(cid:57)(cid:82)(cid:90)(cid:72)(cid:79)(cid:3)(cid:43)(cid:68)(cid:85)(cid:80)(cid:82)(cid:81)(cid:92)(cid:3)(cid:76)(cid:81)(cid:3)(cid:48)(cid:68)(cid:81)(cid:70)(cid:75)(cid:88)(cid:29)(cid:3)(cid:36)(cid:3)(cid:38)(cid:85)(cid:76)(cid:87)(cid:76)(cid:70)(cid:68)(cid:79)(cid:3)(cid:50)(cid:89)(cid:72)(cid:85)(cid:89)(cid:76)(cid:72)(cid:90) (cid:36)(cid:88)(cid:87)(cid:75)(cid:82)(cid:85)(cid:11)(cid:86)(cid:12)(cid:29)(cid:3)(cid:45)(cid:82)(cid:86)(cid:75)(cid:3)(cid:36)(cid:85)(cid:71) (cid:54)(cid:82)(cid:88)(cid:85)(cid:70)(cid:72)(cid:29)(cid:3)(cid:45)(cid:82)(cid:88)(cid:85)(cid:81)(cid:68)(cid:79)(cid:3)(cid:82)(cid:73)(cid:3)(cid:47)(cid:76)(cid:81)(cid:74)(cid:88)(cid:76)(cid:86)(cid:87)(cid:76)(cid:70)(cid:86)(cid:15)(cid:3)(cid:57)(cid:82)(cid:79)(cid:17)(cid:3)(cid:21)(cid:19)(cid:15)(cid:3)(cid:49)(cid:82)(cid:17)(cid:3)(cid:20)(cid:3)(cid:11)(cid:48)(cid:68)(cid:85)(cid:17)(cid:15)(cid:3)(cid:20)(cid:28)(cid:27)(cid:23)(cid:12)(cid:15)(cid:3)(cid:83)(cid:83)(cid:17)(cid:3)(cid:24)(cid:26)(cid:16)(cid:27)(cid:19) (cid:51)(cid:88)(cid:69)(cid:79)(cid:76)(cid:86)(cid:75)(cid:72)(cid:71)(cid:3)(cid:69)(cid:92)(cid:29)(cid:3)Cambridge University Press (cid:54)(cid:87)(cid:68)(cid:69)(cid:79)(cid:72)(cid:3)(cid:56)(cid:53)(cid:47)(cid:29)(cid:3)http://www.jstor.org/stable/4175719 . (cid:36)(cid:70)(cid:70)(cid:72)(cid:86)(cid:86)(cid:72)(cid:71)(cid:29)(cid:3)(cid:20)(cid:28)(cid:18)(cid:19)(cid:24)(cid:18)(cid:21)(cid:19)(cid:20)(cid:20)(cid:3)(cid:20)(cid:21)(cid:29)(cid:24)(cid:20) Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=cup. . Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Cambridge University Press is collaborating with JSTOR to digitize, preserve and extend access to Journal of Linguistics. http://www.jstor.org J. Linguistics 20 (I984) 57-80. Printed in Great Britain Vowel harmony in Manchu: a critical overview JOSH ARD University of Michigan (Received9 August I983) ABSTRACT The history of generativea ccountso f vowel harmonyi n ClassicalM anchu is discouragingT. he earlyt reatments( Vago, I973; Odden,1 978; Finer,1 979) assumedt he wrongt ypeo f vowelh armony( frontv s. backr athert hanr elative height harmony).H ayata (I980) proposedt he correctt ype, but was unable to justifyi t adequatelyw ithint he generativef rameworkT. heses hortcomings are due to deficienciesi n the practice of generative phonologists. To demonstratet he correctt ypeo f vowelh armonyi n ClassicalM anchur equires considerationso f (i) the actual surfacer epresentationa nd/or (2) facts in relatedd ialectsa nd languages.N eithert ype of evidencei s frequentlyu sed in the practiceo f generativep honologists,a lthought he formeri s almosta lways valuedi n theory.A n implicationi s that attemptst o improvep honologym ust weigh practicea s well as theory. One of the majorb attlegroundsb etweenp roponentso f abstractp honology and proponentso f concretep honology has been the descriptiono f vowel harmony in Classical Manchu. Proponents of abstract phonology have claimedt hat vowel harmonyw as a livingp rocessi n this (now dead)l anguage which requiresa bstractr epresentationtso capturet he true generalizations. Proponentso f concrete phonology, on the other hand, have denied that abstractr epresentationas re neededt o describeM anchu,w ith some, but not all, denyingt hat vowel harmonyi s productivei n ClassicalM anchu. There is nothing particularlyn ovel in the argumentst hat have been presentedw ith regardt o abstractnessin Manchu.T he same types of claims and counter-claimtsh at have beenm adei n otherd isputesa bout abstractness have been made in this case as well. The purposeo f this paperi s not to join the fray of this battle, but rathert o providem eta-commentaroyn the entire dispute.W hati s interestinga boutt he controversya, t leastu p untilA rd (I979) and Hayata (I980), is that proponents of both abstract phonology and concretep honologya ssumedt he wrongt ype of vowel harmonyf or Manchu (or at least pre-Manchui f the process had ceased to be productivei n the classical variety of the language). Vago (I973), Odden (I978), and Finer (1979) all assumedt hat vowel harmonyw as of the front vs. back variety,a s is found in Turkish. Ard (I979) and Hayata (I980) point out that the correct type of vowel harmonyi n (pre-)Manchuw as high vs. low. 57 JOSH ARD As will be demonstratedin detail below, there is overwhelminge vidence that the type of vowel harmonyi n (pre-)Manchuw as in fact high vs. low. However,t he purposeo f this paperg oes well beyond this into questionso f phonologicalp racticea nd phonologicalm ethodology.T he evidencet hat is crucialf or decidingt he questioni s of typest hat are not generallyc onsidered in generativep honologicalp ractice.F urthermoret,h e types of evidencet hat areg enerallyc onsideredin generativep honologicalp racticea ren ot conclusive in pointingt o one type of vowel harmonyr athert han the other. The evidencet hat is crucialf or decidingt his questioni s of two primary types. One type involvese videnceZ wicky( I975) categorizesa s 'additional evidence', evidencet hat not all generativep honologistsc onsiderr elevant. Here,f acts fromc loselyr elatedl anguagesa nd anotherd ialecto f Manchua re paramount.H ere one could arguet hat these facts are 'external't o Manchu. For example,a speakero f Manchuw ith no knowledgeo f theseo therd ialects or relatedl anguagesw ould have no access to these facts. The other type of evidencei s clearlyi nternalt o Manchu;i t concernst he actual surface (systematic)p honetic representationo f Classical Manchu words.I n theory,t he surfacer epresentationis crucialf or phonologicalt heory, sincee veryp honologyc ontainst he surfacea s one level of representationI.n this regard, Zwicky (1975: 154) suggests that all generativep honologists consider the 'variant shapes of morphemes't o be among 'the data to be comprehendedb y a phonologicala nalysis'.N evertheless,i n practiceg ener- ative phonologistsh ave not dedicatedm uch effortt o discovert he natureo f the phoneticso f the surfaces hape of morphemes. If we considert he actualp racticeo f generativep honologists,w e noticet hat Schane hardly goes beyond an ordinary phonemic notation and that McCawleyi n his descriptiono f Japaneseo nly includesp honologicalr ules, not featurei nterpretationru les.C homskya nd Halle. .. on the whole... do not go any furthert han an ordinaryp honemicl evel (Fischer-J0rgensen 1975: 217). Indeed,g enerativep honologistsh ave often taken existingp honemicd es- criptionso f languagesa nd deriveda generativea nalysiso nly by considering alternationsi n the phonemicr epresentationos f allomorphs.T his has been especiallyc ommoni n generativer eanalyseso f AmericanI ndianp honologies. Even though evidence about actual pronunciationsi s internal to the language,i t is externalt o the data which modernp honologistsm ust use in describingM anchu, namely written texts. The informationi n the written texts, chieflyo rthographicfa cts and facts about morphologicarl elationships whichc an be abstractedi,s insufficientto determinet he actualp ronunciation. The modernr esearchemr ustr elyo n additionale videncea nd/or assumptions to postulatea pronunciationf or a Manchuw ord. Hence, the terms internala nd external are confusing. To any native speaker of Manchu,i ncludinga speakero f a homogeneouss peechc ommunity,t he VOWEL HARMONY IN MANCHU actualp ronunciationis an internalf act: it is a part of Manchua nd is learned along with the attainmento f competencei n the language. To a modern phonologista nalysingM anchut hroughw rittent exts, the actualp ronuncia- tion must also be internalt o the language,b ut the phonologistsc annotl earn what the pronunciationw as without relyingo n informatione xternalt o the texts themselves.I n other words, to the modern phonologist, the actual pronunciationc ould be said to be ontologicallyi nternalt o Manchu, but epistemologicallye xternal. Externale videncef romr elatedl anguagesa nd a relatedd ialecta nd internal evidenceb ased on the actual sounds in Manchu both supportt he high vs. low analysis.A s will be discussedi n more detail below, therei s no evidence whatsoeverin favouro f frontv s. backv owelh armony.H encet he formert ype of vowel harmonyi s demonstrablyp resenti n (pre-)ManchuH. owever,t he type of evidenceg enerallyf ound in the practiceo f generativep honologists, morphologicala lternationsa nd the nature of phonologicalr ules, does not provide compellinge videncei n favour of either type of harmony.H ayata (I980) suggestst hat these latter types of data do in fact argue for high vs. low harmony but, as will be discussed below, these argumentsa re not convincing. The ascriptiono f the wrong type of vowel harmonyt o a languagew ould certainly have to be considereda grievous error. Yet the methods most commonlyu sed in the practiceo f generativep honologistsh ave been proven to be ineffectivea t preventingt his sort of error from arising.T he types of data usuallyc onsidereda re inconclusivew, hilet he conclusives ortsa re of the type frequentlyu ndervalueda nd under-used. I. ClassicalM anchur eferst o the languageo f the Manchuc ourt, the rulers of the Chingd ynasty,w hichc ontrolledC hina from the late sixteentht o the early twentiethc entury.M ore specificallyt, he languagei n particularr efers to documents written in the seventeentha nd eighteenthc enturies.T hus ClassicalM anchub y its veryd efinitioni s a writtenl anguage,a nd this creates many problems for a phonological analysis. For example, it is not clear exactly how much dialect mixture and standardisationis involved in the writtenf orm. Even though there are a numbero f people of Manchua ncestryi n China (Sunik (I966) notes that officialC hinesef iguresl ist 2.5 million),a pparently all those who are descendantso f clans that formed the original Manchu confederationn ow speak only Chinese.T herei s a group of speakersc alled the Sibe (or Sibo), who still speaka languagev eryc lose to ClassicalM anchu, but the Sibe were not a part of the originalc onfederations, o theirl anguage cannot be considereda descendanto f ClassicalM anchu. Scholarsd isagree on how to classifyS ibe vis-a-visC lassicalM anchu.N orman( I974) calls it a Manchu dialect, while Menges (I97I) refers to it simply as 'gesprochenes Man'u'. 59 JOSH ARD Manchu is a Tungusl anguage,r elatedm ost closely to languagess poken primarilyi n Siberia. The Tungus languagesf orm one branch of the still controversiaAl ltaic family.T he conservativeo pinion is that Altaic consists of Turkic,M ongoliana nd Tungusl anguages.S ome scholarsd eny that these three groupings are geneticallyr elated, suggesting that areal factors are responsiblef or the resemblancesO. ther scholarss upport a more inclusive Altaic family,i ncludingK oreana nd, for many, Japanese. 2. Therei s cleare videncef or residueso f vowelh armonyi n ClassicalM anchu orthographyF. or example,o ne collectives uffixh as the variants- sa/ -sel -so, another- ta/ te: (i) sakdar-sa' old person' da-ta 'chief' gege-se' older sister' eme-te 'mother' (data from Finer, 1979) On the basis of these and other examples,v owels in ClassicalM anchuc an be dividedi nto threeh armonicc lasses: (2) ClassicalM anchuh armonyt ypes I. e II. aoo III. (neutral)i u In generalw ords with vowels of type I requirea ffixesw ith vowels of types I or III, and wordsw ith vowelso f type II requirea ffixesw ith vowelso f types II or III. The vowel harmony system is furtherc omplicatedb ecause, for certain affixes, o is found instead of an expected a due to an additional requiremenot f roundingh armony.T he detailso f the vowel requirementisn affixesa re reportedi n great detail in Avrorin( 1976). The vowel harmonyc lassesa re also requiredi n predictingt he distribution of velar and uvularc onsonants.V elar consonants tend to occur in words containingv owels of type I, while uvular consonantst end to be found in wordsc ontainingv owelso f typeI I. Considert he adjectivizear ffixes- re/ -Na/ -NO and -xun/ -Zon. (3) batUNa' brave' wasixon 'vile' zoroNo 'powerful' wesixun' precious' gexurje' brilliant' (Odden, 1978: 150, I54) Therefore,i f we know that a word contains vowels of types I or II we can predict( i) vowelsi n affixesa nd( ii) the presenceo f velarv s. uvularc onsonants. However,n o predictionc an be madef or wordsc ontainingo nly neutralv owels (i and u): (4) biluNa' calm' mujirje' having a heart' 6o VOWEL HARMONY IN MANCHU 3. Therei s littled oubtt hatv owelh armonyw aso ncea productivep honological process in Manchu. Unfortunately, however, the facts that have been presenteda bove are insufficientf or uniquelyd eterminingth e preciset ype of harmony. Ultan (1973) presentst he results of a cross-linguistict ypology of vowel systems.V owel harmonys ystemsc an be complicatedd ue to variousf actors such as differencesb etweenh armonyi n roots and affixes,s kewingo f vowel patterns,t he presenceo f neutralv owels,e tc. NeverthelessU, ltan was able to discover three primary types of vowel harmony patterns based on the particularp arametert hat predictst he kind of vowels found in each of the harmonicc lasses.H e found that languagesd istinguishedv owelsb asedo n the parameterso f (i) frontness, (ii) tenseness, and (iii) height. Consider the followinge xamples. (5) Major types of vowel harmony A. Front-back Kirghiz( Turkic) I. (a) i (b) ul II. (a) i (b) u e o a o (Kirghiza lso has roundingh armony:v owels must not only all be from I or II, but also must all be from (a) or (b).) B. Tense-lax( a type of horizontalh armony) Nzema (EasternA kan) I.i u II.i o e 0 3a a a (Carefulp honetics tudieso f the parametersin volvedi n this type of harmonyh ave been conductedb y Lindau( I975) and Jacobson (1978).) C. High-low( a type of horizontalh armony) Nanai (Tungus) I. i u II.i 0 a Based only on the symbols in the Manchu words given above it is impossiblet o determinet he preciset ype of vowel harmonyt hat is witnessed in Manchup honology.T he symbolst hemselvesa re basedo n transliterations from the Manchu orthographics ystem, which itself was developed from Mongolians cript.E vent his transliterationis controversials, incet he Manchu scripti s repletew iths ymbolsa ndd iacriticsw hicha llowm ored istinctionst han any phonologisth as been willingt o recognizea s distinctive( see Ligeti (I953) and Melles Whilet he transliterationit selfi s not crucial,t he phonetic (I975)). values of the sounds are. Depending on the exact pronunciationso f the sounds, any of the three types of vowel harmonyc ould be motivatedf or Manchu.F or example,i f the complexs ymbolo representsa front or central JOSH ARD vowel, then it will be difficultt o arguet hat the type of vowel harmonyf ound in Manchuw as of the front-backv ariety,s ince o is groupedw ith the back vowel o. Moreover,i t is importantt o determinei f the other symbols (i, e, a, o, u) representv owels reasonablyc lose to IPA cardinalv owel values. If one or more of them does not, then this could be crucialf or determiningth e correct type of vowel harmony which categorizedC lassical Manchu (or perhapsp re-ClassicaMl anchu). 4. Even though the facts of ClassicalM anchuo rthographyu nderdetermine thet ypeo f vowelh armonyin (pre-)Manchut,h e originalg enerativetr eatments all assumedt hat the type of vowel harmonyf ound in the languagew as of the front vs. back variety. Since Odden (1978) presents a more detailed descriptiono f the actual rules postulatedf or Manchu, his analysis( which differsi n some detailsf rom Vago's but not in detailsr elevantf or this paper) will be presented. Oddenp roposess evenu nderlyingv owels: four high vowels i uii and u, of which the first two are front and the last two are back; and three non-high vowelse , a and o, of whicho nly the firsti s front.V owelh armonyi s accounted for by the followingr ule: (6) [ + syll] -. [aback]/ + syllI CO- [zbackJ The feature[ aback]i s also used to accountf or the presenceo f uvulars,v ia a rule of velar lowering: (7) + cons1 [-high]/-[+ syll 1 L+b ack L+ backJ Since the abstractv owels i and iud o not appearo n the phonetics urface,a rule of absoluten eutralizationis requiredt o eliminatet hem: (8) [+ syll + high [aback] cxround Finally, a palatalizationr ule which appliest o consonantso ccurringb efore surfacei (phonetically[ i]) is required: (9) [ +cons] -+high r + syll L-backJ / +high L- back] Extrinsico rderingi s requiredf or these rulest o operatec orrectly.O bviously, the abstractv owelsm ustb e presentf or the vowelh armonya nd velarl owering rules to operate. Similarly,i must be eliminatedf or the palatalizationr ule to have the desirede ffect. Odden( 1978) offersn o argumentw hatsoevert hat frontv s. backi s the type 62 VOWEL HARMONY IN MANCHU of vowel harmonyf ound in Manchu. In the first paragrapho f his analysis sectionh e states 'a ruleo f Vowel Harmonym ust accountf or the alternation ale in suffix vowels'. He presents no argument that a is back and e is front, but this assumptioni s necessitatedb y the rules he offers. Presumablyt he implicita ssumptioni s that what is transliteratede is a front vowel, because that is what e representsi n most languages. However, there are several languagesw here e representsa central vowel. This is especiallyt rue for languagess poken in the vicinity of Manchu,e .g. Middle Korean (Hayata, I975), Buryat (Bertagaev, I968), and Nanai (Avrorin, I968). Hence, this implicita rgumentb ears little force. Vago (1973) does present some explicit argumentsf or front vs. back harmonya nd suggestsy et other implicit ones, but none of these contains muchf orce,e ither.T het wo explicita rgumentas reb asedo n (i) theo rthography and (ii) the practiceo f earlierg rammariansB. oth are incorrect. Vago asserts( 1973: 584) that the vowels transcribedb y the vowels i, e, a, o and u 'have the usual continentals ound values'. This could be true if the continenth e had in mind was northeasternA sia, but it is definitelyf alse if he means that Manchu e resemblest he cardinalv owel [e]. Vago gave no evidencew hatsoevert hat this remarkw as correct.M odernM anchus cholars are almostu nanimousi n theiro piniont hat the vowel was central.L ie (I972: 48) states that the vowel transliteratede is phonetically[ a],a nalogoust o the a found in other Tungus languages,t he languagesm ost closely relatedt o Manchu. Schmidt (1932: 557) describest he Manchu sound as 'aehnlich dem englischenV okal in but, blood,d oes'. Ligeti (1953: 247) agrees,c alling it 'high, mid-out,b ack-wide[ i.e. central-J.A.], cette voyelle plus ou moins labialisee'.H e furthers tates (Ligeti, I953: 280) that the vowel has much the samev alue as the a in otherT ungusl anguagesa s describedb y Cincius( I949: 78): 'mixed ("central") of mid height'. Paskov (I963: i6) seconds this comparison,a dding that the acoustic value is similar to unstressedb I in Russian,a vowel that is centralt o back in its pronunciationA. vrorin( 1976: 13) also classifiese as a centralv owel. Virtuallyt he only grammarst hat give a frontv aluef or the vowel are the verye arlyo nes (for example,A dam (I873: I5) stated that 'e a la valeurd e notre e ouvert'). These early grammarso, f course,c ould not be based on nearlya s much researcha nd data. Thus, the consensusa mong scholarsi s that the vowel transliteratede did not have the value given to it by Vago. Vago's doctrinairep ronouncement that it was a front vowel is even more surprisingin that one of the studies whichd iscussedt he value of e veryc arefully,L igeti( 1952), is cited in Vago's article. Vago is also concernedw ith the phoneticc orrelateo f the vowel translit- eratedo . He assumest hat it was identicali n pronunciationw ith u at the time the classicalt extsw erec omposed,b ut that it did have a differents oundv alue at an earlierp eriod,a conclusions haredb y most Manchus cholars.H e offers his view of the originals ound value of o as furthere videncef or his analysis: 63 JOSH ARD It is perhapsw orthn oting how the conservativeo rthographics ystemm ay give clues to the descriptive linguist. Recall that. .. phonetic [u] is representedb y two differentg raphs,n amelyu and 6. This can be viewed as evidencet hat,a t one time,t heset wo symbolsh add istinctr epresentations, being later mergedb y a diachronicA N rule. Thus orthographicu could have been phonetic[ ii]a nd orthographico phonetic[ u].( Vago, 1973: 588). Vago was ratherc arelessi n his considerationo f the classicala nd pre-classical phoneticv alue of 6: he states, 'Ligeti assumest hat o is realizeda s a long o' (1973: 584). Presumablyth is comes from a misreadingo f a passagei n which Ligetid iscussese arlierv iews of the sound value of o: On a interprete o long (= 6), u long (= ii), u ouvert, ii, une voyelle entre o et u, o d'un timbre special, sans songer 'ac e qu'elle representait en realite: une simplev arianted u signe u employea presq , y, X (Ligeti I952: 251). As will be discussedb elow, the best evidencei ndicatest hat o did not have a more back articulationt han u. Vago's second argumenti s based on the practiceo f earlierg rammarians: 'Accordingt o the traditionalg rammarse, is classifieda s front, o o a as back, and i u as neutral.' Vago does not enlighten us as to which traditional grammarsh e has in mind, althougha followingf ootnote referst o Gabelentz (I832) and Budenz( I886), again both very earlys tudiest hat could not have benefitedf rom an extensiver esearcht radition.W ithing rammarsI have had access to, the vowels are classifieda s follows: weak (i e u) vs. strong( a o 6) (Harlez, I884: 13): weak (e) vs. strong( a o 6) vs. neutral( i u) (Adam, I873: I 5); hard (a o 6) vs. soft (i u) vs. (e) (Paskov, I963: I6); hard (a o 6) vs. weak (e) vs. neutral( i u) (Lie, 1972: 47); hard (a o 6) vs. weak (e, i, u) (Haenisch,I 96I: 33). None of theseg rammarcs lassifiest he vowelsa ccordingt o the labelso f' front' and 'back'. Therefore, one cannot accept Vago's assertion that this classificationi s typical of traditionalg rammars.T hus, the terminologyo f traditionalg rammarsc annot be used to supportV ago's analysis. The most importanta rgumentf or frontv s. backh armonyi n Vago'sa rticle is an implicito ne. The title of the articlei s 'Abstractv owel harmonys ystems in Uralic and Altaic'. The unmentioneda ssumptioni s that (i) since therei s evidence for some sort of vowel harmonyi n Manchu, and that (ii) since Manchu is an Altaic language, it should have the same type of vowel harmonya s otherA ltaic( andU ralic)l anguages.C lause( ii) containsa hidden assumptiont hat is false: not all Altaic languagest hat have vowel harmony have harmonyo f the front vs. back type. All of the other Tungusl anguages have vowel harmonyo f the high vs. low or relativeh eightv ariety.T his type of harmonyw as also found in MiddleK orean( Hayata, I975). In fact, pure 64 VOWEL HARMONY IN MANCHU front vs. back vowel harmony is clearly justified only for the Turkic languages.I n manyo f the Mongolianl anguagest he vowelst hat shouldc ount as front (e.g. e o) are in fact central vowels (Sanzeev, 1953). Hence, the allegedlyA ltaic patterni s more properlya Turkicp attern. It is not surprisingt hat the Turkicp atternh as been takena s the basic one by WesternE uropeana nd Americans cholars,s ince they generallya re much more familiarw ith Turkic languagest han with other brancheso f Altaic. Moreover,m any scholarsb esidesV ago have made the same assertion.F or example,i n an articleo n Koreanv owelh armony,K im notes that the patterns he gives for Koreand o not correspondt o the Turkict ype: ALL OF THE OTHER LANGUAGES IN THE ALTAIC FAMILY to which Korean is supposedt o belongp ossessf airlyr egularp alatal( vertical)[ frontv s. back] vowel harmony.I f the palatalv owel harmonyi s assumedt o be a common Altaic feature,s houldn'tK oreanh ave had the same in an earlierp eriod? (1978: 227 - emphasis added). Baitchura (I980: I6) even accuses Novikova (I960) of mislabelling X-ray photographso f vowelsi n the Tungusl anguageE venb ecauset he articulation disagrees with what he would expect from his familiarityw ith Turkic languages (the photographs, however, do correspond to her textual description). Therei s nothing wrong in using knowledgeo f relatedl anguagest o infer the nature of a language (the caveats in section 5 below must be noted, however).O n the otherh and, the morec losely relatedt he languagesa re, the more helpfult his type of informationi s likely to be. One would expectt hat the situationi n Tungusl anguagesc loselyr elatedt o Manchuw ould resemble the phenomenao f Manchu more closely than would facts of much more distantlyr elatedT urkicl anguages( recallt hat some scholarsd eny that the Turkica nd Tungusl anguagesa re relateda t all). In short, none of the implicito r explicita rgumentsi n favour of front vs. back harmonyi n Manchu supportst he case. In fact, a closer examination of the evidencem ilitatesa gainstf ront vs. back harmony. It should be noted that the primaryp urposeo f Vago (I973) and Odden (1978) was not to arguef or the TYPE of vowel harmony,b ut rathert o argue that abstracta nalysesa re requiredt o capturet he true generalizationsT. hey totallyi gnoret hep ossibilityt hata differentt ypeo f vowelh armonyc ouldh ave been involved,i ndicatingp resumablyth at this did not occurt o them.H ence, they said nothingt o convincer eaderst hat sucha n alternativew as erroneous. Thereforei, f the argumentsf or front vs. back vowel harmonya s opposedt o highv s. low vowelh armonyw erec onsideredin a legalc ontext,t he proponents of the formerw ould have entereda de facto 'nolo contendere'p lea. In other words,t hey have presentedn o case at all that front vs. back vowel harmony is the type that is required. 3 65 LIN 20