campbell&kaufman&smith.1986.mesoamerica as linguistic area

7/30/2019 Campbell&Kaufman&Smith.1986.MesoAmerica as Linguistic Area

1/42

STOReso-America as a Linguistic AreaLyle Campbell; Terrence Kaufm an; Thomas C. Smith-StarkLanguage, Vol. 62, N o. 3. (Sep., 1986), pp. 530-570.

Stable URL:http://links.j stor.org/sici ?sici=0097-8507%28198609%2962%3A3%3C530%3AMAA LA%3E2.0.00%3B 2-XLanguage is currently published by Linguistic Society of America.

Your use of the JSTOR archive indicates your acceptance of JSTOR' s Terms and Conditions of Use, available athttp://www.jstor.org/about/terms.html . JSTOR' s Terms and Conditions of Use provides, in part, that unless youhave obtained prior permission, you may no t down load an entire issue of a journal or multiple copies of articles, andyou 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. P ublisher contact information may be obtained athttp://www.jstor.org/journals/lsa.html.Each copy of any p art of a JSTOR transmission must contain the same copyright notice that appears on the screen orprinted page of such transmission.

JSTOR is an independent not-for-profit organization dedicated to creating and preserving a digital archive ofscholarly journals. For more information regarding JSTOR, please contact [email protected] .

http://www.jstor.org/Mon May 15 17:16:58 2006


2/42

MESO -AMERICA AS A LINGUISTIC AREALYLE CAMPBELL TERRENCE KAUFMAN THOMAS C. SMITH-STARKSUNY Albanyniversity ofolegio de MexicoPittsburghThat M eso-America constitutes a legitimate l inguistic area h as been questioned. Toaddress this question, con cepts of àreal linguistics' are here survey ed and refined. Pro-posed Meso-American areal traits are reconsidered against these findings, and are com-pared with those of other established linguistic areas. Meso-America proves to be aparticularly strong linguistic area. These results contribute both to the study of Meso-American languages and to an understanding of areal linguistics generally.*In recent years it has been proposed that Meso-America (henceforth MA)

defined basically as a culture area extending from central Mexico through north-ern Central Americais a linguistic area. The first attempts at characterizingthe area were made by Hasler 1959, Kaufman 1973, 1974a,b, Campbell 1971,1977, 1978a, 1979, and Campbell & Kaufman 1980, 1983 (see also Bright 1984,Rosenthal 1981); nevertheless, doubts have been expressed (cf. Hamp 1979,Holt & Bright 1976, Surez 1983a). Some have thought that MA is not a single,well-defined area in the sense of others recognized in the literature, such asthe Balkans or South Asia, but rather may be composed of several smaller,regionally defined areas (cf. Hamp 1979). For that reason, our primary purposehere is to investigate MA in detail from an areal viewpoint. However, to de-termine MA's status requires us first to clarify the nature and definition oflinguistic areas in general. We will do this in 1-3, and then return to thecharacteristics of MA in 4.

1. DEFIN ITION OF AREAL LING UISTICS. As broadly conceived, AL deals withthe results of the diffusion of structural features across linguistic boundaries.As commonly viewed, linguistic areas are characterized by a number of lin-guistic features shared by various languagessome of which are unrelated, orare from different subgroups within a familyin a geographically contiguousarea. The phenomena of the linguistic area are also referred to at times by theterms `convergence area', `Sprachbund', àffinit linguistique', `diffusion area',àdstratum' etc.' However, when it comes to more precise definitions, thereis considerable controversy concerning just what AL is.

* We w ish to thank William Bright and Sarah G. Th omason for helpful com ments on earlierversions of this paper; but we do not m ean to imp ly that they are necessarily in agreement w ithour use of their statements. We also acknowledge the Instituto de Investigaciones Filolgicas,Un iversidad N acional Autnoma de M xico, for providing Lyle Campbell the opportunity to engagein full-time research in 1981-82, during which time h e did the research for this paper and w roteup a preliminary version. Terrence Kau fman and T hom as Smith-Stark have evaluated the originalmanu script and mad e various additions to it. Thus the list of authors reflects not only alphabeticalorder, but relative input to the final product. We three authors are in essential agreement concerningthe arguments and conclusions.Areal phenomena are, to a greater or lesser degree, related to such other areas of study asmultilingualism, substrata, superstrata, linguistic geography, bo rrowing, and language shift or m ain-530


3/42

MESO-AMERICA AS A LINGUISTIC AREA31Many attribute the formal birth of AL to Trubetzkoy's famous proposition

16, presented at the First International Congress of Linguistics (1928:17-18):2`Gruppen, bestehend aus Sprachen, die eine grosse Ahnlichkeit in syntaktischer Hinsicht, eineAhnlichkeit in den Grundsdtzen des morphologischen Baus aufweisen, und eine grosse Anzahlgemeinsamer Kulturwiirter bieten, manchmal auch dusser Ahnlichkeit im Bestande der Laut-systemedabei aber keine gemeinsamen Elementarwiirter besitzenSOLCHE SPRACHGRUPPENNENNEN WIR SPRACHBUNDE: (emphasis in original)

Trubetzkoy's term Sprachbund, roughly a 'union of languages', came to beused as a technical term in English. The name 'linguistic area' (LA), as atranslation of Sprachbund, was first employed by Velten 1943, and was madewell known by Emeneau 1956. Trubetzkoy (1931:233-4) compared AL to tra-ditional dialect geography, but with Isoglosses' which extend beyond theboundaries of a single language. This view of LA's as akin to the featurescharacterizing cross-language linguistic geography is common in later literature(cf. Jakobson 1931, 1938).tenance; in this paper, however, attention is restricted to AL. We can thus, we believe, attain ourgoals of examining the results of language contact which create a linguistic area, without gettinglost in the details of the mechanisms which produced these results.

In this context, certain relevant but less important studies may be mentioned. Some use the term'areal linguistics' or 'area linguistics' in the sense of dialect geography within a single language(cf. Kurath 1972, Goossens 1973); this usage is not relevant to the concerns of this paper. Again,the Italian 'neolinguistic' school has declared itself a champion of AL; however, so far as we candetermine, it has contributed nothing new to the concept. Their dedication to linguistic geographyseems motivated by an extremist reaction against neogrammarian sound laws, where they followSchuchardt (see Spitzer 1922) and Gilliron (see Gilliron & Rogues 1912). See Bertoni 1911, 1923,1925, Brtoli 1925, 1928, 1929, 1933a,b, 1939, and Bonfante 1945, 1947; cf. also Hall 1946.

2We do not wish to suggest that the study of areal phenomena began with Trubetzkoy. Arealconsiderations have been with us almost from the beginning of the formal study of language (cf.Kopitar 1857, who made some Balkan areal features known; see also Miklosich 1861, Sandfeld1930). Schuchardt championed the views that there are no rigid geographical boundaries betweenlanguages, and that language mixture occurs; he antedated Schmidt 1872 as the author of the 'wavetheory' (Wellen-hypothese; see Spitzer 1928:165, 431).

Developments in the study of American Indian languages also played an important role. Powellbecame so caught up in diffusionism that it threatened to prevent him from thinking in geneticterms, causing him to doubt his own classification (cf. Powell [1891] 1966:216-17). His doubtsabout our ability to separate the effects of genetic inheritance from diffusion are echoed in workby Boas, by Dixon & Kroeber, and to some extent by Sapir. While Boas and Sapir apparentlyheld similar views about borrowed structural features and genetic relationships at one time, theybecame quite polarized toward 1920, in what is now a famous dispute (Darnell & Sherzer 1971).Boas was skeptical about the possibility of distinguishing shared similarities stemming from struc-tural diffusion from those resulting from genetic relationship; while Sapir, by contrast, came tobelieve that the effects of borrowing, particularly in morphology, would never be profound. Sapir'sviews prevailed in America, and subsequent study of American Indian languages was characterizedby much reductionism in the number of postulated genetic units. Boas' reservations were almostfully forgotten in America, but received attention in Europe, where they markedly influenced thePrague School (see Trubetzkoy 1931, Jakobson 1938, 1944); from there, interest returned to America(Emeneau [1956] 1980:107).

An important part of the American background is the areal/typological approach characteristicof Dixon & Kroeber 1903, 1913practiced early also by Boas and Sapirin which the descriptionsof individual languages included typological comparisons with other languages of a geographicalregion (cf. Kroeber 1960:17-18; Darnell & Sherzer 1971).


4/42

532ANGUAGE, VOLUME 62, NUMBER 3 (1986)An influential recent definition, especially pertaining to studies of nativeAmerican languages, is Sherzer's (1973:760):`A LINGUISTIC AREA is defined here as an area in which SEVERAL linguistic traits are shared bythe languages of the area and [in which] furthermore, there is evidence (linguistic and non-linguistic) that contact between speakers of the languages contributed to the spread and/orretention of these traits and thereby to a certain degree of linguistic uniformity within the area.It is important to remember that languages which are unrelated or distantly related may verywell and probably do disagree with regard to many traits and yet still [be] in the same linguisticarea according to the above definition, since they share SEVERAL traits (which one might wantto call diagnostic traits).'

Issues over which opinion has been d ivided include the num ber of isoglossesrequired to define a LA, and whether they must bundleor whether a singleisogloss is sufficient. For some scholars, isogloss-bundling is diagnostic (cf.Trubetzkoy 1931:234; Sherzer 1973:760; Masica 1976:179; Katz 1975:12, 16;Holt & Bright 1976; Emeneau 1980:2). Nevertheless, few have insisted on thebundling criterion. Many view AL as akin to traditional dialectology, whereisoglosses frequently fail to group at the borders, but are found more concen-trated around som e core w ithout abrupt bound ariesand where the extensionof individual isoglosses outward from the core may vary greatly (cf. Emeneau[1965b1 1980:128, 136; Ram anujan & Masica 1969:55 0 ; Winter 1973:140; Masica1976:6, 170-71, 179; Henderson 1965:431; Joseph 1983). Moreover, followingthe analogy of dialectology, some linguists have noted that isoglosses of oneLA m ay cross those of another (cf. Jakobson 1944:193 , Becker 1948:23 , Wein-reich 195 8:378-9, Winter 1973 :140, and Haarm ann 1976:24).Given that it is difficult to find isogloss-bundling at the boundaries of LA's,som e scholars have defined a LA as any group of neighboring languages whichshare any diffused or convergent structural features, even a single trait. ThusMasica (172) views a single areal isogloss as the minimum defining feature:`Linguistic areas are apparently phenomena of differing magnitudes, startingfrom the limiting case, the area defined by a single trait' (cf. Bright & Sherzer1976:236, Trubetzkoy 193 1:345 , Jakobso n 1931:139). Katz makes a single, syn-chronic isogloss the basis of his definition (16):

`Von einem Sprachbund kann man sprechen, wenn:(a ) zu einer gegebenen Zeit(b ) ein zusammenhngendes geographisches Gebiet, das(c ) von mindestens einer Sprachgrenze durchzogen ist,(d ) von mindestens einer Isoglosse umspannt wird.'That is, no sharp boundary can be drawn between LA's which share a singlediffused trait and those which share many. One might attempt to justify thispoint of view by considering such analogies as: How m any grains of sand d oesit take to make a heap? How many birds are needed to constitute a flock? orHow many students are required to make a class?The co nclusion that a LA might adequately be defined on the basis of a singleshared feature is disputed by many scholars. For them, a LA defined by asingle isogloss would be trivial. Furthermore, there is a widely-shared feelingthat particular LA's share several traits. In principle, there is no meaningful


5/42

MESO-AMERICA AS A LINGUISTIC AREA33way to distinguish LA's defined on the basis of several features from thosebased o n a single shared trait, if the latter are considered non -trivial. Thus thequestion should be posed not in the form, D oes some entity qualify as a LA ?,but rather, How strong o r weak is a particular LA? T hat is, w e can think of acontinuum o f LA's from those w eakly defined, on the basis of a single sharedfeature, to much stronger areas based on many diffused elements. This ap-proach to defining LA 's also imp lies a m eans of eva luating their strength, towhich we now turn.2. EV ALU ATIO N . O ne strategy for improving the definition of LA's has beento propose criteria of evaluation. For some scholars, this amounts to merecounting of the n um ber of shared traits. For others, it involves ranking themin some evaluative scale according to the varying social, cultural, or historicalcircumstances w hich gave rise to the areas. It is w orth looking into som e ofthese in order to understand A L better.O ne exp licitly stated app roach to evaluating the strength of LA's is that ofK atz (p. 16) wh ich is follow ed (at least implicitly) by several others:

`Es ist zwar Mar, dass durch "near-universals" konstituierte Sprachbnde als solche nichtsehr interressant sind, dieser Man gel lsst sich aber ausgleichen, wen n w ir eine "Wertskala"aufstellen, die besagen soil: Ein Sprachge beit, das von m ehr Sprachbund isoglossen umschlos-sen ist, ist auf dieser Skala hher zu bewerten als eines innerhalb von weniger solchenIsoglossen.'This am ounts to a `mo re-the-merrier' proposal, wh ere the existence of m oreisoglosses is taken as m ore highly valued.Attempts to establish means of evaluating LA's m ake us realize that differenttypes of LA's indeed exist. These differences depend on the circumstancesthat gave rise to them and contributed to their developm ent, as indicated byThomason & Kaufman (1975:27):

`The various areas so identifiable, how ever, are not of a uniform type. In som e of the areasin question, there is current institutionalized m ultilingualism, either m ultilateral or un ilateral.In others there has been m assive shift in the past, with, how ever, some speak ers of the lan-guages shifted-from stil l around. In stil l others there has been gradual diffusion of featuresover long centuries, without high degrees of multilingualism or massive shift. It might beprofitable to try to separate these types, but at present we have n o foolproof m ethod of do ings o .

(Cf. Masica 1976:173, M artinet 195 2:123 , 1956 .) If distinct types of L A's canhave such varied historical backgroun ds, then it follows that they m ay differgreatly in their com position and character and in the w ay that their sharedtraits are interrelated, both w ithin individual languages and across languages.This brings us to the second m ajor approach to defining LA's, which requireshistorical evidence.3 . THE HISTO RICIST APPRO ACH. Perhaps the most important evaluative at-tempts have been based on a realization of the different historical factors whichgo into the creation of L A's. Masica (173) treats as significant the d istinctionbetween LA's wh ich are the relics of past contacts, no longer active, and othersw hich are in the process of formation and extension because of on-going in-


6/42

534ANGUAGE, VOLUME 62, NUMBER 3 (1986)teraction and change. Nevertheless, linguists have been divided in their opin-ions about the need for, or value of, historical facts in areal investigations. Onegroup's approach has been merely to catalog the similarities found in a partic-ular areaallowing these similarities to suggest diffusion, but without carryingout the research necessary to demonstrate the actual borrowing. This is bas-ically a reliance on 'circumstantial evidence'. This `circumstantialisr approach,as we will call it, can be usefulparticularly in the preliminary stages of in-vestigation, or in LA's where reliable historical facts are difficult to obtain.Even so, one would like to be able ultimately to separate real areal featuresthose resulting from diffusionfrom historical accidents, which may resultfrom undiscovered genetic relationships, universals, onomatopoeia, parallel orindependent development, sheer chance, etc. Unfortunately, many circum-stantialists have made no attempt to carry out the historical program (cf. Haas1969, 1976, 1978; Sherzer 1973 , 1976, Sherzer & Bauman 1972; Campbell 1977,1978a; Trubetzkoy 1928, 1931; Bright & Sherzer 1976; Holt & Bright 1976;Kaufman 1973).

The sharpest criticism of this type of AL concerns the selection of featuresto be considered areal. Since nearly everyone considers LA's to be the productsof diffusion, features designated as evidence for a LA should result from bor-rowing, stemming from mutual influence. As already seen, some hold that aLA may be defined by any similarity that happens to be shared among con-tiguous languages. However, if the selected shared features can be explainedequally well by accident, universals, genetic factors etc., then such a LA makesno sense as the product of diffusion; it begins to seem like a mere linguistictypology, which might involve adjacent languages having no relationship.The other main group of arealists link their definition, or at least their re-search practice, more directly to historical proofs, maintaining that featuresdesignated as areal should be demonstrably diffused. We call this the 'historical'approach, followed by the 'historicists'. It is instructive to see why these schol-ars insist on historical evidence, and to consider their criticisms of thecircumstantialists.

In this respect, Jacobsen (1980:205) , speaking of Sherzer 1976, calls for thehistorical program:'The obvious way of making further progress in these matters ... is to go beyond a merecataloging of the presence or absence of a category in a language to a study of the actual meansused for its expression and to a reliance upon the find ings of historical linguistics as appliedto the several languages and families.'

Ham p 1977 also sharply criticizes Sherzer 1976 and the circumstantialist ap-proach. His comments on the relation of AL to genetic classification are directlyto the point:

`... while the comparative method is unquestionably an historical study, the field of areallinguistics is no less so; for it too is occupied with analysing the resu lt of specific, if mu ltiple,linguistic events of the past. Both the comparative method and areal linguistics are historicaldisciplinestwin faces of diachronic linguistics, if you will.' (279)`[Sherzer's] metho dology seem s to make far too little provision for these distinctions [AL andcomparative linguistics] that I consider essential ... his study wou ld lie properly w ithin therealm of typology ... for areal, i.e. ultimately specific historical, questions it m ay be damaging


7/42

M E S O -AMERICA AS A LINGUISTIC AREA3 5in two main ways. The conclusions may result in a listing of a catalog of trivia; and the startingparameters may well have missed the most interesting and crucially tell-tale characteristics.'(281)`Such areal questions can be approached meaningfully and fruitfully only if they are treatedin specific terms for what they arethe results of developments with historical depth andspecificity.' (282)

(Cf. Winter 1973:147, and Silverstein 1978 for similar criticisms.) We conclude,thenwith Bright 1976, Hamp 1977, and Winter 1973that our goal shouldbe to determine the historical facts which explain similarities among languages,regardless of whether they result from common heritage from some proto-language, or from diffusion.

Several other considerations, which we do not take up in detail in the interestof saving space, are also important in evaluating the strength of different LA's.These deal with the weight of individual areal isoglosses; all borrowings arenot equal. Some should count more, given the relative difficulty of borrowing,their degree of integration into the borrowing grammar, etc.

First, highly `marked', exotic, or unique shared traits weigh more than doesmaterial that is more easily developed independently, or found widely in otherlanguages. Nevertheless, such exotic borrowings tend to reside at a more su-perficial position in linguistic structure. Given what is known of linguistic uni-versals, it would be unnatural to expect truly unique or very bizarre borrowedtraits to be found deep inside a grammar. Thus we suspect that, the more deeplyintegrated or interwoven into the basic fabric of a language a diffused featureis, the greater its areal valueand even greater to the extent that it is bothintegrated and marked.

Second, it has at times been claimed that a language can borrow featuresonly when it already has a pre-existing model (cf. Joseph, 205). Related, butin an opposing vein, is the notion that the grammars of certain languages havegaps which somehow reduce their efficiency: when they come into contact withlanguages exhibiting the useful but missing constructions, they incorporatethem,handilyrecognizing what they have been missing, and requiring onlyminimal foreign stimulation to acquire them (cf. Campbell & Mithun 1981). Wedo not argue for either of these claims. However, since such borrowings wouldbe easier to achieve, they should count for less. Such arguments do, in anycase, point out the importance of considering not only what features are bor-rowed, but also whether the borrowing languages are disposed to be receptive.

The integration of diffused traits into borrowing grammars has been called`installation' (Jacobsen 1980) and `naturalization' (Gair 1980). It obviously takestime or intensive contact to install or naturalize foreign constructions; thus wemight feel inclined to take them as strong areal indicators. However, it is oftenthe case that such features are initially borrowed on a more superficial level;the internal forces of the grammar then `snowball' (Joseph, 202) to achieve theintegration, regardless of whether areal influences continue.

Third, the more-the-merrier' principle for evaluating LA's obviously needsrefinement, given that the weight of individual traits depends on difficulty ofborrowing. There is, however, another way in which it fails: Some aspects ofgrammar are universally linked, such that the presence of one may imply the


8/42

536ANGUAGE, VOLUME 62, NUMBER 3 (1986)presence of others. In such cases, should it prove universally impossible orhighly unlikely for a language to have one grammatical featurewithout, atthe same time, exhibiting the others demanded by its presencethen it is hardlyvalid to count phenomena that cannot occur independently as multiple iso-glosses. Such considerations deserve attention; however, passing over themnow will not detract from our general exposition of AL or of MA.

From this discussion of AL, we may conclude, first, that most investigatorsview it as akin to traditional dialectology, but with isoglosses that extend be-yond language boundaries. Second, LA's are entities of differing magnitudes,ranging from the controversial limiting casedefined by a single shared traitto clearly stronger instances, with many shared features resulting from diffu-sion. Such features are of different weights, depending on how 'marked' theyare, and how integrated into the respective grammars. Third, since areal iso-glosses frequently do not bundle (and often overlap), it is of little use to attemptto define LA's based on the coincidence of structural traits at some boundary.Nevertheless, when bundling does occur, it can be taken as strong support forthe LA involved. Finally, since it is generally agreed that meaningful LA's arethe historical product of linguistic diffusion, the stronger LA's are those whoseshared traits can be shown historically to be diffusedand cannot be ascribedto a common ancestor, to chance, or to universals.4. MA AREAL TRAITS. Studies in the Meso-American LA to date have mostly

been circumstantialist in orientation. This is understandable, given that arealstudies in American Indian linguistics have typically been circumstantialist bothbecause of tradition (cf. Darnell & Sherzer) and because of the frequent lackof adequate historical evidence for diffusion of traits. In what follows, wepresent a compilation of circumstantialist traits that have been proposed asareal features of MA, together with an evaluation of their usefulness for defininga LA. The list comes from traits proposed by Kaufman 1973, 1974a,b, Campbell1971, 1977, 1978a, 1979, and Smith-Stark 1982, plus others that have come toour attention. In this evaluation, neighboring languages to the south and to thenorth are used as control cases for checking the areal nature of alleged MAtraits. All relevant languages are shown on Map 1 (pp. 538-9); the accompany-ing Key (pp. 540-42) indicates genetic affiliations and bibliographical sources.3

3 Som e information and exam ples not otherwise referenced are from C ampbell 's records andfield notes; these include Tepehua, Cakchiquel, Quich, Kekch, Nahuatl , Pipil, Xincan, Lenca,Jicaque, Cacaopera, Quechua, and Aymara. Mayan data not otherw ise referenced are from Kauf-man's files.It should be noted that many of the d ictionaries which we cite as sources also contain gramm aticalsketches. For general information on the classification and structure of M A languages, see C ampbell1979; Kaufman 1974a,b; Campbell & Kaufman 1980, 1983.O ur control languages are, of course, not like the control cases of, e.g., biologists; conceivably,an MA areal isogloss could lap over into some of our control languages. How ever, in this instance,several of our controls have no contact with MA (e.g. Bribri, Guaym, Tonkawa, Yuman). Inprinciple, languages found outside an area could once have been inside it, or could otherwise havebeen influenced through transported contact. How ever, in this case, geographical features makeit highly unlikely that significant contact could have occurred. Thus these languages do indeedprovide legitimate controls.


9/42

MESO-AMERICA AS A LINGUISTIC AREA37Since the status of MA as a LA is at issue, we will begin our evaluation ofareal traits by taking the strongest critical stance. That is, we will consider asconvincing only features which are widely distributed within MA; we will elimi-nate traits which are also found beyond MA, or are limited in their distributionto smaller zones within MA. We also discount traits which may easily developindependently in language. Thus we emphasize features peculiar to MA, andgeneral throughout it, but not beyond. We hasten to point out that this is infact not a very realistic view of areal traitsor of their distribution, and mutualinteraction one with anotherbut is taken here only to demonstrate MA as astrong LA, able to survive the most stringent scrutiny. In fact, all these op-erational constraints would tend to disqualify most real LA's. Thus, thoughunique or highly-marked features are especially persuasive, one cannot expecta LA to exhibit any abundance of exotic structural traits unknown elsewherein the world. Similarly, changes that can easily take place independently,through parallel evolution, can also be triggered by the stimulation of languagecontact. Moreover, areal isoglosses often fail to bundle, and show crisscrossingpatternssome beyond the LA's borders, others restricted within the LA.LA's can be the product of such varied local borrowings, with different patternsof distribution. For this reason, it is only in rare and fortunate cases that wemay expect areal features to meet the constraints employed here. To the extentthat MA succeeds in displaying areal traits which meet these strong conditions,there can be little doubt about its validity as a LA. After this strict scrutiny,we will reconsider proposed areal traits of MA which do not meet these con-ditions, but nevertheless lend secondary support to defining the LA.4.1. PH ON OLO GICAL TRAITS. The following are some shared phonological

phenomena in MA:(a ) FINAL DEVOICING OF SONORANTS. A rule which devoices final sonorants(usually 1 r y w, but also nasals and even vowels in some languages), is foundin several Mayan languages (especially Quichean), in Nahuatl, Pipil, Xincan,Totonac, Tepehua, Tarascan, and Sierra Popoluca (e.g. Nahuatl /no-mi:1/[nomi:l] `my field', Quich /axa:w/ [axa:W] lord')as well as in the moresoutherly Cacaopera and Sumu. While this rule seems to be borrowed in atleast some of the MA languages which contain it, its distribution is quite re-stricted within MA; it reaches only a small and discontinuous portion of thelanguages of the area. Moreover, final devoicing of sonorants is not so peculiarphonetically that it could not have happened independently in these languages,though it is sufficiently uncommon to suggest possible diffusion. All these con-siderations make final devoicing of sonorants relatively unhelpful in definingMA as a LA.(b ) VOICING OF OB STRUENTS AF TER NASAL S. This is found in Xincan, Huave,several Mixe-Zoquean languages, Tarascan, and most Otomanguean languages,including Tlapanec-Subtiabaas well as in Lenca and Jicaque; e.g. Guaza-capan Xinca /?ampuki/ [?ambki] `snake', Copainal Zoque /n-tik/ [ndik] `myhouse' (Harrison & Garca 1981:405). This feature has at least two problemswhich prevent it from being a diagnostic trait of the LA. First, its distribution


10/42

....... ........................... .....AP3.............

O hAP I

538ANGUAGE, VOLUM E 62, NUMBE R 3(1986)


11/42

MES O-AMERICA AS A LINGUISTIC AREA39N


12/42

540ANGUAGE, VOLUME 62, NUMBER 3 (1986)KEY (t = extinct)O T O M A N G U E A NOtopamean01 Chichimeco Jonaz (outside Meso-America): De Angulo 1933, Lastra 1984.02 Pame (some varieties outside Meso-America)03 Otom: Echegoyen et al. 1979.04 Mazahua05 Matlazinca06 Ocuilteco (Tlahuica)Popolocan07 Popoloca08 Ixcatec09 Chocho: Mock 1977.010 Mazatec: Jamieson 1978.Mixtecan011 Mixtec: Alexander 1980, Bradley 1970, Daly 1973, 1977, Pensinger 1974.012 Cuicatec013 Trique: Good 1978, Hollenbach 1975.Chinantecan014 Chinantec (several varieties): Merrifield 1968, Robbins 1968, Rupp 1980.Amuzgo015 AmuzgoZapotecan016 Zapotec (several distinct languages): Butler 1980, Pickett 1974.017 Chatino: Pride & Pride 1970.Chiapanec-Mangue018 tChiapanec019 tMangueTlapanec-Subtiaba020 Tlapanec: Schultze 1938, Surez 1983b.021 tSubtiabaAZTE CAN (= Nahuan, of the Uto-Aztecan family)Al Nahuat(l)A2 tPochutecA3 Pipil (includes Nicarao)T O T O N A C A NT1 Totonac: Reid & Bishop 1974.

T2 TepehuaM I X E - Z O Q U E A NMixeanMZ1 Veracruz Mixe (includes Sayula Popoluca, Oluta Popoluca): Clark 1981, Clark & Clark1960.MZ2 Oaxaca Mixe: Lyon 1980, Van Haitsma 1976, Schoenhals 1965.MZ3 TapachultecZoqueanMZ4 Veracruz Zoque (includes Sierra Popoluca, Texistepec Popoluca): Elson 1960, Foster &Foster 1948; Clark & Nordell 1984.MZ5 Chiapas Zoque (Copainal)MZ6 Oaxaca ZoqueMAYANHuastecanM1 HuastecM2 tChicomuceltecYucatecanM3 Yucatec (Maya): Tozzer 1921.M4 Lacandn


13/42

MESO-AMERICA AS A LINGUISTIC AREA41M5 ItzaM6 MopnCholan-Tzeltalan (Greater Tzeltalan)

ChoIanM7 Chol: Aulie 1978, Warkentin & Scott 1980.M8 Chontal (of Tabasco)M9 tCholtM10 Chort

TzeltalanMI1 Tzeltal: Kaufman 1971.M12 Tzotzil

KanjobalanKanjobalan proper

M13 KanjobalM14 JacaltecMI5 Acatec

ChujeanM16 ChujM17 TojolabalMotozintlec

M18 MotozintlecMamean-Quichean (Eastern Mayan)Mamean

M19 Mam: England 1983.M20 TecoM21 AguacatecM22 IxilGreater QuicheanM23 KekchM24 UspantecM25 Pokom (Pokomch, Pokomam)

Quichean properM26 SipacapeoM27 SacapultecM28 QuichM29 CakchiquelM30 Tzutujil

UNAFFILIATED1Tarascan: Foster 1969, Friedrich 1971.2 tCuitlatec: Escalante 1962.3 Tequistlatec (Chontal of Oaxaca): Turner 1971, Waterhouse 1980.4 Huave: Stairs 1981.5 Xincan (four languages: tYupilitepeque, Guazacapn, Chiquimulilla, and Jumaytepeque)

NON-MESO-AMERICAN LANGUAGESYUMAN: Langdon 1970, 1976, Winter 1976.Y1 Dieguerio (Ipai)Y2 Kiliwa

Y3 PaipaiY4 TipaiY5 Yuma (Quechan)Y6 CochimiY7 CocopaY8 MaricopaY9 Yavapai

APACHEAN (Athabaskan family)API Chiricahua Apache


14/42

542ANGUAGE, VOLUME 62, NUMBER 3 (1986)AP2 Mescalero ApacheAP3 Jicarilla ApacheAP4 NavajoAP5 Western ApacheAP6 Kiowa ApacheAP7 Lipan ApacheUTO -AZTECAN: Langacker 1977.SouthernPimanU 1 PapagoU2 Pima (Upper and Lower)U3 Northern Tepehuan: Bascom 1982.U4 Southern TepehuanTaracahiticUS Yaqui: Lindenfeld 1973.U6 MayoU7 TarahumaraU8 GuarihoCora-HuicholU9 Cora: McMahon 1959, Preuss 1932.U 10 HuicholNorthernCupanU 11 LuiseoU12 CahuillaU13 CupeoHopiU14 HopiC H I B C H A NCI CunaC2 Guaym: Alphonse 1956, Payne 1982.C3 BorucaC4 Bribri: Margery 1982.C5 G uatusoC6 RamaC7 Paya: Holt 1974.MISUMALP AN (perhaps part of Chibchan)MS1 Mskito: Lehmann 1920, Thaeler n.d.MS2 Sumu: Heath Ms, Lehmann 1920.MS3 tMatagalpaMS4 tCacaopera

O T H E R N O N - MESO - A M E R I C A N L A N G U A G E SN 1 Seri: Moser 1961.N2 tGuaicurianN3 tTonkawa: Hoijer 1933.N4 tKarankawaN5 tCotonameN6 tComecrudoN7 tCoahuilteco: Swanton 1940, Troike 1981.N8Eastern JicaqueN9 tWestern Jicaque (El Palmar)N10 tHonduran Lenca: Lehmann 1920.N11 tSalvadorean Lenca (Chilanga): Lehmann 1920.N12 Black Carib (Arawakan family)


15/42

MESO-AMERICA AS A LINGUISTIC AREA4 3within MA is quite restricted, and it occurs in non-contiguous locations withinthe area. Second, it is so natural and common that it could have developedindependently.(c ) VOW EL H ARMON Y. Some sort of vowel harmony occurs in Xincan, Huave,Mayan, Copainal Zoque, Mazahua (Amador 1976), and Tlapanecas well asin Lenca and Jicaque. This feature also fails to define MA as a LA, since detailsof the process vary in each individual case. The evidence is not sufficient todemonstrate diffusion, except possibly in Xincan and Lenca (see Campbell1978a). The trait is shared by few languages in the area.(d ) FIXED (predictable) STRESS. Phonemic (contrastive) stress is very rare inMA, though it is known in Tequistlatec and Cuitlatec. A few languages sharethe specific stress rule in which the accent falls on the vowel before the right-most consonant (i.e. V --0 V /C(V)#), e.g. Oluta Popoluca (Mixean),Totontepec (Oaxaca) Mixe, and Xincan, as well as in Lenca and Jicaque. It isquite possible that this rule is shared because of diffusion, but its distributionis too restricted to meet our strict criteria for determining if MA is a LA.Moreover, fixed accent in the majority of the MA languages also does not help,since it is very common in languages outside the area and all over the world.'(e) GENERAL SIMILARITIES OF PHONEMIC INVENTORIES include the following:(e.1) Contrastive underlying voiced stops are absent from MA, except for afew languages where they are of recent origin, e.g. some Otomanguean lan-guages, Huave, Tequistlatec, and Texistepec Popoluca (Zoquean), apparentlyfrom voicing after nasals with subsequent alteration or loss of the nasals. Thelanguages just beyond MA seem to show voiced stops liberallye.g. Sumu,Mskito, and Chibchan languages to the south; and Papago, Northern andSouthern Tepehuan, and Tarahumara to the north.(e.2) Contrastively voiced fricatives do not occur in MA, with the exceptionof a few Otomanguean languages (e.g. Mixtec and Trique), where they are ofrecent development.(e.3) The lateral affricate Al/ is generally absent, except in Nahuatl and To-tonac; Tequistlatec has /t17, the glottalized counterpart of /1/.

(e.4) Uvulars (postvelars) such as /q/ are found only in the Mayan and To-tonacan families. Outside the area to the north, they are reasonably common,e.g. in Yuman and Northern Uto-Aztecan (where they are not original). To thesouth, they are absent until we get to the Andes.(e.5) Aspirated stops and affricates are rare in MA, occurring only in Tar-ascan and in some Otomanguean languages (e.g. Chocho and Otom, histori-cally and phonemically from consonant clusters with h as a member), and inJicaque (on the southern frontier of MA).(e.6) Glottalized consonants are found in Tepehua, Tequistlatec, Mayan,Xincan, and Otopamean (historically and phonemically consonant clusters with

4 Actually, some other languages might fit this rule by defaultby having only final vowels andpenultimate stress, or only final consonants and final stress; e.g., Quichean roots can only end ina consonant, and Quichean languages have final stress. Some Otomanguean languages have pen-ultimate stress, and either no final consonants or /9/ as the only permitted final consonant (e.g.Mixtec and Zapotec); this makes them almost conform to the rule.


16/42

544ANGUAGE, VOLUME 62, NUMBER 3 (1986)PI, Rensch 1978), as well as in Lenca, Jicaque, Coahuilteco, and Tonkawa-which are non-MA languages.

(e.7) Implosives are known only in a few Mayan languages and in TexistepecPopoluca. Several Mayan groups have /bS/ as the labial member of their glottalseries. Some dialects of some Quichean languages have developed /qS/ from*q'. Tzutujil has further changed *t' to /dS/; dialects of Mamean languages alsoshow /147 dS (IS/ (or /gS/). Texistepec Popoluca, in contrast, has changed se-quences of nasal + stop to voiced imploded stops. Given that Texistepec Po-poluca and Mayan are not in contact, and that the Texistepec implosives areof recent origin, these languages clearly do not share implosion via diffusion.(e.8) Tonal contrasts are known in all the extent Otomanguean languages(including Tlapanec), in Huave, and in some Mayan languages (Yucatec, Chon-tal, Uspantec, and the San Bartolo dialect of Tzotzil). Of our control languages,some of the Uto-Aztecan languages above the northern frontier of MA are alsotonal (e.g. Northern Tepehuan and Cora-Huichol), as are Paya, Guaym, andBribri of the Chibchan family.(e.9) Retroflexed fricatives (and affricates) occur in several MA languages:Mamean, Kanjobal, Jacaltec, Acatec (Mayan); Guazacapn and Chiquimulilla(Xincannot contrastive); some Mixean languages; Chocho, Popoloca (Wil-liams & Pierson 1950), Mazatec, Trique, Yatzachi and Guelavia Zapotec (Oto-manguean); and allophonically in Tarascan. Retroflexed consonants are alsofound to the north in Huichol (allophonically), and in Yuman.(e.10) A central vowel, /i/ or /a/, appears in Mixe-Zoquean; in several Oto-manguean languages; in Huave, Xincan, Proto-Aztecan; in some Mayan lan-guages (Proto-Yucatecan, Cholan, and dialects of Cakchiquel and Quich); andallophonically in Tarascan. Such a vowel also appeared in Proto-Uto-Aztecanand persists in many Northern Uto-Aztecan languages; Jicaque also has /4/.

In general, given our restrictions, these considerations of phonemic inven-tories do not serve to define a LA. On the one hand, the traits do not have apan-Meso-American distribution; on the other hand, it is not possible in mostcases to demonstrate that the presence or absence of some phoneme or serieshas resulted from diffusion. Possibly Nahuatl acquired /tl/ from its Totonacanneighbors (Proto-Uto-Aztecan certainly lacked this sound); and the Kanjobalanretroflexed fricatives and affricates apparently derive from Mamean influencewithin the Mayan familybut no such feature is widely enough distributed todelimit a LA. In a less constrained view of areal features, of course, some ofthese similarities might qualify, given that strictly local diffusion can also gointo the creation of LA's.

We have concentrated here on phonemic contrasts, in part because of lim-itations on available information. But of course, subphonemic or allophonictraits are also subject to diffusion. An example is the final devoicing of son-orants (4.11); note also the very widespread phonetic aspiration of final stops(cf. Mayan, Mixe-Zoquean, Pipil, Xincan etc., as well as Lenca and Jicaque).4.2. MORPHOLOGICAL AND SYNTACTIC TRAITS. The following are some gram-matical attributes shared among various MA languages:


17/42

MESO-AMERICA AS A LINGUISTIC AREA45(a ) NOMINAL POSSESSION. The possession of one noun by another typicallyhas the form in MA languages of 'his-noun, (the) noun2', meaning `(the) noun2'snouni'e.g. Quich u-o 'i:2 le: aeih 'the man's dog', lit. 'his-dog the man'.(This order is the more widespread in MA, but the equivalent of 'the nounihis-noun2' also occurs.) The construction is typical of most MA languages, andcan be taken as a diagnostic trait of MA.' The isogloss coincides with the limitsof MA as it has been defined both culturally and linguistically. The trait is notfound in our control languages just beyond MA; i.e., Sumu and Miskito to thesouth lack it, as do Coahuilteco, Tonkawa, and the Uto-Aztecan languages tothe north. Uto-Aztecan seems to confirm both the northern boundary and thediffusion of the trait within MA (cf. Rosenthal). This possessive constructionin Proto-Uto-Aztecan had a pronominal copy of the possessor plus an accu-sative marker on the possessed noun (e.g. `John-Acc his-dog' for 'John's dog'),although some Uto-Aztecan languages have simple juxtaposition of the pos-sessed and possessor without pronouns. (Cf. Langacker, Rosenthal; Cupanlanguages in southern California have the construction 'man his-house'.) It issignificant that Nahuatl has only the MA constructionwhich it acquiredthrough contact with other MA languages, departing from the Uto-Aztecanpattern. It is also interesting that Cora and Huichol, which in some respectsact like transitional languagesoutside of MA, but evidencing a few of itstraitshave a construction intermediate between the original Uto-Aztecan andthe MA pattern, e.g. Cora i nana-ra pari 'the boy's mother', lit. 'the mother-his the boy' (Langacker, 89-90).

(b ) RELATIONAL NOUNS. Another feature shared by nearly all MA languagesis that of relational nounsexpressing locative and related notions, but com-posed of a noun root and possessive pronomial affixes. These correspond toprepositions in English or Spanish. Examples from Pipil, Mam, and Chol are:Pipil

nu-wan 'with me' (nu- 'my')mu-wan 'with you' (mu- 'your')i-wanwith him/her' (i- 'his/her')

(2) Mam (England, 71)n-wkt-a 'on me' (n- 'my')t-wion him' (t- 'his')n-xaq'-a 'below me't-xaq"below him'5 This possessive construction is not in fact unique to MA; but that would hardly be expected,given the limited num ber of possessive types in the world's languages. N evertheless, the construc-tion is rare elsewhere, and is not found in MA's imm ediate neighbors. Ultan (23-4) found only 16languages in his type B.1, characterized by personal possessive m arking of the po ssessed noun.An examination of his examples shows mo st to be quite distinct from the MA pattern, where asalient characteristic is possessive pronominal prefixes on the possessed noun. For exam ple, Turk-ish and Chamorro have suffixes; Cocopa uses affixes on a verb `to have'; Ew e and Albanian employindependent words (`the animal its foot'). Yurok, Wiyot, Karok, an d Acoma (of Ultan's sample),as well as Cupan, Diegueo, Navajo, and Menomini (which should be added to the sample) haveforms equivalent to `(the) man his-house'; but these do not reflect the predom inant MA pattern ofpossessed preceding possessor.


18/42

54 6ANGUAGE, VOLUME 62, NUMBER 3 (1986)(3 ) Chol (Warkentin & Scott, 27)k-ik'otwith me' (k- `my')aw-ik'ot `with you' (aw- `your')y-ik'otwith him/her' (y- `his/her')This isogloss coincides with that of nominal possession at the traditional bor-

ders of MA. The control languages to the south of MASumu, Mskito, Paya,and Guaymdo not contain relational nouns, nor do the northern border lan-guages: Coahuilteco, Tonkawa, Seri, Yuman, or Uto-Aztecan. Since Proto-Uto-Aztecan had postpositions (Langacker, 92-3), relational nouns were ap-parently diffused into Nahuatl. Actually, some Uto-Aztecan postpositions arepreserved in Nahuatl as locative suffixes; but these function differently fromthe relational nouns, which clearly reflect an innovation to the MA pattern.That is, in Nahuatl one says the equivalent of `his-to my-father' rather than`my-father-to', as would be typical in the UA pattern. Cora, again, appears tobe intermediate; it has the basic Uto-Aztecan postpositional pattern with nouns,e.g. mi-kty-hete ùnder that tree' (mi- `that', ktye- `tree', -hete ùnder'); butit uses possessive pronominal prefixes, e.g. ta-het ùnder us' (ta- òur';McMahon, xv), much like the MA pattern. It seems clear, however, that Cora'sMA possessive pronominal type with pronouns is of recent origin, resultingfrom postposing the locative to the pronoun. This is clearer in Northern Te-pehun, where nouns have postpositional locative endingsbut pronominalforms take the same postposed locative endings, with objective pronominalprefixes. Since the object and possessive pronominal prefixes are not distinct,the pronominal locatives are clearly in agreement with the Uto-Aztecan post-positional pattern, with object forms of the pronoun affixes; but they appearalso to resemble the MA pattern (if the pronominal affixes are taken as pos-sessive forms, not distinct from the objective forms; cf. Bascom, 317-18).Mixe-Zoquean languages are similar, in that they too also contain postpositions(see below).(c ) VIG ESIMAL N UMER AL SY STEM. A counting system based on twenty is pan-Meso-American. While it is found in virtually every MA language, it has alsoreached a few languages just beyond the conventional borders of MA, e.g.Coahuilteco, Cora, Mayo, and Northern Tepehun to the north, and Sumu,Mskito, and Guaym to the south. Still, it does not extend much beyond MA;it is not known in Yuman languages, Seri, or Tonkawa; or Northern Uto-Aztecan languages; or in Chibchan languages (e.g. Bribri). We may concludethat this is also a true MA areal trait which was sufficiently strong to reachslightly beyond the conventional boundaries. As noted above, isoglosses typi-cally fail to fall precisely into bundles, but often have varying extensions out-ward from an areal core.

Tequistlatec presents a representative example (Turner, 360):(4 ) 1 anuli2 oge?

3 afan?4 amalbu a5 amage 26 agamts's

20 anusans (anu- 1, - sans man')30 anusans gimbama ?(20 + 10)40 oge ? nusans (2 x 20)50 oge ? nusans gimbama ? (2 x 20 + 10)60 afane?nusans (3 x 20)80 amalbu a nusans (4 x 20)


19/42

MESO-AMERICA AS A LINGUISTIC AREA477agays00 amage2 nugans (5 x 20)8 abaygo00 anugans anus'ans (20 x 20)9 abella00 oge2 nu,fans anugans (2 x 20 x 20)10 imbatnatc .(d) BASIC WORD ORDER. It seems significant that only non-verb-final languagesexist in MA, although the area is surrounded by SOV languages. MA has basicVSO (Mixtec, Trique, varieties of Chinantec, varieties of Zapotec, Mam, Na-huatl etc.); VOS (Xincan, many Mayan languages, Copainal Zoque, Otom,another variety of Chinantec etc.); and SVO (Huave, Mazatec, Tequistlatecetc.) Southern Uto-Aztecan languages characteristically exhibit VSO order;but Proto-Uto-Aztecan was an SOV language (Langacker, 24), and so are most

of the other languages bordering MAe.g. Tonkawa, Coahuilteco, Yuman,Seri; Lenca, Jicaque, Mskito, Sumu, Guaym, and Bribri. Moreover, somecases of clear diffusion seem to be documented within the area; e.g., Xincanand Pipil apparently acquired V O S order from Quichean (Campbell 1978b).

The claim that MA contains no SOY languages bears comment, since it hasbeen suggested that the Mixean languages may present a potential counter-example (Bartholomew 1983). Actually, Mixe-Zoquean languages conform tothe MA non-verb- final pattern in ways that confirm the rule. That is, Mixe-Zoquean has certain constructions typical of verb-final languages, e.g. Noun+ Postpositions, Adjective + Noun order, and Possessor + Possessed wordorder. However, the order of the main constituentsthe verb and its NounPhrase argumentsis not basically SOY. Some would interpret such incon-sistent word-order typology as reflecting an earlier verb-final order; but evenso, the fact that Mixe-Zoquean now departs from that postulated SOY orderconfirms the MA trait. Thus the typical Zoquean language (e.g. CopainalZoque) is VOS; one might suspect departure from supposed earlier (S)0V tothis order under the influence of neighboring V O S Mayan languages.More at issue is the basic word order in Mixean languages. Bartholomew (9)sees it this way:

`Zoque[an] languages tend to have the object after the verb. Mixe[an] does not have the ergativesuffix on the noun and therefore has to use other strategies to identify the grammatical roleof a noun phrase. There seems to be something of a preference to have the object before theverb.'

She actually presents no clear evidence for SOY languages; but she believesCoatln (Oaxaca) Mixe to have relatively free word order, with cases of bothSVO and SOV orders. While the limited Coatln data available to us have noSOY examples (Hoogshagen 1984), an inspection of Tlahuitoltepec Mixe (Lyon1980) suggests that the SOY interpretation is not accurate. Tlahuitoltepec sen-tences are mainly VSO:

(5) yikPompihk mani: ni:.heaedaria water'Maria heated water.'SOV occurs only as a marked, non-basic order when the particle ti 'already'(perfective) is present. Compare exx. 6ab with 7ab:


20/42

54 8ANGUAGE, VOLUME 62, NU MBER 3(1986)(6 ) a. tiwo:n tihk tpo:pi.already Juan house whitened'Juan already w hitened the house. ,b. kwo:n ic n2uk tiyik2ohkiyi .Juan my dog already killed'Juan already killed my dog.'(7 ) a. yik2atuhk kwo:n tihk.closeduan house'Juan closed the d oor. 'b. yikmo22o2hk mani: ma2u2nk.cause.to .sleep Maria baby'Maria put the baby to sleep.'The (S)0V exam ples cited by Bartholomew for Coatln Mixe occur with thesame ti seen in Tlahuitoltepec:(8 ) a. ti jiicy2aaydzimbijcteiy ma tcy.PERF tobacco.leaf wrap.in .bundle LOC mat'He wrapped tob acco leaves in a straw mat.'b. ti cuhuay mooc mucxy.P E R F horse corn chewed.up'The horse chewed up the corn.'Hoogshagen (13) cites only two relevant Coatln examples, both S VO :(9 ) a. he 2ugdeh:ty ti ymo2oy j2uue 2i:k.the fatherF gaveis.son the toyb. he tatpi:te pi2k2ana2 ak he lay yahmo2oy he kway.thelder.man the boyo the girl cause.to .give the horse'The older m an caused the bo y to give the girl the horse.'We conclude that Mixean is NOT a counter-example to MA non-verb-finalpattern, since SO Y is apparently a marked, non-basic order in these languages.N one of the Mixe-Zoquean languages has SO Y b asic order for the main con-stituentsthe verb and its nominal arguments.An interesting case in the opp osite direction is Chichimeco Jonaz, an Oto-manguean language with SO V order. W hile all other Otomanguean languagesare spoken w ithin MA (and h ave non-verb-final order), Chichimeco Jonaz isspoken in Guanajuato, outside MA. It may have acquired its SO Y order fromneighboring languages beyond the borders of MA; if so, it is the exceptionwhich confirms the rule.The isogloss dem arcating non-verb-final languages is thus diagnostic of MA,and is valuable for defining the LA.(e ) A B S E N C E O F S W I T C H - R E F E R E N C E . The languages surrounding MA on bothsides have switch-referencenot found in any MA language, but known inCoahuilteco (cf. Troike), Seri, Yum an, and Jicaque. The absence of sw itch-reference coincides with MA; thus one m ight conclude that this constitutes anadditional isogloss diagnostic of the LA. Ho wever, in view of the hypothesisthat switch-reference exists only in SO Y languages, its absence in MA may be


21/42

MESO -AMERICA AS A LINGUISTIC AREA49a reflex of MA basic word order more than of areal diffusion (cf. Jacobsen1967). This bears more study.(f ) INT IMATE PO SSESSIO N . Typically, kinship terms and body parts are inti-mately possessed (either do not occur unpossessed, or require special mor-phological marking when unpossessed) in MA languages. Since this feature ischaracteristic of many languages throughout the Americas, it is not particularlyuseful for defining a LA in MA. The term ìnalienable' is widely used for thiskind of possession, but it is semantically inappropriate.(g ) LOCATIVE S DERIVE D FRO M BODY PARTS. Locative words in many MA lan-guages are derived in a rather direct and obvious way from body parts, e.g. inMayan, M ixe-Zoquean, Totonac, Tlapanec (Schultze, 245), O tomanguean, Tar-ascan, and Nahuatl:

(10) a. Mixtec (Alexander, 79)cihi `stomach; in(side), under'in iheart; in, inside'nuu `face; to, at, from'sini `back; behind'b. Cakchiquel-pan `stomach; in, inside'-cimouth; to, in, at'-ixback; behind'-wi `head-hair; on, on top of'Actually, it is natural and com m on for languag es to have a relationship be tweenbody parts and locative notions, as in Eng. at the head of, at the foot of, atthe mouth of, in ba ck of etc. N evertheless, Meso-Am ericanists have noted thatMA languages seem to share something beyond these ordinary associations:locative constructions which maintain their nominal character (see relationalnouns above) but involve seman tic associations n ot usually found in other partsof the world, e.g. `stomach' for ìn', `tooth' for `to, at' etc. In spite of thisfeeling of som ething peculiar being shared in M A languages, it cannot be show nthat the trait is diffused, and it is po ssible that som e of these associations resultfrom universal tendencies. Thus, while it coincides in some ways with other

MA features, it is not sufficiently free of com plications to be used for defininga LA.(h) ABSO LU TIV E AFFIX ES. Som e MA languages have a nominal suffix calledthe àbsolutive', borne by nouns that have no other affix, e.g. Nahuatl tlaskal-li `tortilla-ABS', no-tlaskal `my-tortilla' ; Quich xolom-a:x `head-Ass', a-xolo:m`your-head'. The àbsolutive' suffix has no real semantic content, but occurson nouns which are otherwise morphologically isolatedi.e. show no otheraffixes. In MA, most Mayan languages have an absolutive, as does Nahuatl.These are not really equivalent, however: thus, in Quich, the so-called ab-

solutive suffix occurs only on a certain class of otherwise intimately possessednouns, but in N ahuatl the absolutive is much mo re general, occurring with mostnouns w hen they bear no other affixes. In the key languages surrounding M A,an absolutive is found in Paya and the Misumalpan languages on the southernborder, and in the Uto-Aztecan languages to the north. This trait is not sig-nificant as an areal feature, since its distribution is very limited within MA,


22/42

55 0ANGUAGE, VOLUME 62, NUMBER 3 (1986)and it is also found outside the area. Also, there is no evidence that it is diffused,given that the languages containing it are not in contact.

(i) ABSENCE OF 'PLURAL' MARKERS ON NOUNS. In many MA languages, the'plural' is either totally lacking or is limited to 'human' nouns. But this is notvery helpful in areal considerations, since many other American Indian lan-guages share this trait.

(j) NUMERAL CLASSIFIERS. In several MA languages, nouns after numbersappear with a classifiera morpheme which indicates the class of noun beingcounted (somewhat akin to 'three HEAD of cattle', 'four LOAVES of bread').Some examples are:

(11) a. Yucatec (Tozzer, 103)hun-ku:1 ee2 'one tree' ('one-plant tree')ka2-ku:1 h2as 'two banana plants' ('two-plant banana')af-t:1 winik 'three men' ('three-Amm man')ka2-t:1 p:k"two dogs' ('two-Amm dog')os -p'e:1 na 'three houses' (`three-INAN house')b. Tzeltal (Kaufman 1971:100)2o.S" lehe te 2 'three plants' ('three flat-thing wood')Pas' tehk te2 'three trees' ('three plant wood')lahun k'as si2 'ten chunks of firewood' ('ten broken-thing

firewood')c. Chol (Aulie)a2 -koht eihmay 'two deer' ('two-animal deer')

ea2-p'ehl hab"two years' ('two-thing year')Such numeral classifiers are found in many Mayan languages, Sierra Popoluca(Zoquean), Zapotec, Huave, Tarascan, and Nahuatl. Some of these cases m aybe the results of diffusion (e.g. within the Mayan family, cf. Hopkins 1970);however, the distribution within MA is too erratic to be significant in a pan-areal way. Also, noun classification systems both with and without numeralsseem to develop easily and independently in language, and exist in many partsof the worlde.g. in Austronesia, southeast Asia, Australia, Africa, and otherparts of the Americas (including, among our control languages, Guaym andBribri).

(k) NOUN INCORPORATION. Some MA languages have noun incorporation, aconstruction by which a nominal object may enter directly into a verb stem.(12) Nahuatla. ni-k-ei:wa t1asla1-11.

I-it-make tortilla-Assb. ni-tlaaal-ei:wa 'I make tortillas'.I-tortilla-make

(13) Yucateca. t-in-Cak-ahe 2.ASP-I-cut-SUFF WOOdb. e'ak-ee2 -n-ah-enI cut wood'.cut-wood-INTR-SUFF-I


23/42

M E S O -AMERICA AS A LINGUISTIC AREA5 1In MA this trait occurs in Yucatec, Mam, Mixe-Zoquean, Mixtec, Trique,Totonac, and Nahuatl.It is possible that the trait is diffused in some cases; probably Yucatec bor-rowed it from Mixe-Zoquean. However, it does not meet our strict criteria forareal considerations. Its distribution within MA is too limited, and it occurswidely in Am erican Indian languages ou tside M A (cf. Kroeber 1910 , Sapir 1911)and elsewhere in the world (Mithun 1984).(1) BOD Y-PART INCO RPO RATION . Related to noun incorporation is the incor-poration of special forms of bo dy-part term s in the verb som etimes as d irectobjects, more frequently as instrumentals. Pipil has many such cases:

(14) tan-kwa `to bite' (tan- `tooth', -kwa to eat')iksi-ahsi `to reach, overtake' (iksi- foot', -ahsi `to arrive')ma:-tu:ka to touch, feel' (ma- `hand', -tu:ka `to plant, bury')mu-yaka-pitsa `to blow one's nose' (mu- REEL, -yaka `nose', -pitsa`to blow')Some Mixtec examples (Alexander, 49) are:

(15) kata-x 2 `will dance' (kata `will sing', x 2 `foot').unda 2 `will push' (from u 2un `will put', nda2 `hand').Bod y-part incorporation is found in N ahuatl , Toton ac, Tarascan, O axaca Mixe,Sierra Popoluca (and Mixe-Zoquean generally), Tlapanec, Tarascan etc. Out-side MA, it is found widely in western North America, e.g. in Yuman, Uto-Aztecan, Maiduan, Washo, Shasta, Achumawi (Sherzer 1976:125); and to thesouth, perhaps in Mskito and Bribri.This trait has the same limitations as object noun incorporation; it is limitedwithin MA, and well-known in languages outside the area. It is therefore oflittle value in MA areal considerations.

(m) DIREC TIO N AL AFFIX ES. Several MA languages have verbal affixes whichindicate direction, typically `toward' or `away from' the speaker:(16) Nahuatlnee-wa:1-kwi in tlaskal-li.m e-hither-take the tortilla-ABs`He brings me the tortillas.'(17) Cakch iquely-e-b'e-n-kamisaxAsP-them-away-I-kill`I'm going there to kill them.'Some of these are Cakchiquel, Quich, Tequistlatec, Mixe-Zoquean, Totonac,several Otomanguean languages (e.g. Otom), and Nahuatl.This trait too is found with considerable frequency in languages outside

MA e.g. Cora, Tonkaw a, Wap po, Wintu, Y ana (Sherzer 1976:126), Q uechua,and Cashinahua (Montag 1981:574-5)and its distribution within MA is notgeneral.(n) VERBAL ASPECT. Aspect is relatively more important than tense in manyMA languages, e.g. Mayan, Tlapanec, Mixtec, Zapotec, and several other Oto-


24/42

552ANGUAGE, VOLUME 62, NUMBER 3 (1986)manguean languages. This is so common in the world's languages, however,that it can scarcely be considered a strong indicator of a LA, particularly sinceseveral other MA languages have strong tense systems.

(o) INCLUSIVE VS. EXCLUSIVE. The pronominal systems of several MA lan-guagese.g. Chol, Mam, Acatec, Jacaltec; Chocho, Popoloca, Ixcatec (Fer-nndez 1961), Otom, Mixtec, Trique, Chatino, Yatzachi Zapotec, Tlapanec;Huave; and several Mixe-Zoquean languagesdistinguish 1st person inclusiveand exclusive pronouns:(18) Chol (Warkentin & Scott, 29)

honon la 'we' (inclusive)honon lohon 'we' (exclusive)(19) Copainal Zoque (Harrison & Garca, 4 17)ti 'our' (inclusive)tis 'our' (exclusive)

(20) Huave (Stairs, 296)ikora 'we' (subject exclusive)ikor ùs' (object exclusive)iko..0a 'we' (subject inclusive)iko: ùs' (object inclusive)

While some of these may have acquired the contrast by diffusion, it is rela-tively easy for such a distinction to develop independently (cf. Robertson 1983for such an explanation of the Mayan cases). The trait is both limited withinMA, and common outside the area (e.g. Sumu, Mskito, Bribri, Quechua, Ay-mara, and many others have such a contrast). It therefore does not meet ourrigid requirements as an areal feature of MA.

(p) `ZERo' COPULA. An overt copula is typically lacking in MA languages inequational constructions with predicate adjectives or noun complements, as inQuich sag le: xa:h 'The house is white' (lit. 'white the house'), a:x-kar le aeih'The man is a fisherman' (lit. 'fisherman the man'). The great frequency ofzero copula in the world's languages, including many American Indian tongues,makes this feature of little use in the investigation of MA areal phenomena;cf. Hebrew, Russian, Black English (all in present tense), and others.(q ) PRONOMINAL COPULAR CONSTRUCTIONS. A copular construction with pro-nominal subjects takes the form of the complement plus a pronominal affix inseveral MA languages, e.g.: Kekch wing-in 'I am a man' (lit. `man-I'), aq-at'You are a woman' (lit. `woman-you'); Pipil ti-siwa..-t 'You are a woman' (lit.'you-woman'-Ass); Tequistlatec a-s'wde? 'It is a church' (lit. it-church'),n-ondizPa 'I'm a devil' (lit. Ì-devil'; Turner, 327). Elsewhere in MA, this con-struction occurs in Mayan, Nahuatl, Chocho, Chinantec, Mazatec, Otom, Oa-xaca Mixe, Oluta Popoluca, and Sayula Popoluca.It is difficult to evaluate the areal properties of this feature properly. Never-

theless, we have insufficient evidence to demonstrate that it is shared by dif-fusion; and its distribution, to the extent that it can be determined, is not generalthroughout MA. Therefore it is safest not to consider it significant for definingthe area.(r ) ABSENCE OF A VERB 'TO HAVE'. Absence of a verb of possession `to have'has been suggested as an areal feature of MA. Several languages have a con-


25/42

M E S O -AMERICA AS A LINGUISTIC AREA5 3struction equivalent to ìs', `there is', or èxists' plus a possessed noun, as inCakchiquel: k'o xun nu t'i 2 Ì have a dog' (lit. `[there] is one my-dog'). Thisoccurs in Mayan (but not Huastec), Mixe-Zoquean, Tequistlatec, Xincan, Chin-antec, Mazatec, and Trique. Some languages that have a verb correspondingto `to have' are Nahuatl, Huave, Huastec, and Mixtec. The absence of `tohave' is common in the world (e.g. in Finnish, Tamil, Sango, and Old Irish-cf. Ultan); and its distribution in MA is so incomplete that it can hardly countas a significant areal trait.4.3. SEMANTIC CALQ UES OR LOAN TRANSL ATIONS. It has been observed, e.g.

by Smith-Stark 1982, that many compound words, and words having multiplereferents, are shared in MA languages through loan translations. The exampleswhich have been noted are given in Table 1.

1. door: mouth of house2. bark: skin/back of tree3 . knee: head of leg4. wrist: neck of hand5 . calf: excrement/belly of leg6. eye: fruit/seed/bean of face7 . b ile: b it te r8. finger: child of hand9. boa constr ictor: deer-snake10 . moon: grandmother11 . ring: coyol palm-hand12 . witch: owl, sleep(er)13. cramp: (associated in some way w ith)deer14 . fiesta, ceremony: (big) day15 . root: hair of tree16. twenty : man17 . lime: (stone-)ash18 . egg: stone/bone of bird19 . wife: intimately possessed `wom an'20 . porcupine: thorn-opposum, thorn-lion,

thorn-peccary, thorn-pig21 . cougar: red jaguar22 . anteater: honey sucker, suck-honey23 . to kiss: to suck24 . to smoke: to suck25 . branch: arm (of tree)26. to marry: to join, to f ind27 . gold/silver: excrement of sun/god28 . eclipse: eat the sun/mo on; the sun/moo ndies; sun/mo on to rot29. coral snake: mother of d r i ver ant

3 0 . bladder: house (of) urine3 1. vein: road (of blood)3 2. canine tooth: dog-tooth, snake-tooth3 3 . molar: grindstone (metate)3 4. edge: mouth3 5 . thumb: mother of hand3 6. mano (of metate): hand/child of metate3 7. poor: orphan, widow3 8. rainbow: snake, cougar, turtle, squirrel,or weasel3 9. otter: water-dog, water-fox40 . ceda r: god t ree41. medicine: liquor, poison42. to cure: to suck (to smo ke)43. pataxte (non-domesticated cacao): t iger-cacao (jaguar-cacao)44. town: water-mountain45 . soot: nose/mucus of fire46. to write : to paint, to str ipe47. to read: to look, to count, to shout48. alive: awake49. son and daughter : man's aredistinguished, but a single term forwoman's5 0 . head: bottle gourd (tecomate)5 1. thirst: water-die52 . need : want , be wanted53 . enter: house-enter5 4. cockroach: contains the root for `house',often compounded with ìn' orsomething equivalent5 5 . feather: fur

TABLE 1.Smith-Stark has shown that many of these calques have a limited distribution

within MA; some are found in very few languages (e.g. `root: hair of tree').Nevertheless, of the 52 cases which he examined (some of them are combinedunder a single number here, and others which he did not consider have been


26/42

554ANGUAGE, VOLUME 62, NUMBER 3 (1986)added), he found that the following are widely distributed, and coincide withthe borders of MA:(a ) 'Knee: head of leg'. One might suspect that this semantic association isnatural enough for this proposed calque to have developed by itself. Never-

theless, the association was not found in Smith-Stark's control languages, eitheron the borders of MA or beyond; this suggests that it is indeed a valid arealborrowing.(b ) 'Boa constrictor: deer-snake'. The association of deer with snake seemsarbitrary enough to conclude that this legitimately reflects MA areal interaction.(c ) 'Lime: ash, stone-ash'. Given that corn is soaked in water with lime whereit is available (e.g. highlands), but with ashes elsewhere, to soften it for grinding,it is not too surprising to find a linguistic association between the two. Thiscalque, then, derives from cultural facts about the preparation of corn in MA,and for that reason is established as a strong areal feature.(d ) 'Wrist: neck of hand'. Although it is not difficult to imagine a naturalconnection here, such an association is not found in Smith-Stark's controllanguages. This is then an areal trait.(e ) 'Egg: stone of bird, bone of bird'. This is another calque found only inMA, and thus an areal feature of MA.(f ) 'Vein: road (of blood)'. It is also possible to imagine this semantic con-nection developing independently, but the association does not appear in Smith-Stark's control languages.' This too, then, is a MA areal feature.(g ) 'Molar: grindstone (metate)'. Again, it is possible to imagine this semanticassociation developing from the nature of the two objects. Nevertheless, theconnection seems specifically MA, not being found in Smith-Stark's controllanguages.(h ) 'Edge: mouth'. This association is also found throughout MA and not inthe control languages, though one could perhaps expect a natural connection.'Still, given its distribution, it is a feature of MA.(i ) 'Thumb: mother of hand'; 'Finger: child of hand'. These semantic asso-ciations seem sufficiently arbitrary to support the MA area, given that they arefound throughout MA but not in the control languages.'(j ) 'Gold or silver: god-excrement, sun-excrement'. This calque is clearlyMA and not the result of accident.(k) 'Alive: awake'. This semantic relationship could perhaps result fromchance; but distribution within MA, and not outside, makes it seem a legitimateareal feature.

(1) 'Town: water-mountain'. The distribution of this loan translation fits6But in South America, Cavinera (Key 1963) and Sirion (Schermair 1957) appear to have the

equivalent of 'blood-road' for 'vein'.7Compare South American Cashibo (Shell 1965), where 'mouth' and 'edge' are related.8See the discussion in Brown & Witkowski 1981 of 'people' = 'digit' metaphors. They find two

of 23 North American languages to have the 'thumb' = 'mother' and 'finger' = 'child' associations;in South America, one of ten languages has the former, three of ten the latter. While the MAconstruction is perhaps a bit more specific as 'mother/child of hand', these cases need to be keptin mind in evaluating this calque.


27/42

M E S O -AM ER ICA AS A LINGUISTIC AREA5 5Smith-Stark's requirements of occurring in various extremes of the MA geo-graphical area, found in such languages as Nahuatl, Pochutec (Boas 1917),Oluta Popoluca, Sayula Popoluca, (Sierra) Totonac, and Mazatec.

(m) Porcupine: thorn-opossum', or thorn' plus some other animal. Althoughthis distribution makes it seem areal, we cannot rule out that the semanticconnections derive from the physical character of the animal. Eng. porcupineis ultimately from `pig' + `thorn'; Spanish forms such as puerco espn `pigthorn' and zorro espn `fox thorn' attest recent formations. Perhaps, then, thisfeature is to be given less credibility areally, even though its distribution isappropriate for such an interpretation.

Taken as a group, these calques constitute strong evidence of diffusion withinand throughout the MA linguistic area.

Most of the MA traits listed above are shown in Table 2 (pp. 556-7). A `plus'means the language has the trait, a `minus' indicates its absence. A questionmark means that available evidence suggests the plus or minus given, but doesnot demonstrate it conclusively. Parentheses indicate that the language has thetrait as indicated, but perhaps not in its canonical patterni.e. perhaps withoccurrence limited to certain constructions or particular dialects. A blankmeans that no information is available in the sources consulted (listed in thekey to Map 1).

5. MA IN SUMMARY. So far we have considered areal linguistics in general,and some circumstantialist traits that have been proposed in favor of a LA inMA. We now hope to come to some conclusions. Many of the circumstantialistfeatures have turned out not to meet our tight constraints for defining a LA:they were either too restricted in their distribution, or were amply attestedbeyond the area. In several cases, the historical information available is in-sufficient to demonstrate borrowing; but in other cases, fortunately, the his-torical evidence is clear. For example, Proto-Uto-Aztecan is sufficiently well-known to make clear when Nahuatl has changed to become more MA. Thecomparative evidence for several other language families is also sufficient todetermine borrowing in individual cases, as discussed above.

After careful scrutiny, five features are found to encompass the traditionalMA area so conventionally defined by both linguists and anthropologists. Theseare: (a ) Nominal possession (of the type his-dog the man).(b ) Relational nouns.(c ) Vigesimal numeral systems.(d ) Non-verb-final basic word order, to which absence of switch-referenceis correlated.

(e) Several widespread semantic calques.In effect, these five isoglosses coincide at the borders of MAexcept forvigesimal numbering, which extends a bit beyond. For some scholars, a single

shared trait would have been sufficient to define a LA. In our study, fiveisoglosses enclose the area and bundle at its borders. This constitutes extremelystrong evidence for, and confirmation of, MA as a LA.


28/42

556ANGUAGE, VOLUME 62, NUMBER 3 (1986)I 2+ I+I ++++++++ 1 1+11++ 71+

1++I+I+1771717o

g;III++ + + + + +IIII ++III

+++++++III2 1 111+

1II1

I.1 + + ++

p,I++11+1+ ++ ++ + ++ + + + + + ++++1c)+11+1+1 ++++++1+11I+++II++ +1+ + 1 ++++ 1 +

I++++ 2+++++++++++++ ++ + + + +1 + +++11+++++111111+++ ++ +1I++++++1++.)++ +++11+11++++++-7 +2+++II+++1II lii i1 1+++++++ +++++++++.0+++++ ++++++ ++++++++++++++++ ++++++ + ++++++++ + + + + + +V)VI V) VICDOV)000> >>>>> C ) ) v) c) c> C >> C >) C > C > C ) c ", C )Ocncn>. cncnC>>0>>>v) >>>cncncncn>> +++++++++++ +++++++++

+++++++7++++1-7++++ +++++++++-7++++++++++


29/42

CONTROL LANGUAGESCora+) - +VSO++++- - ++++(+) +Northern Tepehuan- +VSO +-++) - +- (+)++Other Southern Uto-

Aztecan- - VSO ++- - ++-)+) - (-)(+)- - - - - -Northern Uto-Aztecan - - - SOV + +(+) - +++)- (-) - -- - - +- - - - +Coahuilteco- +SOV - +- - - -+++- -(+) - +- - +- - -Tonkawa- - SOV - -- + - + - -+- - - - +- - -SeriOV -YumanOV- ++- (+) +++-- -- - - +(+)LencaOV + ++--+++ - -Jicaque- SOV- ++- - - - + + - - +Miskito- +SOV+ - -+ )+(-)+

umu- +SOV++-+-+-aya-- --- +-uaym- +SOV+--+ - -Bribri- - SOV++++++ - -TABLE 2. Numbered columns refer to the following features:1. Nom inal possession ( 'his-N the N ')2. Relat ional nouns3. Viges imal numera l sys tems4 . Basi c word order: no SOV order s5 . Absence of switch-reference6. Inalienable possession of bod y-part and kin terms7. Locat ives der ived f rom body parts8 . Absolut ive nominal af f ixes9. Absence or l imited occurrence of 'plural ' markers on noun s10. Num era l c lass if iers11 . Noun incorporation12. Body-part incorporation13. Verbal di rect ional af f ixes

14 . 'Aspect' more important than `tense'15. Inclusive vs. exclusive pronominal forms16 . 'Zero' copula

17 . Pronominal copu lar constructions w ith affixes18 . Absence of a verb 'to have'19 . Final devoicing of sonorants20 . Vo icing of obstruents after nasals21 . Vowel harmony22 . Presence of the stress rule: V_C(V)#23. Contrast ive voiced s tops24 . Contrastive voiced fricatives25 . Presence of the lateral affricate (tl)26 . Presence of uvulars (q etc.)27 . Presence o f aspirated stops and affricates28. Presence of glottalized consonants29. Contras t ive tones3 0 . Presence of retroflexed fricatives (and affricates)31. Presence of a central vowel (/i/ or /a/)


30/42

558ANGUAGE, VOLUME 62, NUMBER 3 (1986)However, if we look at MA in a more conventional waywithout the strongdistributional restrictions which we imposed abovethen MA has much moresupport as a LA. That is, typical LA's such as South Asia and the Balkans(considered below) are characterized by different sorts of diffusion. Some arerestricted locally, and do not extend throughout the area; some reach beyondthe borders of the area; some overlap, or show criss-crossing isoglosses fromother LA's. Having established the legitimacy of MA as a LA on the bases ofstrict criteria, we feel it safe to re-assess certain of the traits discussed, con-sidering them to provide further support for MA as a LA to the extent thatthey have been diffused across language boundaries. The aggregate of suchfeatures in fact corresponds to the situation in other established LA's. Wesuggest that the following traits listed in Table 2 lend supplementary supportto MA as a LA, either because they appear demonstrably diffused in someinstances or because their peculiar character is broadly MA in nature: 7, 9, 10,15, 17, 19, 20, 22, 30.

While we have limited ourselves to structural features, many MA languagesalso share traits that are perhaps better considered part of an ethnography ofcommunication than of a formal grammarso-called `Sprechbund' features.Among these are the particularly MA form of ritual language with paired se-mantic couplets, called huehuetlatolli in Nahuatl and O'ono:x in Quich (cf.Ocuilteco, Pipil, Xincan, all Mayan languages, Mixe-Zoquean); whistle speech(Mazatec, some Zapotec varieties, Mopn, some Nahuatl dialects, Totonacdialects); reverential or polite vs. familiar contrast for 2nd person address (Na-huatl, Quich, Sipacapetio, Mam, Aguacatec, Ixil, Mixtec).

Also, since we have taken LA's to be characterized by diffused structuraltraits, as most arealists do (cf. Klagstad 1963:180), we have totally neglectedany mention of lexical borrowing. However, the existence of such borrowingis quite natural within LA's. If we include Trubetzkoy's (1928:18) notion thatLA's offer 'eine grosse Anzahl gemeinsamer Kulturwiirter' as part of theirdefining characteristics, then MA fares well. For studies of widespread loanwords in MA, cf Campbell 1976, 1978b, Campbell & Kaufman 1976, Justesonet al. 1985, and Kaufman 1976, 1980. Thus Sprechbund features and widespreadloan words circumstantially strengthen our faith in MA as a L A.This conclusion that MA is a 'strong' LA reflects our approach to definingLA's in terms of their strength, determined by the number and weight of sharedtraits. The true strength of MA, with five bundling isoglosses, can be under-stood only when MA is seen in comparison with the best-established LA's inthe literature, i.e. the Balkans and South Asia. Obviously a detailed treatmentof these two areas is far beyond the scope of this paper; nevertheless, someconsideration is important in order to see just how strong MA is as a LA.6. COMPARISON WITH THE BALKANS AND SOUTH ASIA. The languages whichbelong to the Balkan area are Rumanian, Bulgarian, Macedonian, Albanian,Greek, and perhaps Serbo-Croatian. Turkish, though not a member of the LA,

also shares some Balkan areal traits. In Albanian, a member of the LA, theGeg (northern) dialect shares fewer Balkan areal traits than the Tosk (southern)


31/42

M E S O -AM ER ICA AS A LINGUISTIC AREA5 9dialect. Serbo-Croatian is controversial, in that some include it here, but othershold it to be a non-member (cf. Joseph, 131, 147; Comrie 1981:198). ThusKlagstad (179) includes Serbian (and Turkish), but excludes Croatian. This wasroughly the position of Weigand (1925:8), who regarded Greek, Serbian, andTurkish as Balkan languages geographically, but not linguistically. Similarly,Georgiev (1977:9) recognizes `core' Balkan languages, with the same threewhich 'restaient a la priphrie'. Schaller (1975:103) lists Greek and Serbo-Croatian not as `first level', but rather as `second level' Balkan languages, withTurkish at a third level. Birnbaum (1965:20) includes Serbo-Croatian only 'con-ditionally'. Scholars generally maintain that Serbian has more Balkan char-acteristics than Croatian. Asenova (1977:29) and Sarah Thomason (p.c.) holdthe Torlak dialects of Serbian to be genuine core Balkan. The following fea-tures, usually accepted as characteristic of the Balkan area, are scrutinizedlike those of MA:(a ) A CENTRAL VOWEL (/i/) or (/a/). This is not found in Greek or StandardMacedonian, though it is in some Macedonian dialects.(b ) VOWEL HARMONY (or umlaut). This trait's history is clear in Rumanian,Bulgarian, and Greek, where a stressed vowel has been influenced by the stress-less vowel of the following syllable (e.g. Rum. o > oa before a non-high vowelin the following syllable; otherwise it stays o; Alb. u > ii and a > e with i inthe post-stressed syllable; Bulgarian shows an alternation between /ya/ and/e/ under similar conditions). These changes are sufficiently different in theselanguages, and are natural and widespread enough, to allow independent in-novation to compete with the proposed areal explanation.

(c ) SYNCRETISM OF DATIVE AND GENITIVE. In Bulgarian, Albanian, Rumanian,and Greek, the dative and genitive have fused in form and function. While thisis generally considered a strong areal feature, such syncretism is not unusual.For example, many languages have possessive constructions where genitiveand locative functions alternatee.g. `John's bicycle is' or `To/at John is abicycle' for `John has a bicycle' (cf. Ultan). As Joseph (241) puts it, `the dative/genitive merger ... surely must be viewed within the context of a general driftwithin Indo-European away from highly developed synthetic case systemsviewed from such a perspective, the convergence is perhaps somewhat lessstriking.'(d ) POSTP OSED ARTICL E. W ith the exception of G reek, Balkan languages post-pose the definite article; this is perhaps the best-known Balkan feature. It isnot unique; cf. the Scandinavian languages.(e ) PER IPH RA STIC FU TUR E. Balkan languages have periphrastic futures withan auxiliary verb corresponding to want' or have'. Such a feature could evolveindependently without difficulty, as it did in English (`future' from a verb thatmeant `to want') and Vulgar Latin (from habere `to have').(f ) PERIPH RASTIC PERFECT. Except for B ulgarian (and Macedonian?), Balkanlanguages have a periphrastic perfect with an auxiliary verb corresponding to`have'. Historical evidence suggests that this is a borrowing in Greek and Al-banian, probably from Latin. This same construction has diffused throughoutmuch of Europe. Superficial similarity between Macedonian and Albanian per-


32/42

56 0ANGUAGE, VOLUME 62, NUMBER 3 (1986)fects formed with the verb 'have' have long been pointed out; however, 'theseforms function in an entirely dissimilar manner in the hierarchical arrangementof verbal categories in their respective language systems, with the result thatthe forms are superficially comparable but the systems are not' (Joseph, 241).(g ) No INFIN ITIVE. Balkan languages are said to lack infinitives, having insteadconstructions with finite verb forms, e.g. 'I want that I go' for 'I want to go'.This is considered a strong Balkan areal feature. Nevertheless, 'the Balkanlanguages differ rather dramatically in the extent to which they show the lossof the infinitive' (Joseph, 242). Macedonian lacks it, as does Greek, except fora productive remnant in xo-perfects. In Bulgarian, too, it is basically absent,though the 'short' form of the earlier infinitive is used in restricted contexts(Joseph, 243). The situation in Albanian, Rumanian, and Serbo-Croatian is morecomplicated. In Albanian, the original infinitive has been replaced; of the twoprincipal dialects, Geg has an infinitive which does not overlap with finite formsat all, while Tosk shows re-emergence of the infinitivethough the categorymay have additional finite functions (Joseph, 91, 100, 243). Rumanian's geo-graphically separated dialects differ. Istro-Rumanian of Yugoslavia maintainsthe infinitives 'to a wide extent' (Joseph, 174, 177). The data on Arumanian,spoken in parts of Greece, Albania, and Macedonia, are unclear: its historicalinfinitive seems to function as a nominalization, but is used verbally in someinstances (Joseph, 174-5, 196). The infinitive is a recognized category in Daco-Rumanian (of Rumania and Moldavia), but its use is quite restricted (Joseph,161); it is lacking from Megleno-Rumanian, spoken in an area of Macedonianand Bulgaria (Joseph, 177). While it is often broadly reported that Croatian hasinfinitives, and that Serbian lacks them, the facts are not so simple. Only theTorlak dialects show complete absence of infinitive, rivaled by Banat of theYugoslavRumanian border. Otherwise, eastern dialects vary greatly, rangingto comparatively wide use in the standard language. The western dialect grouphas the infinitive to a greater extent. Infinitive replacement is an on-goingprocess, spreading from east to west (cf. Joseph, 132-7).(h ) PLEONASTIC USE OF PERSONAL PRO NO UN S. Balkan languages employ per-sonal pronouns in sentences with animate objects, thus marking the objecttwice; e.g. Rum. Lam scris lui Ion 'I wrote (to) John', lit. 'to-him I wrote himJohn', i- 'dative singular'.

(i) IDENTITY OF LOCATIVE AND DIRECTIONAL EXPRESSIONS. Greek, Rumanian,and Bulgarian do not distinguish formally between stationary locatives anddirectionals (with motion), e.g. Gk. stin e13a `in(to) Greece'. The syncretismis clear in Greek at least, where historically these were distinct. The lack ofdistinction between cases of location vs. motion perhaps results from inde-pendent development; cf. Eng. He went in the house vs. He is in the house,or Sp. en meaning 'in', either with or without motion (similarly Eng. on withand without motion). Moreover, to quote Joseph (241), 'the question of a dative/locative merger among Balkan languages changes when it is viewed in termsof the morphosyntactic alternations ... at one level there is such a merger inGreek, at another level the dative and locative are kept distinct from oneanother.'


33/42

MESO-AMERICA AS A LINGUISTIC AREA61(j) NUMBERS 11-19. Except for Greek, the Balkan languages count 11 to 19with a construction corresponding to `one over/on ten', two over/on ten' etc.,e.g. Rum. unsprezece `eleven' from *unu-supre-dece. While this feature is dif-fused, it seems almost lexical in character and therefore not difficult to borrow,with no particular impact on the grammatical fabric of the language. (The MAvigesimal system is much more profound; it required switching from the base-five or base-ten systems that historically lay behind the numbering systems ofmany of these languages.)

In summary, few Balkan isoglosses bundle at the LA's borders; some failto reach all the Balkan languages, while others extend beyond. Of the strongestBalkan features, the postposed article is not in Greek; and the absence of theinfinitive is highly varied in its distribution within the dialects of the languagesfew qualifying completely with full absence, and several failing categorically(Geg Albanian, Istro-Rumanian etc.) Depending on Serbo-Croatian's Balkanstatus, loss either extends beyond the LA to several Serbian dialects, or failsto reach other dialects within the LA. The concept of an areal core from whichisoglosses expand outward (Masica, 170-71; Joseph, 245) seems inappropriatefor the Balkans. The language with the greatest number of areal features isRumanian, on the northern border of the area; but Macedonian, the languageconsidered most typically Balkan (Hamp 1977:281), lacks several of the arealtraits. In sum, Traits (a)(c) are strong areal indicators, but are not shared byall Balkan languages. Traits (d)(f) may be good areal features, but are lesspersuasive, and are not distributed throughout the area in every case. Traits(g)(j) seem weak. MA certainly compares well to these data. (For Balkaninformation, see Asenova 1977, Berns

campbell&kaufman&smith.1986.mesoamerica as linguistic area

Documents