3
LINGUISTICS 159 of morphophonemics are today more widely discussed than they were when first intro- duced. But Sherzer is perhaps wise in utili- zing data from “classical phonemics,” even though such data are no longer discussed in modern linguistics; to utilize morpho- phonemic formulations would have forced Sherzer to restate the available phonemic lists of yesterday in terms of today’s phono- logical theory-a difficult, if not impossible, task without further fieldwork. The contents for each of the dozen culture areas are discussed under three head- ings: Phonology, Morphology, and Observa- tions and Conclusion. Under the last of these three headings there is an absurdly inter- esting summary comparing the number of “traits” found in each area with the number of “traits” absent in that area (e.g., for the Arctic, 17 present versus 43 absent; half of the “traits” present are phonological, half morphological). But what is the basis of the present-absent list? Among absent sounds Sherzer does not list are ingressive vowels in Maidu, even though they may well be non-phonemic,” as we used to say in bygone days. But as sound features implo- sives do occur in North America. If a grammatical feature, as incorporation of noun after verb stem, is found widespread in such a family as Salish and also found in Tsimshian (a neighboring language which is in a different family), contact with Salish is said to be the source of diffusion to Tsimshi- an, without mentioning whether Tsimshian instances are borrowings of strictly linguistic forms, or whether the diffusion is of the kind that Kroeber called “stimulus diffu- sion”-i.e., for language, similarity, or com- parability in pattern combined with dis- similarity in form (for example, calques appear to be found in all languages). The phonological traits discussed for each area are typologized, but sometimes listed as single phonemes occurring throughout an area versus others which are entirely nonoc- curring in the area. Thus, the stop /q/ and the lklql opposition are found in some Western Subarctic languages, but /qW/ is not found at all in this area. The “traits” discussed under Morphology (without reference to syntax) are not ty- pologized at all! Is Sherzer to be censured for this? Not really, since his frame of reference is a subsystem typology that treats aspects of phonology, as in Hockett’s work and in mine (and in Pierce’s also, even though numbered “language types” are in- voked, but each “language type” is actually a subsystem type). In Europe, whole- language typologies of the last century ap- peared to threaten the more successful com- parative method. These types were, for the most part, morphological types. In this century Sapir restated them, stripped of cultural associations, as parallel to the com- parative method; and Greenberg, following Sapir , quantified them. Sherzer breathes not a word about this. What he does do for e,ach of his dozen areas is to state under Morphology what some- thing in the morphology marks or means, together with the means by which the grammatical feature is marked. Thus, the nominal case system in Eskimo is “marked by nominal suffixes.” Sometimes the means by which the grammatical feature is marked is given before the grammatical feature. Thus, the marker reduplication is given before what grammatical features it often marks (distributive, plural, diminutive) and followed by the statement “No Eskimo dialects’’-naturally not, since the gram- matical features of Eskimo words are all expressed by suffixes. But this rather incipi- ent hint at morphological type for Eskimo is not mentioned. What Sherzer does present is an areal- typological study of the way sound systems were typologized at midcentury, followed by a nontypologized treatment of morphol- ogy-only somewhat reminiscent of Edward Sapir’s Language (1921). What effect will all this have, i.e., this combination of phonological typology with grammatical features and the way they are marked in American Indian languages? Also, how do these linguistic aspects of the lan- guages combine with the generalized cultures in which they are spoken? Will the effect be to stimulate work to “unscramble” (in Driv- er’s sense) the elements of culture areas? Or will it be to stimulate research in relating sound systems to the rest of grammar, and thereby make possible a more useful whole- language typology with constituent subsys- tems? If the latter, will the subsystems only be relatable, or will different types of whole languages also be relatable to whole culture areas? Linguistic Subgrouping and Lexicostatistics. Isidore Dyen. Janua Linguarum Series Mi- nor, 175. The Hague: Mouton, 1975. 251 pp. Dfl. 64.00 (paper). Lyle Campbell State University of New York, Albany This book is a collection of Dyen’s writings on subgrouping and lexicostatistics, with a 27-page preface and Swadesh’s 200-word list as an appendix. This review

Linguistics: Linguistic Subgrouping and Lexicostatistics. Isidore Dyen

Embed Size (px)

Citation preview

Page 1: Linguistics: Linguistic Subgrouping and Lexicostatistics. Isidore Dyen

LINGUISTICS 159

of morphophonemics are today more widely discussed than they were when first intro- duced. But Sherzer is perhaps wise in utili- zing data from “classical phonemics,” even though such data are no longer discussed in modern linguistics; to utilize morpho- phonemic formulations would have forced Sherzer to restate the available phonemic lists of yesterday in terms of today’s phono- logical theory-a difficult, if not impossible, task without further fieldwork.

The contents for each of the dozen culture areas are discussed under three head- ings: Phonology, Morphology, and Observa- tions and Conclusion. Under the last of these three headings there is an absurdly inter- esting summary comparing the number of “traits” found in each area with the number of “traits” absent in that area (e.g., for the Arctic, 1 7 present versus 43 absent; half of the “traits” present are phonological, half morphological). But what is the basis of the present-absent list? Among absent sounds Sherzer does not list are ingressive vowels in Maidu, even though they may well be

non-phonemic,” as we used to say in bygone days. But as sound features implo- sives do occur in North America.

If a grammatical feature, as incorporation of noun after verb stem, is found widespread in such a family as Salish and also found in Tsimshian (a neighboring language which is in a different family), contact with Salish is said to be the source of diffusion t o Tsimshi- an, without mentioning whether Tsimshian instances are borrowings of strictly linguistic forms, o r whether the diffusion is of the kind that Kroeber called “stimulus diffu- sion”-i.e., for language, similarity, or com- parability in pattern combined with dis- similarity in form (for example, calques appear to be found in all languages).

The phonological traits discussed for each area are typologized, but sometimes listed a s single phonemes occurring throughout a n area versus others which are entirely nonoc- curring in the area. Thus, the stop /q / and the lklql opposition are found in some Western Subarctic languages, but /qW/ is not found at all in this area.

The “traits” discussed under Morphology (without reference to syntax) are not ty- pologized at all! Is Sherzer t o be censured for this? Not really, since his frame of reference is a subsystem typology that treats aspects of phonology, as in Hockett’s work and in mine (and in Pierce’s also, even though numbered “language types” are in- voked, but each “language type” is actually a subsystem type). In Europe, whole- language typologies of the last century ap- peared to threaten the more successful com-

parative method. These types were, for the most part, morphological types. In this century Sapir restated them, stripped of cultural associations, as parallel to the com- parative method; and Greenberg, following Sapir , quantified them.

Sherzer breathes not a word about this. What he does do for e,ach of his dozen areas is to state under Morphology what some- thing in the morphology marks or means, together with the means by which the grammatical feature is marked. Thus, the nominal case system in Eskimo is “marked by nominal suffixes.” Sometimes the means by which the grammatical feature is marked is given before the grammatical feature. Thus, the marker reduplication is given before what grammatical features it often marks (distributive, plural, diminutive) and followed by the statement “No Eskimo dialects’’-naturally not, since the gram- matical features of Eskimo words are all expressed by suffixes. But this rather incipi- ent hint a t morphological type for Eskimo is not mentioned.

What Sherzer does present is a n areal- typological study of the way sound systems were typologized at midcentury, followed by a nontypologized treatment of morphol- ogy-only somewhat reminiscent of Edward Sapir’s Language (1921).

What effect will all this have, i.e., this combination of phonological typology with grammatical features and the way they are marked in American Indian languages? Also, how d o these linguistic aspects of the lan- guages combine with the generalized cultures in which they are spoken? Will the effect be t o stimulate work to “unscramble” (in Driv- er’s sense) the elements of culture areas? Or will it be to stimulate research in relating sound systems t o the rest of grammar, and thereby make possible a more useful whole- language typology with constituent subsys- tems? If the latter, will the subsystems only be relatable, or will different types of whole languages also be relatable t o whole culture areas?

Linguistic Subgrouping and Lexicostatistics. Isidore Dyen. Janua Linguarum Series Mi- nor, 175. The Hague: Mouton, 1975. 251 pp. Dfl. 64.00 (paper).

Lyle Campbell State University of New York, Albany

This book is a collection of Dyen’s writings o n subgrouping and lexicostatistics, with a 27-page preface and Swadesh’s 200-word list as an appendix. This review

Page 2: Linguistics: Linguistic Subgrouping and Lexicostatistics. Isidore Dyen

160 AMERICAN ANTHROPOLOGIST [ S O , 19781

will consider migratory theory, lexicosta- tistics, and subgrouping (the book’s major themes), but not its considerable content dealing with Austronesian (Malayo-Poly- nesian) languages.

Dyen’s formal statement of migratory theory based on maximum divergence and minimal moves of the languages and sub- groups of linguistic families is undeniably an important contribution to the repertory of linguistic techniques for getting at prehis- tory. Although Dyen indicated when the article was written (1956) that “instances in recorded history should provide the proper testing of the formulae” (p. 50), migration theory has not really been tested. Many sources of information can converge to help confirm or disconfirm its inferences, i.e., the ecological and geographical aspects of the reconstructed vocabulary of a protolan- guage, loan words, correlations with ethno- history and archaeology, etc. Such conver- gence makes inferences about homelands and migrations much stronger than those based o n migration theory alone.

A n o t h e r significant contribution is Dyen’s forceful reiteration of exclusively shared innovations as the strongest criterion for linguistic subgrouping. The criterion had largely been neglected after its early neo- grammarian formulation, and Dyen called it again to the attention of modern linguists.

In the evaluation of the rest of the book two questions must be faced: (1) Is lexi- costatistics valid? ( 2 ) Is it useful for sub- grouping?

Lexicostatistics is a t best highly contro- versial, with perhaps the bulk of current opinion on the negative side. A reiteration of the strong challenges leveled against all its basic assumptions, however, would hardly be helpful, since Dyen seems aware of these, confronting criticism with optimism for re- solving the difficulties expressed in the critiques. For example, it has been asserted that the rate of retention is not constant through time or cross-linguistically, based on both very conservative languages (Icelandic, Armenian, Georgian) and high-loss languages (English). Dyen confronts these apparent counterexamples optimistically :

There are n o satisfactory explanations thus far for a significantly higher than normal rate of retention. Since this par- ticular variation is limited t o three in- stances, it appears a t least provisionally the better hypothesis that they exemplify the operation of a n unknown factor o r . . . factors whose nature is perhaps to be identified by studying each case thor- oughly [p. 211.

Lexicostatistics is often criticized when similar but noncognate forms are taken to be cognates. A less frequent, but no less telling, criticism is that in its application one may fail to recognize true cognates that have undergone far-reaching sound changes.

One can also criticize lexicostatistics for its erroneous results. To mention just one example, Swadesh, the founder of lexi- costatistics, found Uspantec (in the Mayan family) to be more closely related to Mam (13mc) than any other language of the Mamean subgroup. Nevertheless, it is uni- versally recognized that Uspantec belongs to the Quichean subgroup and bears no close relationship t o Mamean, and this is obvious by any other method.

Lexicostatistic practices can also be criti- cized. For example, lexicostatistics does find or prove distant genetic relationships be- tween languages because it assumes the relationship and attaches a date to the separation.

Is lexicostatistics valid for subgrouping? I believe we can agree with Dyen that “regard- less of its weaknesses the list procedure would provide a useful first approximation to such a subgrouping” (of the more than 500 Austronesian languages) (p. 16). But in conceding the potential usefulness in pro- viding a first approximation (especially in families with a large number of languages), I d o not concede that lexicostatistics is valid for subgrouping; the criterion of shared innovations is the only one valid for sub- grouping. Dyen is inconsistent, but seems to concede this also: “Lexicostatistics offers a perhaps less reliable, but a quicker procedure for determining these interrelationships. The fact that we use a less reliable method need not prejudice the later use of more reliable determinants” (p. 159). But by casting aspersions on the traditional method which employs the criterion of shared innovations, he seems to argue that lexicostatistics may actually provide a more reliable method of subgrouping. He says that that method is difficult to apply (not a real criticism) and that the agency of individual scholars and incomplete data may produce varying results (p. 22). On the other hand, he takes lexi- costatistical results t o be “justifiable to the extent that they can be reproduced inde- pendently with insignificant variation. . . . In a large number of cases different compara- tivists have independently reached lexicosta- tistical percentages which differ by amounts that appear not to be important” (p. 178). This is often not true, since different schol- ars make different judgments in their appli- cation of the method, obtaining quite differ-

Page 3: Linguistics: Linguistic Subgrouping and Lexicostatistics. Isidore Dyen

LINGUISTICS 161

ent results, particularly in judgments of cognacy.

If lexicostatistics is invalid generally and frequently produces erroneous results, then it cannot be valid for subgrouping. The concession that i t may be useful as a first approximation is not inconsistent with this conclusion Some linguistic relationships are so obvious that they show u p in any method, including just gross scanning. Lexi- costatistics may provide useful hypotheses to be tested by the standard method, but such hypotheses are worthy of little faith until they have been so tested.

To conclude, I personally am the kind of disbeliever (see Dyen’s characterization of disbelievers [p. 1761) who finds the chal- lenges to lexicostatistics’ basic assumptions compelling; who finds it frequently to pro- duce varying, erroneous, and misleading results; and who finds it abused in practice. Dyen’s book contains some important con- tributions and much useful information o n lexicostatistics. I t will not, however, sway the skeptics to embrace lexicostatistics nor use it for subgrouping.

Defining a Linguistic Area: South Asia. Colin P. Masica. Chicago: University of Chicago Press, 1976. xiii + 234 pp. $16.00 (cloth).

Joel Sherzer University of Texas

There are two types of linguistic classifi- cation which are useful in the reconstruction and interpretation of culture history-the genetic and the areal-typological-diffusional. Most classificatory research by linguists, whether or not they are anthropologically oriented, has been of the genetic type. This is unfortunate since each of these types of classification provides different sorts of in- formation, complementarily crucial in his- torical interpretation. Genetic classification reflects common identity and ancestry of disparate groups of people, original home- land and subsequent migrations, and aspects of earlier habitat and culture. Areal classifi- cation reflects contacts among groups of people and the nature of this contact. A valid linguistic history must reconstruct the interplay of the genetic and diffusional through time.

The value of Masica’s book to historical linguistics is that it is a precise and careful study of a single linguistic area, South Asia, in terms of language contact and diffusion. It is an experiment in theory and method and is thus of scholarly interest beyond the

South Asian limits. There are a number of issues that must be dealt with in any areal-typological research:

(1) What constitutes a linguistic area? One can focus on either a single trait o r a complex of traits. Thus i t is possible to talk of a retroflex stop area, i.e., a n area in which all languages have retroflex stops. Or can one delimit geographic areas in terms of clusterings of linguistic traits which cut across genetic boundaries. In this case a crucial question is what con- stitutes a significant clustering of traits, significant enough to postulate a lin- guistic area. ( 2 ) What kinds of traits should be map- ped in a n areal linguistic study? Are certain traits more diffusable and more likely to provide interesting results? Masica chooses t o map relatively few linguistic traits (see below), but investi- gates each of these in great detail. He argues that it is relatively superficial traits (surface expressions of deeper categories) that should be studied. In a study of North American Indian languages [re- viewed in this issue-Ed.], I investigate a much larger number of traits, in less detail for each. My results agree with Masica’s in showing that it is relatively marked linguistic traits which show the most interesting areal distributions. ( 3 ) How can one distinguish shared traits due t o genetic relationship, from those due to diffusional relationship, from those due t o chance? Again more areal research, as well as attention to such matters as markedness, is needed here. (4) I t is not mere contact in space that causes the diffusion of linguistic traits, but rather the existence of certain com- municative conditions-density of popula- tion, type of social organization, degree of bilingualism, trade, intermarriage, etc. While there is no doubt that it is the existence of such factors which enables the diffusion of linguistic traits, the inter- pretation process is a challenge for future research; i.e., given the linguistic evi- dence-linguistic areas and areal distribu- tions of t r a i t s h o w can one infer the nature, the degree, and the direction of sociocultural influences?

Masica provides a particularly useful dis- cussion of the linguistic issues involved in areal-typological studies. He points out the difficulties involved in interpreting data from grammars written from distinct theo-