Phonology in the twentieth century: Theories of rules and theories of representations 0226019152, 9780226019154

Citation preview

Phonology in the Twentieth Century

STEPHEN

R.

ANDERSON

Phonology in the Twentieth Century Theories of Rules and Theories of Representations %URXJKWWR\RXE\ /LQJXLVWLFV/LEUDU\

The University of Chicago Press Chicago and London

Stephen R. Anderson is professor of linguistics at the University is the author of The Organization of plwnology (1974)· The University of Chicago Press, Chicago 6063i The University of Chicago Press, Ltd., London © I~85 by The University of Chicago

AI.I flgh~s reserved. Published 1985 Printed In the United States of America

94 93 92 91 90 89 88 87 86 85 5 4 3 2

1

Library of Congress Cataloging . in Publication Data Anderson, Stephen R. Phonology in the twentieth century. Bibliography: p. Includes index. C omparanve . . I. Grammar _, History-s-zoth centu I

ry.

P217·3·A53 1985 0-226-01915-2

ISBN

and. . general-Phonology-

. TItle 414 8' 2 5- 773

of California,

Los Angeles. .

He

For Janine

CONTENTS

ACKNOWLEDGMENTS INTRODUCTORY REMARKS

Why study the history of twentieth-century phonology? Motivations for the present book The historical origins of modern views: A concrete example The organization of the present book 1 FERDINAND

DE SAUSSURE

The Saussurean view of language, languages, and linguistics The linguistic sign The relation of languages to their history 2 SAUSSURE'S

VIEWS ON SOUND STRUCTURE

Sounds, sound images, and their study 'Phonemes' and 'phonetic species' The linguistic representation of signifiants Some approaches to the study of phonological differences Saussure's description of alternations Saussure and the phonological tradition

1 1 7 10 14 17 23 26 28

33 34 37 40 43 50 55 i

3

THE KAZAN

SCHOOL: BAUDOUIN

DE COURTENAY

AND KRUSZEWSKI

Biographical remarks The study of sound systems in the Kazan school The nature of phonological structure Kruszewski's theory of alternations Baudouin's development of the theory of alternations The later history of 'Kazan phonology'

56 57 61 65 68

73 79

•

viii / 4

CONn~NTS

'G Y FROM l:H:E Moscow CIRCLE PRAGUE SCHOOL PHONOL?, srvr.i .. , T UBETZKOY S GRUNDZUGE TI1ROVGH R d f th '. Prag', ue Circle and the life of Trubetzkoy

The backgroun 0 e ". Units in phon

DE. SAlJSSlJRF.

seem valid in the study of langlla~e,. and one would ex~e(;t that the prority of strictly synchronic studies in linguistics would be established by merely point~ ing this out, In arguing for the centrality of synchronic considerations, however, SausSl:lre was challenging the central doctrine of the then-current neogrammarian view of explanation in linguistics: that historical study was not only important but, indeed, the only genuinely 'scientific' approach to the facts of language. Interestingly enough, the temptation of this view can be seen as being based On an essentially Saussureau insight: the arbitrariness of the linguistic sign. If the signs of a particular language are indeed completely arbitrary, then their present reality can have no possible present explanation. If we thus seek an explanation of the way things are, the best we can do is to show how they got to be that way: to establish such antecedent stages as we can, and a chain of sound "Iaws" relating them to one another and to the forms presently in use. This was the view of scholars such as Hermann Paul and Karl Brugmann, and the spectacular success of the neogramrnarian methods in the study of IndoEuropean resulted in its overwhelming acceptance at the time. For Saussure, however, such, a theory was completely unsatisfactory as an explanatory account of the nature of language (or of particular languages), An obvious objection to the historical view is that it simply pushes the problem back: if we account for a present stage in terms of an orderly series of changes undergone by an earlier system, we are still left with no account at all of that earlier system itself. Where did it come from? The chicken-and-egg aspect of this problem is self-evident, but we might regard the difficulty as even more pernicious than that. This is because, in seeking antecedent stages from which to derive a present etat de langue, we continually push the problem back into reconstructed systems which cannot even be observed (except inferentially, through the testimony of their modern reflexes), A second, and even more fundamental difficulty with the historical notion of explanation was that, for Saussure, it completely falsifies the object of study. As we saw above, Saussure saw the locus of historical change (and thus the domain of operation of sound laws) as exclusively in parole. If we look to historical chan~e for an explanation of a synchronic state, however, we are thereby attempting to reduce the facts of langue to facts of par I hi h ' ·11 .. . h . 0 e , w IC IS 11 tota y. 1, egrnmate, given t e basic conceptual distinctness. 0,f th,e two. ThIS IS not to say, of course, that Wecannot study lingui ti hanz , HT . " • . IS IC c' ange systematically. we can recognize that diachronicallv related t. ' .. represent distinct hats de langue which a' .. s heles of a given language lated, and thus that change does affie tIre nonet eless systematically re, . . c angue over ti . A however, the link between such relat d . ,l~e. s we saw above, of the domain of the study f l. ~ systems IS, stnctly speaking outside . . . 0 angue Itself. Th'h . , ' .tendenCies which affect acts of I . roug the operatIon of phonetic ferent primary linguistic experie~~~~ e, a Subsequent generation (having difo go on) may well' du. . . m uce a different sys-

RELATION

or

LANGUAGES

TO THEIR HISTORY

I 31

tern o~ tangue ~n th~ ~asis of the observations of parole which they makerc~ulttng 111 a linguistic cha~ge apparently affecting langue. The important noint to. make .here, ..though, IS that the motivation I'for the c hange IS. never 111 . r. lallg~e Itself. The st~dy of c~ange is entirely dependent on the prior understandlllg of synchronic states 111 themselves, together with facts from a discipline exter~al t~ the stu~y of langue (i.e., phonetics, a branch of the study of parole). Historical studies thus can never yield an explanatory theory of the nature of langue. One might ,:ell suggest that, while Saussure's view seems cogent with regard to phonet~c change, the phenomenon of analogy surely represents a type of change motivated by the system of langue directly, and thus belongs to the study of langue. Saussure anticipated this objection, however, and provided an answer. According to his conception of its nature, analogy constitutes an aspect of langue. all right, but not change in langue, because analogy is claimed not to constitute a change at all. Rather, when we create an apparently novel analogical form, we are doing so (by the definition of .analogy) by applying some rule of the system of langue: a rule which already existed prior to its application. We are thus simply realizing a latent possibility of the system, rather than effecting a change in it. Although it will be seen that this view of analogy saves Saussure's claim that the study of historical change is never a proper part of the study of langue, it is apparent that if it is carried to its logical conclusion, it results in such a broad a notion of the rules of the system that it is probably not satisfactory. Nonetheless, since Saussure gave very little attention to the problem of how to formulate the rules of a synchronic system, this consequence did not arise for him. The central point of the Saussurean notion of linguistics, then, is that there is nothing that historical investigation can contribute to the study of synchronic linguistics, and it is this synchronic study alone that can yield explanatory answers to the central question of the field: the nature of a language (and, in general, the nature of langue). It might seem that in order to establish the priority of synchrony in the study of language, it would suffice to point out the considerations just discussed; it is therefore a bit hard to see why so much space is devoted in the Cours and in Saussure's notes to repeatedly exorcising the spirit of a historical approach from ling.uisti~s. . . A consideration of the predominance of such hlstonca.lIy ?flented views ~t the time, however, quickly shows us why so much attention IS ~evoted to this issue. When we recall the extent to which, at the end of the nineteenth century, historical linguistics was considered to provide a genuine e~planation (and, indeed, the only scientifically valid one) for the facts of ~ar~lCularla?guages, we can see why Saussure felt compelled to retu.rn to. this l~s~e agam • and again, in every conceivable context. It is quite posslb.le, m.add~tlOn,th~t the purely personal factor of his own somewhat tense relatIonship with pronu-

321

FE~DIN"N()

DE SAllSSVRE

nent figures of the ncogrammarian moverne~t during h~s student days in Leipzig had (at least subconsciously) something to do with the fervor with which he pursued this end. . . " Whatever the reasons, Saussure's work IS most assiduous In ehminating from his formulations anything that bears the slightest resemblance to a his. torical approach to central linguistic problems. He probably realized that this attitude was in some ways slightly exaggerated, but felt that in the then, current climate of opinion there was "no danger in insisting above all on the nonhistorical side" (quoted in Godel 1951: 45). This was no doubt true, and perhaps any less vigorous defense of the priority of synchronic considerations would have been less effective in reorienting the field toward its central problem. Nonetheless, Saussure's categorical rejection of anything with even an appearance of a historical basis had profound consequences for the delimitation of problems and possible solutions in the study of sound systems: consequences which Saussure might or might not have accepted, but which go well beyond the scope of his fundamental objections. We will see some of these in the next chapter.

2 Satlssurc's Views on Sound Structure

There is a decided Jack of concrete evidence to be found in Saussures own work for the way one. ought to apply his general views to the specific problem of the sound structure of language. Although the nature of synchronic linguistic syst~ms occupied his attention during most of his teaching career (at least after his return to Geneva), he only discussed such questions in his very last classes. Furthermore, he produced essentially no actual descriptions of individual languages, and we are thus deprived of most possible sources of evidence for his views. The Cours itself is largely devoted to the general semiological problem of the nature of the linguistic sign, and says little that is very specific about the character of sound systems. There is, however, one important source of evidence in this work: the appendix to the introduction, which deals with (what we would now call) phonetics. This appendix represents a curious interpolation in the rest of the text, since unlike the remainder of the book, it is based not on Saussure's lectures on general linguistics but on a series of lectures given in 1897 on the theory of the syllable. Bally's own notes on these lectures, together with student notes on a similar presentation at the beginning of the first series of lectures on general linguistics in 1906, form the basis of the text. It might appear that if this material is our primary source for Saussure '8 picture of sound structure, its comparatively early date compromises its relevance to his later views on general linguistics. The reappearance of essentially the same presentation in his later lectures, however, suggests that his ideas on these topics remained relatively stable. In addition to this appendix, it is also possible to cite some few notes of Saussure's own (GOOel 1954), and some additional notes (evidently for a book on the subject) in a manuscript in the Harvard library that has been studied by J~kobs()n (I9~O) .. Finally, some.conclusions can be drawn (especially concermng Saussure S VIewof altern~tlons) from his courses inI909- 10 on Greek and Latin phonology (see. ReichlerBeguclin 1980) and morphology (see Godel 1957)· The overall pic..ur~ that emerges is a coherent one, and does not suggest that the 1897 matenal In the Cours was unrepresentative of Saussure's later views. . .' . The reason for dwelling on these purely textual Issues. IS that, while Saussure's name conveys a sense of almost ultimate authority (at least for

34 /

I SAUSSlJRE'S VIEWS ON SOUND STRUCTURE

some), finding out what his actual opinions were on concrete issues is ofr . . . ., en all too much like the mterpretation of the ancient oracles. A very sparse and Ii ired text, full of suggestion but lacking in specifics, allows each interpreter~' find what he wants, and thus to legitimize his own picture of the probJem. N° doubt the presentation in this chapter does not avoid such traps either, bu~ there seem to be some fairly clear points in Saussure's presentation of phonetic problems, and we will attempt to stay close to these.

Sounds, sound images, and their study We can recall that Saussure took linguistics to be the study of a certain class

of signs, and that the signs in question have the character of uniting a (signified) concept with a (signifying) sound image. Most interpreters of Saussure have attempted to downplay the linguistic relevance of the sound images, but it seems to me that to ignore the question of what specific character they have is in a way to miss the point of Saussure's conception of language. It is common; for instance, to cite the Saussurian doctrine that "dans la langue il n'y a que des differences ... sans termes positifs" as evidence for the view that the particular elements composing the sound system are not legitimately the object of linguistic study. But to say that the linguist must be primarily concerned with differences between sounds is not by any means to reject entirely the study of the sounds themselves. While the linguist's main interest is in the system of oppositions between signs, these oppositions rest on the differences among sound images, and these differences themselves reside in the character of the sounds that are differentiated. Saussure thus stresses that the study of the formation and positive, physical character of sounds (the content of traditional phonetics) is not in itself a linguistic study: it is only when we consider the relations between sound images that We are studying the system of language. But his insistence on the sound image as one of the two inseparable' faces of the sign makes it clear that insofar as their nature supports their differential function, these sound images are indeed an aspect of the object studied in linguistics. We can perhaps make this issue more concrete by posing it in terms of our usual conception of a gra~~ar today. Within such a grammar, we can identify two aspects of the description of the sound system of a language. First the grammar provides a ..set o~ representations for linguistic forms , in the forr rmo f a system 0 f transcription . together with the principles for its In · terpr:e tation. " '. Such a system of transcription is generaJIy taken to be fu dam 11' d - f anv narn . . . J' n· amenta y In ed pen ent 0 any particular language and its definiti ' . " . . applicabJe terms based 0 hr .'. '. . . '. on IS given In universally of an . di ld I I n uman lIngUIstIccapaCItIes (rather than on the facts . In 1VI ua anguage). ' Second, however, a grammar . id ticular to an individual langua pro~~ ~s a system. of rules, or principles partations as (potentially) belon i~e, w ~c,.chara.ctenze some of these represeng g to different signs, and others as (potentially)

SOUNI>S,

SOUN[)

IMAGES, AND THEIR STUDY

I 35

belonging to the same sign in Saussurean terms. 'Redundancy rules' , for example, specify that if the representation corresponding to a given sign has some property p, it must also have (or, in some cases, may not have) some other property P'. Such rules specify the range of permissible variation in the realizations of a given sign, and thus (by implication) the characteristics that necessarily differentiate distinct signs. There are, of course, other sorts of regularity than those expressed in redundancy rules, and these are described by other sorts of rule. The general point should be dear: the rules of language (as opposed to the transcriptional system employed to represent forms in the language) are particular to that language, and, taken together, they characterize the system by which sound differences correspond to oppositions between SIgns. Saussure's point, formulated in these terms, is clear: it is the business of the linguist to study not the nature of (phonetic) representations but the system of rules which underlies the differentiation of signs and thus constitutes a particular language. Seen in this light, however, the sound system itself is anything but irrelevant to the task of the linguist. Indeed, it is only on the basis of an understanding of the nature of sound images that the task of formulating the rules making up any particular system of signs can even be approached. We must arrive at a proper conception of these sound images in order to have an appropriate basis for the study of the system. For one thing, as we will note below, they are identified less with an articulatory characterization of utterances than with a somewhat more abstract and 'timeless' perceptual one. Yet they constitute the elementary units whose differentiation is the basis of the linguistic system. It is sometimes suggested, nonetheless, that Saussure felt signs to be such abstract entities that the connection between the signifiant of a sign and a sound image per se is a completely accidental and contingent fact, unconnected with the nature of language. Indeed, it is insisted in the Cours that the material sound itself does not belong to langue but is simply the substance which supports linguistic expression ("phonation ... is only the execution of the sound images"-cited by Godel 1957: 82), and thus is a matter of parole. What is at issue here is the irrelevant or accidental character not of sound images as is sometimes suggested, but of sounds. These latter, for Saussure, are particular physical, articulatory implementations of linguistic possibilities, and thus belong to the study of parole. Sound images, on the other hand, have a timeless character as perceptual archetypes (1916: 98); and while these serve as the basis of concrete acts of production or perception, they are not to be identified with them. These sound images, as essential (though not independent) components of the linguistic sign, are thus not excluded from langue.

To this interpretation it might be objected that Saussure explicitly says that phonetic implementation of sounds is not necessary, since the signs can be evoked by other means. It is, however, revealing to note the example Saussure

36 I SAlJSSURE'S

Vinws ON

SOUND STRUCTURE

~ses to make this point: to wit, the possibility of transposing Jinguistic . Into writing. On the face of it, the expression of signs In writing has no:~Tns ~hatsoever to do with sound images, since it 'involves a compJetely differe~g visual rather than auditory medium. t, If we look at the context of this example in the notes on which the Cours is based, however (see Godel 1957: 193-94), the issue appears in a somewhat different light. In fact, for Saussure, writing bears a more or less direct relation (depending on the particular system) to the system of sound images, taken in its essential (rather than external, articulatory) character. The segmentation imposed by alphabetic writing systems, he feels, corresponds to a fundamental property of sound images. This is a point he makes explicitly in the Cours with regard to the Greek writing system: its segmentation reflects the parallel segmentation of sounds in perception, which is imposed as a part of the essential character of hUman speech perception. Alphabetic writing thus provides, a sometimes imperfect, but largely accurate, representation of sound images, just as the articulatory formation of concrete, sounds does. The relevance of the example of signs realized in writing, then, is not that sound images are inessential to langue but that (physical) sounds are. We must conclude, then, that an appreciation of Saussure's views on the system of language must be founded, in part, on his conception of the nature of sound images. In the study of sound images, Saussure distinguishes essentially three approaches which can be seen as characterizing three distinct fields. To a considerable extent, these divisions correspond to those of later linguists, but although he often uses the same terms as those appearing in later work, he USes them in radically different ways. For the modern reader, the Saussurean terminology thus requires at least a note of clarification. The nature of the linguistic sign (especially its linear character) and its realization in syntagmatic combinations leads directly to the study of morphology, in approximately the sense of subsequent linguistic theory. The spoken chain can be divided into discrete signs, and morphology is the study of the principles underlying such division. In various places, it is made clear that this segmentation of the chain is based on (synchronic) proportional analogies, which establish the relations between morphologically related words. As noted in chapter I, Saussure is reluctant to attribute illdependent existence to the subparts of words isolated in this way, preferring to concentrate on the relations which underlie the divisions. For our purposes here, it is sufficient to note that divisions corresponding to different signs at least the size of the word are as~u.medt~ exi~t and to be real to the speakers of the language . . Indlvldual.sJg?s 10 the spoken chain can be studied in terms of the rnechanl~ms and ,pnncl~]es by which their sound images are realized in speech, but this study IS, by Its nature, part of the linguistics of parole rather than of langue. Saussure caJJs this syn h '. . d f . "'" . c_rome stu y 0 the artIculation and acoustics of concrete sounds phonologie· it . , . II . ". IS essenna y the same as what most linguists

'PIIONflMLS' ANn 'PIION,mc specIEs' I

37

would today call phonetics: 1~1the discussion below, we will use the modern [t'rlllino)ogy except ~hcn~ It JS essential to call attention to Saussure's usage (in which case we, w~lI rC,'er to this discipline as phonvlogie), Saussure al~o dl~tmgu~S!les a discipline which he calls phonetique, but he uses that word to ~lllte a dlflerent sense than we do today, Saussure's phonetique is not a synchronic study at all; rather, it is the study of the historical evolution and chan~e o~ ~ounds. ~ike his phonologie, it is an aspect of the study of parole. Sll~ceIt IS ess~nhally based on the mechanisms by which speakers realize the signs of their language in concrete acts of speaking. Saussure had considerable faith (as had the neogrammarians) that the detailed study of the facts of speech would yield a comprehensive explanation of the mechanisms of sound change. To conform more closely to modern usage, we will refer to this study below as historical phonetics (except where Saussure's terminology itself is in question). Saussure's usage of phonologie and phonetique is somewhat confusing to the modern reader, since essentially no one other than his student M. Grammont (from his years in Paris) followed his terminology. Nonetheless, both terms correspond to well-established aspects of the study of speech. Neither one provides us with a name for the study of sound images considered as a part of langue, however. The sound images that form one aspect of the linguistic, sign differ from concrete sounds in essential ways (they are timeless rather than being realized in time, they are neutral between production an? pe:ception, etc.) and thus are not directly accessible to either phonetic or historical phonetic study. In fact, there is no reason to believe Saussure had any w?rd f?r,the, study of the role of sounds in langue: this is simply an aspect of linguistics. Indeed, since he emphasizes many times in his lectures that the ,stu~y of the I, . tic sign must be based on the simultaneous study of the, slgnifia~t a~d mguis ichare h.., 1ifie the pedagogical concerns which are so so evid evi ent everyw here ere ir 10 IS the slgm e", ' 'hir " oid presentation of fundamental problems would p,robablyha~e led urn ~o avoi ' ld sug-gest separatIon of one face of the sign from any term, w h'IC h WOU ' ' , an illicit . the other.

'Phonemes' and 'phonetic species' ,' d images let us contrast them with the understan,d the nat.ure of .soun 'not ~cifical1y linguistic) methods of object of study in phonetics. Using the ( . Pfund in speech. These units hvsi . .. , n study the umts 0 so P, ysical mvesngation, we. ca 'd-' speer (called 'acoustic' by an au, rtory a have an articulatory SI'd'ie, an' d also " . ',h. "