) 14, 212f., pre-auxiliary position (of adverbial)
220f., 225–227, 230, 240, 244f., 41, 70f., 84–88, 91, 95, 144
250 predicate raising 180f.
predicational adverb(ial) 116–118,
O 127
ObjP (Object Phrase; and object predicative (adjective/DegP/QP) 35,
movement) 13, 29, 207, 38, 173–179, 182, 196–198, 201
222–235, 240–242, 245f., 250 (see also post-copula position)
obligatory adverbial 8, 11, 27 PredP (Predication Phrase) 117f.
of-insertion 46, 63f. prenominal position (of DegP/QP)
Optimality Theory 15, 38, 67ff. 37f., 170, 174–177, 182–188, 192,
order (of adverbials) (see adverbial 195–201
classes, base position, hierarchy) present perfect (see existential
orphan (syntactic ∼) 32, 315f., reading, up-to-now reading)
321f., 328 preverbal position;
overlap (temporal ∼) 21, 35, ∼ of adverbial 5, 15, 21, 26f.,
137–141, 150f., 163 29f., 41, 49f., 142f., 146f.,
206, 231, 281
P ∼ of DegP/QP 37, 170,
parenthetical 39, 92, 113, 163, 246, 174–178, 182–185, 192–199
283, 314, 315 (see also right Principle C (data indicating adverb
periphery, left periphery) position) 17, 258, 313f.
participle movement 5f., 92f. Principle of Full Interpretation 172
passive (and adverbials) 143, 158f. probe 23, 54–60, 63
 Subject index

process-related/oriented adverb(ial) S
16, 220, 244, 262, 264, 269, 280, scope 25, 28, 41, 67ff., 107–109,
282f. 114f., 125, 180f., 240ff., 253ff.,
proposition-related adverb(ial) 9, 289ff. (see also quantifier scope,
16, 77f., 90f., 254, 272–277, 282f. scope ambiguity, Scope Principle)
(see also sentence adverbial) scope ambiguity 16f., 22, 31, 206,
proto-patient and manner adverbial 247, 260, 265, 271, 277
22, 261f. Scope Principle 76, 96, 247, 258
pseudocleft and depictive 135f. scrambling 13, 15, 17, 30, 135,
psych-verb 46, 150 205ff., 253ff.
pure domain adverb(ial) 105, 124f. SDP (Strong Determiner Phrase)
Purity of Extended Projection 81, 97 180–182, 192
selectional restriction 19, 116
semantic type (semantic entity)
Q 15f., 77–79, 88, 121, 147, 162,
QP (Quantifier Phrase) 7f., 37f., 254, 272, 277f., 282f.
167ff. sentence adverb(ial) 16, 272–277,
quantificational adverb(ial) 13, 171, 282, 285 (see also
206, 209, 242, 248, 294 proposition-related adverbial)
quantificational force 173f., 181, 198 sentence-final position (see also right
quantifier scope (data indicating periphery);
adverb position) 16f., 22, 24, ∼ of adverb(ial) 12, 24f., 98,
31, 257f., 260, 270 135, 234f., 246, 329
quantization 24, 56–61 ∼ of result clause 192f., 197
∼ of verb 221f.
snowball movement 14, 29, 234f.,
R 241, 247 (see also pied-piping,
reduced relative 176, 191f. VP-remnant movement)
reference time; relation of adverbial to so-pronominalization 171
∼ 137, 284, 290, 304–313, 328 speaker-oriented adverb(ial) 20
(see also event time) specifier (see functional projection,
relativized minimality 173, 209 specifier approach)
repetitive adverb(ial) 14, 206, 220, specifier approach (to the syntax of
318f., 321 adverbs) 9–12, 17, 45ff., 123f.,
result clause 36, 38, 167ff. 205ff., 278f.
resultative predicate 135, 259 speech act adverb(ial) 18, 283
rheme (see theme-rheme structure) speech act operator 118, 123
right adjunction 11f., 15, 25, 30, 41, Split-DegP Hypothesis 37, 167f.,
51, 98, 135f., 280f. 173, 196 (see also DegP
right periphery; Hypothesis)
∼ of the NP and VP 37, 167ff. stacking (of adverbials) 26
∼ of the sentence 29–31, 234f., stage level 148, 177 (see also
246, 289ff. (see also individual-level)
parenthetical, sentence-final subject-auxiliary inversion (see
position) inversion)
 Subject index 

subject-oriented adverb(ial) 16, 20, U

28f., 68–75, 85–96, 99, 184, 220, unaccusative 23, 46, 160
265f., 273–275 Uniformity Condition on Chains
subject-verb inversion (see inversion) 173
SubjP (Subject Phrase) 213–217, uninterpretable feature 54–56, 59,
229–231, 241f. 63, 207, 242, 244
up-to-now reading of present perfect
31, 290, 296–299, 313, 326–328
T
T-associate relation 60
telicity 12, 24, 45–48, 56–59, 138f. V
temporal adverb(ial) 8f., 16f., 24f., V2 constraint 238
28f., 31–34, 71–75, 85–88, 92–95, verbal complex 22, 258–264, 269f.,
113, 229, 234–236, 245, 266–268, 282f.
270f., 275f., 278, 280, 284, 289ff. Vorfeld 12f., 26, 211f., 236–242 (see
thematic hierarchy (see hierarchy) also left periphery, topicalization)
theme-rheme structure (data VP adjunction 78, 136
indicating adverb position) 17, VP-adverb(ial) 41, 155, 212, 225,
24, 256f. 239, 244, 250
theta-binding (θ-binding) 173f. VP ellipsis 28f., 67ff., 313
though-attraction 36, 169 VP fronting 28f., 67ff., 135 (see also
time-span adverbial (see in-adverbial) fronting, left periphery)
topic-comment stucture (see also VP movement 205ff.
discourse and the interpretation VP-remnant movement 13, 29f., 41,
of adverbials) 13f., 19, 25, 30, 205ff. (see also pied-piping,
39, 81, 104, 111–115, 119f., 124f., snowball movement)
127, 228–230, 236–238, 248, 272,
VP shell (see Larsonian shell)
275–277, 285
topic rule 25, 30, 112f., 125
topicalization 7, 11, 13f., 28, 30, 39, W
67–69, 121, 123, 222, 239–242, weight 27, 34 (see also light adverb)
248–250, 259 (see also fronting, when-clauses 137
left periphery) wh-movement 11, 14, 28, 67ff., 104
TopP (Topic Phrase) 30, 212, 222, w-pronoun (data indicating adverb
236, 241, 248f. position) 258, 260, 267f., 270
TP (Tense Phrase) 59, 118, 212f.,
225f., 231–235, 241, 250
transparent adverb(ial) 20f., 35, 38, X
131ff. X-bar theory 9–11, 40, 243
In the series Linguistik Aktuell/Linguistics Today the following titles have been published
thus far or are scheduled for publication:

1 KLAPPENBACH, Ruth (1911–1977): Studien zur Modernen Deutschen Lexikographie. Auswahl

aus den Lexikographischen Arbeiten von Ruth Klappenbach, erweitert um drei Beiträge von Helene
Malige-Klappenbach. (Written in German). 1980. xxiii, 313 pp.
2 EHLICH, Konrad and Jürgen REHBEIN: Augenkommunikation. Methodenreflexion und
Beispielanalyse. 1982. viii, 150 pp. With many photographic ills.
3 ABRAHAM, Werner (ed.): On the Formal Syntax of the Westgermania. Papers from the 3rd
Groningen Grammar Talks (3e Groninger Grammatikgespräche), Groningen, January 1981. 1983.
vi, 242 pp.
4 ABRAHAM, Werner and Sjaak De MEIJ (eds.): Topic, Focus and Configurationality. Papers from the
6th Groningen Grammar Talks, Groningen, 1984. 1986. v, 349 pp.
5 GREWENDORF, Günther and Wolfgang STERNEFELD (eds.): Scrambling and Barriers. 1990.
vi, 442 pp.
6 BHATT, Christa, Elisabeth LÖBEL and Claudia Maria SCHMIDT (eds.): Syntactic Phrase Structure
Phenomena in Noun Phrases and Sentences. 1989. ix, 187 pp.
7 ÅFARLÍ, Tor A.: The Syntax of Norwegian Passive Constructions. 1992. xii, 177 pp.
8 FANSELOW, Gisbert (ed.): The Parametrization of Universal Grammar. 1993. xvii, 232 pp.
9 GELDEREN, Elly van: The Rise of Functional Categories. 1993. x, 224 pp.
10 CINQUE, Guglielmo and Giuliana GIUSTI (eds.): Advances in Roumanian Linguistics. 1995.
xi, 172 pp.
11 LUTZ, Uli and Jürgen PAFEL (eds.): On Extraction and Extraposition in German. 1996. xii, 315 pp.
12 ABRAHAM, Werner, Samuel David EPSTEIN, Höskuldur THRÁINSSON and C. Jan-Wouter
ZWART (eds.): Minimal Ideas. Syntactic studies in the minimalist framework. 1996. xii, 364 pp.
13 ALEXIADOU, Artemis and T. Alan HALL (eds.): Studies on Universal Grammar and Typological
Variation. 1997. viii, 252 pp.
14 ANAGNOSTOPOULOU, Elena, Henk van RIEMSDIJK and Frans ZWARTS (eds.): Materials on
Left Dislocation. 1997. viii, 349 pp.
15 ROHRBACHER, Bernhard Wolfgang: Morphology-Driven Syntax. A theory of V to I raising and pro-
drop. 1999. viii, 296 pp.
16 LIU, Feng-Hsi: Scope and Specificity. 1997. viii, 187 pp.
17 BEERMAN, Dorothee, David LEBLANC and Henk van RIEMSDIJK (eds.): Rightward Movement.
1997. vi, 410 pp.
18 ALEXIADOU, Artemis: Adverb Placement. A case study in antisymmetric syntax. 1997. x, 256 pp.
19 JOSEFSSON, Gunlög: Minimal Words in a Minimal Syntax. Word formation in Swedish. 1998.
ix, 199 pp.
20 LAENZLINGER, Christopher: Comparative Studies in Word Order Variation. Adverbs, pronouns, and
clause structure in Romance and Germanic. 1998. x, 371 pp.
21 KLEIN, Henny: Adverbs of Degree in Dutch and Related Languages. 1998. x, 232 pp.
22 ALEXIADOU, Artemis and Chris WILDER (eds.): Possessors, Predicates and Movement in the
Determiner Phrase. 1998. vi, 388 pp.
23 GIANNAKIDOU, Anastasia: Polarity Sensitivity as (Non)Veridical Dependency. 1998. xvi, 282 pp.
24 REBUSCHI, Georges and Laurice TULLER (eds.): The Grammar of Focus. 1999. vi, 366 pp.
25 FELSER, Claudia: Verbal Complement Clauses. A minimalist study of direct perception constructions.
1999. xiv, 278 pp.
26 ACKEMA, Peter: Issues in Morphosyntax. 1999. viii, 310 pp.
27 RŮŽIČKA, Rudolf: Control in Grammar and Pragmatics. A cross-linguistic study. 1999. x, 206 pp.
28 HERMANS, Ben and Marc van OOSTENDORP (eds.): The Derivational Residue in Phonological
Optimality Theory. 2000. viii, 322 pp.
29 MIYAMOTO, Tadao: The Light Verb Construction in Japanese. The role of the verbal noun. 2000.
xiv, 232 pp.
30 BEUKEMA, Frits and Marcel den DIKKEN (eds.): Clitic Phenomena in European Languages. 2000.
x, 324 pp.
31 SVENONIUS, Peter (ed.): The Derivation of VO and OV. 2000. vi, 372 pp.
32 ALEXIADOU, Artemis, Paul LAW, Andre MEINUNGER and Chris WILDER (eds.): The Syntax of
Relative Clauses. 2000. vi, 397 pp.
33 PUSKÁS, Genoveva: Word Order in Hungarian. The syntax of Ā-positions. 2000. xvi, 398 pp.
34 REULAND, Eric J. (ed.): Arguments and Case. Explaining Burzio’s Generalization. 2000. xii, 255 pp.
35 HRÓARSDÓTTIR, Thorbjörg: Word Order Change in Icelandic. From OV to VO. 2001. xiv, 385 pp.
36 GERLACH, Birgit and Janet GRIJZENHOUT (eds.): Clitics in Phonology, Morphology and Syntax.
2001. xii, 441 pp.
37 LUTZ, Uli, Gereon MÜLLER and Arnim von STECHOW (eds.): Wh-Scope Marking. 2000.
vi, 483 pp.
38 MEINUNGER, Andre: Syntactic Aspects of Topic and Comment. 2000. xii, 247 pp.
39 GELDEREN, Elly van: A History of English Reflexive Pronouns. Person, Self, and Interpretability. 2000.
xiv, 279 pp.
40 HOEKSEMA, Jack, Hotze RULLMANN, Víctor SÁNCHEZ-VALENCIA and Ton van der
WOUDEN (eds.): Perspectives on Negation and Polarity Items. 2001. xii, 368 pp.
41 ZELLER, Jochen: Particle Verbs and Local Domains. 2001. xii, 325 pp.
42 ALEXIADOU, Artemis: Functional Structure in Nominals. Nominalization and ergativity. 2001.
x, 233 pp.
43 FEATHERSTON, Sam: Empty Categories in Sentence Processing. 2001. xvi, 279 pp.
44 TAYLAN, Eser Erguvanlı (ed.): The Verb in Turkish. 2002. xviii, 267 pp.
45 ABRAHAM, Werner and C. Jan-Wouter ZWART (eds.): Issues in Formal German(ic) Typology. 2002.
xviii, 336 pp.
46 PANAGIOTIDIS, Phoevos: Pronouns, Clitics and Empty Nouns. ‘Pronominality’ and licensing in
syntax. 2002. x, 214 pp.
47 BARBIERS, Sjef, Frits BEUKEMA and Wim van der WURFF (eds.): Modality and its Interaction
with the Verbal System. 2002. x, 290 pp.
48 ALEXIADOU, Artemis, Elena ANAGNOSTOPOULOU, Sjef BARBIERS and Hans-Martin
GÄRTNER (eds.): Dimensions of Movement. From features to remnants. 2002. vi, 345 pp.
49 ALEXIADOU, Artemis (ed.): Theoretical Approaches to Universals. 2002. viii, 319 pp.
50 STEINBACH, Markus: Middle Voice. A comparative study in the syntax-semantics interface of
German. 2002. xii, 340 pp.
51 GERLACH, Birgit: Clitics between Syntax and Lexicon. 2002. xii, 282 pp.
52 SIMON, Horst J. and Heike WIESE (eds.): Pronouns – Grammar and Representation. 2002.
xii, 294 pp.
53 ZWART, C. Jan-Wouter and Werner ABRAHAM (eds.): Studies in Comparative Germanic Syntax.
Proceedings from the 15th Workshop on Comparative Germanic Syntax (Groningen, May 26–27,
2000). 2002. xiv, 407 pp.
54 BAPTISTA, Marlyse: The Syntax of Cape Verdean Creole. The Sotavento varieties. 2003.
xxii, 294 pp. (incl. CD-rom).
55 COENE, Martine and Yves D’HULST (eds.): From NP to DP. Volume 1: The syntax and semantics of
noun phrases. 2003. vi, 362 pp.
56 COENE, Martine and Yves D’HULST (eds.): From NP to DP. Volume 2: The expression of possession
in noun phrases. 2003. x, 295 pp.
57 DI SCIULLO, Anna Maria (ed.): Asymmetry in Grammar. Volume 1: Syntax and semantics. 2003.
vi, 405 pp.
58 DI SCIULLO, Anna Maria (ed.): Asymmetry in Grammar. Volume 2: Morphology, phonology,
acquisition. 2003. vi, 309 pp.
59 DEHÉ, Nicole: Particle Verbs in English. Syntax, information structure and intonation. 2002.
xii, 305 pp.
60 TRIPS, Carola: From OV to VO in Early Middle English. 2002. xiv, 359 pp.
61 SCHWABE, Kerstin and Susanne WINKLER (eds.): The Interfaces. Deriving and interpreting
omitted structures. 2003. vi, 403 pp.
62 CARNIE, Andrew, Heidi HARLEY and MaryAnn WILLIE (eds.): Formal Approaches to Function in
Grammar. In honor of Eloise Jelinek. 2003. xii, 378 pp.
63 BOECKX, Cedric: Islands and Chains. Resumption as stranding. 2003. xii, 224 pp.
64 BOECKX, Cedric and Kleanthes K. GROHMANN (eds.): Multiple Wh-Fronting. 2003. x, 292 pp.
65 MANNINEN, Satu Helena: Small Phrase Layers. A study of Finnish Manner Adverbials. 2003.
xii, 275 pp.
66 GROHMANN, Kleanthes K.: Prolific Domains. On the Anti-Locality of movement dependencies.
2003. xvi, 372 pp.
67 MIŠESKA TOMIĆ, Olga (ed.): Balkan Syntax and Semantics. 2004. xv, 474 pp. + index.
68 BREUL, Carsten: Focus Structure in Generative Grammar. An integrated syntactic, semantic and
intonational approach. 2004. x, 432 pp.
69 KISS, Katalin É. and Henk van RIEMSDIJK (eds.): Verb Clusters. A study of Hungarian, German and
Dutch. 2004. vi, 514 pp.
70 AUSTIN, Jennifer R., Stefan ENGELBERG and Gisa RAUH (eds.): Adverbials. The interplay between
meaning, context, and syntactic structure. 2004. x, 343 pp.
71 GELDEREN, Elly van: Grammaticalization as Economy. 2004. xvi, 320 pp.
72 FUß, Eric and Carola TRIPS (eds.): Diachronic Clues to Synchronic Grammar. viii, 218 pp. + index.
Expected Fall 2004