Review of Language in the Americas, by Joseph Greenberg

26
Linguistic Society of America Review: [untitled] Author(s): Lyle Campbell Reviewed work(s): Language in the Americas by Joseph H. Greenberg Source: Language, Vol. 64, No. 3 (Sep., 1988), pp. 591-615 Published by: Linguistic Society of America Stable URL: http://www.jstor.org/stable/414535 Accessed: 24/02/2010 19:54 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=lsa. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. 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]. Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access to Language. http://www.jstor.org

Transcript of Review of Language in the Americas, by Joseph Greenberg

Linguistic Society of America

Review: [untitled]Author(s): Lyle CampbellReviewed work(s):

Language in the Americas by Joseph H. GreenbergSource: Language, Vol. 64, No. 3 (Sep., 1988), pp. 591-615Published by: Linguistic Society of AmericaStable URL: http://www.jstor.org/stable/414535Accessed: 24/02/2010 19:54

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unlessyou have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and youmay use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/action/showPublisher?publisherCode=lsa.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

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

Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access to Language.

http://www.jstor.org

REVIEW ARTICLE

Language in the Americas. By JOSEPH H. GREENBERG. Stanford, California: Stanford University Press, 1987. Pp. x, 438.

Reviewed by LYLE CAMPBELL, SUNY Albany* I begin by anticipating the conclusion: Greenberg's Language in the Amer-

icas (henceforth LIA) has a detrimental impact on the field; its classification should not be accepted; the record should be set straight. This is a strongly negative evaluation, but necessary. G's well-deserved eminence draws un- warranted attention to LIA, as evidenced by frequent positive reports of it in the popular press. Given this, it will be difficult for any review, no matter how well reasoned, to balance the scales towards a more defensible classification of American Indian languages and towards more realistic methods of investi- gating distant genetic relationship. Precisely this, however, is my intent. I con- centrate on three aspects of LIA: the background claims, its methods, and its substance, together with considerations of distant genetic methods in general.

LIA's main thesis is that 'all the indigenous languages of the Americas, except those of the Na-Dene and Eskimo-Aleut groups, fall into a single vast assem- blage' (38). G claims to demonstrate the genetic unity of this group, called 'Amerind', and this is presented as a major breakthrough in the classification of New World languages.

BACKGROUND

1. It will be helpful to begin with a clarification of claims which attempt to position LIA with respect to the general field of American Indian linguistics.

1.1. Is IT NEWS? G's classification of American Indian languages is NOT new; it has been known in essence since 1956 (cf. Greenberg 1960, 1979, Greenberg et al. 1985, 1986), though proposed supporting evidence is published here for the first time. Nevertheless, unlike his African classification (of which the LIA reader is frequently reminded), this classification has not 'won general accep- tance' and has not 'become the basis for a considerable body of comparative work on [American] languages' (1)-it has not stimulated any significant sub- sequent research.

Several Americanists, in particular the participants in Campbell & Mithun, (1979b, henceforth LNA), are portrayed as representatives of a recalcitrant establishment which will receive G's work 'with something akin to outrage' (43; see also Ruhlen 1987:219, where LNA is characterized as 'linguistic cynic- ism'). This is puzzling, since the history of American Indian linguistic study is characterized largely by 'lumping', attempts to reduce the great linguistic diversity to an ever smaller number of more far-flung genetic groupings. It is squarely within this 'lumping' tradition that G's own work fits, together with

* I would like to thank William Bright, Adolfo Constenla, Ives Goddard, Terrence Kaufman, and Paul Newman for helpful and thoughtful comments on versions of this paper; I hasten, however, to absolve them of any unwanted association with the claims I make herein.

591

LANGUAGE, VOLUME 64, NUMBER 3 (1988)

such other 'super-groupers' as Sapir, Radin, Whorf, and Swadesh (and many others listed in the bibliography without further comment here; cf. Bright 1984:15, Campbell & Mithun 1979a). Quite in contrast to the 'establishment' characterization given it, LNA represents the iconoclasts and calls for a break with the lumper tradition and sober assessment of the evidence in cases of remote genetic proposals.1

1,2. OUT OF AFRICA. G emphasizes the fact that his classification of African languages (cf. Greenberg 1948, 1955[1949-54], and particularly 1963[1966]) was originally disputed by Africanists, but subsequently 'has won general accep- tance' (1). The claim is that Americanists may reject G's LIA classification, but they, too, will be shown wrong and G's view will prevail, just as in Africa (cf. Ruhlen 1987, J. D. Sapir 1987).

There are good reasons, however, why this position should be rejected. First, the African languages that Greenberg grouped together are simply more de- monstrably related than American languages are. Furthermore, received opin- ion concerning African classification at the time was particularly bad, based on misguided and now discredited racial principles, on cultural traits, and on shared typological features. In contrast, much progress has been made in un- derstanding the linguistic diversity of the Americas, and this research is based on valid historical linguistic principles. Moreover, portions of G's African clas- sification have not been demonstrated and are still in dispute.

Most obviously, success elsewhere does not guarantee success in a new area. Sapir's work, which underlies G's to a large extent, is a good example. Sapir's

1 A number of Americanists, particularly those of LNA, have been characterized as being un- willing to entertain any proposal of distant genetic relationship (43; cf. Ruhlen 1987, Darnell 1987, Fleming 1987:24). This is, however, false. Several proposals of remoter relationships were sup- ported by LNA authors (cf. Campbell & Mithun 1979a:38). While not all of these proposals are of equal strength, the attitude, contrary to that attributed to LNA, has not been one of resistance to any 'lumping' proposal in principle; rather, such scholars emphasize the need for reasonable sup- porting evidence.

An 'attitude problem' might better be charged to G than to the LNA authors. LIA shows nearly total disregard for most of the work by American Indian specialists over the last twenty years or so. That is, a perusal of LIA's references reveals few from this period; the only citations from the 1980's are to G himself, or to nonlinguistic works, save one review which mentions G favorably. Most references cited from the 1970's are not about American Indian languages, but rather concern the philosophy of science, African or other non-American languages, or other anthropological themes. The articles of LNA are listed, but it is clear from LIA and the Greenberg notebooks that these were largely ignored, or utilized only selectively. Of the few American Indian linguistic works cited from the 1970's (fewer than ten), most treat South America. Even in the few that are employed, distortions appear. For example, G draws from Campbell 1977 in both LIA and his notebooks for 'Proto-Mayan' forms, but these forms are actually from Quichean, a Mayan subgroup which has undergone many changes. Here is a list of just a few of the major studies that are neglected by LIA: Berman 1983, Bright 1976 and 1984, Campbell 1973, 1975, 1976, 1978a, 1978b and 1980, Campbell & Oltrogge 1980, Campbell & Kaufman 1981 and 1983, Constenla 1981, Golla 1984, Justeson et al. 1985, Kaufman 1973, 1974a and 1974b, Klein & Stark 1985, Sherzer 1976, Shipley 1980, Sorensen 1973, Suarez 1974, 1975, 1979 and 1983a, Whistler 1977. The many omissions are all the more unfortunate since some of the authors criticize methods such as G's directly (cf. e.g. Callaghan & Miller 1962, Campbell 1973, Goddard 1975). Ruhlen (1987:216-7) is very much mis- taken when he asserts that 'there was little activity in Amerindian classification following ... 1964.'

592

REVIEW ARTICLE

Ritwan hypothesis-that Yurok and Wiyot of California are related to Algon- quian-has been proven to the acceptance of all; it is a successful case of demonstrated distant genetic relationship (see Goddard 1975, Haas 1958a, Tee- ter 1964). Nevertheless, Sapir's 1925 Subtiaba-Hokan paper, one of the best known proposals of distant genetic relationships and an article that contains the clearest exposition of Sapir's methods, has proven unsuccessful-though it is accepted and frequently cited in LIA (46, 47, 132, 380, etc.). Subtiaba- Tlapanec has been shown to constitute an ordinary branch of the Otomanguean family (Suarez 1979, 1983b, 1986), while considerable doubt has been raised about the genetic unity of Hokan in general (cf. Goddard 1979, Langdon 1979). If there is a connection between Subtiaba-Tlapanec and any of the so-called Hokan languages, it must embrace all of Otomanguean; but no such connection has shown any promise.

To take another case, Sapir was certain of the Na-Dene affiliation with Sino- Tibetan: 'If the morphological and lexical accord which I find on every hand between Nadene and Indo-Chinese is "accidental", then every analogy on God's earth is an accident' (Golla 1984:374). Needless to say, no specialist today embraces this claim; in fact, the genetic relationship among putative members of Na-Dene itself has been seriously questioned (Krauss 1979, Levine 1979). If a linguist of Sapir's calibre can prove right on some occasions but wrong on others, then an African success for G bears no necessary implication of success in the American arena.

METHODS

2.1. HISTORICAL PERSPECTIVES. The historical study of American Indian lan- guages is characterized by attempts to reduce the vast diversity of the Americas to manageable genetic schemes. Many hypotheses were proposed initially as hunches, or long shots, to be tested in subsequent research. As Kroeber put it, 'Others will be convinced as to the fact of relationship [of his broader Hokan hypothesis], and if they want the "how", that is another story for later, or let them work it out themselves' (Golla 1984:178). Or, to cite two typical comments of Sapir's, 'A certain amount of groping in the dark cannot well be avoided in the pioneer stage of such an attempt as this [his putative Hokan-Coahuiltecan alignment] (1920:289); and 'Such a scheme [Hokan] must not be taken too literally. It is offered merely as a first step toward defining the issue, and it goes without saying that the status of several of these languages may have to be entirely restated' (1925:526).

Unfortunately, too often such preliminary proposals were taken as demon- strated and subsequently were repeated uncritically. This means that many of these widely accepted but poorly founded proposals of remote affinity are in need of reassessment. Kroeber and Sapir were both extremely influential, and both were involved in attempts to reduce the ultimate number of genetic units to ever-larger groupings. Later scholars attempted to relate unaffiliated lan- guages to these groups, heaping even more languages onto the Hokan or Pen- utian pile.

The spirit of these reductionist times is revealed, for instance, by Dixon &

593

LANGUAGE, VOLUME 64, NUMBER 3 (1988)

Kroeber's 1913a one-page summary announcements reducing California to four stocks (from the 22 of Powell 1891); by J. P. Harrington's 1917 'one-liners' (e.g. 'Mr. John P. Harrington announces that he has found genetic relationship between Washoe and Chumashan'; see Harrington 1913 for a similar an- nouncement about Yuman and Chumashan); and by quotes such as these:

Sapir to Radin 1913: 'The process of slaughter of linguistic families, upon which several of us seem to have embarked of late, is going on apace' (Golla 1984:113). Kroeber to Sapir 1914: '... this family [Hokan] now stretches from Oaxaca to Oregon ... I should not be surprised if it were to grow far north and east also, or we may discover new relatives in Mexico ... the eighty-two families in ... Mexico have already shrunk to sixty-four. I believe it will be a very few years only before we are positively down to half that number. I very much wish you could take a few evenings off and dispose of Beothuk' (Golla 1984:150).

Unsurprisingly, many of these rough-and-ready proposals have not panned out, and several are demonstrably wrong. But G's classification relies on the proposals made by Kroeber and Sapir and is squarely within this same reduc- tionist tradition, even employing their methods. G goes to great lengths to associate himself with their practice and classifications (v, 25, 40-1; cf. Darnell 1987).

2.2. AN EXAMPLE. An example of one of these often-repeated long-shot pro- posals is instructive. The Xinca-Lenca hypothesis was first proposed by Leh- mann 1920 (though he also included Mixe-Zoquean, Tequistlatec, and Chumash-Salinan) and accepted by Sapir; it became entrenched in the literature and is reflected in LIA as part of G's Chibchan (cf. 382). Nevertheless, Leh- mann's evidence (1920:767) is the only data ever published in support of the hypothesis. I assess his evidence in its entirety (actual Xinca forms are pre- sented in brackets, from Campbell & Kaufman 1976):

XINCA LENCA (1) water uy Pu:y] cuy 'invierno' (2) night suma [sa-siha] 'in the dark' ts'ub 'Nacht' (3) dark, black ts'ama [siha] ts'ana-uamba 'Morgen grauen' (4) shade ti-tsuma [ti-siia 'in the dark'] saba (5) dog xusu [x = [s]] shushu (6) cough ojo Poho] hoo, oiguin (7) maize au, aima Payma] ama, aima (8) bean xinak [$inak] shinag

(9) one ical Pik'al] etta, ita (10) two bi-al, pi-ar, pi [pi?] pe (11) three vuaal-al, hual-ar [wal(a)] laagua, lagua (12) four in-ar [(h)irya] heria, erio; sa, eslea

In ex. 1, 'Water' and invierno 'winter' are not semantically equivalent, not even if by 'winter' one might understand 'rainy season'. The Xinca and Lenca forms are not phonetically equivalent (there is no account of Lenca (c) in view of Xinca 0).

Exx. 2-4 'night; dark; shade' are all based on a single Xinca root; at best there can be only one correct etymon here, not three. Ex. 3 lacks semantic equivalence ('black, dark' vs. Morgen grauen 'dawning').

Both forms in 5 are borrowed from Spanish chucho, the local word for 'dog' (Xinca and Lenca both lack an alveopalatal affricate).

594

REVIEW ARTICLE

The forms in 6 are onomatopoetic (cf. Quechua uhu-, Mayan ox-). As for 7, words for 'maize' are borrowed throughout Mesoamerica and Cen-

tral America (cf. Cacaopera-Matagalpa aima, Sumu ama, Subtiaba ima, Proto- Mayan *eAm, Tarascan ema, etc.).

'Bean' (8) in both languages is borrowed from Mayan (cf. Western Mayan *cinak', Proto-Mayan *kinaq').

The forms in 9 are not really phonetically similar. The numerals in 10-12 are widely borrowed in this part of Central America. In sum, not one single cognate set proposed by Lehmann is without serious

problems; the Xinca-Lenca hypothesis thus has no support. Unfortunately, the evidence for many other proposed remote relationships

is not unlike that for the Xinca-Lenca hypothesis (for discussion see Callaghan & Miller 1962, Campbell 1973, Campbell & Kaufman 1981, 1983, Goddard 1979, etc.). LIA is in this tradition. Since this is a strong claim, I consider G's methods and contrast them with standard criteria required for plausible proposals of distant genetic affinity.

2.3. Some proposed procedures for dealing with potential distant relation- ships are of little real value, while others must be qualified. I now turn to these flawed procedures.

2.3.1. LEXICAL COMPARISONS. The use of lexical comparison (as opposed to grammatical evidence) in determining distant relationships has been contro- versial (Haas 1969). Lexical information has often been employed exclusively because other data were unavailable, or because the tradition in American Indian studies of relying on word lists was followed unquestioningly. Several linguists adopted Duponceau's 1838 assumption that all American Indian lan- guages shared a common grammatical typology (polysynthesis), and that shared grammatical patterns would therefore not be useful for comparison or classi- fication (since it was believed that American tongues would all exhibit the same sort of grammar). Others, such as Powell 1880, believed that grammar was correlated with the supposedly fixed order of social evolution, making gram- matical similarities among languages a consequence of their stage of social evolution, not of genetic affinity. Whatever the reason, lexical comparison came to be the sole evidence offered for many distant genetic proposals in the Americas. Nevertheless, lexical considerations alone will usually not prove sufficient to satisfy skeptics or to establish the plausibility of such proposals. Goddard (1975:254-5) comments as follows:

'It is widely believed that, when accompanied by lists of the corresponding sounds, a moderate number of lexical similarities is sufficient to demonstrate a linguistic relationship ... However, ... [t]he criteria which have usually been considered necessary for a good etymology are very strict, even though there may seem to be a high a priori probability of relationship when similar words in languages known to be related are compared. In the case of lexical comparisons it is necessary to account for the whole word in the descendant languages, not just an arbitrarily segmented 'root', and the reconstructed ancestral form must be a complete word ... The greater the number of descendant languages attesting a form, and the greater the number of comparable phonemes in it, the more likely it is that the etymology is a sound one and the resemblances not merely the result of chance. A lexical similarity between only two languages is generally considered insufficiently supported, unless the match is very exact both phonologically and

595

LANGUAGE, VOLUME 64, NUMBER 3 (1988)

semantically, and it is rare that a match of only one or two phonemes is persuasive. If the meanings of the forms compared differ, then there must be an explicit hypothesis about how the meaning has changed in the various cases. Now, if these strict criteria have been found necessary for etymologies within KNOWN linguistic families, it is obvious that much stricter criteria must be applied to word-comparisons between languages whose relationship is in question.'

Note that most LIA examples are lexical, and precisely of the sort against which Goddard warns.

2.3.2. GLOTTOCHRONOLOGY. Glottochronology does not find or test distant genetic relationships; it merely assumes them and proceeds to attach a date. It is to be eliminated from considerations of proposed remoter affiliations. Nevertheless, G would apparently out-Swadesh Swadesh himself in his ex- tension of glottochronology and in his application of it to distant genetic comparisons: '... through multilateral comparison we can extend glottochronological theory to account for resemblances not between two lan- guages, but among any number of languages ... with only 10 languages, even after 10,000 years about 42 percent of the original vocabulary is recoverable. More languages will, of course, greatly increase these values' (28; see also LIA Appendix A). However, as Chafe (1987:652) observes, 'the addition of more languages also greatly increases the incidence of chance resemblances'. Need- less to say, skeptics will not be encouraged by G's extensions of glottochronology.2

2.3.3. MULTILATERAL COMPARISONS. G makes much of his technique of mul- tilateral comparisons, which consists not of 'looking at a few languages across many words [but] rather ... at many languages across a few words' (23), calling it a 'revolution in methodology' (3). This strategy is not new with Greenberg.

2 Some scholars have tried to establish other statistical means for deciding issues of distant genetic relationship, such as what proportion of some particular list of vocabulary items may turn out by chance to be similar. Bender 1969 is perhaps the best-known example (see also Collinder 1947). Such studies make the assumption that when the vocabularies (of the sort specified by the method) of compared languages show a significantly higher number of similarities than those ex- pected by chance (by their calculations), the languages in question are (probably) genetically re- lated. There are many reasons why these tests typically fail, but the most compelling is that in practice they simply fail to distinguish plausible cases from blatantly improbable (or impossible) ones. For example, comparisons of Finnish with Cakchiquel (Mayan) and Quechua, and with proposed Proto-Mayan and Proto-Zoquean forms (Campbell 1973, Campbell & Kaufman 1981, 1983) show a much higher level of similarity and matching than that permitted by chance in Bender's claims. A method that fails so thoroughly in applications can hardly warrant much faith.

In a related vein, it is worth mentioning that many scholars take basic vocabulary to be necessary to successful distant genetic proposals. It is generally accepted that basic vocabulary tends to change less drastically and is preserved longer than more abstract or culturally charged vocabulary (cf. LIA 22). Therefore, distant genetic proposals which lack basic vocabulary among the postulated potential cognates are suspicious. This criterion, however, is independent of glottochronology; it does not assert that a universal, culture-free vocabulary list exists or that core vocabulary cannot suffer severe changes or that the rate of change cannot be speeded up or slowed down by socio- cultural factors. Rather, the general tendency for basic vocabulary, broadly defined, to be less subject to change suggests that plausible proposals of distant affinity should show a number of basic vocabulary items.

596

REVIEW ARTICLE

Kroeber employed it in America: 'the resemblances between any two individual languages are never so convincing as those which appear when an entire group is examined' (Golla 1984:88). In fact, G (25) may have learned the notion from Americanist practice: 'Kroeber himself once told me that after fruitlessly mak- ing individual paired comparisons, he one day conceived the notion of looking at all of them simultaneously. He and his co-worker Dixon immediately noted that two major groupings accounted for a large majority of these families, to which they gave the names Penutian and Hokan; and these groupings, later expanded to include many languages outside of California, have stood the test of time.'

But notice that this is exactly the problem-Kroeber's Penutian and Hokan, framed on the basis of multilateral comparison of short lexical lists, are disputed (cf. Langdon 1979, Shipley 1980). The question of whether the languages of these groupings will ultimately prove to be genetically related is still open, but the technique of multilateral comparison, which gave us these controversial proposals in the first place, cannot decide the issue for us. Nevertheless, G builds on these shaky Hokan and Penutian foundations for more far-flung and even shakier groupings.

Finally, multilateral or mass comparison was apparently NOT really employed to arrive at the LIA classification. Most of the data of these 'multilateral' com- parisons are not published in LIA, but Greenberg assembled the data for the LIA classification in handwritten notebooks, which are available for inspection in Stanford University's Green Library. The arrangement of language groups in these notebooks corresponds to the classification in LIA, which suggests that G did not use the notebooks actually to perform multilateral comparison, but rather that he had a predetermined classification which is reflected in the way the notebooks were assembled. Moreover, a review of G's African papers (see Greenberg 1955[1949-54]) reveals that G's African classification for the most part predated his method of multilateral comparison (or, in G's words, 'I was only dimly aware of the significance of mass comparison at the time of writing' [Greenberg 1955:106]). The method as employed in Africa has also been criticized for laxity in the interpretation of possible resemblances.

Mass comparison is not a 'method' (as is the comparative method) but is rather a strategy, or a heuristic practice. Similarities unearthed by it must still be evaluated against standard criteria, and those similarities which may pos- sibly have an explanation other than common ancestry must be set aside.

CRITERIA FOR DISTANT GENETIC RESEARCH

3. To evaluate LIA properly, legitimate methods in remoter genetic research, together with the proper cautions that their use dictates, must be considered. I take these up here, concentrating on criteria needed to establish a proposed remote relationship as plausible; these same criteria serve also to evaluate such proposals, those of LIA among them. (For more detailed discussions see Cal- laghan & Miller 1962, Campbell 1973, 1975, 1978a, Campbell & Kaufman 1981, 1983, Campbell & Mithun 1979a, Goddard 1975, Teeter 1964.)

To begin, one must assemble a number of similarities and matchings, and

597

LANGUAGE, VOLUME 64, NUMBER 3 (1988)

then one must take care to eliminate (or at least qualify) all which lack sufficient semantic or phonetic similarity, or which could be explained by factors other than common ancestry-e.g. borrowing, chance, onomatopoeia, sound sym- bolism, and universal or typological interconnections. Convincing proposals must minimize these other possible explanations of the compared material, leaving the genetic explanation the most probable. Unsuccessful proposals do not usually fail because of lack of similarities or matchings, but because of lack of care in distinguishing genetic from nongenetic potential accounts of these. Such is the case with LIA.

3.1. SOUND-AND-MEANING ISOMORPHISM. Greenberg 1963 recommended the useful procedure of allowing only similarities involving both sound and mean- ing. Resemblances involving sound alone (e.g. the presence of a tonal system) or meaning alone (e.g. morphemes of sex gender) can be independent of genetic relationships-the results of diffusion, typological tendencies, etc. Unfortu- nately, this criterion is not always exercised. For example, Sapir's six super- stocks for North America, of which G is fond, were based on gross morphological and typological similarities, as 'suggestive but far from demon- strable ... at the present time' (Sapir 1925:526). McQuown's (1942:37-8) version of the Macro-Mayan hypothesis (accepted in LIA, 143) is based on gross pho- nological similarities: 'The only other language family besides Totonacan of Mexico that has this glottalized series is Mayan, and this fact together with other significant details suggests to us the probable genetic relationship of To- tonac-Tepehua with Mayan; but the relatively small number of coincidences in vocabulary indicates to us that this kinship is quite distant' (my translation). Since glottalization can be created internally from clusters of C + ? (as in Otopamean) or can be diffused (as in Armenian dialects from Caucasian lan- guages), it is easy to see the wisdom in G's isomorphism principle. However, since several other Middle American languages are now known also to have glottalization (Tequistlatec, Otopamean, Jicaque, Xinca, and Lenca), clearly Macro-Mayan had a shaky origin at best. G should have heeded his own counsel in this case.3

3 While the principle clearly is aimed in the right direction, it may be overstated. Some kinds of patterned matchings in grammatical subsystems, even without phonological correspondences, might prove to be persuasive evidence of relationship. This reflects Sapir's emphasis on 'sub- merged' features, considered in detail in ?3.6.

Another of Greenberg's (1963) valid procedures was to permit only linguistic information as evidence, eliminating nonlinguistic criteria from consideration, e.g. shared cultural traits, my- thology, or folklore. Though the value of this practice should be obvious, it has not always been followed. At worst, one finds proposals such as Kennedy's 1856 linking of Carib with West African languages, whereby he wished to prove that American Indians were not the result of a separate Creation and were therefore immigrants from other continents; as evidence he presented fourteen Carib words with vague resemblances to certain West African equivalents. We might also recall proposals to link Mayan with Natchez and 'the Chahta-Muskoki family' based on assumed Mayan pyramid similarities with Mississippian mounds, and suggestions that Mayan is related to languages of the Antilles based on mythological accounts of the first settlers as having arrived from the East by sea (cf. Brinton 1869). Given the many mistaken cases of proposed linguistic connections based on nonlinguistic information and on mistaken assumptions about how language, race, and culture may reflect each other, it is quite clear why such information cannot form the basis of a claim of genetic relationship.

598

REVIEW ARTICLE

3.2. BORROWING. Diffusion is a well-known source of potential nongenetic similarity among languages. Loanwords can not only result in lexical similar- ities, but can also mock regular sound correspondences in many cases-a point made well in LIA (20-3: cf. also Goddard 1975). For this reason, lexical com- parisons alone are seldom totally convincing. One might suspect that such a realization would lead to greater caution in the interpretation of similarities and correspondences, but this is not G's response. Instead, he takes it as an argument for the admission of phonologically similar forms of whatever sort and against sound correspondences in general.4

More to the point, some of G's proposed etymologies suffer from a failure to eliminate borrowings. Examples are Cuitlatec navaxo 'knife' (9, p. 108), from Spanish navajo 'knife, razor'; Tunebo baxi-ta 'machete' (9, p. 108), from

4 One widespread claim bears comment-namely, that lexical borrowings are highly compatible with identical or very similar sound correspondences, but that 'real' sound correspondences, held to be good indicators of genetic relationship, tend to be 'exotic' and not so similar phonetically (cf., e.g., Witkowski & Brown 1981). This claim is misleading. It may be true that sounds in borrowed forms often tend to be similar to the sounds of the donor language, but they may also be very different. They may be modified to accommodate the structure of the borrowing language, e.g. Texistepec Popoluca (Zoquean) kunus 'cross' from Spanish cruz, where n substituted for r, which is absent from Mixe-Zoquean phonology, and where u breaks up the unpermitted consonant cluster kn. Loans also undergo the normal sound changes of the borrowing language: the older the loan, the greater the possible number of such changes. Moreover, the donor language may also undergo changes which modify the vocabulary source of the loan. Both these sorts of changes, in the donor and in the borrower, may make the loanword less similar to the source word in the donor language. For example, Finnish hake-a 'to fetch, get, look for' is to be paired with English seek /sik/ and German such-en Izu:x-I, yet one could hardly call Finnish h, English s, and German z, or a, i, and u: 'very similar' or 'nearly identical'. Nevertheless, the Finnish word is borrowed from Pre-Germanic *sa:keja (PIE *sa:g-), which by the time of Proto-Germanic had changed to *so:kja via the regular shift of *a: to *o:. Later changes produced the current Germanic forms; among those changes were the German voicing of s, the German shift of k to x, and vowel changes in both English and German. The form was borrowed as Pre-Finnic *sake-, which became hake- by the late Proto-Finnic change of *s to h (Campbell 1986). This example shows that the sounds of borrowed forms need not be particularly similar to the sounds of the donor language. Such examples are not oddities, but are frequently encountered.

The other side of the claim is that identical or very similar sound correspondences are weak and should be distrusted in questions of distant genetic relationship on the ground they may reflect borrowings (Witkowski & Brown 1981:906). We have just seen that radically different sounds in correspondence sets do not necessarily negate the possibility of borrowing. On the other hand, similarity in true correspondences depends on many factors. Some languages may change little over time. Pipil, separated for over a thousand years from other varieties of Nahua, has virtually identical sound correspondences with Central Mexican Nahuatl. English, on the other hand, in roughly the same interval, has come to differ radically from Old English, with extensive changes in the vowel system, the consonant inventory, consonant clusters, distribution of allophones, etc. Even English, however, after its many changes, generally retains unchanged PIE *r, *1, *m, *nj, *s, *w, and *y. Extensive change over time is possible, but not necessary. Therefore, persistent identical or very similar correspondences should not be shunned or denied, nor should excessive weight be given to non-similar correspondences, given that loans can also exhibit them. It should now be clear that borrowings may involve 'exotic' corresponding sounds and that identical real correspondences due to a common ancestor may persist.

This being the case, it is easy to understand why many scholars insist that a remote genetic relationship can be demonstrated only through the application of the comparative method, where the reconstruction of an ancestral system is worked out to some degree, accounting for the recurrent sound correspondences in a plausible historical framework.

599

LANGUAGE, VOLUME 64, NUMBER 3 (1988)

Spanish machete (note that in Tunebo [x] alternates freely with [s], and nasal consonants do not occur before oral vowels); Cayapa kacu 'horn' (104, p. 114), from Spanish cacho 'horn'; and Kekchi po, Pokomchi poh, po 'moon' (166, p. 239), which are correctly indicated as borrowings from Mixe-Zoquean *po2l *poya in G's source (Campbell 1977:110); also LIA 44 (p. 148), 198 (p. 247), 240 (p. 258).

3.3. SEMANTICS. Semantic shifts are a part of linguistic history. Neverthe- less, any necessary but undocumented assumption of semantic shift increases the possibility that chance, rather than common ancestry, explains a phono- logical similarity. Therefore, in its initial stages, any hypothesis worthy of at- tention should be able to muster a goodly number of semantically equivalent forms to establish its plausibility. If a hypothesis cannot be established as prom- ising with semantic equivalents, then the admission of nonequivalent forms will do nothing to strengthen the hypothesis.

G's forms are quite permissive in semantic latitude. Semantic equations such as the following are not convincing: 'excrement/night/grass', 'ask/wish/seek/ pleasure', 'bitter/to rot/sour/sweet/ripe/spleen/gall', 'body/belly/heart/skin/ meat/be greasy/fat/deer', 'child/copulate/son/girl/boy/tender/bear/besmall', 'deer/ dog/animal/silver fox/lynx', 'earth/sand/sweepings/mud/dirty/field/devil/bad/ underneath / bottom', 'earth / island, forest / mud / village, town / dust / world / ground', 'feather/hair/wing/leaf', and 'hole/mouth/ear/listen/chin/nose/smell/ blow nose/sniff'. And these are only some of G's unconvincing semantic equations.

3.4. SHORT FORMS AND UNMATCHED SEGMENTS. Another criterion concerns the length of proposed cognates and the number of matched segments within them: the greater the number of matched segments in a proposed cognate set, the less likely it is that accident accounts for the similarity. Thus, while mono- syllabic CV or VC forms may represent true cognates, they are so short that the similarity could easily be due to chance. And if only one or two segments of longer forms are matched, with no explicit proposal to account for the un- matched segments, then chance becomes a strong possibility. It is important to account for the whole word, not just for some arbitrarily segmented part of it. A match of only one or two segments will not be persuasive (cf. Goddard 1975:254-5). Many of G's proposed cognates involve such forms, e.g. 217 (158) Cakchiquel sa 'be' (actually 'only') compared with noncognate Quiche us 'become'.

3.5. ONOMATOPOEIA. Onomatopoetic forms should not be taken as potential support for a hypothesis of genetic relationship, since the similarity may be due to the fact that the different languages have independent approximations of the sounds in nature, rather than to common ancestry. This criterion calls into question a considerable number of LIA forms. For example, 'blow' (phu, pui, puhi, pho-, 196) exhibits similarities cross-linguistically, as Tylor (1871:229) showed in his long list of 'blowing' words with the elements pu, puf, bu in various languages, e.g. Balto-Finnic puhu- [Finnish puha(htaa)] and En-

600

REVIEW ARTICLE

glish puff. Some other onomatopoetic LIA examples are 'breast' (cf. 'nursing', 198), 'cry' (209), 'drink' (uk, ku2, 2uhk, Wuki, 213; similar forms are found throughout the world), 'duck' (e.g. Proto-Central Algonquian *si:s i.:pa; cf. Nahuatl pisisi 'duck', where initial p was lost except for onomatopoetic forms, 22, p. 167), 'swallow' (261), 'wind' (122, 162, p. 99).5

3.6. GRAMMATICAL SIMILARITIES. Not everyone has relied solely on lexical comparisons, and it is generally recognized that good grammatical evidence may offer special support for hypotheses of remote kinship. Sapir (1925:491) spoke of 'submerged' features, by which he apparently meant similarities so arbitrary that they would resist explanation by chance or diffusion. According to Darnell & Sherzer (1971:26), Sapir 'believed it possible to separate out the superficial elements of grammar which probably resulted from diffusion from a "deeper" and more '"profound" kernel of grammar reflecting genetic origins'. Sapir's (1925:491-2) claim was as follows:

'When one passes from a language to another that is only remotely related to it, say from English to Irish or from Haida to Hupa or from Yana to Salinan, one is overwhelmed at first by the great and obvious differences of grammatical structure. As one probes more deeply, however, resemblances are discovered which weigh far more in a genetic sense than the dis- crepancies that lie on the surface and that so often prove to be merely secondary from a historical perspective. In the upshot it may appear, and frequently does appear, that the bulk of what is conventionally called its grammar is of little value for remoter comparison, which may rest largely on submerged features that are of only minor interest to descriptive analysis.'

Teeter's 1964 'contextualization' or 'depth hypothesis' is quite similar. He believes that much work in detecting genetic relationships is wasted on the manipulation of lexical items, which are not very 'contextualized' in terms of their role in the grammar. In Goddard's words, 'It is necessary to show not only that the resemblances are so numerous and detailed as to exclude the possibility of chance as an explanation but also that they are so tightly woven into the basic fabric of the languages that they cannot be explained simply as borrowings' (1975:259). Plausible distant genetic proposals should have support from such 'submerged' or 'contextualized' grammatical correspondences (cf. Bright 1984:12, Campbell & Mithun 1979a, Goddard 1975, Teeter 1964).

This strategy of seeking arbitrary, submerged grammatical similarities may be a very effective means of establishing the plausibility of distant genetic proposals. Still, it is not foolproof. Some extremely arbitrary, striking, and seemingly 'contextualized' similarities may nevertheless have nongenetic ex- planations. For example, Quechua and the Mayan language Quiche might be thought to share two such submerged features: different forms of first person singular pronouns and a discontinuous negative construction. Both languages

5 As with onomatopoeia, the relationship between sound and meaning in cases of sound sym- bolism is also not totally arbitrary. Sound symbolism (ideophones, phonoesthetics) refers to sounds which symbolize directly some semantic attribute of the referent, typically size, shape, or extent. It has important consequences for remote relationships because it can create both irregular sound correspondences among related languages and greater apparent but nongenetic similarity among unrelated languages. In the Americas there are frequent symbolic interchanges among the sets qlk, ?/e, s/sil, round/nonround vowels, front/back vowels, etc. (cf. Nichols 1971, Campbell & Kaufman 1983).

601

LANGUAGE, VOLUME 64, NUMBER 3 (1988)

have two sets of pronominal affixes found in distinct contexts. Proto-Mayan had *i and *w-, while Quechua A has -ni- and -wa-. This is a striking corre- spondence. Closer inspection, however, reveals the apparent correspondence to be spurious. The -ni- form in Quechua A came historically from the empty morph that is automatically inserted between two morphemes which would otherwise create an unacceptable consonant cluster. The first person singular morpheme was originally *-y (Parker 1969:150), which took the empty morph -ni- when attached to consonant-final roots; ultimately -y in this context was lost (-ni + y > -ni) and -ni was reinterpreted as the first person singular marker. Moreover, the -wa- of Quechua A comes from Proto-Quechua *ma, as evi- denced in the Quechua B dialects (Parker 1969:193). Thus, what seemed to be a striking submerged correspondence turns out to be an accidental one instead.

The second example also involves specious resemblance. Negation in both Quechua A and Quiche employs discontinuous constructions that are phonet- ically similar in the two languages: Quechua mana ... -cu, Quiche man ... ta. Given the similarity, one would expect a historical connection, but there is none. Proto-Mayan had only *ma. Quiche acquired the discontinuous con- struction by making the optative particle *tah obligatory in Quiche conditional sentences in non-negative contexts; man appears to be from ma 'negative' +

na 'now, still, yet, first, well, later'. Accidentally similar negatives of the form ma are found in Afroasiatic, Sino-Tibetan, Thai, Indo-European, Algonquian, etc., so that this phonetic similarity is not sufficient by itself to argue for a connection between Quechua and any Mayan language. Again, then, what seemed like a submerged similarity proves to be an accidental correspondence. The moral is clear: while grammatical information is important to proposals of remote relationships, care must be exercised in interpreting potential cases of submerged features.

In Chap. 5, G presents what he considers to be grammatical evidence in support of his claims, though none is of the submerged type. Given that the interpretation of even seemingly submerged similarities can be deceptive, still greater care ought to be exercised in evaluating the kinds of grammatical ex- amples presented in LIA. Such resemblances should fare better than equations with some unrelated language selected at random. I have compared G's gram- matical evidence with Finnish data and have found that Finnish grammatical forms fit most of G's forms; this makes it difficult to accept the G grammatical data as convincing evidence. In the interest of saving space, I give here only Finnish comparisons with G's first twenty-five examples (273-289) [others available upon request]. Only the stronger matchings are presented here, al- though reasonable matchings are available for nearly all. The example numbers are G's; I have skipped LIA forms which are limited to only two languages or one group or area (4, 5, 7, 8, 9, 17, 20, and 21). LIA 'Amerind' forms are presented first, followed by Finnish (F) forms.

(1) n '1st person': F -n 'I' (mene-n 'go-I'), -ni 'my' (talo-ni 'house-my'). (3) m '1st person' (Chibchan-Paezan; Penutian-mainly p1.): F me 'we', -mme 'we, our' (me

mene-mme 'we go-we'; talo-mme 'our house'), mind 'I'. (10) Macro-Tucanoan, Macro-Ge, Chibchan-Paezan ka-, kai-, ka, ga, ka-, -ke- 'second person

sg.': F -kaa '2nd person p1. imperative', -k (> -2) '2nd person sg. imperative'.

602

REVIEW ARTICLE

(11) s '2nd person': F -si 'your (sg.)' (talo-si 'house-your'), sa 'you (sg.)'. (13) t 'demonstrative, 3rd person pronoun': F ta- 'this', tuo 'that'; cf. Estonian ta 'he, she'. (14) m 'demonstrative, 3rd person pronoun': F ta-ma 'this', nai-ma 'these', muu 'other'. (15) n 'demonstrative, 3rd person pronoun' (Andean, Equatorial): F ne 'they, these', nuo

'those', na- 'these'; -nsa, -Vn '3rd person possessive' (talo-nsa [house-his] 'his, her, their house' (cf. also Itene -n 'sg. verb object': F -n 'accusative sg.').

(16) p 'demonstrative, 3rd person' (Macro-Panoan, Andean, Hokan): F -p, -pi (< *-pA) '3rd person sg.' (dialect forms: -p, -vi, -e, -V), -va-t (< *-pa-t) '3rd person pi.' (tule-pi 'comes-he', tule-vat 'come-they').

(18) s '3rd person', Almosan-Keresiouan: F se 'it [3rd person nonhuman pronoun]', -nsa '3rd person possessive' (talo-nsa 'his/her/their house').

(19) k- '1st person dual' (Carib, Algonquian [actually Algonquian 'inclusive']): Finno-Ugric *-k 'dual' (cf. Ob-Ugric k-), which is reflected in some frozen F plural pronouns and verb forms; cf. also -k 'plural'. G also lists k '1st person pi. (inclusive)' in some other cases; cf. Yokuts mak' 'first person dual inclusive': F me-k, -mme-k 'we' (Eastern dialects, Standard F [SF]), me(2), -mme(2) (-2 < *-k) (mek mene-mmek 'we go-we', SF me( 2) mene-mme( 2)). This form is said by G to have developed from 'two' (291); cf. F kaKte- 'two'.

(22) k 'near (proximal) demonstrative': F takd(-lainen) 'one from here', jo-ka 'who, which, that, each, every', mi-ki 'what, which, that', ku-ka 'who, that', -k 'lative' (SF > -(?)).

(23) ni 'this', ni 'hither', noni 'there (near you)', nina 'here', naa 'this', eni 'this', ani 'here', ana 'this', nan 'that', 'demonstrative of near-deixis' (SA): F ni- 'these (here)', ne 'these', -na 'locative (in, here, koto-na 'home-at')', -nne 'hither' (ta-nne 'here, into this, toward this, hither').

(24) Macro-Panoan h-demonstrative (Chacobo ha 'that, he, she'): F hdn 'he, she', he 'they' (han tuli 'he came', he tulivat 'they came').

(25) t 'reflexive' (Macro-Ge tu- '3rd person reflexive', Tacana ti 'reflexive marker on verbs'): Balto-Finnic *-te- 'reflexive verb marker', F -tu-l-ty- (pese-yty- 'wash oneself, pese- 'wash').

3.7. DISTRIBUTIONAL CRITERIA. In the comparison of two families it is the existence of cognates in the sister languages which ensures an etymon's an- tiquity within its family and hence provides for its potentially greater antiquity in even remoter comparisons. Proposed cognates based on a single language or with a limited distribution within a language family are not strong evidence for proposed distant relationships. A form found in only a single language or subgroup may be old, but it may also be an innovation, the result of lexical change, borrowing, affective or symbolic formation, analogical reformulation, replacement, shift, etc. Therefore, only forms with a legitimate ancestry within their own families can be given strong consideration in hypotheses of distant relationship. For any word selected randomly, there will normally be a number of forms accidentally similar in sound and meaning in the vocabularies of the different languages being compared. This radical increase in the possibility of accidental agreements confirms the philological saw that it is easy to find what one is looking for. In fact the number of expected accidental matchings will be roughly proportional to the number of languages consulted; each language has a large vocabulary which comprises potential targets for accidental match- ings. Like shooting fish in a barrel, one can hardly help but find similarities from among the vast number of lexical items available when many languages are compared. For this reason, confidence is warranted only for forms attesting cognates with a wide distribution in a language family, while caution is in order

603

LANGUAGE, VOLUME 64, NUMBER 3 (1988)

for the assessment of distributionally limited forms. This is a problem for G, since several of his proposed Amerind etymologies are of this distributionally limited sort. (A few examples are cited below in ?3.9.)

3.8. REPEATED COGNATES. Many proposals of distant relationship repeat the same etymon under different proposed cognate sets (as in the Xinca-Lenca example). A single form from one language cannot be cognate to multiple forms in another, unless the multiple forms are derived ultimately from a single orig- inal etymon. Therefore, the total number of proposed cognate sets will always have to be reduced by at least the number of repeated etyma. LIA has many such examples, among them Jacaltec ita 'food' (i.e. itah) (149) under the Pen- utian gloss 'eat'-but its cognates, Tzeltal itax 'verdure' and Kekchi icax 'grass' are given under the Amerind gloss 'dirty' (212). In fact all three are regular reflexes of Proto-Mayan *itYax 'vegetable, edible plant' (Tzeltal 'edible vegetable', Kekchi 'edible vegetable, pasture plants', Jacaltec 'food, edible plant').6 Another example is the form listed under two separate etyma as Shaw- nee alikatwi 'cloud' (46, p. 168) and 'shadow' (165, p.177); actually this is a mislabelled Miami form, from Proto-Algonquian *a:laxkwatwi '(be a) cloud', based on *-axkw- 'sky'-which G compares to both Bella Bella anqw 'cloud' and Bella Bella ne:kk 'night'. Similarly, the same Algonquian etymon for 'liver' (*-Okwen) is listed in two separate 'etyma': 'bitter', 25, p. 167, and 'liver', 119, p. 179 (Goddard 1987:657).

3.9. NONCOGNATES COMPARED. Another problem involves false cognates, where unrelated forms from languages within a particular family are put to- gether in the belief that they are (or may be) related and can be used as evidence for more distant connections. Obviously, if reconstructions are based on non- cognate forms within a family, their value in remoter comparisons is ques- tionable. G's proposed etymologies contain many examples of this sort. Here is an instance from Mayan: under 'liver' (154) G cites Kekchi c'oc[']el and Quiche kus 'heart'; but actually these are from two separate etyma, Kekchi c'oc' 'liver, earth' (from Proto-Mayan *c'o:c' 'earth') and Quiche k[']u[2]s 'chest, heart' (Proto-Quichean *k'u2us). Other examples are 16 (p. 146), 33 (p. 147), 60 (p. 149), 167 (p. 155), 184 (p. 156), 217 (p. 158), 246 (p. 160), 273 (p. 161), 274 (p. 161), 20 (p. 188), 49 (p. 198), 68 (p. 206), 215 (p. 251), 221 (p. 253), 234 (p. 256), 254 (p. 262), 269 (p. 266), etc. (For Algonquian cases, see Goddard 1987.)7

3.10. ERRONEOUS RECONSTRUCTIONS. A related problem is that of false re- constructions taken as the basis for more remote comparisons. That is, some intrafamily reconstructions, in spite of valid cognates, are wrong and the mis- takes may make the reconstructions appear more similar to forms from other

6 The true Jacaltec cognate of Uspantec ti[ ?] 'eat' (149) is ti? from Proto-Mayan *tjp'to eat meat'. 7 It might be argued for such cases, where multiple cognates have been lumped together in a

false set, that some items could prove to be correct cognates in broader comparisons even if other forms from the false sets did not pan out. Though this is possible, one would still be faced with the distribution problem and with the issue of chance similarities.

604

REVIEW ARTICLE

languages or families for which a remote relationship is being sought. Several such examples from Brown & Witkowski's 1979 attempt to demonstrate a ge- netic relationship between Mayan and Mixe-Zoquean are discussed in Campbell & Kaufman 1983. While G is not engaged in family-level reconstructions, one might suspect that he has chosen certain alternative forms in order to create a greater appearance of similarity. For example, for 'light' (154) G lists forms such as 2eki, akke, 2oko, k'wi2i-s, qe, qai, k1cu, kjc, to which an alleged Proto- Mayan form *q'i:xJ 'day, sun' (actually Proto-Quichean, from Campbell 1977) is compared. The real Proto-Mayan form is *q'i:o (cf. Campbell 1977:38; see also Fox 1978:220, G's other source for Mayan), but this does not exhibit the same degree of similarity to the compared languages as does the Mayan form chosen for citation. (See also 160, p. 155.)

3.11. SOUND-CORRESPONDENCE PROBLEMS. A central feature of the compar- ative method is regular, recurrent sound correspondences in cognate forms, and it is widely believed that a sufficiently select list of lexical similarities accompanied by proposed sound correspondences, better called 'matchings' until kinship is proven, is adequate to prove a genetic relationship (for criticisms of this view, see above, Goddard 1975, Campbell & Kaufman 1981). It is true that rigorous sound correspondences in plausible potential cognates may con- stitute a reasonable argument in support of a proposed remote relationship. Nevertheless, the search for sound correspondences sometimes lacks the req- uisite rigor. A proposed cognate set will provide little support if it involves irregular correspondences. Moreover, it is useless to propose 'regular' sound correspondences which are illustrated by but a single example. These consid- erations may seem self-evident; nevertheless, such dubious correspondences occur much too frequently in actual proposals. (For examples, see Olson's Maya-Chipaya hypothesis [1964, 1965], discussed in Campbell 1973.)

G's approach to 'irregularity in correspondences' is not to doubt the proposed cognates, but rather to accept the proposed 'etymology, along with the rela- tionship', seeking 'to explain what appears to be a discrepancy' (8); he rejects the stricture against 'unique correspondence', maintaining that a 'resemblance may be unique but regular... [or] it can be recurrent, yet...not a genuine cor- respondence at all' (9). This is like playing poker with all the cards wild; any intuited similarity goes. Without a detailed application of the comparative method and a plausibly reconstructed sound system, it will be impossible to assess the status of such matched phonological resemblances.

3.12. ERRONEOUS MORPHOLOGICAL ANALYSIS. In lexical comparisons, it is necessary to account for the entire word and not just for some arbitrarily seg- mented portion thereof. Where morphological elements are etymologized, one must demonstrate that the segmented morphemes actually exist (or existed) in the grammatical system. Unfortunately, unmotivated morphological assump- tions are all too prevalent in postulated remote relationships. When exposed, these assumptions usually show the similarities upon which the comparisons are based to be vacuous. For example, in LIA, the Mayan forms Uspantec cak-ex and Tzeltal taki-n 'dry' are erroneously segmented and compared with

605

LANGUAGE, VOLUME 64, NUMBER 3 (1988)

Kekchi ceki (actually caqih) and Chol tixinis (Chol root t3kin; 47, 197). All, however, are from Proto-Mayan *tYaqi2Q 'dry'; the -x, -h, -n are regular reflexes of PM *-g. In examples 109, 120, G lists Seneca (njota)-re 'lake' (compared with Siouan *re), an arbitrary segmentation (cf. Mohawk -nyatar(e)- and Seneca -nyotae-, with regular loss of *r; Goddard 1987:657). Tunebo baxi-ta 'machete' (108) is an arbitrary segmentation of the loan from Spanish machete.

Unrecognized morpheme divisions can also complicate matters. For ex- ample, in G's 'etymology' table for 'hand' (58), he lists Rama mukuik, saying, 'What makes this evidence [this table] particularly cogent is the existence of variant forms of a similar type-namely, with the suffix -k or sometimes -ka attached to a base ma or the like' (57). However, 'hand' in Rama is kwi:k; the mu- is the possessive prefix; it has nothing to do with G's targeted *ma-ki (57). Similarly, Tzutujil 2ak'ual (actually ak'wal 'child' [37, 148]) is composed of ak' 'new', w-'my', and al 'woman's child' (cf. Proto-Mayan * al 'child'); it is com- pared to Alsea kJ'i:la, Yuki kil, and Huave kwa:l. Tzotzil ti2il 'hole' (115, 152) is actually ti 'mouth' (cf. Proto-Mayan *tYij) plus -il 'suffix of abstract pos- session'. This root tijis not so similar to the compared forms, talokh, tol, tan, telkan. In 175 (156) Tzeltal and Tzotziljat 'penis' is compared with Patwin jot; but the Tzeltal and Tzotzil form is y-(G'sj) 'his' + -at 'penis' (body parts are inalienably possessed; cf. w-at 'my penis'). In 220 (158) a Cakchiquel form paruwi 'above' is compared with Tunica 2aparu 'heaven, cloud'; but the Cak- chiquel word is composed of pa 'in, on', ru- '3rd person possessive prefix', and -ivi'head, hair', i.e. literally 'on his head'. In 240 (258) 'snake', Kekchi k'anti is compared to other Mayan forms can, kan, etc., and to other languages, e.g. Hitchiti sinti, Creek cetto, etc. The true Proto-Mayan etymon is *ka:n 'snake', but Kekchi k'anti2 (the correct form) is a Cholan loan for a species of snake, from Chol k'an 'yellow' plus tij 'mouth'. Without the noncognate loan and segmented morpheme -ti 'mouth', the compared forms do not appear so similar. (For other Mayan cases, see 1 (145), 7, 8 (146), 88 (150), 198 (157), 221 (253), 254 (262), etc.)

3.13. SPURIOUS FORMS. Occasionally nonexistent 'data' show up in LIA. For example, G (155) lists Proto-Mayan *tsu:k 'navel', citing Campbell 1977 as the source. There is, however, no such form in Campbell 1977, or in Quichean, or in Mayan generally; the form given was Proto-Quichean *musu 2s 'navel' (Campbell 1977:155). G's *tsu:k is simply spurious.

3.14. RANDOM SAMPLE COMPARISON. In spite of the many methodological problems charted above, one could perhaps still be positively inclined toward LIA under the belief that the sheer number of resemblances in LIA might suffice as evidence for G's proposals. It is therefore important to point out that the collective weight of the similarities adduced in support of any distant genetic proposal should appear stronger than comparisons from some unrelated lan- guage chosen at random. I have chosen Finnish to compare with LIA forms, not completely at random, but because it is a language about which I know something; it is not seriously taken by anyone of reason to be closely related to any American Indian language. The following comparison of G's data with

606

REVIEW ARTICLE

Finnish shows that LIA proposals do not, in fact, fare better than some un- related language which might be chosen at random.

In order to save space, I present below comparisons from only three pages chosen at random of G's putative Penutian etymologies (145-147). I applied the following strictures. The Finnish comparisons should not involve more semantic latitude than LIA allows (thus G's main gloss is given in small capital letters, together with other relevant glosses in small letters from individual languages that he cites). This is followed by some of the relevant forms from G's sets, given without the language names (for detail, see pp. 145-7). Fol- lowing G's apparent practice, in CVC forms the two Cs ought to be phonetically similar (though at times only one C seems to match for G), while the V may vary; in CV or VC forms, both the C and V ought to match in some way. Suffixes are given in parentheses. I accepted comparisons of Finnish to G's Amerindian forms involving 1, r, n to any of 1, r, n; k, h to k, h, x, q; k to k, c; t, s to t, s, s, c, c; v to w, u, b, p; Finnish final or preconsonantal -n to both m and n; and, since Finnish h comes from *s, h is compared to s, s, h, and x. More persuasive matchings are marked with an asterisk (*). After the LIA data, I list in brackets methodological problems with the examples: [SEM] for lack of semantic equivalence, [SH] too short, [PH] lack of sufficient phonetic simi- larity, and [ON] onomatopoetic.

(1) BE ABLE, know, strong, male: tuman, temma, ndom, etamax. [SEM]. F[innish]: *tomera 'strong, staunch'; *tuima 'strong'; toimi- 'to work, function, serve, perform'.

(2) ABOVE, up, over, heaven, hill: liile, 2ol, olil, la:?, lalli-s [SEM, PH]. F: *laki 'top, summit, ceiling'; *ylhd- 'above, up'; *y/l 'up, over'.

(3) ALL(1), finish: k'is, k'as, kus, kis [SEM]. F: *kasa 'lots, heap, multitude'; kaikki 'all'; kesto 'lasting'.

(4) ALL(2), many: moma, mu2e, maj [PH]. F: *mone- 'many'. (5) ALL(3), each: mon, man, emen, hemen, ma:no, amali, muli. F: *moni 'much, many'. (6) ANGRY, anger, be angry: aka2a, 2ika [SH]. F: *dka- 'anger, be angry'. (7) ARM, hand, finger, take, nail: se, sem, isi, esi, ij.s, isu, wasi, -was, wos, wis, vis, 2asi,

is(k'aq), asa [SEM, SH, PH]. F: *sorme- 'finger'; *(kdsi-)varsi 'arm', sakara 'claw, horn'. (8) ARRIVE, come!, come, come from: wila, war, ala, hul-, el, wulatP, uli. F: *ulo(-ttua-a)

'to come to, reach'; vaeltaa 'to walk, wander'. (9) ARROW, bow, nettle: nok'o, nuk'on, nuk', onoxk, naki, la(:)h, laki [SEM]. F: *nokko(-nen)

'nettle'; nuoli 'arrow'. (10) ASHES, dust: put, potpot, pootel, p idusi, p otil [SEM]. F: poly 'dust'; puru 'sawdust'; pui-

'thresh'; puu(ska) 'gust'. (11) BACK, beaver's tail: wahk, waxi [SEM; only two languages compared]. F: valjaa- 'harness

strap, traces', vieri 'side', vatkaa 'to beat, whip'. (12) BAD, ugly, badness: katse, a'atxal, kasq, qoit5, kacum [SEM]. F: *katala 'bad, wretched,

miserable'; kalsea 'coarse, dreary, gloomy'. (13) BARK (SKIN): su2u, suwa2, (ol-)so [SH, PH]. F: suomu 'scale'; suikale 'strip'; sulppu 'wood

pulp'. (14) BEAT, kill, knock: peka, pak, pa:k, bok, buktu [SEM]. F: *piekse- 'beat, whip',

pako(-ttaa) 'to coerce, force, drive', tappaa 'kill'. (15) BEAUTIFUL, good, healthy, happy: kalli, kul-, sili, cal, cul [SEM]. F: *kaunii- 'beautiful';

*kallii- 'dear, beloved, expensive'. (16) BITE, eat: k'ust, kehca, kat5, ku2t, kasat [ON]. F: kesti 'feast, banquet', kuohi(ta) 'castrate'

(note: reindeer are castrated by biting off their testicles), kutiaa 'to itch', katko- 'to break off, tear into pieces'.

(17) BLACK, blue, bruised, gray, green: kat'i, kajtet, kasa:ht, k,ta, ku:ta, ka2ahs [SEM]. F: katve 'shade, shaded'.

607

LANGUAGE, VOLUME 64, NUMBER 3 (1988)

(18) BLOOD(1), red: pojts, misk, pos, posk, paccan, pajEikin. F: punaise- 'red'; paisu- 'swell'. (19) BLOOD (2): issis, es, 2its, icikci, as, Rate [SH, PH]. F: suoni 'blood vein'. (20) BLOOM, flower: puk, pux, pak, mbah. F: *puhkea- 'bloom'. (21) BODY, shell: wil, iwal [SEM; only two languages compared]. F: *vartalo 'body'. (22) BOIL(I), to cook: se 2t, soxt, sutusutu. F: sihise- 'to sizzle'. (23) BOIL(2): uk, khohkhoh [SH, PH; only two languages compared]. F: *kuoha(htaa) 'to boil';

kiehu- 'to boil'. (24) BOY, son: eppuce, -pec [SH, PH]. F: poju 'boy', *poika 'boy, son'. (25) BREAST(I): 2uCU, icic, cu [SH, ON]. F: utare 'udder' (cf. 26.) (26) BREAST(2), milk: mu, musu, mus, mas, mas-x [SEM, SH]. F: *ime(ttaa) 'to nurse, suckle,

imu 'sucking'; maito 'milk'. (27) BRING(l): U:i, u [SH; only two languages compared]. F: *vie- 'bring, take'; ottaa 'take,

bring' (cf. -ttaa 'transitive active/causitive'). (28) BRING(2), carry in hand, bring water: maha, muhi, mah [SEM]. (cf. 27, 29.) F: *muu(ttaa)

'carry off, remove'. (29) BRING(3), carry, take, bring people: ci?', ic'-el, i:x, eq [SH, PH, SEM(?)]. F: saa- 'bring,

get'. (30) BROAD: dapdape, taphe [only two languages are compared]. F: tovi: 'interval, while,

stretch of time, quite some time, continuously'; tappaii 'to fill, to stuff; tapo- (as in tapo-taynni, 'crammed, crammed full, tight, chock-full, overflowing'); typo- 'all, com- pletely, absolutely' (as in typo-tyhjd 'all empty, completely empty'); tyvi 'broad end, base'; tuppaa- 'to crowd, throng'.

(31) BROTHER (OLDER), sibling, younger brother: it5ik, it5i, 2eta, 2atta, axc, atsik, ha2t, -hta- [SH]. F: esi(koinen) 'first-born child'; seta 'uncle'; serkku 'cousin'.

(32) BURN, blaze, take fire, scorch, fire: hel, heh, xul, he/si, h/lt3, xul. F: *hiili 'embers, charcoal'; *karve(ntda) 'to scorch, singe'.

(33) BURY: muk, muq, mak, mukey, mos. F: *muokaa- 'to work ground, till'. (34) CHEEK, jaw, face, in front of: ila, al-, al, a/wi2 [SEM, SH]. F: leuka 'jaw'; ilme '(facial)

expression'. (35) CHEST, belly, heart: -tm-, tomna, ti mi:ne, tunan, tome, tu:n, tena, tempo [SEM]. F: *tuma

'nucleus'; *ytime- 'heart, core'. (36) CHILD(l), boy: ska, sak, sax. F: *sikio 'offspring, brat, the young'.

In short, Finnish offers strikingly similar matchings to many of G's proposed Penutian forms in this set and reasonable resemblances to nearly all. The Fin- nish matchings are all the more remarkable since G had a large number of languages from which to seek similarities (potential cognates), while all the similarities just cited come from a single language, Finnish. Given the large number of target languages, G could hardly miss finding what he was looking for.

Given that the etymologies proposed in LIA fare no better than resemblances from some non-Amerindian language that might be chosen at random, LIA's claims should be rejected. In fact, on G's approach one would be hard put to distinguish plausible relationships from more marginal proposals. How, for example, are the LIA groupings to be segregated from such marginal proposals as, for example, Huave-Uralic (Bouda 1964), Mayan-Altaic (particularly Turkic; Frankle 1984a, 1984b; Wikander 1967, 1970, 1970-1971), Carib-African languages (Kennedy 1856), Uto-Aztecan-Polynesian (and other implausible proposals; Key 1981a, 1981b, 1984), Miwok-Costanoan with Indo-European (Sadovsky 1984), and Huave-Algonquian (Suarez 1975)?

Also in this vein, it should be mentioned that LIA forms frequently show greater similarity to forms in other American Indian languages outside of the

608

REVIEW ARTICLE

groupings to which the languages bearing the forms have been assigned in the LIA classification. This leads us to question how the proposed 'etymologies' were determined and why certain groupings and not others were proposed. For example, G's cognates for HAND (table 9, p. 58; 137, p. 229) exhibit forms approximating maka phonetically, spread throughout his groupings; neverthe- less, other languages within each of these groups lack forms with k and/or with these vowels. Thus, the maka approximation is seen in Bororo mako in Macro- Ge, Amahuaca maka in Macro-Panoan, Quechua maki in Andean, Uto-Aztecan maka in Central Amerind, and Totonac makan in Penutian.8 However, forms lacking k or exhibiting other vowels (u is frequent) would seem to present unaccounted-for cross-group similarities, e.g. Botocudo um in Macro-Ge, Agua- runa umu in Equatorial, Auake ume in Macro-Tucanoan, Sinacatan Xinca mux in Chibchan-Paezan, and Achomawi -mu in Hokan. Notice that in casual observation G's Xinca form (mux) seems to pattern more closely with his Achomawi form (-mu), though the former is placed in Chibchan-Paezan while the latter is in Hokan. One such cross-group similarity would perhaps be of little significance. However, LIA has many. For example, in 11 ASHES (185), Xincan, with malh, mali, is placed in Chibchan-Paezan, whose other forms are bun, bura, puru, mono, poggo, muli, mukne, emukn, plua; and yet the Xincan forms are much more similar to G's Hokan forms, i.e. mal, mari-p'a, mwa:r, -mak. In these two examples, HAND and ASHES, Xincan appears to pattern more closely with G's Hokan than with the languages of G's Chibchan-Paezan group, where he placed it. G might insist that en masse the languages exhibit more similarities within the groupings than across them, but unfortunately, in many cases his data do not confirm such a claim.

LANGUAGE NAMES

4. The list of language names upon which G makes his subgroupings and classification contains many errors; I mention but a few. Some of the languages named are not languages at all. Membreno, putatively Lenca (194, 293, 382, 425), is not a language, but rather a person-a reference (Membreno 1897, not in LIA's bibliography) which contains several Lenca word lists. Also, there are only two known Lenca languages, not the six of LIA; Guajiquero, Intibucat [sic], Opatoro, and Similaton (382 and elsewhere) are merely names of towns where Honduran Lenca was spoken. Papantla (380 and elsewhere) is not a third Totonacan language, but simply one of the main towns where Totonac is spo- ken. Chiripo and Estrella, putatively Talamancan (382 and elsewhere), are town names where Cabecar is spoken. Viceyta, also listed as Talamancan (382 and elsewhere), is a colonial name applied to both Bribri and Cabecar; it is definitely not an independent language. Similarly, Terraba, Tiribi, and Tirub are not sep- arate languages: Terraba is the name given to the christianized Tiribi whom the Spanish brought from Panama to Costa Rica after 1700. Tirub is merely

8 G's 'guess ... that -ki is a very old Amerind noun-forming suffix' (59) does not save the day; it still remains to be explained why these languages-relatively isolated within the groupings to which they were assigned-might retain such a suffix (about whose possible existence G specu- lates), while all the others lost it.

609

LANGUAGE, VOLUME 64, NUMBER 3 (1988)

the language name some scholars have preferred, based on the native name [tifrub]. These constitute one language, not three. Moguex is the older name given to Guambiana, so these are not two separate languages (382 and else- where). Nonama and Waunama, given as Chocho members (382 and else- where), are alternant names for the same language.

Yupultepec (110, 184, 209, 241, 245, 261, 382, 435) should be Yupiltepeque; Borunca (382, etc.) should be either Boruca or Brunca-these are the two known names for this language. Tetontepec is really Totontepec (380 and else- where). One telling example, Corobisi (382), is merely the name given to the Rama speakers of Costa Rica; Rama and Corobisi are not distinct languages.9 Nevertheless, on the basis of only a single cited Corobisi form (111), G groups his Corobisi language with Guatuso, Cabecar, and Rama. Such examples sug- gest that caution should be exercised with respect to both the classification and the methods lying behind it.'0

CONCLUSION

5. In light of the considerations presented here, it seems necessary to con- clude that LIA is, indeed, an unfortunate book-it misleads nonspecialists and, for serious scholars, complicates the task of getting on with the business of working out American Indian linguistic relationships.

The similarities (so-called etymologies) assembled in LIA do not eliminate possible (and in some cases real) nongenetic explanations for the resemblances. Moreover, nearly every Americanist who has examined G's book finds shock- ing distortions in the LIA data from languages he or she knows well-e.g., 'Errors in the Algonquian data alone invalidate 93 [of G's 143] equations' (God- dard 1987:656; see also Chafe 1987). But G would claim insulation from prob- lems in the data: 'The fact is, the method of multilateral comparison is so powerful that it will give reliable results even with the poorest of materials' (29). G's method is 'powerful', but it is not reliable, as we have seen here. There is an enormous number of lexical (and grammatical) targets in the hundreds of American Indian languages. G's system of multilateral comparison and his very liberal interpretation of resemblances translate perceived similarity directly into 'etymologies'. Thus it is capable of generating vast numbers of intuited similarities that G takes as support for his proposed distant genetic relationships; but his method is not capable of distinguishing plausible proposals from highly fanciful ones (such as, for instance, Finnish-Amerind).

This being the case, the LIA classification and its attendant methods must be rejected. I can only hope that this review will constitute something of a

9 The fuller story is that Corobisi was the name of a language mentioned in Spanish chronicles from the sixteenth and the earlier part of the seventeenth centuries; not a single word of this language is known. Conzemius 1930 equated with Corobisi a word list from Upala-not the same as the area attributed to colonial Corobisi, but nearby-demonstrating that it was Rama. Whether in fact this represents the colonial Corobisi is unknown.

10 Some languages are also neglected in the classification (379-86). For example, of the 31 Mayan languages, only 18 are named; of 26 (some count many more) Otomanguean languages, only 11 are listed.

610

REVIEW ARTICLE

counterbalance to LIA and also a call for a more realistic orientation to the investigation of distant genetic relationship and to the task of working out the relationships among American Indian languages.

REFERENCES

ANDREWS, E. WYLLYS V. 1970. Correspondencias fonologicas entre el lenca y una lengua mayanse. Estudios de Cultura Maya 8.341-87.

BENDER, MARVIN L. 1969. Chance CVC correspondences in unrelated languages. Lg. 35.519-31.

BERGSLUND, KNUT. 1959. The Eskimo-Uralic hypothesis. Journal de la Societe Finno- Ougrienne 61.3-29.

BERMAN, HOWARD. 1983. Some California Penutian morphological elements. IJAL 49.400-12.

BONNERJEA, RENE. 1978. A comparison between Eskimo-Aleut and Uralo-Altaic de- monstrative elements, numerals, and other related semantic problems. IJAL 44.40- 55.

BOUDA, KARL. 1964. Huavestudien I und II: Uralisches im Huave. Etudes Finno-oug- riennes 1.18-27, 167-75.

BRIGHT, WILLIAM. 1976. The first Hokan conference: Conclusions. Hokan studies, ed. by Margaret Langdon and Shirley Silver, 361-3. The Hague: Mouton. . 1984. The classification of North American and Meso-American Indian languages. American Indian linguistics and literature, 3-30. Berlin: Mouton.

BRINTON, DANIEL G. 1869. [The nature of the Maya group of languages.] Proceedings of the American Philosophical Society 11.4-6.

BROWN, CECIL H., and STANLEY R. WITKOWSKI. 1979. Aspects of the phonological his- tory of Mayan-Zoquean. IJAL 45.34-47.

CALLAGHAN, CATHERINE A., and WICK R. MILLER. 1962. Swadesh's Macro-Mixtecan hypothesis and English. Southwest Journal of Anthropology 18.278-85.

CAMPBELL, LYLE. 1973. Distant genetic relationships and the Maya-Chipaya hypothesis. Anthropological Linguistics 15:3.113-35. . 1975. Constraints on sound change. The Nordic languages and modern linguistics, ed. by Karl-Hampus Dalstedt, 2.388-406. Stockholm: Almqvist & Wiksell International. . 1976. The linguistic prehistory of the southern Mesoamerican periphery. Fronteras de Mesoamerica. Mesa redonda 14.157-84. Mexico: Sociedad Mexicana de Antropologia. . 1977. Quichean linguistic prehistory. (University of California Publications in Linguistics, 81.) Berkeley & Los Angeles: University of California Press. . 1978a. Distant genetic relationship and diffusion: A Mesoamerican perspective. Proceedings of the International Congress of Americanists 52.595-605. Paris. . 1978b. Quichean prehistory: Linguistic contributions. Papers in Mayan linguistics, ed. by Nora C. England, 25-54. Columbia: University of Missouri.

--. 1980. Coahuiltecan. Encyclopedia of Indians of the Americas 4.265-7. St. Clair Shores, Michigan: Scholarly Press. . 1986. Cautions about loan words and sound correspondences. Linguistics across historical and geographical boundaries: In honor of Jacek Fisiak on the occasion of his fiftieth birthday, ed. by D. Kastovsky and A. Szwedek, 221-5. Berlin: Mouton. , and TERRENCE KAUFMAN. 1976. A handbook of the Xincan languages. University of Pittsburgh, MS. . 1981. On Mesoamerican linguistics. American Anthropologist 82.850-7.

--. 1983. Mesoamerican historical linguistics and distant genetic relationship: Getting it straight. American Anthropologist 85.362-72.

611

LANGUAGE, VOLUME 64, NUMBER 3 (1988)

CAMPBELL, LYLE, and MARIANNE MITHUN. 1979a. North American Indian historical lin- guistics in current perspective. In Campbell & Mithun 1979b, 3-69.

---, -- (eds.) 1979b. The languages of Native America: Historical and comparative assessment. Austin: University of Texas Press. [LNA]

CAMPBELL, LYLE, and DAVID OLTROGGE. 1980. Proto-Tol (Jicaque). IJAL 46.205-23. CHAFE, WALLACE. 1987. Review of Language in the Americas, by Joseph H. Greenberg.

Current Anthropology 28.652-3. COLLINDER, BJORN. 1947. La parente linguistique et le calcul des probabilites. Uppsala

universitets arskrift 1949.1-24. CONSTENLA, ADOLFO. 1981. Comparative Chibchan phonology. University of Pennsyl-

vania dissertation. Ann Arbor, MI: University Microfilms. CONZEMIUS, EDUARD. 1930. Une tribu inconnue de Costa-Rica: Les indiens Rama du

Rio Zapote. L'Anthropologie 40.93-108. DARNELL, REGNA. 1987. Review of Language in the Americas, by Joseph H. Greenberg.

Current Anthropology 28.653-6. --, and JOEL SHERZER. 1971. Areal linguistic studies in North America: A historical

perspective. IJAL 37.20-8. DIXON, ROLAND B., and ALFRED L. KROEBER. 1913a. Relationships of Indian languages

in California. Science, n.s. 37.225. . 1913b. New linguistic families in California. American Anthropologist 15.647-55.

DUPONCEAU, PETER STEPHEN. 1838. Memoire sur le systeme grammatical des langues de quelques nations indiennes de l'Amerique du Nord. Paris.

FLEMING, HAROLD. 1987. The Stanford conference: As seen by Hal Fleming. Mother Tongue: Newsletter of the Long Range Comparison Club 4.24. Boston University.

Fox, JAMES A. 1978. Proto-Mayan accent, morpheme structure conditions, and velar innovations. University of Chicago dissertation.

FRANKLE, ELEANOR. 1984a. Las relaciones externas entre las lenguas Mayances y al- taicas. Investigaciones recientes en el area maya. XVII Mesa Redonda Mexico: Sociedad Mexicana de Antropologia, 1.209-25. . 1984b. Los morfemas vocalicos para derivaciones verbales en los grupos mayance y turquico. Investigaciones recientes en el area maya. XVII Mesa Redonda Mexico: Sociedad Mexicana de Antropologia, 2.517-24.

FREELAND, LUCY S. 1930. The relationship of Mixe to the Penutian family. IJAL 6.28- 36.

GODDARD, IVES. 1975. Algonquian, Wiyot, and Yurok: Proving a distant genetic rela- tionship. Linguistics and anthropology in honor of C. F. Voegelin, ed. by M. Dale Kinkade, Kenneth L. Hale, and Oswald Werner, 249-62. Lisse: Peter de Ridder Press. . 1979. The languages of south Texas and the lower Rio Grande. In Campbell & Mithun 1979b, 355-89. . 1987. Review of Language in the Americas, by Joseph H. Greenberg. Current Anthropology 28.656-7.

GOLLA, VICTOR. 1980. Some Yokuts-Maidun comparisons. American Indian and Indo- european studies: Papers in honor of Madison S. Beeler, ed. by K. Klar, M. Langdon, and S. Silver, 57-65. The Hague: Mouton. (ed.) 1984. The Sapir-Kroeber correspondence. (Survey of California and other

Indian languages, 6.) Berkeley & Los Angeles: University of California Press. GREENBERG, JOSEPH H. 1948. The classification of African languages. American An-

thropologist 50.24-9. . 1955. Studies in African linguistic classification. Branford, Connecticut: The Lan- guage and Communication Research Center, Columbia University, and The Pro- gram of African Studies, Northwestern University. Reprinted from Southwestern Journal of Anthropology 5.79-100, 190-8, 309-17 (1949), 6.47-63, 143-60, 223-37 (1950), 10.405-15 (1954). . 1960. The general classification of Central and South American languages. Men

612

REVIEW ARTICLE

and cultures: Selected papers of the 5th International Congress of Anthropological and Ethnological Sciences, 1956, ed. by Anthony Wallace, 791-94. Philadelphia: University of Pennsylvania Press.

--. 1963. The languages of Africa. IJAL vol. 29, no. 1, Part II (Indiana University Research Center in Anthropology, Folklore, and Linguistics, Publication 25.) . 1979. The classification of American Indian languages. Papers of the 1978 Mid- America Linguistics Conference, ed. by Ralph E. Cooley et al., 7-22. Norman: University of Oklahoma Press.

-, and MORRIS SWADESH. 1953. Jicaque as a Hokan language. IJAL 19.216-22. GREENBERG, JOSEPH H.; CHRISTY G. TURNER II; and STEPHEN L. ZEGURA. 1985. Con-

vergence of evidence for the peopling of the Americas. Collegium Antropologicum 9.1:33-42.

--. 1986. The settlement of the Americas: A comparison of the linguistic, dental and genetic evidence. Current Anthropology 27.477-97.

HAAS, MARY R. 1958a. Algonkian-Ritwan: The end of a controversy. IJAL 24.159-73. --. 1958b. A new linguistic relationship in North America: Algonkian and the Gulf

languages. Southwest Journal of Anthropology 14.231-64. --. 1969. Grammar or lexicon? The American Indian side of the question from Du-

ponceau to Powell. IJAL 35.239-55. HAMP, ERIC P. 1967. On Maya-Chipayan. IJAL 33.74-6.

--. 1970. Maya-Chipaya and typology of labials. CLS 6.20-2. . 1971. On Mayan-Araucanian comparative phonology. IJAL 37.156-9.

HARRINGTON, JOHN P. 1913. [Announcement of relationship of Yuman and Chumashan.] American Anthropologist 15.716. . 1917. [Announcement of relationship of Washo and Chumashan.] American An- thropologist 19.154.

HOLT, DENNIS. 1976. La lengua paya y las fronteras linguisticas de Mesoamerica. XIV Mesa Redonda de la Sociedad Mexicana de Antropologia, 1.149-56. Mexico. . 1986. The development of the Paya sound-system. Los Angeles: UCLA dissertation.

HYMES, DELL H. 1964. Evidence for Penutian in lexical sets with initial *C- and *S-. IJAL 30.213-42.

JUSTESON, JOHN; WILLIAM NORMAN; LYLE CAMPBELL; and TERRENCE KAUFMAN. 1985. The foreign impact on lowland Mayan language and script. (Middle American Re- search Institute, Publication 33.) New Orleans: Tulane University.

KAUFMAN, TERRENCE. 1973. Areal linguistics and Middle America. Current Trends in linguistics, ed. by Thomas Sebeok, 11.459-83. The Hague: Mouton. . 1974a. Idiomas de Mesoamerica. (Seminario de Integraci6n Social Guatemalteca, pub. 33.) Guatemala: Editorial Jose de Pomada Ibarra. . 1974b. Meso-American Indian languages. Encyclopaedia Britannica, 15th edn., vol. 11.956-63.

KENNEDY, J. 1856. On the probable origin of the American Indians, with particular reference to that of the Caribs. Journal of the Ethnological Society 4.226-67.

KEY, MARY RITCHIE. 1978. Araucanian genetic relationships. IJAL 44.280-93. 1981a. Intercontinental linguistic connections. (Humanities Inaugural Lecture Se-

ries.) Irvine: University of California. . 1981b. North and South American linguistic connections. La Linguistique 17:1.3- 18. . 1984. Polynesian and American linguistic connections. (Edward Sapir monograph series in language, culture, and cognition, 12; supplement to Forum Linguisticum 8:3.) Lake Bluff, IL: Jupiter Press. , and CHRISTOS CLAIRIS. 1978. Fuegian and central South American language re- lationships. International Congress of Americanists 52:4.635-45. Paris.

KLEIN, HARRIET E. MANELIS, and LOUISA R. STARK. 1985. South American Indian lan- guages. Austin: University of Texas Press.

613

LANGUAGE, VOLUME 64, NUMBER 3 (1988)

KRAUSS, MICHAEL E. 1979. Na-Dene and Eskimo-Aleut. In Campbell & Mithun 1979b, 803-901.

LANGDON, MARGARET. 1979. Some thoughts on Hokan with particular reference to Po- moan and Yuman. In Campbell & Mithun 1979b, 592-649.

LEHMANN, WALTER. 1920. Zentral-Amerika. Berlin: Museum fuir Volkerkunde zu Berlin. LEVINE, ROBERT D. 1979. Haida and Na-Dene: A new look at the evidence. IJAL 45.157-

70. MATTESON, ESTHER. 1972. Toward Proto Amerindian. Comparative studies in Amer-

indian languages, ed. by Esther Matteson, 21-92. (Janua Linguarum, Series Prac- tica, 127.) The Hague: Mouton.

McQuowN, NORMAN. 1942. Una posible sintesis linguiiistica macro-mayence. Mayas y olmecas, 37-8, Segunda Mesa Redonda sobre problemas antropologicos de Mexico y Centro America. Mexico: Sociedad Mexicana de Antropologia.

-. 1956. Evidence for a synthetic trend in Totonacan. Lg. 32.78-80. MEMBRENO, ALBERTO. 1897. Hondureniismos: Vocabulario de los provincialismos de

Honduras. 2nd edn. Tegucigalpa: Tipografia Nacional. NEWMAN, STANLEY. 1964. Comparison of Zuni and California Penutian. IJAL 30.1-13. NICHOLS, JOHANNA. 1971. Diminutive consonantal symbolism in western North Amer-

ica. Lg. 47.826-48. OLSON, RONALD D. 1964. Mayan affinities with Chipaya of Bolivia, I: Correspondences.

IJAL 30.313-24. . 1965. Mayan affinities with Chipaya of Bolivia, II: Cognates. IJAL 31.29-38.

PARKER, GARY J. 1969. Comparative Quechua phonology and grammar IV: The evo- lution of Quechua A. University of Hawaii Working Papers in Linguistics 1:9.149- 204.

POWELL, JOHN WESLEY. 1880. Introduction to the study of Indian languages, with words, phrases, and sentences to be collected. 2nd edn. Washington: Bureau of Ethnology. . 1891. Indian linguistic families of America north of Mexico. Seventh annual report, Bureau of American Ethnology, 1-142. Washington, D.C.: Government Printing Office.

RADIN, PAUL. 1919. The genetic relationship of the North American Indian languages. University of California Publications in American Archaeology and Ethnology 14:4.489-502.

RANKIN, ROBERT L. 1981. Review of The Caddoan, Iroquoian, and Siouan languages, by Wallace Chafe. IJAL 47.172-7.

RUHLEN, MERRITT. 1987. A guide to the world's languages: Vol. 1: Classification. Stan- ford: Stanford University Press.

SADOVSKY, OTTO J. 1984. The discovery of California: breaking the silence of the Siberia- to-America migrators. The Californians 2:2.9-20 (Nov/Dec.).

SAPIR, EDWARD. 1920. The Hokan and Coahuiltecan languages. IJAL 1.280-90. . 1925. The Hokan affinity of Subtiaba of Nicaragua. American Anthropologist, n.s. 27.402-35, 491-527. . 1929. Central and North American languages. Encyclopaedia Britannica, 14th edn., vol. 5.138-41.

SAPIR, J. DAVID. 1987. Review of Language in the Americas, by Joseph H. Greenberg. Current Anthropology 28.663-4.

SAUVAGEOT, A. 1923. Eskimo et ouralaien. Journal de la Societe des Americanistes de Paris 15.279-316.

SHAFER, ROBERT. 1952. Athapaskan and Sino-Tibetan. IJAL 18.12-9. . 1957. A note on Athapaskan and Sino-Tibetan. IJAL 23.116-7.

SHERZER, JOEL. 1976. An areal-typological study of the American Indian languages north of Mexico. Amsterdam: North-Holland.

SHIPLEY, WILLIAM F. 1980. Penutian among the ruins: A personal assessment. Berkeley Linguistics Society 6.437-41.

SORENSEN, ARTHUR P., JR. 1973. South American Indian linguistics at the turn of the

614

REVIEW ARTICLE

seventies. Peoples and cultures of native South America, ed. by Daniel R. Gross, 312-41. New York: Doubleday.

STARK, LOUISA R. 1970. Mayan affinities with Araucanian. CLS 6.57-69. --. 1972. Maya-Yunga-Chipayan: A new linguistic alignment. IJAL 38.119-35.

SUAREZ, JORGE A. 1974. South American Indian languages. Encyclopaedia Britannica, 15th edn., vol. 17.105-12.

--. 1975. Estudios huaves. (Coleccion cientifica lingiistica, 22.) Mexico: Instituto Nacional de Antropologia e Historia.

--. 1979. Observaciones sobre la evolucion fonologica del tlapaneco. Anales de An- tropologia 16.371-86. Mexico. . 1983a. The Mesoamerican Indian languages. Cambridge: University Press.

--. 1983b. La lengua tlapaneca de Malinaltepec. Mexico: Universidad Nacional Au- tonoma de Mexico.

--. 1986. Elementos gramaticales otomangues en tlapaneco. Language in global per- spective: Papers in honor of the 50th anniversary of the Summer Institute of Lin- guistics, 1935-1985, ed. by Benjamin E. Elson, 267-84. Dallas: Summer Institute of Linguistics.

SWADESH, MORRIS. 1953. Mosan I: A problem of remote common origin. IJAL 29.26- 44. . 1954. Perspectives and problems of Amerindian comparative linguistics. Word 10.306-32. . 1959a. Mapas de clasificacion lingiistica de Mexico y las Americas. (Instituto de Historia, pub. 51; serie antropologica, 8.) Mexico: Universidad Nacional Autonoma de Mexico. . 1959b. The mesh principle in comparative linguistics. Anthropological Linguistics 1:2.7-14. . 1966. Porhe [Tarascan] y maya. Anales de antropologia 3.173-204. . 1967. Lexicostatistic classification. Handbook of Middle American Indians, ed. by Norman McQuown, 5.79-115. Austin: University of Texas Press.

TEETER, KARL. 1964. Algonquian languages and genetic relationship. Proceedings of the 9th International Congress of Linguists, 1026-33. The Hague: Mouton.

THALBITZER, W. 1928. Is there any connection between the Eskimo language and the Uralian? Atti del XXII Congresso Internationale degli Americanisti, 551-67. . 1944. Uhlenbeck's Eskimo-Indo-European hypothesis: A critical review. Etudes linguistiques 1944.66-96.

TYLOR, EDWARD B. 1871. Primitive culture: Researches into the development of my- thology, philosophy, religion, language, art, and custom. London: Murray.

UHLENBECK, C. C. 1905. Uralische Anklange in den Eskimosprachen. Zeitschrift der Deutschen Morgenlandischen Gesellschaft 1905.757-65. . 1942-5. Ur- und altindogermanische Anklange im Wortschatz des Eskimos. An- thropos 37-40.133-48.

WHISTLER, KENNETH W. 1977. Wintun prehistory: An interpretation based on linguistic reconstruction of plant and animal nomenclature. Berkeley Linguistics Society 3.157-74.

WIKANDER, STIG. 1967. Maya and Altaic: Is the Maya group of languages related to the Altaic family? Ethnos 32.141-8.

--. 1970. Maya and Altaic II. Ethnos 35.80-88. 1970-71. Maya and Altaic III. Orientalia Suecana 19-20.186-204.

WITKOWSKI, STANLEY R., and CECIL H. BROWN. 1978. Mesoamerican: A proposed lan- guage phylum. American Anthropologist 80.942-4. . 1981. Mesoamerican historical linguistics and distant genetic relationship. Amer- ican Anthropologist 83.905-11.

Department of Anthropology [Received 15 October 1987; SUNY at Albany revision received 13 May 1988.] Albany, NY 12222

615