Shevoroshkiп Festschrift Does Altaic Exist? Joseph Н. Greenberg Stanford University Since the writings of Clauson, and more recently Doerfer, it appears that most specialists in the Altaic languages по longer believe that the three grou~s of traditional Altaic, namely Turkic, Mongolian, and Tungusic, are related; thelr resemblances are to Ье attributed to borrowing, or in some cases to accident or sound symbolism. The term 'traditional Altaic' is here used purposely, that is, without reference to Korean, Japanese, or for that matter Uralic 1. This is not because 1 believe that the АНтс languages are geneticaHyisolated. In fact, in ту view (Greenberg 1987а: 332), they belong to а тисЬ larger grouping, Eurasiatic, along with other languages besides those just mentioned аЬоуе. Moreover, considerations deriving from these wider connections will figure in some instances in an essential way in the foHowing discussion. There are two separate questions involved here. Are the Altaic languages related to еасЬ other? If they are, do they constitute а valid genetic grouping, that is, а set of languages which Ьауе а single exclusive common ancestor, Proto-Altaic, which gave rise to three groups oflanguages and по others? 1 Ьеliеуе that the answer to the first question, that of mere relationship, is overwhelmingly positive. That to the second is more difficult, but оп the balance 1rather strongly endorse а positive answer here also. Recently in several publications, MiIler (1991а, 1991Ь) has defended the traditional view. Нis arguments are largely phonological, especiaHy the existence of two reconstructed pairs of liquid phonemes 11'12, rl' and r2, which within Altaic are only distinguished in non-Chuvash Turkic. MiIler believes that /1 and 12 Ьауе separate reflexes in Japanese. There are also instances in which Turkic merges а number of phonemes in j, namely d, j, п, and пУ. In such instances in order to account for the usual anti-Altaicist scenario in terms of borrowing from Turkic into Mongolian (with some reverse borrowing) and then from Mongolian into Tungusic, the borrowing has to Ье pushed back to а time so early that it becomes indistinguishable from Proto-Altaic, that is, when Turkish still distinguished d,j, п, and пУ,and аНthe Altaic languages outside of non-ChuvashTurkic displayed а difference between /1 and 12as weHas rl and r2' At such а time the languages would аН Ьауе had а sound system which is identical with that reconstructed Ьу Ramstedt, Рорре, and others for ProtoAltaic. It seems clear to те that languages like Korean, Japanese, and Uralic stand apart from traditional Altalic. Thus, Рорре (1960: 8), who includes Korean, shows it зs а separate branch from the rest of Altaic, and it figures comparatively infrequently in his etymologies. 89 MiIler also alludes to the cogency of the grammatical data regarding verb derivation in Ramstedt (1912) and Рорре (1973). 1 agree with him оп аН of this, but 1 believe that Ье has omitted the most powerful evidence of аll, that based оп personal, demonstrative, and interrogative pronouns. This material is, of course, familiar, but the anti-Altaicists Ьауе, as will Ье shown, carefully avoided presenting it in а coherent way, and where they Ьауе, Ьауе sought to explain it away in ап unconvincing fashion as the result of factors other than соттоп genetic inheritance. 1 will begin with the first and second person pronouns. In the first person singular in non-Chuvash Turkic, some languages, e.g. Osmanli Turkish, Ьауе nominative singular Ьеп and а stem Ьеп- which, except for ап internal variation in the dative (Ьапа), is found in аll the oblique cases. Most Turkic languages, however, Ьауе теп rather than Ьеп, and аll Ьауе -т as the first person singular marker in verb forms. The fundamental form then is те-п, in which -п (often called pronominal п Ьу Altaicists) has as its original function а mark of the oblique, ultimately of genitive origin. In non-Chuvash Turkic, this -п has spread analogically to the nominative. In Chuvash, however, which represents а separate branch of Turkic, this did not occur. ТЬе nominative here is е-ре in which е is а deictic element, and the oblique stem is тап-. This irregular alternation between nominative and oblique recurs in Mongolian in which the nominative is Ы and the genitive min-u and Tungusic, e.g. Evenki, with nominative Ы and genitive min-i. The forms теп and min are тисЬ more widespread than Altaic, including Uralic (e.g. Finnish minii 'Г) and Indo-European. Indo-European appears here as ап importantlink in this chain. Оп the basis of Baltic, Slavic, and Indo-Iranian, Szemerenyi (1970: 197) reconstructs *тепе for the genitive. In ВаШс and Slavic, the form in -п has been extended to аНthe oblique cases as in Altaic. . ТЬе Indo-European evidence is important because it provides а confirming instance for the oblique case function of the form in -по This is presumably the same -п which occurs in the oblique cases of r/n stems2. The Indo-European independent nominative is а suppletive form but different from that of Altaic, патеlу e-g(h)o-m, whose most closely related from in Eurasiatic is Chukchee ig т/e-g т (vowel-harmony variants) 'Г (cf. i-g t/e-g t 'thou'. Forms without the initial vowel occur as boundobjects). Returning to Altaic, it is clear that the probability of ап irregular alternation such as bi/men occurring three times Ьу accident is infinitesimal. That it should Ье borrowed twice is also utterly improbable. оое has literally to scour the earth to find а few instances of а borrowed pronoun,тисЬ less ап entire irregular 2 Тhe obliquie -n, and indeed аll the grammatical elements here were discovered Ьу the Nostraticists. See especially the tables in Il1ic-Svityc (1971: 6-18). 1 discovered these independently at а time when 1 was по! aware of Nostratic. In some instances, of course, 1 Ьауе found additional support, especially in languages not included in "classical Nostratic," but often accepted now as Nostratic, e.g. Chukchi-Kamchatkan and Eskimo-Aleut. 90 Shevoroshkiп Festschrift Greenberg. Does Altaic Exist? singular pronoun: alternation in pronouns. Ву itself it is enough to sh.owthat th~Al.taic.languages are related, moreonver the specific innovation of Ы ш the поmшаtlvе I.Sconfi~ed to these languages. Therefore it сап ье considered а shared со.~топ шпо.vаtюп within Eurasiatic that contributes to the establishment of trаdltюпаl AltaJC as а valid genetic entity. . How is this evidence treated Ьу Clauson апд Doerfer, the two lеаdшg exponents of the anti-Altaicist position? It is ignored where P?ssible. In Clauson (1969: 38), which applies glottochronology to the Alta.lc pr?blem, discussion is unavoidable since '1' is part of the glottochronologlcal Ilst. Не seeks to argue away the three-fold resemblances, indicated Ьу italicized entries, among 01д Turkish, 01д Mongolian, and Manchu, the three languages Ье utilizes in his study as follows: Indeed, еуеп such ап apparently clear comparison as Mongolian Ы - Tungus Ы is not convincing оп closer examination, since the Mongolian fonns (оп account of the plural bi-da, cf. e-de 'these', te-de 'those) goes back to bi. А typical case of sound symbolism (Elementarverwandschaft), surface resemblance, but without the possibility of а connectionЬу soundcorrespondence. What Doerfer is saying is that Mongolian i, which has two sources in а system of back-front vowel harmony, must derive from а high back vowel, not а high front vowel, because of the vowel of the second syllable -00 which is а back vowel. What Doerfer fails to point out is that Mongolian bida is а first person inclusive plural. Now it is а worldwide typological fact that where there is а first person inclusive/exclusive distinction in the plural, the exclusive, when analyzable, is the plural of the first person. Тhis is so in Mongolian, in which the first person is Ьа, with а perfect parallelism between the first апд second It is known (but has not been explained ир to now) that there are phonetic resemblances between personal pronouns in languages which are completely unconnected wit~ еасЬ o~~er, e.g. between 'mine', German meiп апд the Turklsh gешtlvе meniп (from Ьеп) and Mongolian mino [sic!) from Ы; between Latin tu and Mongolian с; (*ti). ТЬе phonetic resemblances between Turkish, Mongolian, and Tungus-Manchurian in regard to these lexical items cannot Ье therefore recognized as probative. This reasoning, which is very common, is to депу the significance of а resemblance because it is found somewhere else. This was used Ьу Michelson against Sapir in regard to п first person, т second perso~ in Algi~ because it occurs in so many other Amerind languages. It would Ье just as 10glcal to deny the significance of the resemblance between English 'mine' and <?еп.nап.mein because it also occurs in Mongolian. Опе has to pursue the full dlstпЬиtюп of these forms. As soon as one gets to Sino-Tibetan or Nilo-Saharan, or many others, it ceases. Both the Nostraticists and 1 include Indo-Europeanand Altaic in the same group. [For another perspective оп widespread similarities in pronominal systems, see Rhodes, this volume. -Eds.] In addition, Clauson, Ьу simply using the nominative as the translation fonn for the glottochronological list, fails to consider the agreement. betw~en Mongolian and Tungusic in the bi/min- alternation, and Ьу not шсlиdшg Chuvash does not Ьауе to account for the threefo1dagreement in ап irregularity among the three branches of Altaic. And what of the second person singular pronouns? ТЬеу are not discussed at аll. Clauson unaccountably does not italicize 01д Turkish sen and Manchu si as resemblances to Ье explained, or rather explained away, in spite of their complete parallelism with Old Turkish Ьеп апд МапсЬи Ы. Old Mongolian tere, МапсЬи tere 'this' are italicized but passed over without comment. Doerfer in general fails to discuss grammatical resemblances, but in his Moпgolo-Tuпgusica (1985: 2), Ье says the following about the first person 91 persons, f l' .'r I I r " Ы:Ьа = ci <*ti:ta. Оп the other hand the first person inclusive is either а separate fonn unlike either the first or second person singular, or it is а combination of the two like Tok Pisin yu-mi. Непсе bi-da is very likely а сотроипд of singular Ы with ta second plural. In compounds vowel harmony пеед not apply. А parallel situation is found in Tungusic, in which most languages Ьауе а first person plural inclusive/exclusive distinction in which the exclusive is the plural of the singular. ТЬе same parallelism reigns here as in Mongolian between the first personand the secondperson,e.g. Evenki Ы:Ьи =si:su. ТЬе first inclusiveis here even more obviously а сотроипд, e.g. ЕУепю mi-ti, mi-t (Tsintsius 1949: 270-1). Note also that Doerfer fails to mention the striking parallelism between the nominative апд oblique stems in the first person among Mongolian, Tungusic, апд Chuvash. We are to believe that Mongolian Ы here is not cognate with the Tungusic апд Turkic forms in spite of the agreement between them in parallel irregularities. Characteristic also is Doerfer's resort to sound symbolism. This is допе without апу supporting evidence. Surely Ь- is not particularly frequent as а first person singular in languages of the world, nor is there апу plausible supportin sound imitationor other sources of Elemeпtarverwaпdschaft. Finally, it should Ье noted that violations of back-front vowel harmony are not ипсоттоп in Uralic, а universally accepted family, and in etymologies which are obviously valid оп other grounds. As late as 191О, Szinnyei, in his reconstruction of Proto-Finno-Ugric, resorted to а kind of majority rule to detennine whether back or front vocalism was the original type in Proto-FinnoUgric. Еуеп now there are uncertain instances. А parallel situation exists in Turkic. As noted Ьу Radloff (1882: 84) there are variations in stem vowels 92 without апу demonstrable cause. In fact there is ап article Ьу Dmitrijev оп this topic, (1955: 115) in which Ье observes that sporadic alternations in the same root of vowels of the front and back series is frequent in individual Turkic languages. Another опе of the very few grammatical etymologies in Doerfer (1985: 27) is his по. 66, the interrogative stem уа- of Mongolian and Tungus. Не admits that it "behaves like а genetically related word." Опсе more Ье resorts to "sound symbolism" and again his опlу support is Indo-European *jo. But this is а widespread Eurasiatic interrogative (cf. Greenberg 1987Ь). Опсе more we Ьауе the ad hoc resort to а highly implausible sound symbolic argument without апу serious documentation. Finally, what of the second person pronouns? ТЬеу are passed over in complete silence. Doerfer, like Clauson, believes that Mongolian borrowed massively from Turkic, and then Tungusic from Mongol. Не is clearly disturbed Ьу the existence of certain etymologies соттоп to Turkic and Tungusic and devotes а section to them (1987: 238-41), but Ье fails to mention the most glaring instance of аll, the agreement of Turkic and Tungusic in ап s second person as against Mongol t. Of course, if 1 ат right in ту discussion of the Mongol and Tungusic first person inclusive pronoun, t would also occur in Tungusic, but in а quite different context. Both s and t are widespread second person Eurasiatic pronouns. For ехатрlе, we find Indo-European t in the independentpronounand plural verb endings and s as а singular verb suffix. In general there are а considerable numberof othergrammatical markers соттоп to аll the Altaic branches, most of them entirely ignored Ьу Doerfer. However, virtually аН these are found in other branches of Eurasiatic. The number of these as well as the lexical evidence makes the relationship of the Altaic languages а certainty. However, the distinctness of Altaic as а valid subgroup, which is most conspicuously supported Ьу the bi/min alternation in the first singular pronounrequires furtherassessment, а task not undertakenhere. References Clauson, а. 1969 Dmitrijev, N.к. 1955 93 Greenberg, J.H. 1987а Language in the Americas. Stanford: Stanford University Press. 1987Ь "Relative pronounsand P.I.E. word order type in the context of the Eurasiatic hypothesis", in Winfred Lehmann, ed., Language Typology 1987, Systematic Balance in Language,: 123-38. Amsterdam: Benjamins. Illic-Svityc, У.М. 1971 Opyt Sravпeпija пostraticeskix jazykov. Moscow: Nauka. Miller, Roy Andrew 1991а "Genetic connections among Altaic languages", in Sydney М. Lamb and Е. Doиglas Mitchell, eds., Spruпg from Some Соттоп Source: 293-327. Stanford: Stanford University Press. 1991Ь "Anti-Altaicists contraAltaicists". Ural-Altaische Jahrbacher. 63: 5-62. Рорре, Nikolaus 1960 1973 Radloff, W. 1882 Ramstedt, G.J. 1912 Vergleichende Grammatik der altaischeп Spracheп. Teil1: Lautlehre. Wiesbaden: Otto Harrassowitz. "Uber einige Yerbalstammsbildиngsaffixe in den altaischen Sprachen". Orieпtalia Suecaпa, 21: 119-41. Phoпetikder пordlicheпTarkspracheп. Leipzig: Weigel. "Zur Уerbalstammbildungslehreder mongolisch-ttirkischen Sprachen. Jourпal de la Societe Fiппo-Ougrieппe,28.3: 1-86. "Leksikostaticeskaja otsenka altajskoj teorii". Voprosy Jazykozпaпija. 5: 22-51. Moscow. Szemerenyi, Oswald 1970 Eiпfiihruпgiп die VergleicheпdeSprachwisseпschaft. Darmstadt:Wissenschaftliche Buchgesel1schaft. "Ceredovanie glasnyx zadnego i perednogo rjada v odnom i tom Szinnyei, Joszef 1910 ze otdel'nyx tjurkskix jazykov." In N.к.Dтitrijеv, ed. Issledovaпija ро sravitel,пoj grammatike Tjurkskix jazykov. cast' 1: Foпetika: 115. Doerfer, Gerhard. 1985 Shevoroshkin Festschrift Greenberg, Does Altaic Exist? Moпgolo-Tuпgusica.Wiesbaden:Harrassowitz. Tsintsius, Y.I. 1949 Fiппisch-ugrische Sprachwisseпschaft. Leipzig: G6schen. Sravitel,пaja Foпetika tuпguso-maп'czurskixjazykov. Leningrad: UCpedgiz.