Methods for the Descriptive Analysis of Archaeological Material

Author(s): J. C. Gardin
Source: American Antiquity, Vol. 32, No. 1 (Jan., 1967), pp. 13-30
Published by: Cambridge University Press
Stable URL: https://www.jstor.org/stable/278775
Accessed: 27-09-2018 15:14 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Cambridge University Press is collaborating with JSTOR to digitize, preserve and extend
access to American Antiquity

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
 METHODS FOR THE DESCRIPTIVE ANALYSIS OF
ARCHAEOLOGICAL MATERIAL

J. C. GARDIN

ABSTRACT generally available in mimeographed copies

Most studies on the use of punched cards and com- from the Centre d'Analyse Documentaire pour
l'Archeologie, where they have been devised,
puters in archaeology seem to take for granted that scien-
tific standards exist to express the data upon which and where they are undergoing experiments.
algorithms are to be performed, for retrieval or classifi-
The list concerns only non-verbal data; how-
cation purposes. The author's view is different; exam-
ples are given of descriptive codes which have been de-
ever as early as 1958, an application of the same
signed under his direction since 1955 for the storage methods to archaeological written documents
was also demonstrated (see Revue d'Assyriolo-
of archaeological data (artifacts, abstract or figured rep-
resentations, buildings, etc.) on punched cards of various
gie, Vol. 52, 1958, p. 36, and Orientalis, Vol.
kinds (marginal, peek-a-boo, IBM, etc.). In order to
27, 1958, pp. 417-8). More recently, a full text
obviate the shortcomings of natural language, three cate-
gories of rules are required: orientation, segmentation, has been submitted to a semantic analysis of
differentiation. The concluding remarks concern the the same inspiration (Allard and others 1963).
relation of the descriptive languages which are thus ob-
tained to scientific language in general; differences are THEORY
stressed, as well as reasons for postulating a continuum
from the former to the latter. A general outline of the procedure was given
a few years ago (Gardin 1958); it can now be
THE CONCERN for standards of descrip- formulated in a more systematic fashion by
tion in archaeology is not new; it seems, considering three kinds of rules, applicable to
however, to have been revived out of a growing the analysis of any of the foregoing data: rules
interest in information-processing methods of orientation, rules of segmentation, rules of
which require a break-up of the data to com- differentiation.
binations of well-defined features. Those meth- Orientation. Artifacts, buildings, and abstract
ods are usually envisaged in connection with or figurative design are essentially motionless
the use of punched cards or digital computers, when examined by the archaeologist; in order
and with a view to applications in information that they should give rise to a "representation"
retrieval and/or automatic classification; for ref- of any sort, it is necessary that some rule or con-
erences in English, see the bibliography by vention be adopted as to the position which the
Tugby (1965), and also the current projects observer should take in relation to them. A
mentioned in Chenhall (1965), to be comple-
few examples will serve to illustrate the im-
mented by Kovalevskaja (1965), Kamenetskijportance of such an arrangement before one can
(1965), Marshak (1965), Krug (1965), in Rus- begin the process of segmentation and differen-
sian, and Ihm (1961), Elisseeff (1965), De La tiation.
Vega and others (1965), in French. Most of Figure 1 shows two axes, a and b, which
those papers, however, deal with the mathe- share exactly the same profile-and suppos-
matical aspects of information processing- edly, for the sake of the argument, the same
methods of classification, distribution in space,cross-section so that they can be considered
arrangement in time -rather than with the lin- as identical, the apparent difference resulting
guistic or semiological question: in which terms only from two alternative conventions as to
should the data be analyzed, or expressed, so their representations (in the physical and/or
that they can be stored on punched cards, and psychological sense of the word): "take as upper
handled with computers? Such was our first edge the straight side" (a), or "the concave
goal, when we undertook in 1955 the design side" (b). Statistically, most archaeologists seem
of various codes for the description of artifactsto have a preference for the former, a, probably
(Gardin 1956). The present paper is intended because of a tacit assumption that the haft,
to provide an illustration of the methods whichwhich is usually represented "below" the blade
have been followed for different categories of for obvious functional reasons, is located on the
archaeological data (indicated in Table 1). The concave side, as is apparent in axis c. Counter-
unpublished codes mentioned in this table are examples, however, can also be found, as with
13

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
14 AMERICAN ANTIQUITY [ VOL. 32, No. 1, 1967

the transcription of a given design in terms of

the descriptive language should be unique.
I ) H1 ) Segmentation. Segmentation rules, the sec-
a I b ond step in our procedure, have the same pur-
pose: once the object has been set in a standard
position, from the observer's point of view (lit-
erally speaking), a new set of rules is now re-
quired to impose a unique form of reduction in
c i d terms of morphological components.
Two preliminary remarks should be made at
FIG. 1. Two possible orientations o
this point. One concerns the apparently privi-
according to parallels with c or d
leged status which morphological analysis will
occupy in the following paragraphs, as opposed
axis d; so the tacit assumption has to give way to functional, technological, or any other spe-
to an explicit convention, which imposes an cies of analysis. The reasons for this approach
arbitrary orientation -either a or b -when are many: on the one hand, it often happens
no clues are available as to the supposedly nat- that not much is known about function of
ural, that is, functional position of the object,objects when found by an archaeologist, as dis-
in relation to the observer or to the user. tinct from the cultural anthropologist. Simi-
Figure 2 illustrates the same kind of posi- larly, the kinds of chemical or physical indica-
tional ambiguity in the case of a piece of pottery tions which may occasionally be found in the
when observed out of its functional context, for present archaeological literature do not warrant
it may be conceived either as a cup, a, or as a an extensive use of such data for the discrimina-
lid, b. It may even happen that the object is tion of individual objects. Conversely, morpho-
functionally ambivalent, in which case a con- logical features are the easiest to observe and
ventional orientation would be all the more are those which usually show the widest varia-
necessary to achieve a standardized description. tions; furthermore, the more refined species of
analysis, for instance stylistic, are based on form,
The same problems naturally arise in analyz-
rather than on any other criteria. For these rea-
ing iconographical documents. Even in abstract
sons, it seems best to concentrate first on mor-
design, motifs are seldom wholly symmetrical;
phological elements.
and their customary interpretation -or desig-
Our second remark concerns the very proce-
nation - may change according to the position
dure of reduction in terms of components, mor-
from which the observer looks at them (Fig. 3,
phological or otherwise. It is important to re-
a). Figurative elements may also be ambiguous,
member that our motivation for adopting such
either accidentally (Fig. 3, b), or voluntarily,
a procedure is here entirely practical, that is:
when seen from different angles (anamorpho-
(a) the need to substitute analytical expres-
sis). A frequent species of ambiguity occurs
sions, made up of well-defined terms, for syn-
when the contrast between black and white sur-
thetical, ill-defined designations; (b) the desira-
faces, or more generally decorated and undeco-
bility of using a descriptive system in which
rated parts, is such that the eye can successively
individual terms can be addressed separately,
focus on two distinct representations (Fig. 3, c).
or in any combination 2 by 2 . .. n by n, for
In all such cases, rules are necessary for the
processing purposes. This pragmatic attitude
guidance of the analyst. They may be of many
toward the problem of segmentation should
different kinds (for example, "index all perceiv-
always be kept in mind when considering the
able interpretations of a given pattern, by vary- kinds of rules which have been adopted, and
ing its orientation to the observer"; or "adoptwhich we shall now characterize briefly.
such and such criteria for deciding upon a ca-
nonical orientation, which in turn will deter-
mine a unique interpretation of individual
motifs"; or conversely, "select such and such
b
canonical interpretations of individual motifs, a

irrespective of orientation," etc.); all that mat- FIG. 2. Two possible orientations of the same vase,
ters, at this stage, is that such rules exist, so that as cup a or lid b.

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
GARDIN ] DESCRIPTIVE ANALYSIS 15

a b c d e
. 1 1

FIG. 4. Variations in the delimitation of neck and

body, for vase c, according to parallels with a, b, or d, e.
a c

FIG. 3. Possible variations in the designation of orna-

from one observer to another, so as to preclude
ment, following changes in orientation a, b, or relation
between figure and ground c. a unique reduction. Figure 4 illustrates this
point: in the range of vases a to e, should vase
Only one question need be answered at this c be analyzed in terms of a concave body, as
juncture: what are the criteria for the segmen- vases a and b, or in terms of a convex body
tation of a given object -artifact, monument, with a concave neck or rim, as vases d and e?
No formal definition of a neck is available to
iconographical representation - into a unique
set of parts or components? decide which description, that is, segmentation,
One way to approach the problem is to should be preferred. This kind of ambiguity is
examine the way in which it has been solved common with all categories of artifacts: tools,
-even if only in a very approximate fashion weapons, etc. (Fig. 5). It even arises in the case
- by the natural language habits of archaeolo- of the apparently more obvious figurative repre-
gists, past and present. sentations. Figure 6, for instance, can be de-
A rapid inspection of the literature (or intro- scribed in several ways - (a) two persons greet-
spection of one's own habits) suffices to show ing, next to a standing deity; (b) a pair of wor-
that most natural partitions are based on func- shipers, one of whom faces a standing deity;
tional distinctions. A pot has a foot or base on (c) a worshiper facing a deity with attendant,
which it stands, a body which contains what- etc. - according to the segmentation which is
ever is to be contained, sometimes a neck to implicitly adopted.
help in holding or pouring, etc. Similarly, a tool When confronted with such problems,
or weapon is usually described as consisting of archaeologists frequently maintain that it is
two main parts, one of which is for prehension possible to solve them by considering external
(haft, handle, tang, etc.), the other for action criteria. One can indeed imagine that some
(blade, point, cutting-edge, etc.), and so forth. knowledge of the way in which vase c (Fig. 4)
The case of iconographical documents may not was actually handled is enough to decide
seem so clear; the parts which are named in whether or not the description should mention
natural language are usually those motifs whicha neck; similarly, that a reference to the histori-
cal antecedents of tool c (Fig. 5) should show
are implicitly taken to have a kind of existential
autonomy in that they can be found unchanged whether it is to be segmented as tool a-b, or
or with minor variations in other representa- d-e; and also that the interpretation of the
tions. But it could be shown that this intuitive scene on Fig. 6 will in fact be uniquely deter-
criterion is in some sense also functional: what mined, if other sources - mythological, literary,
changes, from one representation to another, etc. - are brought to bear. Nothing can be ob-
is the role which each such "part" plays in jected to in this hopeful attitude, except, natur-
relation to the whole design, that is, its func-ally, the fact that it is hopeful. More often than
tion in the context -a purely decorative one
if the design is abstract, and/or a narrative func-
tion in the case of figured representations.
No reason occurs, a priori, why this kind of
segmentation, based as it is on a long range of
experience, should be discarded in devising
rigorous analytical procedures. Unfortunately, a b c d e
it soon becomes clear that many tacit assump- FIG. 5. Variati
tions underlie functional distinctions and vary according to

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
16 AMERICAN ANTIQUITY [ VOL. 32, No. 1, 1967

As for iconographic documents, arbitrary cri-

teria will also have to be given to ensure that,
in a given cultural area, no single representation
can be segmented in more than one axiomatic
way. Conventions, in that case, are not metri-
cal, as for the shapes of tools and pots, but are
definitely to be included in the same category
of rules of precedence. For the kind of compo-
x y z sition illustrated by Fig. 6, for instance, such a
rule could be the following: "symmetrical
groups should be considered as independent
a: (x < i y) --- z
parts of the picture, irrespective of their rela-
b: x - y z tion to neighboring individuals or groups"
(which imposes segmentation b, as above); or
"human beings should be arranged into (at
C :
x ---(y z) least) two parts, according to their role as sub-
FIG. 6. Differences in the interpretation of the
jects or objects with respect same
to the action of the
scene, according to parsings a, b, c.
representation" (which imposes segmentation
a or c following the interpretation of the central
not, in the study of "dead"
figure objects
as a co-worshiper observed
or as an attendant of
out of their cultural context, little is known the deity), etc. Formulations of this kind are
about function, genetic filiation, intended mean-nothing but a priori hierarchies, or rules of
ing, etc.; or when something is known, it is to
precedence, which impose arbitrary structures
a large extent a matter of speculation, whichto iconographic representations, much in the
hardly provides a ground for the objective, im-
same way as metrical ratios determine arbitrary
personal description which is here required. segmentations of objects.
An alternative solution would be to approach Differentiation. What has been achieved at
the problem of segmentation in a purely metri- this stage, through orientation and segmentation
cal fashion, which might consist, for instance, in
rules, is only preliminary to the description it-
scanning the physical data according to one orself: our objects have first been positioned in
several absolute grids. This method is only men-relation to the observer, then partitioned accord-
tioned for the sake of the argument, since its ing to grids which now have to be filled, so to
practical drawbacks are obvious: complexity ofsay, with actual information, in the etymologi-
analysis, non-comparability of expressions, losscal sense of the word. Each part of the grid
of the more apparent homologies, etc. Yet, it can be considered as a variable in the analytical
indicates a direction in which one can go in system; the third step is to determine the differ-
order to deal with ambiguities such as those of ent values which each variable may take for
Figs. 4 to 6, after one has made the most of the description of a given set of data - hence
customary, empirical criteria of segmentation. the name "differentiation" which we have given
Let us first take Fig. 4. Since neither internal to that last aspect of our procedure.
nor external criteria are always available to de- The problem of differentiation can in fact be
cide upon the partition of vase c, as compared considered as an iteration of the segmentation
respectively with a-b and d-e, a convention hasprocess for each part or variable taken sepa-
to be provided, of the same a priori character rately. Let us imagine that the values of any
as above. It might consist in selecting for the variable-that is, the different forms which
ratio h/H a given value, say 1/5, below which
the segment corresponding to "h" would be
assimilated to a "neck" (or to a "rim"), as
with d-e, and above which the same segment
would be included in the description of the
"body." I I I _

Similar conventions will be needed to solve o | b |c d I e

difficulties of the kind illustrated by Fig. 5, in FIG. 7. Differentiations on a co
the description of other categories of artifacts. geometric profiles.

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
GARDIN ] DESCRIPTIVE ANALYSIS 17

logical community. Weaknesses are equally

obvious, however: one of them is that, as in the
process of segmentation, little is known of the
c
exact boundary between any two consecutive
units in a given scale of denomination. In other
words, whereas the relative difference between
b such units is generally clear, there is no consen-
sus as to the absolute value of each, with re-
spect to the physical continuum. Thus, on Fig.
7, where is the dividing line between "strongly
convex" and "slightly convex"7 Or between
"mule" and "donkey," "horse" and "mule," on
FIG. 8. Differentiations on a continuous scale of
Fig. 9? Another weakness of natural scales of
S-shaped linear ornament.
differentiation is that they often lack the resolv-
ing power which is needed for information pur-
the corresponding part may exhibit on actual poses. The opposition between "strongly con-
objects - belong to a continuum, as in Figs. 7 to vex" and "slightly convex," for instance, may
9. It is clear that some groupings will have to not be sufficient to differentiate a large number
be made, so as to keep the descriptive vocabu- of pottery shapes, with regard to profile; a finer
lary within reasonable limits; in other words, set of conventional designations may have to be
the continuum will have to be "quantified," invented in such cases.
through explicit rules of differentiation, just as The problem of uncertain bounds between
whole objects had to be quantified through neighboring units of denomination will there-
explicit rules of segmentation. The same ques- fore have to be solved in the same way as
tion then arises: what criteria regulate the dis- above, for segmentation: that is, by adding rules,
crimination process? metrical or otherwise, to rationalize the use of
Here again, the seemingly easiest answer is to such units in systematic descriptions. An exam-
rely on natural differentiations - that is, those ple of a metrical convention is given on Fig. 7,
which are expressed in natural language-or where groups of curves are differentiated accord-
on the specialized vocabulary of archaeologists. ing to predetermined boundary values of a giv-
What will then be found is a range of terms en ratio, h/1. Using this simple kind of axiom,
which can be ordered along the same contin- however, is not always practicable. It would be
uum as above and brought into correspondence unwieldy, for instance, to try to regulate the
with different sections of the latter. For the differentiation of ornaments, let alone figured
curves of Fig. 7 for instance, a possible scale ofrepresentations such as on Figs. 8 and 9, on a
denominations would be, from left to right: a,purely formal basis. An alternative solution,
strongly convex-b, slightly convex-c, straightthen, is to accept equivoques in the descriptive
- d, slightly concave - e, strongly concave; forvocabulary, that is, to leave criteria of differ-
the ornaments of Fig. 8, from top to bottom: a, entiation unspecified, but providing at the same
horizontal S, interlocking - b, Laufenhunde (atime a mechanism to disclose all such equi-
technical term borrowed from the German) voques to the user of the vocabulary. Cross-
c, line of spirals; and for the figured representa-references are a usual means to that effect; a
tions of Fig. 9, from left to right: donkey-mule,more systematic solution is to group the pro-
onager-horse. posed discriminating symbols according to se-
As in the case of segmentation, natural de-mantic affinities, so as to materialize the differ-
nominations of that kind have both merits and ent scales along which indecision may occur
weaknesses. Their main merit is to exist and to
have gained by force of habit a popularity which
is a useful quality in the context of information
systems. For that reason alone, it is obvious
that one should take advantage of existing lin-
guistic differentiations, rather than invent radi- b C

cally new ones, which would impose burden- FIG. 9. Differentiations on a continuous scale of
some learning processes on the whole archaeo- similar figured representations.

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
18 AMERICAN ANTIQUITY [ VOL. 32, No. 1, 1967

TABLE 1. KINDS OF OBJECTS STUDIED

Categories Cultural Group State of elaboration

A. ARTIFACTS

1. Pottery Bronze age, Europe Unpublished code with sample punched card
and Western Asia index, for ca. 1000 pots.
2. Tools & Weapons Bronze age, Europe Printed catalogue with punched card index, for
(metal) and Western Asia ca. 4000 tools (Christophe and Deshayes 1964)

B. ICONOGRAPHY

1. Ornament (universal) Unpublished code.

2. Numismatics Greek Coins from Crete, Unpublished code and punched card index,
6th-4th cent. B.C. for ca. 1000 coins.
3. Glyptography Oriental cylinder seals, Printed catalogue with punched card index under
3rd-lst mill. B.C. preparation, for ca. 4500 cylinder seals.
4. Paintings Greek Vases, Unpublished code and sample punched card
Classical Period index, for ca. 200 vases.
5. Mosaics Roman mosaics Unpublished code, and experimental punched
from Gaule card index, for ca. 400 mosaics.

C. ARCHITECTURE

1. Religious Medieval churches, Unpublished code, under experiment.

France

2. Civil Private residences, Unpublished code, under experiment.

France

from one symbol to the next. The result is a different according to which interpretation is
structured vocabulary, samples of which will be chosen by the analyst. In order to unify the
illustrated in the following section. description, a convention may be chosen that
conical or quasi-conical lids should be consid-
EXAMPLES ered upside down, the end or knob downward
-an orientation which in this case is at vari-
Two examples will be given, one in the fieldance with archaeological usage. The result will
of pottery, the other one in glyptography. Both
be that identical objects, so far as their shape
examples are based on existing codes, in use
goes, will be segmented and differentiated in the
at the Centre d'Analyse Documentaire pour
same way, whatever distinctions may be alleged
l'Archeologie (see Table 1).
as to their function. A counterpart to this un-
Pottery. Even a summary of the complete orthodox, but apparently needed, rule will be
code of pottery would take too much space for
to indicate elsewhere in the description, under
the purport of this paper. It seems more appro-
a distinct category, the observer's view, if he
priate to illustrate the general descriptive pro-
cedure through one relatively simple example,has any, regarding function.
such as that on Fig. 10.
First, the object has to be properly oriented,
c
or positioned, in relation to the observer. The
more common prescribed position is that which .... -. . Al
is usual in archaeological publications: the pot AE{-
is viewed as standing on its "base," with the A2

"opening" upward. In the present case, how- B _

__ _._ _ _ _ _~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

ever, a problem arises insofar as the pottery in

Fig. 10 may be interpreted as a lid, the repre-
sentation of which should then be the other
FIG. 10. Conventional
way round, with the "knob" upward. The seg-
body A (upper part Al, lower part A2), base B, and
mentation which is to follow will naturally be rim C.

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
GARDIN ] DESCRIPTIVE ANALYSIS 19

Neck Rim PROFILE

SLOPE VI concave |d, straight X, convex

divergent i

Body UL ..... ..
paralleeL [
vu du xu i-

0, /

Base converge
vo " do xot..
FIG. 11. Ambiguities in the separation of pottery parts,
taken by pairs; neck/rim, neck/body, etc. FIG. 13. Analysis of simple profiles: body (upper or
lower parts), neck, base.

The next step is segmentation according to

three parts: Body, Base, and Rim (other sec-
pre-determined rules. The natural segmenta-
tions of the code concern Neck, Handles, Lugs,
tion in this case is into three parts, A, B, C, as
Spouts, and also non-morphological aspects of
shown on Fig. 10; how is it to be related to the
conventional division of the code into "Body," the description, such as Decoration, Tech-
niques, Archaeological origin, etc.).
"Base," "Neck," and "Rim"? Formal rules are
given to distinguish between those four parts, A. Body
when necessary (that is, when any two of them, a. Upper part (Fig. 13)
linked in the diagram on Fig. 11, seem equally
1. slope ("divergent," "vertical," "con-
acceptable to describe a given segment of a
vergent")
given vase). Empirical analysis, however, is
usually sufficient, as in this example, to separate 2. profile ("concave," "straight," "con-
the "base" B from the rest of the profile on Fig. vex")
10. An ambiguity nevertheless arises for part C: b. Lower part
should it be described as the "upper part of the 1. slope (same terms)
body," or as a "rim," distinct from the body?
2. profile (same terms)
Both interpretations are morphologically ac-
ceptable, in view of the existence of vases such c. Junction between upper and lower
as on Fig. 12, a and b respectively. It is then part ("continuous," "angle," "rib,"
necessary to refer to the formal rule which is etc.) (Fig. 14)
provided in the code as a basis of the empirical d. Ratio H/L (Fig. 15) (for example, 5
distinction between "rim" and "body." The re- ranges of values, arbitrarily separated
sult in this case is that part C will convention- by numbers 1/2, 1 and 2)
ally be analyzed as a "rim."
e. Ratio o/a (Fig. 15) (same differentia-
The object now stands ready for description
tion)
proper, that is, the differentiation of features
for each of its various parts. For lack of space, f. Ratio l/L (Fig. 15) (for example, eith-
we shall have to restrict ourselves to a few er "< 1/2" or ">1/2")
extracts from the code, showing the kind of
traits which have been proposed just for those
c I qx qv zi Zo

0z
a
b

FIG. 12. Two vases showing possible ambiguities in the FIG. 14. Analysis of junc
denomination of part C, on Fig. 10. parts of the body, bod

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
20 AMERICAN ANTIQUITY [ VOL. 32, No. 1, 1967

L e. Bottom, lower side (same terms)

f. Ratio h/H (same principle as above,
T- A, d)
o,
g. Ratio h/1 (same)
h. Ratio l/L (same)

FIG. 15. Measurements for the analysis of C. Rim (Fig. 18):

pottery shapes: Body.
a. Junction with upper part of the body
(or with the neck)
Notes:
b. Inclination, relative to the body ("in-
(1) The distinction between "upper" and turned," "out-turned," etc.)
"lower" parts of the body is to be interpreted as
c. Inclination, absolute (from the verti-
an arbitrary one for which formal rules are cal: for example, "near zero," "near
given; it applies to all kinds of vases, including 90?," etc.)
those for which no dividing line is apparent
(for example, the vase on Fig. 10, where the d. Sides, relative inclination (for exam-
separation A1-A2 is purely formal). ple: "divergent," "parallel," "conver-
gent")
(2) The lists of features in brackets are those
which are provided in the code which we are e. Inner side, profile (for example: "con-
now using; but other differentiations could be cave," "straight," "convex")
adopted, either finer or broader, according to f. Outer side, profile (same terms)
needs.
g. End, profile (same terms)
(3) If finer distinctions were to be used, how-
h. End, absolute inclination (as under c,
ever, one should be careful not to introduce
thereby unnecessary redundancy. For instance, above)
under "profile" (of upper and lower parts), one
Thus, three groups of eight features will be
might wish to distinguish between "strongly
used to characterize the vase on Fig. 10, respec-
convex," etc.; but it should be borne in mind
tively for the body, the base, and the rim. Each
that this additional specification is to a large
feature is designated by a short symbol - usu-
extent correlated with other features, such as
ratios H/L, o/a, and l/L, single or combined. ally a mnemonic letter for recurrent features
(Fig. 13) or a number for more specific ones
(4) Nothing is said here of absolute dimen-
sions; these are stated under a separate heading,
(Fig. 17) -preceded by an alpha-numeric
following the same quantification procedure as group which indicates the segment or part of
for relative dimensions, above. the description to which it belongs (example:
Al, Body, upper part, etc.). The final product
B. Base - Similar features are used for the
following components of the base (Fig. L
16):
a. Junction with lower part of the body
(same as Fig. 14)
b. Sides (same as Fig. 13, to which can
be added other profiles, if necessary -
for example: "angle, in-turned," "an-
gle, out-turned," "S-shape," "inverted
S," etc.)
c. Junction with bottom (Fig. 17)
d. Bottom, upper side ("convex,"
"straight," "concave," "depression,"
L _ l I
etc.) FIG. 16. Pottery, variables for the analysis of bases.

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
GARDIN 1 DESCRIPTIVE ANALYSIS 21

tation of an artifact and that of an iconographi-

0 1 2 3 4 5 6
cal document. In both cases, segments are indi-
vidualized on the basis of their functional rela-
tion to the whole context -a physical func-

i J2 KW3y tion when the context is an object which is to

perform a given task, and a logical function
FIG. 17. Pottery bases: junctions between sides when the context is a representation which is
and bottom (examples). to convey a given meaning (or at least a given
pattern, if the representation is non-figurative).
of the description is then a chain of symbols of The problem of iconographical segmentation
the following type: then becomes relatively easy to state, if not to
solve: what are the logical functions to be taken
"Al/d, i; A2/d, i; A3/...; Cl/6; C2/3,.... into account in the analysis of figured represen-
etc."
tations of the kind shown on Fig. 197
In actual fact, all alpha-numeric groups (pre- Several systems can be proposed. We shall
fixes) are prearranged on a descriptive sheet illustrate only one, which has been adopted for
the composition of a punched card index of
which the analyst only has to fill up with very
short symbols, most of which are mnemonic and engraved gems from the Near East, to be pub-
are recurrent from one group to the other. lished shortly.
Thus, the full analysis of any object requires The diagram on Table 2 summarizes that
very little writing; and yet, the obtained for-system. The underlined symbols indicate the
mula is both richer and more precise than any different functions or "parts," properly speak-
of the usual descriptions of pottery in naturaling, which may be allotted to elements in the
language. Also, it is immediately amenable to picture. The initial partition is of the broadest
punched card and computer techniques for kind. It concerns the breaking up of the docu-
documentation or other purposes. ment-a physical unit, in the present case, a
cylinder seal- into autonomous sub-units,
Iconography. The scene on Fig. 19 will serve
from a logical standpoint: that is, scenes in
to illustrate our method of iconographical
which iconographical elements are explicitly
analysis, following the same principles as for
interrelated (through tangible association or evi-
pots, tools, and weapons. The motifs on that
dent interaction), but which conversely have no
picture, and also the general features of com-
necessary connection with one another. This
position, are borrowed from Oriental glyptog-
first aspect of segmentation is one of the most
raphy, as it developed in the Near East from the
challenging in any kind of formal analysis,
third to the first millennium B.C.; however,
whether of iconographical or other documents.
their present combination is artificial and is
Though no mention was made of it earlier in
only meant to illustrate as many elements as
this paper, it is present in the analysis of arti-
possible for the purpose of this demonstration.
Let us bypass the first step in the descriptive
procedure, orientation, since its function has
been abundantly commented upon in the last , i/c
pages; moreover, the natural orientation of a
picture does not usually give rise to contentions, ......./ - -- --
so that only a few rules are needed in that
respect.
r~T'~~~~~~~~~~~~~~~~~~~~~~~~~~
The case of segmentation is much more in- e I

volved: what criteria shall we use to divide the

X -- <9~\
scene into a system of components? A general
answer has been given above: as a set of (physi-
o,I
cal) parts, we shall consider the (logical) parts,
or roles, which individual elements may take
in a picture, figurative or other. This deliberate I
play on the word "part" suggests that there is
in fact a complete analogy between the segmen- FIG. 18. Pottery, variables for the analysis of rims.

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
22
AMERICAN ANTIQUITY [ VOL. 32, No. 1, 1967

TABLE 2. ANALYSIS OF FIGURED REPRESENTATIONS: PARTS AND THEIR RELATIONS

Document

Individual scenes or sequences I

Group Adjuncts
(G) (H)

e0

Not interacting Interacting Interacting: Not interacting

App. (S,O)

Subjects : S Independents Ind.

with attributes L, I,Q/S Appositions .App
L, Locative with attributes L,I,Q/H
I, Instrumental
Q, Qualificative

Subjects ; : Objects :
(S) (0)

I
S St 'i" S2 0 O'

with attributes L,I,Q/S with attributes L,I,Q/O

or with attributes L,I,Q/G

facts, when a single object has two axe,

or more
etc.). But it is naturally better known in
functions, each performed by a different sub-
the context of linguistic analysis, where a simi-
lar distinction
unit (for example: hammer-axe, axe-adze, pick- is common between the larger

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
GARDIN ] DESCRIPTIVE ANALYSIS 23

1 i 2-- 3 4- c. when interacting in the broadest sense of

the word (example: greeting, meeting,
etc.), including the case of a purely pas-
sive participation in the action (example:
spectators, attendants, etc.); and so forth.
Any number of so related beings form a
X"group" in the picture. By definition, a scene
5 or sequence is limited to one group, G. It can,
however, include "adjuncts," that is, additional
motifs which are neither related to that group,
1Kf a^ in terms of the preceding conventions, nor to
one another. For the description of Oriental
cylinder seals, the category of adjuncts has been
Q I L, divided into two parts: "Independent" beings,
SS S S,
\? 0_ which apparently have the same status (or sta-
I 8 _ ture) as actors in group G, but turn away from
FIG. 19. Iconography (oriental engravin
that group (example: the central figure, no. 2,
of a segmentation scheme: separate scenes of intmerre on the top line of Fig. 19); and "Appositions,"
lated Subjects/Objects (here two: nos. 1nd
ar 7-8), with which are smaller beings usually located in
possible "attributes" (symbols L.I.Q., in gro )up 7-8), and gaps within or outside group G. A third case
occasional "adjuncts" (symbols Ind. and App., as in occurs when appositions themselves form a
scene no. 1, with other adjuncts in the s
nos. 2,4,5), which may in turn form a se aame register, minor group in the picture, according to the
(symbol "App. S,O," no. 3). same criteria as above; the corresponding scene
is then mentioned as a special kind of adjunct
physical unit of discourse and logicoal sub-units to group G (symbol "App. S, O") but after-
such as sentences, propositions, etc It would wards is analyzed as a group in its own right
require much space, here as elsewhelre, to enun- (example: motif no. 3, on Fig. 19).
mitation of Let us now return to group G itself. It is,
ciate truly formal rules for the delii
sentences or their iconographical e !quivalents. in turn, to be segmented into components, ac-
Let us rather admit that the proce 'ss remains cording to the different "parts" which individual
largely empirical, its only basis being the follow- motifs may play in the picture. There is natur-
ing quasi-definition: a scene or seque nce may be ally no such thing as a universal "cast," equally
formed of (a) a group of explici itly related suited to the analysis of all kinds of icono-
motifs, to which may be added un related ad- graphical scenes; the one which is illustrated in
junctions; (b) one or several not eN Kplicitly re- Fig. 19 and Table 2 is to a large extent the re-
lated motifs, when a group such as (a) is pres- flection of specific compositional features in ori-
ent on the document. A sizeable 1 uncertainty ental glyptography. Some distinctions, how-
remains, however, as to the criteria for appre- ever, are of a general order: such is, first, the
ciating explicit relatedness; the indi ications be- opposition between "subjects" and "objects,"
low will throw a dim light on this undeniably and secondly, that between both categories of
dark aspect of segmentation. actors on the one hand and their so-called
The only objective method foi r deciding "attributes" on the other. "Subjects" and "ob-
whether given beings should bensidered
cor as jects" are conventionally determined by depend-
related or not, in iconographical analysis, is ing upon the kind of interaction, the essence or
again by a priori conventions, suchas the fol- appearance of the beings concerned, etc. As for
lowing:
"attributes," they are distinguished on the basis
a. (beings will be taken as relat
Led) when of their apparent subordination to those beings,
physically in contact with or
ie another "subjects" and "objects" alike -a subordina-
(example: man sitting on a horrse, woman . .
tion which may express itself through
with child, etc.);
patterns:
b. when engaged in the same i intransitive
process (example: walking, procession, (1) the location of x in a subjugated posi-
running, etc.); tion under y ("Locative," L; example:

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
24 AMERICAN ANTIQUITY [ VOL. 32, No. 1, 1967

TABLE 3. ICONOGRAPHICAL SEGMENTATION

Related beings G

Subjects (S) Objects (0) Adjuncts H

S S' S" S2 0 0' Ind. App.
L/S L/S' L/S" L/S2 L/O L/O' L/Ind. L/App.
I/S I/S' I/S" I/S2 I/OI/O' I/Ind. I/App.
Q/S Q/S' Q/S" Q/S2 Q/O Q/O' Q/Ind. Q/App.

ample: S', S", etc.);

"deity standing on a griffin," others co-occur
"king tread- so seldom
that they need
ing on crouching griffin," not be separated in actual prac-
etc.);
(2) the use of x in a given tice (example: L/Ind., L/App.,
action by y etc.); others
("In-
strumental," I; example: "deity standing still, when applied to inanimate objects, hardly
in a chariot drawn by a griffin," etc.); bear the distinction (example: I/... and Q/...,
(3) the carrying of x by y ("Qualificative," etc.). The segmentation system can therefore
be simplified, up to the point where fewer
Q; example: "woman holding child on
knees," "hero lifting dead animal over components are separated (Tables 4, 5).
shoulders," etc.); and so forth. Let us now turn to differentiation: what kind
of discriminations shall we make for each of the
The three attributive cases - L, I, Q - apply
proposed "parts" of the picture? A general
to both animate and inanimate beings. In order answer has already been suggested above when
to avoid ambiguities, one should further specifyillustrating some of the difficulties of a formal
to which "part" of the picture they are each differentiation of figurative elements (Fig. 9).
related: "subjects" or "objects" within group G Just as in the case of artifacts, one should in
("L, I, Q/S," "L, I, Q/O"), or adjuncts to the principle draw up lists of such elements as
group ("L, I, Q/Ind.," "L, I, Q/App."). In might be observed in each of the predetermined
actual practice, however, the segmentation canOccurrences of the same elements in
parts.
be simplified, to make up only for the broader different parts, however, are here more likely
distinctions (example: "L, I, Q/G," "L, I, than ever: a lion, in oriental art, may occupy
Q/H"). practically all the positions laid down in Table
Finally, for "subjects" and "objects," further 3. It is therefore advisable to disregard segmen-
partitions are provided, if necessary: tation altogether when compiling the lists of
(1) S and 0, primary Subject and Object of differentiated motifs. An alternative method is
the action in group G; to present them according to a general classifica-
(2) S' and O', secondary Subjects and Ob- tion scheme, which is then independent from
jects (example: "followers," "attend- the particular part or position which each motif
ants," etc.) may take in various compositions. The follow-
(3) S", the "presented" subject, brought by ing table is a summary of the classification
S or S' to the presence of 0 (example: which has been adopted in the field under
study; Fig. 20 illustrates the actual differentia-
an offering to a god, a prisoner to a king,
etc.) tion in a given class (Hybrids).
(4) S2, the "chain-subject," when a being be- Animates
longing to group S is himself the object
Humans (including human-shaped gods, etc.)
of an action by S2, etc.
Animals Attitudes, humans
Formal rules are given to help in separating Hybrids Attitudes, animals
"primary" from "secondary" figures, etc., but
Inanimates
they need not be consulted very often, thanks
to the usual convergence of empirical analysis Buildings Ornament
in that respect. Furniture Inscription S
Table 3 illustrates the theoretical segmenta- Emblems
tion which may result from the preceding dis- Plants Unidentifi( ed
tinctions. Some logical parts in this system, Containers
however, can only apply to animate beings (ex-Instruments (including weapons)

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
GARDIN ] DESCRIPTIVE ANALYSIS 25

TABLE 4. SIMPLIFIED SEGMENTATION, ANIMATEcal codes.

BEINGS In the preceding lines, no provision
has apparently been made for the description
Related beings G
of action. More precisely, whereas the existence
Subjects (S) Objects (0)
of action is implicit in the very segmentation
system which has been adopted, the essence of
S S' " S 2 0 0' Ind. App.
L/S L/O L/H action does not seem to have any bearing on
I/S I/O I/H differentiation, as one might normally expect.
Q/S Q/O Q/H An expression such as "Sm.Oa," above, tells us
nothing about the way in which the "human-
subject(s)," Sm, is or are, acting toward the
The analysis of an iconographical document "animal-object(s)," Oa. The reason for not
thus consists in identifying individual motifs, specifying the action should by now be clear to
as designated in such lists, and allocating to the reader. It is essentially that only explicit
each one of them a position or a "part" in the features find their way in the codes or differen-
segmentation system. The process may seem to tiation systems which we have described, con-
be the reverse of that which has been described cerning the external shape of beings or things,
for artifacts, where one first identifies the con-their attitude or posture, other formal details
ventional parts of the object before designatingof their representation, etc. As for actions, they
the various features in each part. That differ- are here considered as implicit meanings, at-
ence is, however, largely superficial. It is true tached to some particular combinations of such
that the symbols which are used in each cate- features. The interpretation "humans fighting
gory are formally symmetrical: for artifacts animals," for instance, certainly implies many
they include a positional prefix, followed by the explicit features as to the attitude of the former
name of a given trait ("Blade, upper edge/con- (raising hands, holding weapon, etc.) and/or
cave"), whereas for iconographical documents, the latter (lying, crouching, etc.). There are
the prefix indicates the trait, followed by the then two reasons, in principle, for omitting such
positional symbol ("palm-tree/Independent," interpretations: one is that they are redundant
outside the main group in the picture). Yet, (since the explicit features upon which they are
the alternative is only for pragmatic reasons: based will anyhow be mentioned in the mor-
on tools, pots, weapons, etc., morphological fea- phological description); secondly, to the extent
tures do not make much sense when dissoci- that they are interpretations they tend to be
ated from their location, whereas in figuredsubjective,
rep- or simply changing from one observ-
resentations, motifs can be envisaged in isola- er to another, so that little reliance can be
tion, regardless of their position in the picture. placed upon them for our purpose. Here again,
Moreover, counter-examples exist in both cases: the analogy with the analysis of artifacts is com-
in the code for Tools and Weapons, some very plete: saying that a given tool is a "sickle" is
specific features are designated by symbolsbut summarizing under one single name an
where the positional indication is only secon- empirical set of morphological features, with-
dary ("holes/on tang, on blade," etc.); con- out going to the pains of formally defining the
versely, most iconographical codes contain a sec-range of admitted variations for each part of the
tion in which structural information comes set (hence the occasional diversity of names
first with symbols of the following kind:for "Sm"the same object: "sickle," "pruning-knife,"
(Subject: humans), "Oa" (Object: animals), "scythe," "knife with curved blade," etc.). The
"Ip" (Independent: plants), etc. Such symbols rejection of the category of action, in icono-
are strictly comparable to those which have graphical analysis, would therefore be in keep-
been described above for the analysis of arti-
facts: a combination of the kind "Sm.Oa" indi-
cates a broad category of iconographical scenesTABLE 5. SIMPLIFIED SEGMENTATION, INANIMATE BEINGS

(humans [fighting, tending, etc.] animals), in

Related beings G Adjuncts H
the same way as a combination "Functional
part: point. Prehension part: shaft-hole" indi-
S 0 Ind. App.
cates a broad category of tools or weapons.
L/G L/H
That analogy may help in understanding an
I, Q/S I, Q/O I, Q/H
important negative characteristic of iconographi-

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
26 AMERICAN ANTIQUITY [ VOL. 32, No. 1, 1967

FIG. 20. Iconography (oriental engravings), example of a differentiation scheme, for Hybrids. Each unit is desig-
nated by a three-symbol expression, indicating: a) upper part of the body; b) lower part; c) absence or presence
of wings (symbols - and +). Additional somatic anomalies may be expressed by a fourth symbol, etc.

ing with a very general rule in our descriptiveever, is largely a matter of practical concern,
method. which will not be debated here.
In actual practice, however, it has been found
CONCLUDING REMARKS
convenient to provide some reference to that
category, through the use of limited, rather Let us recall the object of analysis, from pre-
formal dynamic designations, which serve as ceding examples: its primary aim is to make up
short substitutes for complex combinations offor some deficiencies of archaeological descrip-
static features. For example, "striking, about totions, when set in terms of natural language,
strike," "seizing, grasping," "treading," etc., are and to provide for a more satisfactory communi-
useful specifications of the relation between cation between archaeologists about given sets
Subjects and Objects, which do not imply muchof data. The descriptive codes which have been
interpretation but only an empirical summariza- sketched are only means to that end; yet, there
tion of different combinations of attitudes or is a danger that they may be regarded as final
attributes, on both sides. Such summarizations products, to the extent that they constitute a
may, but do not necessarily, dispense with thescientific terminology superior to any existing
explicit enumeration of components; this, how-one in the field of archaeology. Our conclud-

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
GARDIN ] DESCRIPTIVE ANALYSIS 27

ing remarks will be concerned with some of the

b) Second, the descriptive languages which
theoretical issues which are involved in that
have been proposed are not universal. The
confusion. semiological quantification of perceptual data
To say of a given terminological systemisthat
unavoidably influenced by the observer's lin-
it is scientific may mean either of two things: guistic, or more generally cultural, habits, even
it may indicate that there exist formal ruleswhen as the objects of perception are concrete
to the relation between symbols and observa- things, static monuments, and it is likely that
tional data; it may mean that the system is a other authors would have constructed other
useful tool to establish relations of quite an- codes, for the same objects. For our purpose,
other kind between the elements of the ob-however, this objection is self-defeating; any
served data themselves, as reflected by symbolicartificial language would fall by the same stand-
relationships within the system. Let us considerard, so that the real alternative is between cul-
the merits of our codes from both standpoints turally-bound codes, or no codes at all. In addi-
(Sections A and B, respectively). tion, it must here again be emphasized that the
A. As far as the first criterion is concerned,proposed quantification is based on ad hoc con-
there seems to be little doubt that the prerequi- ventions which stand midway between the
sites of scientific method have generally been usually tacit assumptions of a purely "cultural
met. The obtained formulas do not claim to order of things," and a more reasoned "natural
express any so-called objective order of things, order," in turn based on conventions, only of
but only to provide inter-subjective paraphrases another purport. Descriptive codes, in this
of such things, in terms which can be regularly sense, are both universal and culturally-rooted,
and without ambiguity correlated with physical just as other well-known artificial codes such as
observations. This regularity and lack of ambi- the metric system, the gregorian calendar, the
guity of the sign-thing relationship may decimal
be system. Table 6 is an attempt at a
accepted as a minimum condition of accepta- rationalization of the foregoing distinctions.
bility for a language of science. Various objec- c) Third, the reduction is not unique, by
tions can, however, be made, even from that which one means either that the same data can
limited viewpoint, to the codes which we have be analyzed in different terms, within the same
described: descriptive system - another formulation of ob-
a) First, that the codification process is not jection a, above - or that other systems are
entirely formal; there are cases, especially in conceivable for the description of the same data
iconography, when empirical differentiations are- a corollary of objection b. Explanations have
accepted at their face value, only because for- been given for both cases; let us only add that
mal discrimination would seem unnecessarily the plurality of descriptive systems is not in
complex from a pragmatic standpoint. The only itself an objection to their applicability, any
answer to this objection is that anomalies of more than the diversity of scientific theories is an
this kind are deliberate, and that explicit indi- obstacle to their validity. Different theories serve
cations are then provided to make up for the different ends, in the interpretation of a given
indecision which may result ("see also" refer- set of data, just as different systems may serve
ences, semantic proximity of descriptive fea- different needs, for the description of the same
tures, double notations, etc.). data. But the criteria are obviously not the

TABLE 6. THREE KINDS OF SEMIOLOGICAL QUANTIFICATIONS OF PERCEPTUAL DATA

Natural language Artificial codes Scientific language

based on no explicit system based (in part) on a priori based on experimental findings,
or rules conventions,

therefore not universal, which provide for universally which lead to a tentatively universal
acceptable descriptive systems, cognitive system,

but rather indicative of different intermediate, see Fig. 21 indicative of a supposedly "natural"
"cultural" orders of things. order of things.

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
28 AMERICAN ANTIQUITY [ VOL. 32, No. 1, 1967

same in deciding between conflicting theories in 1 D Cognitive

Perceptual
science, or conflicting descriptions for documen-
data systems
tation purposes.
d) Last, the kind of analysis which has been
suggested has no cognitive value, in that it does
not help us to understand the phenomena
L 2 Descriptive _____3
under analysis, to discover structure, meaning,
systems
genetic filiations, etc. There is no disagreement
on that point either, insofar as the primaryFIG. 21. Two paths in the transition from perceptual
data to cognitive systems.
purpose of analysis is here quite different.
Rather than state again the difference, let us
reformulate the objection in another way, for phrases of empirically observed distributions.
the sake of the discussion: since cognitive sys- The a priori approach to segmentation and dif-
tems are the admitted goal of science, is it neces-ferentiation is at variance with the methods of
sary to pass through the intermediate stage structural analysis; and any pseudo-structure
which is suggested in our table above? which may be found in the code - for example
There is no general answer to that question. the system of physical or logical "parts" illus-
In the diagram on Fig. 21, both paths can be trated above - should naturally be rejected, or
followed, from perceptual data to their scien-at least provisionally ignored, when the goal of
tific interpretation in terms of a cognitive sys- analysis changes from the description of indi-
tem: either direct (1), or indirect, through anviduals to the discrimination of groups, which is
intermediate descriptive system (2-3). Admit-but another way of defining structural, or for
tedly, there is in fact no such thing as a strictlythat matter, scientific analysis at large.
direct or indirect strategy, in those terms; and
c) In particular, one of the drawbacks of a
the only purpose of our over-simplifying dia-
priori segmentation procedures is that they seem
gram is to suggest that there may exist cases
to imply some kind of kinship between identi-
when the construction of a systematic termi-
cal "parts" of different objects, for which, how-
nology, for the sake of description alone (path
ever, no historical evidence has been provided.
2), actually opens the way for further semio-
In other words, the description is based upon
logical refinements, culminating in the design
superficial analogies, rather than upon reasoned
of a truly scientific language (path 3). A sub-
homologies, and one therefore has no right to
stantiation of that viewpoint will be given in
use it otherwise than as a mediate system be-
the concluding paragraph (Section C).
tween natural language and truly scientific lan-
B. Let it first be emphasized that the possible guage, as indicated in Table 6.
existence of path 3 does not warrant the desig-
d) One of the possible interpretations of that
nation of descriptive terminologies or codes as
intermediate position is to think of a descrip-
"scientific," according to the second criterion
which was considered above. The reasons for tive code as a machine for the automatic (and
denying this qualification are many: if one knows no better, random) generation of
types, at least in the definition of types which
a) The symbolic representation of archaeo-
logical data in terms of such codes answers no is implied by what archaeologists do, not by
experimental criterion. It is not guided, for what they say, about typology. Notwithstand-
instance, by the knowledge of existing cultural ing the bulk of critical reflection on types and
distributions, or genetic filiations, or stylistic typologies, the more usual de facto basis of so-
affinities, etc.; nor does it aim at a systematic called types, in archaeological publications, is a
elimination of correlated features, which are classification scheme involving: 1) a largely a
certainly many, insofar as the analysis aims priori selection of discriminating features
at the multiplication of isolated features rather (which in turn implies tacit decisions as to
than at a reduction of redundancy. orientation, segmentation, and differentiation);
b) Conversely, however refined such an 2) a still more capricious ordering of such fea-
analysis may be, it does not necessarily contain tures - also implicit, many times - which de-
the elements which would be required to build termines the hierarchical organization into types,
up structural models, that is, symbolic para- subtypes, etc. Taxonomies of this kind - for

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
GARDIN ] DESCRIPTIVE ANALYSIS 29

pottery, tools, etc. - seldom reflect a systematic

associating given objects with one and the same
consideration of external data, other than groupthe
is their proximity, in terms of their sym-
more obvious (though often unreliable) bolic archae-
representation. Different functions have
ological origin (site, stratum, stade, etc.). To been proposed to measure proximity (or dis-
that extent, they are formally comparable to tance) (De la Vega and others 1965); what is
any of the milliard typologies which one can important for our present purpose is that a
immediately derive from a digital index based possible by-product of the calculus is a reduced
upon the methods which we have described. descriptive system, in which some units have
C. Many other arguments could be found to been eliminated, others fused into higher units,
show that descriptive methods do not lead to etc. Whether such models should be called
a scientific language, in the more restricted sensestructural is still open to question; but it is not
of the word. They would, however, all amount doubtful that they have then acquired some
to different formulations of one and the same cognitive value, insofar as they can help in
commonplace, namely, that one should not con- identifying origins of objects, forming hypotheses
fuse two kinds of descriptive units: the so-called as to cultural distributions, suggesting progress
"-etic" and "-emic" units of Pike (1954: 8-28),in descriptive systems, and so forth.
for instance, or any other phrasing of a long- Supposing now that automatic learning meth-
debated dichotomy. "-Etic" inventories can be ods are added to the process, with a proper
compiled in their own right, and need no justi-feedback from archaeological evidence; it will
fication; whether they are at all necessary, orthen become increasingly difficult to lay a
even useful, for the design of "-emic" systems is boundary between the initial denotative analy-
a debatable point. Some anthropologists tend sis and its successive structural outcomes as the
to oppose it, but with arguments which connote experience of automata grows. This, however,
the higher status of "-emic" models to them, is yet for a remote future, but it obviously has
rather than any theoretical objection to the some bearing on the present issue.
possibility of a continuum from an "-etic" analy-
Acknowledgments. I wish to thank E. A. Hammel
sis to the latter (Gardin 1965). Others, lin- for suggesting that this paper should be published as a
guists and logicians, have, on the contrary, un- shortened version of a manuscript which was first pre-
derlined that possibility, either as a practical or sented at the Symposium on Componential Analysis
as a wishful proposition (Hjelmslev 1961: 95-7; organized by him at the Center for Advanced Study in
the Behavioral Sciences, Palo Alto, 1964. Peter Pick was
Shaumian 1960: 127). The whole problem
also kind enough to make some useful editorial com-
then is to ascertain whether a given code elab- ments.
orated at the first level (denotative, "-etic"), is
sufficiently refined, if not strictly exhaustive, to
ALLARD, M., MAY ELZIERE, J. C. GARDIN, AND F. HOURS
justify its use in arrangements of the second 1963 Analyse conceptuelle du Caran sur cartes
perforees (Conceptual Analysis of the Coran on
order (structural, "-emic"). This is not a ques- Punched Cards). Mouton & Co., Paris-La Haye.
tion which we are in a position to answer for
any of the codes which we have outlined in thisCHENHALL, ROBERT G. (EDITOR)
paper. It must, however, be confessed that we 1965 Newsletter for Computer Archaeologists, Vol.
1, No. 1. Department of Anthropology, Arizona
hope to use some of them in a direction which State University, Tempe.
would seem to imply a positive answer, namely,
CHRISTOPHE, J. AND J. DESHAYES
for automatic classification. Two procedures
1964 Index de l'Outillage sur cartes perforees: outils
have been considered, both of which lead to de l'age du Bronze, des Balkans a l'Indus (Tool
some reorganization of the basic descriptive fea- Index on Punched Cards: Bronze Age Tools
from the Balkans to the Indus). Centre Na-
tures: a) in one case, the goal is to calculate, tional de la Recherche Scientifique, Paris.
within the description matrix, the minimum set
DE LA VEGA, W. F., S. REGNIER, AND MONIQUE RENAUD
of combined features which can account for a
1965 Techniques de la classification automatique,
known distribution of objects (on any empirical mimeographed report. Maison des Sciences de
basis: geographical, chronological, stylistic, etc); l'Homme (Centre de Calcul) et Centre National
de la Recherche Scientifique (Section d'Auto-
b) in the other case, the objective is conversely matique Documentaire), Paris.
the arrangement of the set into so-called natural
ELISSEEFF, V.
groups, when no cultural distribution is avail-
1965 Possibilites du scalogramme dans l'6tude des
able (Gardin 1963: 146-50). The criterion for bronzes chinois archaiques, Compte-rendus du

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms
30 AMERICAN ANTIQUITY [ VOL. 32, No. 1, 1967

Seminaire sur les Moddles Mathematiques dans KOVALEVSKAJA, V. B.

les Sciences Sociales, ann6e 1964-5. Ecole
1965 The use of statistical methods for the study
Pratique des Hautes Etudes, 6e Section, Paris. of large archaeological collections. Archae-
ology and the Natural Sciences (in Russian),
GARDIN, J. C. pp. 286-300. Nauka, Moscow.
1956 Le Fichier mecanographique de l'outillage. KRUG, G. K. AND 0. Ju
Institut francais d'Arch6ologie, Beyrouth.
1965 Mathematical methods for the classification of
1958 Four codes for the Description of Artifacts, an ancient pottery. Archaeology and the Natural
Essay in Archaeological Technique and Theory. Sciences (in Russian), pp. 318-25. Nauka, Mos-
American Anthropologist, Vol. 60, No. 2, pp. cow.

335-57. Menasha.
MARSHAK, B. I.
1963 Problemes d'analyse descriptive en archeologie.
1965 On the finding of criteria for establishing
Etudes arche'ologiques, edited by P. Courbin, pp. analogies and differences in pottery collections.
132-50. Service d'Edition et de Vente des Pub- Archaeology and the Natural Sciences (in Rus-
lications de I'Education Nationale, Paris. sian), pp. 308-17. Nauka, Moscow.

1965 Analyse documentaire et analyse structurale en PIKE, K. L.

archeologie. I'Arc, n?26 (on Claude Levi- 1954 Language in Relation to a Unified Theory of
Strauss), pp. 64-8. Aix-en-Provence. the Structure of Human Behavior, Pt. 1, Pre-
liminary Edition. Glendale.
HJELMSLEV, L.
SHAUMIAN, C. K.
1961 Prolegomena to a Theory of Language. Uni-
1960 Linguistic problems of cybernetics and struc-
versity of Wisconsin Press, Madison.
tural linguistics (in Russian), Voprosy Filosofii,
9. Moscow.
IHM, P.
1961 Classification automatique des objets de l'age TUGBY, DONALD J.
du bronze. Compte-rendus du Seminaire sur 1965 Archaeological Objectives and Statistical Meth-
les Modeles Mathematiques dans les Sciences ods: a Frontier in Archaeology. American An-
Sociales, annee 1960-1, pp. 28-33. Ecole tiquity, Vol. 31, No. 1, pp. 1-16. Salt Lake City.
Pratique des Hautes Etudes, 6e Section, Paris.
CENTRE D'ANALYSE DOCUMENTAIRE
KAMENETSKIJ, I. S. POUR L'ARCHEOLOGIE (C.N.R.S.)
1965 Datation of layers from percentages of pottery Marseille, France
types. Archaeology and the Natural Sciences
(in Russian), pp. 302-7. Nauka, Moscow. April, 1966

This content downloaded from 161.9.203.10 on Thu, 27 Sep 2018 15:14:13 UTC
All use subject to https://about.jstor.org/terms