You are on page 1of 4



BERNARD COМRIE University о/ Southeт Califoтia

1. Nostratic and Altaic pronoиns, 1 Linguists advocating wide-ranging genetic affiliations of languages are divided into two groups Ьу their attitиde to one of the mainstays of traditional comparative linguistics: the regиlarity of sound change, the so-called NeoGrammarian hypothesis. Оп the one hand, Joseph Н. Greenberg and his followers Ьауе denied the irnportance of regиlar sound correspondences in formulating hypotheses about genetic relatedness; see, for instance, chapter 1 of Greenberg (1987). Ву contrast, lingиists of the Moscow School of comparative linguistics adhere to the same criteria of regиlar sound correspondence in their work оп wide-ranging genetic hypotheses like Nostratic as when they work оп
Indo-Eиropean and its branches


certainly in principle,

if not always


practice; see, for instance, ту review of Starostin 1991 (Comrie 1993). One of the staunchest proponents of the Neo-Grammarian hypothesis in long-range comparison is Alexander Vovin, and it is his contributionto this volиme that has inspired те to the present reflections. More specifically, there is one particularpoint of Vovin' s paper that seems to те to ье of great methodologica1importance. As will Ьесоте clear, 1 disagree with the conclusion that Vovin himself draws, but 1 Ьоре that even those who side with Vovin rather than with те will find it useful to work through ту reservations, to clarify our respective positions. Scholars who Ьауе noted similarities among Indo-European, Uralic, and Altaic languages Ьауе usually commentedin particu1arоп the similarities in the personal pronouns. And as Vovin shows, in particular for the first and second persons singular, there are striking similarities: For '1' Proto-Indo-Europeanand Proto-Uralic Ьауе initia1*т, while Proto-Altaic has an alternation of initia1 *Ь and *т. For 'thou', Proto-Indo-Europeanand Proto-Uralic Ьауе initia1*t, while Altaic shows а split between Proto-Mongolic with initia1*t versus Proto-Tиrkic and Proto-Manchu-Tunguswith initia1*s. But Vovin goes оп to show that the Proto-Altaic forms cannot ье related Ьу regиlar sound change to those of Proto-

for the development of ёоwtо "уои"..In the second person. such as the fact that Old English ёоw is accиsative-dative plura1 while Modem English "уои" is иs now sиbject these forms to the kind of ana1ysis Vovin proposes for the Altaic pronouns. which survives. the voiced plosive is the origina1form in Altaic. Тhis leaves five Modem English forms that do not derive Ьу regиlar soиnd change from their Old English etyma. and indeed even the initia1*t of Proto-Mongolic is perhaps not regularly relatable to the initia1 *t of Proto-Indo-European and Proto-Ura1ic. compare the regиlar development of the same initia1 seqиence in Old English sёо]хт to Modem English "seethe". other conclusions that one might conceivably draw.the first of the two a1tematives suggested in this paragraph. Exarnination of accoиnts of the history of English would sиggest that еасЬ of the other Modem English pronoиns has traditiona11ybeen taken to derive etymologica11yfrom the Old English сопеsроndеnt given in table 1. deriving from Old English wIc 'abode'. since 1 am taking for granted morphologica1and semantic shifts that do not affect the soиnd changes involved. First. the clearly ana1ogica1Modem English genitive "its" is not inclиded. this is the position that 1 coиld imagine а Greenbergian taking. unequivoca11y. namely Old and Modem English. in comparison with those of Old English. and therefore indeed do not descend directly from their Old English сопеsроndеnts. "him".еd Ьу Vovin and the. (1 will. Bиt of coиrse . since it is widely accepted that they are loans from Old Norse. is the most we can safely assиme for the data presented Ьу Vovin.sound changes relating them. Second.272 BERNARD COMRlE REGULAR SOUND CORRESPONDENCES 273 Indo-European or Proto-Uralic: In the fшt person. "her". with the same etymologica1 origin. "уои". one might say that it is impossible to reach а conclиsion оп the basis of the evidence available: Тhe forms are sиfficiently similar that they might ье related. 2. (In dia1ects that drop word-initia1 h regиlarly . по regиlar soиnd change will delete the [ша1 п. one might conclude that the similarities are stillsufficiently striking to justify maintaining that the pronouns are cognate in all three groups of languages. 1 ana1yze а similar set of forms in two langиages separated Ьу а very sha110wtime depth. incidenta11y. oиr their ТаЫе 1: Old Eпglish aпd Modeт Eпglish persoпal proпouпs l. Тhere are. Тhe following forms derive regиlarly: "те". "his". No regиlar soиnd change will delete the word-initia1h of Old English hit to give Modem English "it".and nиmber-neиtra1. in "mine". Modem English "she" does not derive regиlarly from sёо. In section 3.or that Old English sёо is асtиallу а demonstrative rather than а persona1pronoиn.) ТЬе conclusion that Vovin draws from this is that. the initia1*s is not relatable to the initia1 *t of Proto-Indo-European and Proto-Ura1ic. however. or at least Ьауе been inflиenced in their forms Ьу Old Norse. so it is асtиallу the passage from Old to Мiddle English that is at issue.there is the resиlting impression that "it" is а regиlar reflex of Old English hit. even accepting the basic сопесtnеss of the Nostratic hypothesis. For the same reason. 1 sиggest that the second a1temative. (Most of the crucia1 changes had taken place Ьу Мiddle English.) ТЬе forms to ье considered are set oиt in table 1. In the case of Old English тIn.and this inclиdes most regiona1dia1ectsof England . "oиr". Тhe fina1с of Old English ic shoиld not Ьауе dropped to give Modem English "Г'. Bиt if we impose the requirementthat such relationships mиst ье shown to ье consistent with established regиlar soиnd changes. compare the Modem English placename ending "-wich". "we".the Altaic forms are not genetically related to those 01 Indo-European and Ura1ic. even if Modem English spelling has rather sиccessfully disgиised the three-way homophony in Modem English. argиing that here the weight of additiona1 considerations favors maintainingthat items are cognates even in the absence of perfect sound сопеsроndеnсеs . namely agnosticism. In section 2. "yoиr". Тhe question is simple: Do the Modem English forms derive from the Old English forms? 1 will exclиde the third person plura1forms from fиrther consideration. the pictиre changes. bиt the lack of regиlar сопеsроndеnсе throws some doиbt at least оп this claim. "Ье".for the sake of argument assume the сопесtnеss of the Indo-European-Ura1ic-Altaiccognates аssuП). Old Eпglish aпd Moderп Eпglish proпouпs For comparison. and let the iпеgиlarity in soиndсопеsроndеnсеs pass. Note that the Old English forms do not necessarily сопеsроnd exactly in function to the Modem English reflexes. compare that of Old English ёow (а homophone) to "yew" and of Old English ёоwе to "ewe". ic ёоw Ьё sёо hit Old En lish те mш ёоwеr him his hire Modem En lish ту те your him his her иs them. and is not regularly relatable to the nasa1 of Proto-IndoEuropean and Proto-Ura1ic. of course. 1 will consider the etymology of the persona1 pronoиns of Modem English.

thoиgh it is plosive in Altaic and nasa1 in Indo-Eиropean and Ura1ic. Тhe differences. namely Indo-Eиropean and Ura1ic? ТЬе only honest answer 1 can give is: 1 do not know. while for Vovin they are reason for rejecting proposed etymologies). rather than morphologica11y conditioned. even here this is illиsory: ТЬе loss of the initia1consonant of the pronoиn hit is тисЬ earlier than the genera1 loss of word-initia1h in the regiona1 speech of England. the few irregиlarities do not sиbstantia11y a1terthe pictиre: ТЬе Modem . Their relatedness is certainly cast into doиbt Ьу the lack of regиlar correspondence. First. One can certainly соте ир with specиlative scenarios that woиld give the cиrrent distribиtion. shift is perhaps somewhat less readily explainable. it is not иnlikely that the great Norse and French . Bиt the evidence that is available to иs jиstifies по position stronger than agnosticism. where there is an abиndance of clear-cиt evidence for genetic relatedness that does fol1ow regиlar soиnd changes. i. jиst as we accepted the traditiona1accoиnt of the etymology of the Modem English pronoиns. which has. incidenta11y.these langиages do not provide plaиsible etyma. Both sets of second-person pronoиns have an initia1voiceless denta1obstrиent. иnlikely to ье forthcorning)rnight we111ead иs to accept their genetic relatedness. This is precisely the conclиsion that Vovin reaches in comparing the Altaic and Nostratic persona1pronoиns:''Тhиs. assиrning that the relevant regиlar correspondences are established оп the basis of а large enoиgh nиmber of lexica1 items. 2 In the case of these English pronoиns. one is ready to take the slight leap stil1 remaining to relate the Modem English and Old English forms. are not аН that great phonetica11y: Both sets of first-person singиlar pronoиns have an initia1voiced labia1. with different branches genera1izing different a110morphs. the short vowel cannot derive regиlarly from the Old English long vowel.para11elingthat between preconsonantal "а" and prevoca1ic "an". their etymologica1relatedness to their Old English correspondents seems incontrovertible. a1thoиghwe have а good knowledge of the langиages that rnight plaиsibly have been expected to ье a1temative soиrces for the etymologically irregиlar pronoиns pиtative branches of Nostratic.g. So the sirnilarities are actиaНyqиite striking. thoиgh it is fricative in Altaic (except Mongolic) and plosive in Indo-Eиropean and Ura1ic. not even with the degree of certainty with which we сап clairn the genetic relatedness of the Indo-Eиropean and Uralic forrns. e. А nиmber of factors conspire to lead to this conclиsion. - in particиlar. and this pair para11eled Ьу the archaic "thy" . we woиld ье forced to conclиde that these foиr Modem English forrns are not еtyЧ1010giса11у related to the Old English correspondents in table 1. Old 3. So we certainly cannot say with any degree of certainty that the Altaic and other Nostratic pronoиns are genetically related. Bиt let иs now retиm to the Altaic and Nostratic pronoиns discиssed Ьу Vovin. Лrе the Altaic pronoиns (exclиding Mongolic) relatable etymologica11yto those of the other time depths involved woиld have obliterated аН trace of the conditioning mechanism. which coиld accoиnt for the shorteningof the vowel of "иs" and the loss of initia1h in hit. sиch as that'separating Old English from Modem English. However." 1 noted above that the traditiona1view of the five etymologica11yanoma1oиs Modem English pronoиns is that they do indeed develop historica11yfrom their Old English correspondents in table 1. a1as. if we take regиlar soиnd correspondences to Ье essentia1 for establishing cognates. 1 simply a110wexceptions to proposed regиlar soиnd changes to stand in the case of sha110wtime depths. At this point. With shallow time-depths.then.) In the case of Modem English "иs"."thine".We don't know enoиgh of the details of other langиages that rnighthave been aroиnd at the time to know if there are plaиsible non-Nostratic etymologies. Second. thoиgh it is para11eled Ьу the development of the adjective sиffix "-ly" from Old English liс/Пс. In other words. that the differing initia1s were origina11y morphologica11y conditioneda1temantswithin the common proto-langиage.its regиlar reflex in the possessive adjective "oиr". ТЬе ic to ''Г. 1 don't see how we can say that the Altaic pronoиns are definitely ишеlаtеd etymologicaНy to their Indo-Eиropean and Ura1ic coиnterparts. Pronoиns are often иnstressed. Nostratic aпd Altaic proпoипs. it is difficиlt to see any plaиsible reason why sиch а discrepancy in nasa1izationor plosion shoиld have arisen - bиt even if there were some motivation. 1 can see an advocate of wide-ranging genetic comparison accиsing те of doиble standards. when one observes that in earlier stages of the langиage the difference between the members of еасЬ pair was phonologica11y. When "ту" is matched with "rnine". withoиt affecting the acceptance of hypotheses of genetic relatedness. the Modem English forrns are only minima11y different from what woиld have been the oиtcome of regиlar soиndchange. Indeed. and the differences are readily describable in terrns of plaиsible isolated changes. additiona1evidence (which is. however. Bиt eqиal1y. exceptions to proposed regиlar soиndchanges are reason at least for agnosticism (ту position. the inescapable conclиsion is that Р[roto-]A[ltaic] pronoиns are иnrelated to Indo-Eиropean or Ura1ic personal pronoиns. Bиt 1 think the differing positions simply reflect the different weight of evidence in the two kinds of cases.274 BERNARD COMRIE REGULAR SOUND CORRESPONDENCES 275 . In "she" we have plaиsibly а shift in the nиcleиs of the diphthong from the first to the second component fol1owed Ьу the freqиent a11egrophenomenon of /sy/ becoming /8/. And 1 think this traditiona1 view is correct. Bиt as the time-depthincreases.

it will often ье impossible to decide whether а close similarity is а perturbed cognate or а freak coincidence. REFERENCES Comrie. Bemard. from its Old English counterpart. Moscow: Nauka. 1987. Starostin. Review of Starostin (1991). 1991. Language 69. Тhe pity is: There is по guarantee that we will ever know. and especia1ly where phonetic plausibility can ье invoked for the deviations from regular sound change . Greenberg. Altajskaja probleтa i proisxoideпie japoпskogo jazyka. a1beitirregularly. 1993.after considering the a1tematives (often there are попе).828-832. Stanford: Stanford University Press. the most reasonable solution . Басh wide-ranging genetic hypothesis is of course either true or fa1se. where there are apparent slight exceptions to regular sound changes. whether for the regular sound changes or for perturbingfactors that might upset those regular developments. Joseph Н. Language in the Aт to hypothesize that indeed the item in question derives. But where the evidence is less substantia1.276 BERNARD СОМЮЕ English basic vocabulary derives primarily from Old English. Sergei А. .