CHAPTER7 PROTO-JAPANESE BEYONDTHE ACCENT SYSTEM ALEXANDER VOVIN Un iversi t.t,o/' Huu'ui' i at lllc1nou l. Introduc'tion This chapter deals with tu'o problems: the controversyover the archaic natureof thc 2.5 accentclassof norninalslbund only in Kansai Japanesc($2) and the possiblccorrelationbctween low registerand vowel lcngth (\3). Although at first glancc thesetwo problcms look unrelatcd,I will attemptto show bclow that they are both relevant lbr reconstructing ccfiain segr.r.rental features of pJ that either have bccn consideredcontroversialuntil now, or havc passedaltogether unnoticed. The traditionalreconstructionof pJ accentfbr binroraicnouns is presentcd in thc fbllowing table. Thc follor,l'ingsourcesare used: the Ruiju rr.r,rigi.riri (RM, l0ll l; a dictionary which rnarkspitch accent),and also dialect data frorn tnodern Kyolo, Tokyo, and Kagoshima. Segmcntal fonr-rsof nouns are given in their Middle Japanese(M.f) form. Accent classand reconstruction 'cdgc' pJ*lltl 2. | /a.ii ' b r i d g c ' 2 . 2l i r i i p.l*tll '1lou'cr' 2.3/Lntt p.l +LL 'rhopsticks' 1 . . 1/ . r i pJ *LIJ 2 . 5/ i r t r ' s p r i n g ' p . l* L F RM Kvoto Tokl'o Kagoshima HH HL LILH LFiLHT HH. HH-H IlL. HL-L IIL. HL-L LII. LL-H LF, LH-t- LII. Ltl. LII. IlL. IIL. IlL. Ltl-L HL. LII-L LII. LL-H t-H. LL-ll LH. LL-TI LH-H LH-L LH-L I lL-L IIL-L TableI T h i s r e c o n s t r u c t i o nw,h i c h c a n b e t r a c e dt o K i n d a i c h i( 1 9 7 5 ) ,a n d i s a l s o adoptedin iLTT, presentsthe pJ accentsystem as practically identical witli the one found in Lhefutiju nt.t,ogishct.lt is also quitc closeto the rnodemKyoto systcrn Howevcr, not everyoneagrces:this systernis consideredcontroversialby some linguists(Tokugawa 1972,Ramsey 1979,1980),who proposedthat the p.l sysrern I P i t c h s h a p c si n n t o d e m d i a l e c t sa r c g i v c n b o t h i n i s o l a t i o n a n d w i t h a f o l l o u i n g p a f i i c l e D e p e n d i n go n p a r t i c u l a rm a n u s c r i p t s . 142 A L I J X A N D E RV O V I N was closer to the rnodern Tokyo systetr. Nevcrtheless,the dctarled arguments prcsentedby Marrin in favor c.l1'thetraditionalsystern(.ILTT:162-15) havc been lcti virtually uncontesteduntil today. Twclvc years ago I also provided a detailed refutationof thc Tokugawa-Ramseyhypothesis(Vovin 199-5).basedon internal, p h i l o l o g i c a l , a n d e x t e r n a l c v i d e n c e ,t h e r e f o r e I w i l l n o t d i s c u s s t h e w h o l c e o n t r o \c r s y a g a i nh e r c . 2. The segmental source ofoccent class2.5 However, there still rcmains one controversialpoint, which is in some scnsc peripheralto the controversybetwccn traditional and Tokugawa-Rarnsey rcconstructions,but still deservesour attentionand needsto be resolvedone way or another.This point concernsthe archaicnatureof the accentclass2.5, which is tbund only in modern Karrsaidialects.The philological evidencefbr its existencc is rrrorecontroversial,since not all manuscriptso1-the Rlrflr nt.togishohavc the 'eastern dot' (to-ten),believedto ntark the distinctivefalling pitch on the sccond r.noraof nouns belonging to the accent class 2.-5.Naturally, supporterso1'the traditionalrcconstruction,who are 'Kyoto-oriented'.bclieve in its archaicnaturc, u,hile Tokugawa and Rarnscy,who are 'Tokyo-oriented', insist that this accent class representsa local innovationin Kansai dialects.The r.nostrecentchallenge to the archaic nature ol- thc 2.5 class was presentcdby Unger who writes ( 2 0 0 0 a : 2 0 -I2) : T h c R u s s i a t tl i n g u i s t P o l i v a n o v s p c c u l a t e dt h a t l i n a l l a l l i n u p i t c h i s a t r a c c o 1 ' a " t r u n c a t c d "n a s a lp h o n e m c .M a r t i n ( 1 9 8 7 . 3 6 1 n . I t o C h a p t s r . { . s \ l , l ) t h r n k sh c got thc idca by contparing Korean rrclolr'morning'rvith.lapanese r - r . s ri d r .. which is 2.5 and coulcl bc taken back to an earlier *asa.rl. Howevcr. Martin fbund that a handful of Ryukyu *'ords u,ith flnal nasalsunrcflected in nraini s l a n d s c o g n a t c sa r c s c a t t c r e dr a n d o r n l y a c r o s sr e c o n s t r u c t c da c c e n t c l a s s c s .a n d H i r a t a a n c lI h a v c h u n t c d i n v a i n l b r a d d i t i o n a l K o r e i r n n t a t c h e sr v i t h f i n a l n a s a l s l b r . l a p a n e s e2 . 5 n a s a l s . P o i l i v a u o v ' s e x p l a n a t i o n f o r c l a s s 2 . 5 i s t h c r c f b r e d u b i o u s . ( l n a n y c a s e .h e s u g g c s t sn o l i n k b c t w c c n c l a s s e s1 . 2 a n d 2 . - 5 .a n d a p a r l f r o n r t h c l a l l i n g p i t c h i n K y o t o - t y p c d i a l c c t s .t h e r c s c c m s t o b e n o n c ) . . . t h e f - a c t that class 2.5 is rslativcly sntall and distinctivc only in one dialcct group s u g g c s l s t h a t i t s o r i g i n s s h o u l d b c s o u g h t i r . rr c l a t i v c l y r e c e n t d c r c l o p t n e n t s within that group. Hcrc I assunlc that Unger confuses two different issucs that have no connectionbctweentl.rernwhatsoever;flrst, the correspondence of 2.5 classnouns to nouns with final -lr in Korean, and second,the conespondenccof final -ly', found in Haterulna, to -O in other Japanesedialects. The archaic nature of Haterurna-N hasbeen persuasivelyrefutcd in the literaturc(JLTT:74, Oyler l997), and it will not conccnt us here. The flrst problern,however,remains,and I will d i s c u s si t b e l o w i n d e t a i l . First. thc idea that nouns in accent class 2.5 havc parallets in Middle Korean (MK) words rvith flnal -nr indeed belongs to Polivanov (1924), who, however, givcs not just onc exarnple but two. Second, Unger and Hirata are apparentlyunawareof the othcr literatureon the subjectwhcrc additionalparallels PROTO-JA reflectinga correlationbet MK words with final -m a have been in vain. The folt sourceswherethey werefir Old Japanese'r r r . v a2 . - 5' n t o m i n g llll 2.5 'crane' M.l lilr 2.5 'pigweed' ltunr 2.5'spring' .sllxrr'rvhitc' a d . j c c l i v ca c c e n t : B rloun acccnt:2.5 < +siru'o-m Azrnrrr'black' a d j c c t i v c a c c e n t :B n o r - l na c c c n t :2 . 5 < * k u r w o - m i'rlz 'dark' a t i j c c t i v c a c c c n t :A n o u n a c c c n t :2 . 5 < * k u r a - m uka 'rctl' a d . j c c t t v ca c c e n t :A l r o u n i l c c c n t :2 . 5 < * a k a - m rrx o 'bluc/grecn' ad.jcctiVcaccent: fl n o u n a c c c n t : 2 . - 5< * a w o - m The table above rer J a p a n e s cn o u n s .H o w e v e ri. distribution of thesewords i K o r e a n c x t e r n a le v i d e n c ei s I Acccnt notations arc givcn acc acccntual data on OJ arc not suffic a T h e c o m p a r i s o n o l - . l a p a n e saen d ( 1 9 6 6 : 2 3 t i ) .b u t u , i t h o u t r h e d i s c u 2.5 in Japanesc. PROTO-.IAPANES B E Y O N DT I I E A C C E N TS Y S T E M tents been riled mal. hole ,ome nsey way chis r the cond I the ture, )cent enge 'ntes t43 rcflccting a correlationbetweenJapanesenounsbelongingto accentclass2.5 and MK words rvith final -m are provided, other-wisetheir hunt would probably not havc been in vain. The following table summarizestheseknown parallelsand the sourccswhere they were flrst suggested: Old Japanese'r I\liddle Korean sourcc zr.srr 2.5 'rrorning / r / / 12/ . 5 ' c r a r l c ' r)r'hrirr'tnorning' 'crirnc /l ri/rlrirl-i P o l i ra n o r l 9 l - l : l - 5 1 M.l lilir 2.5 'pigrvccd' 7riftiri 'pigvn eed'r punr 2.5'spring' 7;lriri'spring' .rlrlr.r'u'hitc' chri a d l c c t i v c a c c c n t :B n o u n a c c e n t :2 . 5 . ' - + s i n v o - m th-vti-nt'dancc'(n.) 'dancc'(r'.) \ \ ' h i t n r e nl 9 t l 5 : l 0 l Vovin1994:150 (pre-p.lnominalizcr*-r.nfor ad.jcctivcs dcnotingcolors) .tlu7r-'hclp'(r,.) lrrnrz 'black' a d j o c t i v c a c c c r t t :B lrolln ilcccnl:2.5 '- *kunvo-rrr t v t ) l l ' - t ' r - n t ' h e l p '( n . ) .r'lr.r- 'laugh' (r.'.) r t i r : - t t t i - t r t 'l a u g h ' ( n . ) irri a 'dark' a d . j c c t i v ca c c c n t : A .xrl-'cry'(r,.) n o u n a c c c n l : 2 . - 5< * k u r a - r . n v i t l - v ' r i - n t ' c r 1 , '( n . ) rrla 'rcd' a d - j c c t i v ca c c c n t : A lril- 'change" (r'.) kil-i-nt'changc'(n.) n o u n a c c e n t :2 . 5 < s a k a - n r arrn'blue/grecn' a d j c c t i v c a c c c n t :B n o u n a c c c n t :2 . 5 < + a r v o - m T a b l c2 eno ,ouns I -1/, re of 9e1), will The table above representsexternal evidence in favor of *-m in some nouns.However. it does not irnr-nediately tell us of the possibleage and .lapanese distributionof thesewords in Japaneseitself. Thus, we still have to see wliether Korean externalevidenceis applicableto pJ as a whole or is localizedin Kansai iddle who, a are allels ' Accenl nolalions are givcn according to EMJ as rcf'lcctcd in thc RrrlTl nt.vitgish6. sincc thc a c c e n t u a ld a t a o n O J a r e n o t s u f f l c i c n t a n d . / o ra r c c o n t r o v c r s i a l . 'The c o n r p a r i s o no f . l a p a n e s ca n d K o r c a n r v o r d s f o r ' p i g i v c c d ' w a s p r o p o s e do r i g i n a l l y b y M a r t i n ( 1 9 6 6 : 2 3 1 3 )b.u t w i t h o u t t h c d i s c u s s i o no 1 ' c o r r c l a t i o nb e t w e e n t i n a l - n r i n K o r e a n a n d a c c e n t c l a s s 2.5 in Janancsc. t44 A L E X A N D E RV O V I N alone. Traditionally all thc comparisons above, including the four lexical comparisonsand the correspondenccof accent class 2.5 in nouns designating colcrrsto the norninalizer-m in MK were consideredto be examplessuppofiing the genetic rclationshipbetween .lapaneseand Korean. In order to confim or disprovethis position.I will offer a detailedanalysisof all etyrrologiesbelow. ( l ) O J u s u 2 . 5 ' r r o r n i n g ' - M K d c h 6 n r ' i d . ' .A l t h o u g ht h i s c o n r p a r i s oins usr"rallyused to support thc theory of geneticrelationshipbetwecn Japancseand Korean. (cf. Martin 1966:236,Whitrnan 1985:244),therc are problelnswith this etyrrology. First. as whitrnan notcs, it is difflcult to account for the aspirated /-ch-/ in the MK form (Whitman 1985:244).MK oc'hrim< proto-Korean(pK) *acoKom or *aKocorn, none of which could comespond pJ *asa to 2 5 < pre-pJ 'moming' *asam.Second,Japancser.r.ra has a Imited distributionin Japancse:it is f b u n d i n R y u k y u a n o n l y i n S h u r i a n d S h i t o i d i a l c c t sa s ? a . s a( O G J l 9 9 l : 1 2 3 . Hirayarr.ra1966284). lt is significantthat the word does not presentitself in the southernRyukyus.where the rnainlandJapaneseinfluencewas nrinimal.Thus. thc lirnited attestationexclusively in Shuri and Shitoi suggeststhat in thcse two dialects it is a loanword fiom rnainlandJapanese,especiallygivcn that another word, pR *tutor.nuti 'rnoming' (a cognate of M.l tutomete 'carly rnoming' discussedbelow), is attcstodthroughoutthe Ryukyus,and is also presentin Shuri ( l { i r a y a r n a1 9 6 6 : 2 t 3 4H, i r a y a r n a1 9 6 72 0 2 , O G J l 9 9 l : 1 2 3 ) . I n a d d i t i o n ,t h e w o r d n.ra'nro-rning'is also attcstedin u:trrrtupoems,but rnostof the poems in which it appears'do not have very distinctivcEOJ features.(' I suspcctthat EOJ nsn rnay be also a loan front WOJ. Givcn all this and the fact that it is irnpossibleto connect pJ *asa < pre-pJ *asam with pK *acoKorn or *aKocom, I think that OJ a.sais a loan liorn some lbrm of old Korean like *acharnthat already had an aspirated +-ch- (and not a sequencc*-Koc- or *-coK-) that would be expected to be borrowcd as o.f or pJ *-s-. Even though oJ a.saand MK ach6rnare not genetically related,the accentpattern2.5 is still confirmed by the Korean word as deriving fiorn *-n, but since the Korean loanword is wcll attestedonly in WOJ, it can confirrr the archaicnatureof *-m only fbr central Japanesc(cJ),/but not for all of Japanesc. 'id.'. (2) O.f tunr 2.5'cranc'- MK rr.r,zilx,tint-i Thc correspondences seem to bc irnpeccable,and there are no problems that would point to a loanword scenario,unlikc the etymology for 'rnorning' discussedabove.The -i in thc MK forrn is likely to be a dirrunitive suffix, cf. tv'ulem without this suffix in the d i v e r g c n tc e y c w u d i a l e c t( c h o y 1 9 8 7 : 8 7 8 )I.n a d d i r i o n ,a l t h o u g ht h e w o r d i s n o t attested in EoJ, it docs not suff-er frorn limited distribution in Ryukyuan: Kikaijima (Amarni), Sani (Arnarni) tsuru; Tokunoshima (Arnami) Lwruntui; ' Scc cranrples c o n v c n i c n t l y p r o r , i d c di n M i z u s h i m a ( 1 9 1 3 4 : 6 4 6 ) . " 1 ' h c o n l y r e a l c x c c p t i o n i s M Y S 3 - 5 0 2 .w h i c h h a s s o r n cr y p i c a l E O J v c r b a l m o r p h o l o g y . - 'C'entral .lapanese' rcI-ersto a branch of Japancscwhich cornprises WOJ, MJ, and rhe ntajority of n t o d e r n r n a i n l a n dc l i a l c c t s b . u t i t d o e s n o t i n c l u d c I r O . l a n d t h e n r o d c m H a c h i j c l - j i m ad i a l e c t . PROTO Y o r o n . I z e n a .K u m e j i m t.|ilru: Shirobe (Miyako) Takebu (Yaeyama) tfiru Even if the Japanese wor prohably liom the perio this etymology fully su from a lost final *-m. (3) MJ ./i.vu2.5 , attestationon the Japane H,,n:1|y4r.r.1 r i9 1 8 1s.o i t ils OJ lbnn was*piyt wi /-wi/ goesback to pJ *qi diffi cult. However,since countelpart /-wi/, there Ncvcrtheless,there is a namely the corresponde work wc must first assum of -1- to -1-,which is pos assurre that MJ /i.vu < y supportcd,but it is entir Yonaguni dialect which i rnedial -.t- correspondsI ternporal and geographi elsewhere(JLTT:20).Thr this chapter,but thereare on Yonaguni r/- (Whitma proto-fom of the Japan correspondence of pK *-t. MK) or a corresponden and both require a great etymology conditionally etymology. and therefore Japaneseaccentclass2.5 t (4) OJparu 2.5 ,s grave phoneticproblerns.I /i/. Second,even if this w xaCu corresponding to M merely coincidentalresem t Marlin proposed to reconsrruc nowadays it seernsthat this prop '' l dcdicatc a substanlial chaprer SYSTEM BEYONDTIIE ACC'[]NT PROTO-JAPANESE 145 Yoron, Izena, Kuntejirna(Okinawa), Ishigaki (Yaeyama)tsuru, Hirata (Miyako) (Yaeyama).[r4: frrrr; Shirobe(Miyako), Kawahira (Yaeyarna)l^sirrr;Kuroshima Takebu (Yaeyanta)t.firu; (Nakamatsu1987:4t3,117, 165, 201, 248, 281. 326)' Even if the Japaneseword is a loanwordfrom Korean,it rnustbe a very early onc, probably fiom thc period of rnutual cocxistenceon the Korean peninsula.Thus, this etyr-nologyfully supportsthe interpretationof accent class 2.5 as rcsulting from a lost final *-nt. n A 'id.'. First, there is a problem of (3) MJ /i1,u2.5 'pigweed' - MK piltim attestationon the Japaneseside. The word appearsfbr the first time in the herbary Honzou,amt'ri(9ltt), so it is MJ, not OJ. This leavesroom for unceftaintywhether l-wil . Sinceolsu-rai its OJ fbrm was *piyu with ko-ruil-il or *pwiyu with r.rl^stt-rui /-wi/ goesback to pJ *oi or *ui, this would make a comparisonwith Koreanrather ilifficult. However, sincek6-ruil-il appearsmore fiequently in OJ than its ttlsu-nri *piyu. counterpart/-wi/, there is a strong possibility that it was actually OJ Neverthcless.thcre is another possiblc phonetic problcrr.rin this comparison. of MJ --t,-to MK -/-.nIn order for this conrparisonto namely the correspondcnce work wc must flrst assurlethat MK pililm < pK *pitum, with well-known lenition of -l- to -1-,whrch is possible,but again cannot be proven. Second,we also must *d is widely assumethat MJ /i.t'u < pJ *pidu. The hypothesisthat O.l .l' < pJ supported,but it is entirely based, I am afraid, on a mistaken theory that the Yonagur.ridialect which indeed has initial z/- correspondingto QJ 1' (though oJ medial -.1'-correspondsto Yonaguni --)', not -zl-) preserves,alone anlong all a pJ *d which shifted to } ternporal and geographicalvarieties of Japar.rese. elsewhere(JLTT:20). The full discussionof this problem is outsidethe scopeof *d based this chapter,but there are some good prelirninaryargulnentsagainstpJ * p i d u b e i n g r e j e c t e da s a p o s s i b l e o n y o n a g u n i r / - ( w h i t m a n 1 9 8 5 : 1 8 ) . ' ' W i t hp J proto-fbrn-rof the Japaneseword, we are faced with two possibilities:either corespondenceof pK *-t- to pJ *-y- (assumingthat a lenition took place prior to of pK *-l- to pJ *-y-. Neither is an enticing solution. MK) or a correspondence and both rcqr.rirea great deal of explanation,but I ar.nwilling to accept this etymology conditionally for tire time bcing. Still, it is not a very reliable of etymology,and therelbre,it cannotprovide full supportfor the correspondcnce Japanescaccentclass2.5 to MK -rl. l: ' s p r i n g ' - M K p w 6 m ' i d . ' . H e r e . I b e l i e v e .w e h a v e t w o (4) Oi paru 2.5 grave phoneticproblcrns.First, OJ -/1 never correspondsto MK -O- exceplbefbre /i/. Second.even if this was to be confinned, there are no other instancesof pJ *aCu correspondingto MK u,o. Thus, I think that thcse two words represcnta i; mcrely coincidentalresetnblance. ( e )t ' of + i o n t h c b a s i s o f t h i s c o r r c s p o n d c n c c( M a r t i n | 9 6 6 : 2 I I ) . b u t M a r t i n p r g p o s c dt o r c c o n s t r u c tp JK that this proposal is no longer supported. nowadays it scct.t.ts " l d c d i c a t ea s u b s t a n t i a lc h a p t e rt o t h i s i s s u c i n a f b r l h c o m t r r gb o o k ( V o v i n . f b n h c o n l i n g ) . 146 A L E X A N D EV RO V I N ( 5 ) . l a p i l n e saec c c n tc l a s s2 . 5 f o r n o m i n a lf o r m s o f a d j c c l i v e s ld" e s i g n a l i n g colors MK verbal norrinalizer-rr. Phonctically,the conespondenceseetnsto bc impeccablc.However, thcrc are trvo problcms.First. this particularshift from an original accent class to accent class 2.5 when a word is uscd as a nominal is lirnited in Japaneseto adjcctivesonly. Second,we cannot verify that the sal"ne processwas applicableto other varietiessimply becauseaccent 2.5 class is not attcsted in other varieties of Japancseoutside MJ and the Kansai dialccts of r-nodcrnCentralJapanese(CJ). Thc limited attestationwithin Japanesein this case (quite sirnilar to the caseof the word for 'n-rorning'discussedabove), I bclieve. pcrintsrnore to thc borrowing scenariolrom some varicty of Old Korean to CJ than to colnmon genetic inheritance.This is even rlorc likely in the light of the lact that we arc dealing here with derivationaland not irrflectionalrnorplrology; and thc fbnner is nrorc casily borrowed than the latter. Thcrcfore, I believe,this corespondenceconfims the archaicnaturc of accentclass2.5 within CJ, but not rvithin Japaneseitsell'. Thercfbre,an attemptto rcsolvethis controversywith the help of external data leavesus with only a partial solution.lt could be confinned that accentclass 2.5 has an archaicnaturein CJ. so Tokugawaand Ramsey'spoint of view that this accentclass originatedin the post-MJ era can be rcliably refuted.However, it is irnpossiblcto confirrn the existenceof accentclass 2.5 in pJ. bccauseout of the five exarnplesdiscussedabove only (2) has a potentialfor being explainedeither as a comlron inhcritanceor as a very old borrowing. Neverlhcless, external evidence should ncver take precedcnce in cxplaining internal data, otherwiseone runs the very high risk of falling into a Moscou'Nostratic trap. where intcrnal etymologiesarc disregardedand external etyrnologiesoverridc the intemal data. I bclieve that there is substantialinternal Japaneseevidencc supportingthe archaic nature of acccnt class 2.5 not only on thc level of CJ, but on the pJ level. This evidenceis bascdrnainly on dcrivational morphology.Considerthe fbllowing examples: 'cldcr' a. O.l x,rlsn2.5 'to rule' < *wosarnaCi-< *wosatn-(r)a-Ci-. cf. O.f v'o.iarlc'blamc, offence' b . O J t o g u2 . 5 'to cf . C)Jtoguntc- blarne,to reproach'< *togamaCi-< *togarn-(r)a-Ci-. r" S o - c a l l c d a d j c c t i r e s i n M J a n d o t h e r l a t c r v a r i e t i e so f J a p a n c s c a r e i n { a c t q u a l i t y v c r b s . a n d n o t " a d . j e c t i r , e si"n t h c s c n s c o f t h c t c r n r a s s i g n e dt o a d j e c t i v c so f E u r o p e a n l a n g u a g e s T . hus. e.g..MJ 'dark' but rathcr'to bc dark'. liirtr- docs not lnean PROTOc. OI tutwo(_ni cf . MJ tutom 'to stnve, to unclear. d. OI pa.t,a2.5 '< Examples(a-d)al some denominalverbs u ulLtldnlL't'u "to wann up' kl'omeru 'to purify' (< t nagame- 'to prolong vot analyzestheseverbsas cr (ILTT 792). At first gtan that intransitiveverbs cr which Martin reconstru alalarnlru'to bc warmed are lhrce argurncntsagain ( l ) * - m a -i s a s u unnecessaryentity. By C with the nominalizer*-m ( 2 ) M a r t i n ' sa n a l y producesyet one more su pafi as norninalizer*-mr c c e i v eq u i t ea s y m m e t r i transitives: adjec i n r r a n s i t i v e as d : je ( 3 ) I n M a r t i n ' sa n br.rllznouns and adjecttve trcatedas a final phonem after adjectives. One rnight argue i and acccnt class 2.5, be or -ment bclong to the 2. rnake go to thc end', frrpa " N o t l i s t c di n Z d b . O h n o ( 1 9 9 o n l y o b l i q u c c v i d e n c ef b r t n a t t thc basis ofthc fact that the vo, a s u ' c l l a s p o s s i b l y o n t h e b a s r sr rr V c r b s l i k c h o m a r u ' p r a i s c so ' t PROTO-.IAPANESE BEYONDTHEACCENTSYSTEM n g "D be an is me 1ot of tse 147 c . O J t u t w , o ( - n i ; r2r. 5 ' e a r l y i n t h e r n o r n i n g , cf. MJ rulotnere'id.', believedto be a derivationfrom the oJ tutv,ome_ 'to strivc, to tnake effbrt', although thc sernanticconnectionrernains unclear. ,Io quicken.. d. OI pa.tu 2.5 'quickly' , cf . MJ /'o.t,unte_ ASS Exanrples(a-d) allow us to reanalyzcthe derivationofall deadjectivaland sorle denominal verbs in -me- such as aralotllent .to renew' (< arala-.ncw,). a t a t e m e r u" t o w a r m u p ' ( < o r a t a k u - ' w a n r ' , i l u m e r u . t o h u r t ' ( < i t u _ . p a i n f u l ' ) , k i . v o m e r u ' t op u r i f y ' ( < k i ) , o - ' p u r e ' )s, e b a m e r u ' t on a r r o w ' ( < s e m l b a - . n a r r o w ' ) , n a g u m e - ' t o p r o l o n gv o i c e w h e n r e c i t i n gp o e t r y ' ( < n a g a - ' l o n g ' ) , e t c . r r Marlin analyzesthescverbsas consistingof stem+suffix*-rna- + transitivityflipper *_Ci_ (ILTT 792) At first glance,his point of view scernsto be supportedby the facr that intransitiveverbs comespon<1ing to trarrsitivesin -meru all cnd in -ntut.tt. whiclr Martin reconstructsas *-ma-ra- (19g7:792):uratamartL.to be renewed'. atulutnunt 'to be warmcd up', sehamunt 'ro becomenatrow', etc. However. there are thrce argumentsagainstMartin's analysis: his trs the her ( l) *-rra- is a suffix with an unclearfunction, thus it represenrs an extra unneccssaryentity. By Ockham's razor ir is, therefore,pref-erable to replace it with the norninalizer*-m to which verbalizer*-(r)a_is added. CJ the gv; his not nal ln oa nal mal on rnal ci- d not ., MJ (2) Martin's analysisof intransitives in -ntaru as having a stcm in *-ma-raproducesyet one more suffix *-ra- with an unclcarfunction. Iicanalyze tl.re-ntar pan as nominalizer *-m * ysl6slizer *-(r)a- + passivizingsuffix *_r_. Thus, wc recelvequite a synlmetricalpictureof how transitivesand intransitivesarc derived: transitives: adjectivalstern+ *-nr * *-(r)a- + *-Ciintransitives:adjectivalstern l *-1r * *-(r)a- + x-1(3) In Mafiin's analysisit is unclearwhy his verbalizer*-mals addedto bolh nouns and adjcctives.This strangc fact can be easily explained if *-rn is treatedas a final phonerneof the nominal root, but as a nominalizing suffix *_rn a f t c ra d j e c t i v e s . one rnight argue againstthe connectionbetweenthe final r-m rn nouns and accent class 2.5, becausenot all nouns with denominal vcrbs in _nrcrur or -tnant belong to the 2.5 class.Thus, oJ kipa 2.3 'brink', cf. oI kipa,t?e-.to m a k e g o t o t h e e n d ' , k i p a m a r - ,k i p a r n -' t o r e a c ht h e e n d ' , o J . s i u , ' 2 . 1 . w r i n k l e ' . " N o t l i s t e d i n Z d b . O h n o ( 1 9 9 0 : t t t t 4 )t r a n s c r i b c si t a s t u t v ' o r v i t h type A /-u,o/, although thcre is o n l y o b l i q u e o ' i d e n c e f o r t h a t ( s i n c e t h c w o r d i t s c l f i s n o r a t t e s t e dp h o n o g r a p h i c a l l y in O.l) on t h e b a s i so f t h c t a c t t h a l t h e v o w e l s / u / a n d 1 o i d o n o t n o r m a l l y . u , r . r b i n . r o g . t n . r within one root. as wcf f as possibly on thc basis of the type A r,'owcl in the vcrb ruttone- 'to strive, to makc cflbr1. lr Verbs llkc h.nrt'r'tr 'praiscs' ctrhu:irnLertr'bcgins'. of coursc, do n.t bclonq hcre. n L I T X A N D EVRO V I N l4u O J . s l n n n l - l r ' t ow r i n k l e ' ( M o d e m . l a p a n e s es i v ' o n t e - )T.h e s e c o n dc a s el o o k s l i k e a real exception.but I rvill try to show later that there is evidenccfor OJ kipa 2.3 ' b r i n k ' h a v i n ga n o r i g i n a l2 . 5 a c c e n t . Therc are also sornebird narnesbelongingto accentclass2.5 besides1lo"l ' c r a n e ' a l r e a d y r n e n t i o n e da b o v e : . s r g i ' s n i p e ' ,O J t c t t l u ' c r a n c ' ,l r . r b l ' k i t e ' , 'rbis'. lukulzrrllr'owl'. Cf. sufflx -me in other bird names: stt:ume tokilnki ' h a r t ' f i n c h ' .M J l u b u k u r o n t e' s w a l l o w ' . I t i s 'sparrow'. k u m o m e ' s e a g u l l '. r i n l c interestingthat that none of thc bird narles found in acccnt class 2.5 has this sufflx. Thus. it is possible to speculatethat accent class 2.5 in these words originatcdas a truncationof thc sr,rfflx-rie. The evidencecitcd abovedirectly pointsto *-m as a sourceofaccent class 2 . 5 . l n a d d i t i o n .t h e r e i s i n t c r n a le v i d e n c et h a t i n d i c a t e st h a t s o m e k i n d o f n a s a l ( n o d i r e c tc v i d e n c ef o r * - m ) u n d c r l i e sa c c e n tc l a s s2 . 5 : 'a breath cxhaled through a narrou'cd rrouth', cf. OJ a. OJ lr.s[llr.r2.5 'to usfir)obuk-lu,;lu'lontuk- exhalea breathmaking one's ntouth narrowcr' < *usu'onpuka-< ?*uswom-puk-(r)a-. ' c g g p l a n t ' , M J n u s u b i3 . 5 b ' i d . ' , w i t h u n c l e a rs u f f l x a t i o n< cf. b . J n u s u2 . - 5 *nasunpi< ?*nasumpi.which can be an altcrnationof an earlier *nasutnl. Finally, thcrc are some exafirples that constitute additional external evidence(whethertheseare to be taken as cognatesor as early loanwords): 'snakc' has *-m preservedintact due to the fact that it is a. OJ pemi 2.5 Marrin believesIhaI OJ pemi followed by obsolctcsufflx -r (dirninutive'?). i . 6 ( J L T T : 4 0 4 )C . f. MK p61'dm'snakc'. 2 . 5 < * p a C i i m3 'rnaster'"-'OK *rrVlim 'lord' (> MK '.nint).The vocalism in b. OJ nlrsl 2.5 thc first syllablemay bc a problernlthe vocalismof the first syllablein OK is unclear. ' t r u l y ' , p o s s i b l yf r o r n m u - ' t r u e ' * - . s o' ' ! ' , w h i c h c a n b e c . O J m o s u ( - n i )2 . 5 'truly, really'. comparedwith MK c'hom 'monkcy'- WM sannay't'in'id.'. d . O J . s a r u2 . 5 'snake' 'harc'. cf. developmentof 3.6 to 2.5 in OJ pemi c. OJ u.sagi3.6 'hare'. *osigam above.Cf. Koguryo The last cxarnplebrings us to an intcrcstinginternalissuc.On the basisof Koguryo *osigam we would expect that the p.l xusagi would bclong to accent 'Zdb l i s t s . s r u r r i l - a s a c o n s o n a n l a lv c r b . T h c c r a m p l e s . b o t h i n l o g o g r a p h i c s c r i p t i n O J a n d i n p h o n e t i c i n M J . p r o l i d c n o b a s i s l b r c o n c l u s i o n t h a l . v i u a l r - u s e d t o b e a c o n s o n a n tr a t h e r t h a n a 'has is in thc fom silrrbi/arrlit'r'ir vou,eI r,crb in O.l. The earliest knovn'nattcstation of this "'erb *,rinkfed' (Konjuku nutnoguturi 26.2) that shows that it is either a consonantal or upper-bigradc vou'cl vcrb (Ohno 1990:695). '' I a n r i n d e b t e dt o B l a i n c E r i c k s o n l b r n o i n t i n s t h i s o u t t o l I c . PROTO- c l a s s3 . 7 ( L H L - L ) t h a tc a c c e n tc l a s s3 . 6 ( L H H _ have a number of case accentuattoncould make although the bulk of th 'salmon' 2 . 5 1 2 . 3o' l r a b o typical for accentclasse nrade above regardingth suggestsa hypothesistl attrition: a numberof wc u n d e r2 . 4 , 2 . 3 , a n d 3 . 6 . r lirrrhcrsupporlrhisaflriti OJ asa 'hemp' 2 Kyoto and Shuri is inegu accentualclass,althoughl Now we can retur 2 . - 5 .O J k i p a m i , l i m i t ' , d r Kyoto has irregularaccen and not 3.6, suggesting, hr Finally, I am no ntonotroralcaccentclass words. I would ratherfol ( l J H - L ) a n d 2 . 2 b( H L - L ) . the fall of pitch that Unge thc fall within Ihe last mc ./it I I ov,i ng casepa rtic le, bu The isolationforrn for mo to the fact that Kyoto auto a l s o n c c c s s a r tyo k e e pi n r registerclass.Thus,one st belongingto accentclassI Therelorc,I believ solne extentexternal,to lir final consonant*-rn. It rs tt o1'final *-r.r.rfor every nour no surprise,becausenot er fiom these nouns.It is not explanationis even smalle i s t h a t a c c e n tc l a s s2 . 5 r s rightfully belongsro pJ. H because its specific shape r5 M a r t i n n o t e c lt h a t . . a n u n r b e ro l s h i f i c d t o 2 . 1 " ( J L T T :I 7 7 ) . PROTO-JAPANE BS EE YOND T H EA C C E N T SYSTEM c a s el o o k sl i k e -or OI kipct2.3 .5 besides/lo'lr :', lr.rbl'kite', alnes"sllzltme ; w a l l o w ' .I t i s ;s 2.5 has this r these words o f a c c e n tc l a s s r kind of nasal routh', cf. OJ ruth narrower' r suffixation < arlier*nasum- ional external vords): fact that it is s that OI pemi re vocalismin syllablein OK which can be pemi 'snake' )n the basisof ong to accent ript in OJ andin rant rather than a vabiturikeru 'has I4g c l a s s3 . 7 ( L H L - L ) t h a t c o n e r a t e sw i t h a c c e n tc r a s s2 . 5 ( L H _ L ) , b u t r t b e l o n g s to a c c e n tc r a s s3 . 6 ( L H H - H ) , w h i c h c o r r c r a t e s w i t h 2 . 4 ( L H - H ) . F . u r t h e n n o r ew, e have a number of caseswhere either a lack of accentualdutu o. rrregurarity in accentuationcourd make a choice between 2.5 and 2.4 0r 2..50, i : qucrtionabre, a l t h o u g ht h e b u r k o f t h e e v i d e n c e n o r m a ' y p o i n t s t o c r a s sz . i ( c f . , 'salmon' e . g . ,. s a r e 2.512.3'ror above-rnentioned u,,^ra.elder, 2.512.4,r\. The sa'e prcturers t y p i c a l f o r a c c e n tc r a s s e 3 s .6 and 3.7.This, in combination with the observation madc above regardingthe potentiarKoguryo cognatc *osigar.rfor pJ *usagi 3.6, suggestsa hypothesisthat the srnall Jasses 2.5 and 3.7"a/bbecarnesrrall by attrition: a number of words that used to bclong to those classesare now fbund u n d e r2 . 4 , 2 . 3 . a n d 3 . 6 . r ' T s h e r e i s o n e a d d r t r o n aer x t e m a l c o m p a . s o nt h a t c o u r d lurther suppor-t this attritionhypothesis: OJ asu 'hernp' 2.3 _ MK sam ,id., . The accentuationin both modcrn Kyoto and Shurr is irregular(JLTT:3g4), possibrysuggestinga shift frorr another acccntualclass.althoughKyoto and Shuri contradicteachother. N o w w e c a n r e t u r nt o t h e i s s u eo f w h y o I k i p a , b r i ' k , i s 2 . 3 r a t h e rt h a n 2 5. oJ kipctni 'rimit', derived frorn kipatn- .to reach the end,, is 3.6 (arthough Kyoto has irreguraraccent).the 3-mora accentclassthat correrateswith 2.3 is 3.4, and not 3.6, suggesting,however,that the original crassof tr4r,,rruy rruu" becn 2.-5. Finally, I arn, not. srrre that Unger is right when t lrr",-t, that the ' o n o l r o r a r c a c c e n tc l a s s 1 . 2 ( H - L ) c o r r e l a t e s w i t h a c c e n tc r a "s s2 . 5 l b r b i m o r a i c words' I wourd rathcr follow the traditionar position that it corrcratesrvith 2.2a (llH-L) and 2.2b (HL-L). The principar diff'erencebetween 1.2 and 2.5, regarding the fall of pircrrthat Unger rnentions, is thai in Kyoto-type diarectsclass 2.5 has thc fall v'ithin rhe last mora i.vtratetr/brnt, and r4ie,:ihe last mora ,in he/itre tha ./bllov'ingt'aseparric/e,but crass1.2 in hoth.a.rer has the fall ct/ierrhefirst rnora. The isolationforrr fbr modem Kyoto is HL, the sameas to, u...nt class 2.2, due to the fact that Kyoto automaticallyrengthens one-morawords to two_moras.rt is a l s o n e c e s s a rryo k e e pi n m i n . t n a i z . s i , u t o * - r e g i s t e rc l a s s ,w h i l e I . 2 i s a high_ registerclass.Thus. one shouldnot expcctto find any final *-,n unJc.tyrng words b e l o n g i n gt o a c c e n tc l a s s1 . 2 , Therefbrc.I berievcthat there is suflrcient evidence,rnainly internalancrto sorneextentextcrnal.to link the origins of accentclass2.5 in pl wrtn a loss of thc final consonant*-rn. rt is true that it-is not fossible to dernonsiratethe probabirity of final *-rn for every noun belonging to accentcrass2.5, but that shourd come as no surp'se, becausenot every noun in this crasshas verbal .oun*,pu.,' dcrivcd from thesenouns' It is not surprising that the number of words with an cxternal explanationis even smailer.Thus, th*e concrusionfor the trrst pu.t ortnr,, chapter i s t h a t a c c e n tc r a s s2 . 5 i s n o t a r e c e n t i n n o v a t i o ni n K a n s a id i a r e c t s b, u t that it rightfully belongs to pJ. However, within the pJ accent systenri1 is secon<lary, becauseits specific shape (the fall of pitch wrthin the last rnora) rs due to a or upper-bigrade :;Y,::ili:j:ijl1l;X#i:r., of words in crass 2.1/5withnnaropcrr vower in ord.rapancsc harc 150 PROTO-J ALI]XANDER VOVIN *-tn. sinplification of segmcntalphonetics,nanlcly the loss of the final consonant This should come as no sulprise. becausc pitch acccnts. likc lones. do not represent an original characteristicof a language; they rnust colne from 'sorncwhere'usually tulxs out to bc certain segmentalf'catures somcwhere.This that cither mergedw'ith other features,or have been lost altogethcr,leaving their tracesin the shapeof tones or pitch accent.This brings us to the secondproblcnl to be discussedin this chapter:the conelation bctween low registerand vowel length. The segmental sources oJ low register Martin suggestedthat pj low registerprobablyreflectsinitial vowel length in pre-p.f (ILTT:249-53). I later argucd that this solution is likely to be only partially plausible,becausethcre are morc words with initial low registcr than words with high registcr,and this createsa typologicallyunjustificdsituationlvith long vowels occurring more frcquently in initial syllables than short vowels ( V o v i n 1 9 9 5 : 1 3 1 )I. p r o p o s e dt o r e m e d yt h i s b y p o s i t i n gt w o s o u r c e sf o r w o r d s with low register:pre-pJvowcl length.as suggestcdby Martin, plus prc-pJ initial voiced consonants.and offcred somc external cvidence fbr the origin of low register from initial voiccd consonants(Vovin 1995 124-31). This solution, from threernaiorproblems: althoughtypologicallymore clegant,suf-fers 3. (1) Until recently it was impossibleto tell on thc basis of the internal Japaneseevidencewhich low registerwords would retlcct initial vowel lengthand which u'ould reflectinitial voiced consonantsl bascd on correspondenc and Onna, as vowel len H a t r o r il a b e l e d2 . 1 - 2 . 5n o sarneclasseswithout vow 2 . 3 b - 2 . 5 bb e l o w ) .T h e a p arc a number of example ( N S N T 2 l : 1 0 3 - 0 5 )t ,h ee x presentatrons are very sca out. llattori's correspon signilicantrefinementof F by Moriyo Shirnabukuro Shirnabukuro,this volume rnostly corclate with the rnost u,ords belongingto otherwise secondary (S Shirnabukuro managed t reconstructionof vowel l dialectsrather-thanjust on data are not consistent.I Nakijin dialec regularlyc 2 . 3 a - 2 . 5 aw , h i l e t h e r ei s n c lcngth in Shuri (subclasse Gloss& rnortar 2. (2) t was still adheringat this point to thc traditionalreconstructionthat posits pJ *b- and *d- on the basis of the SakishimaRyukyuan evidence.This *d-: one on the basis crcatedtwo different reconstructedseriesof voiced *b- and of the Sakishimaevidence.and the other on the basisof the lou'register; brcath 2., n e e d l e2 . shadow 2 b u c k e t2 . bridegroc (3) It was unclear what to do with high register words that start with vowels or nasalsonorants*m- and *n-: surely theseare also voiced. Thus. on rny earliercxplanation,they shouldbe all expectedto be found in low registerclasses, which ccrtainly also have their shareof vowel and nasal sonorantinltial words. My tentativesolution was to posit an initial glottal stop or *H- for high register words startingwith vowels and nasalsonorantsversussmoothvocalic ingressand ) .s I w i l l t r y t o p u r e n a s a ls o n o r a n t sf o r l o w r e g i s t e rw o r d s ( V o v i n 1 9 9 5 : 1 3 1 - 3 2 A demonstratebelow, this solution may sttll be partially true. althoughat that time without a solutionto ( I ) it was lrlerespeculation. Regardingthe major problern ( I ) above, Hattori was the first linguist to suggestthat Ryukyuan preservcsvowel length in somewords belongingto accent sf c l a s s e s2 . 3 , 2 . 4 , a n d 2 . 5 ( N S N T 2 1, 2 2 ) . H e a l s o a d d e da c o u p l eo f e x a r n p l e o words with vowel length from accent class 2.2. Hattori's proposal was largely flower 2.. nountaln cloud2.3 s h o u l d e rI board 2.4 rain2.5b s w e a t2 . 5 Shimabukurorecon 2 . 5 a i n t h i s t a b l ea n d i n t h e 't' F n . I i n S h i m a b u k u r o ' sc h a p t e r P R O T O - j A P A N E SBEE Y O N DT I I E A C C I I N TS Y S T E M l5l within severalOkinawan dialects,in particular Shuri basedon correspondences and Onna, as vowel length does not seem to be prcservedoutside Okinawa. H a t t o r il a b e l e d2 . 3 - 2 . 5n o u n sw i t h v o w e l l e n g t ha s s u b c l a s s( l ) a n d n o u n si n t h c sameclasscswithout vowel lengthas subclass(2) (theseare labeled2.3a-2.5aand 2.3b-2.5bbclow). Thc apparcntweaknessof Hattori's proposalis that while there are a nurnberof examplcswith and without vowel lcngth fbr classes2.4 and 2.5 ( N S N T 2 l : 1 0 3 - 0 5 ) .t h e e x a m p l e sw i t h v o w e l l e n g t hf b r c l a s s e s2 . 3 a n d 2 . 2 i n h i s presentationsare vcry scarce(NSNT 22'.100-0l). In addition, as Martin pointed o u t . H a t t o r i ' s c o n c s p o n d e n c esso m e t i m e sa r e n o t c o n s i s t e n (t J L T T : 2 5 2 - 5 3 ) .A significantrefinerncntof Hattori's proposalhas beenmade in a reccntdissertation by Moriyo Shimabukuro.who demonstratedtwo impo(ant facts (2002: see also Shirnabukuro,this volutne).First, words with lirst syllablevo\\'el length in Shuri rrostly corrclate with the low pitch classes2.3-2.5, and initial vowel length in most words bclonging to classes2.1-2.2 is eithcr found in colnpounds or is othenvise sccondary (Shimabukuro 2002:293). Second, most lmportant, Shirnabukuro rnanagcd to find a strong intcrnal colTclation supportilrg a rcconstructionof vowel lcngth in Ryukyuan on the basis of two Okir.rawan dialectsrather than just onc, thus taking carc of Martin's objectionthat Hattori's data arc not consistent.Shirnabukurodemonstratedthat initial accent in thc Nakijin dialcct regularly correspondsto vowcl length in Shuri for thc subclasses 2.3a-2.5a,rvhile thcre is no such colrcspondcncein the caseof absenccof vowcl l c n g t hi n S h u r i( s u b c l a s s e2s. 3 b - 2 . 5 b()S h i r n a b u k u r2o0 0 2 : 2 0 3 ) : " ' Gloss& accentclass nrortar2..la brcath2.4a necdlc2.,la shadou'2.5a bucket2.5a 2.5a bridegroonr ; t ) Nakiiin 'lulsi(:) ilci(:): plali halgi(:) nultni(:) mulhu(:) llorver2.3b 2.3b nrountain _phirna: c l o u d2 . l b 2.4b shoulder board2.4b r a r n2 . 5 b su'cat2.5b KUIIU: yarrlil: hata: hica: Shuri '?u:si - '1i:.^i ha:i ka:gi u:ki ,nru.ku hana Jarlla KLllllu kata '1ica '/anri _'?arli hasi: '?asi Table 3 o It )f i] Shimabukuroreconstructsvowel length for the words belonging to 2.4a2.5a in this table and in the absenceof any other cogentexplanationI would like ' " F n . I i n S h i r n a b u k u r o ' sc h a p t c r r n t h i s b o o k e x p l a i n s t h e s y r n b o l su s e d i n T a b l e s 3 - 6 PROTO-JA A L I ] X A N D E RV O V I N 152 to agreewith him. What rernainsto be explained,though,is the absenceof words belonging to 2.3a in this table; in this regard we achieved no significant irnprovementovcr Hattori. Shirnabukuroalso presentstl.rreeexceptionsto the o1'long vowel in Shuri to initial acccnt in Nakrjin, whcrc initial correspondence acccntin Nakijin correspondsto short vowel in Shuri (Shimabukuro2002:204): initial vowel lcngth, with registcr. It is rnost likely I Gloss t o r l o i s e2 . 3 a olfshore2.3a jar 2.3a G l o s s& a c c e n lc l a s s Nakiiin bonc 2.-la phLrlni( :) phu-lni(: ) 'iulnri( :) boat 2.4a sca2.4a bcan 2.3a huni huni ?umi ball 2.3a root 2.3a llowcr f .Jb n t o u n t a i n2 . 3 b Table 4 cloud2.3b dog 2.3b Shirnabukuroarguesthat the vo\vel of the first syllable "must have been long earlier in its history. but it has bcen irregularly shortened"(Shimabukuro 2002:204).I-le further suppofisthis point with cxarnplesof long vowels in these threc words found in thc Kamishiro. Onna, and Matsuda dialects(Shimabukuro 2002:204), but this is still reminiscent of Hattori's irregular correspondences criticized by Martin. Although Shimabukuromay be ultimately right. for the sake as tight as possible,I will omit thesethree words of kecping the corrcspondences from tufiher consideratton. Therefore. although the issuc of regularity within Okinawan raised by Martin has bcen" I believe. rcsolved answered in Shimabukuro'sdissertation. there renrains a problem as to whether it is possible to reconstructseparate subclasscs(a), involving vowcl length, and (b), without vowel lcngth for accent class 2.3. I believe that in spite of the fact that both Hattori and Shimabukurcr failed to providc a decisive answerto this problem (both give iust one cxample each for subclass 2.3a (NSNT 22:100. Shirnabukuro 2002:204, 369)). their intuitit,c solution is in fact correct,becausethere are uncontroversialexamples belonging to subclass2.3a that I add in Table 5 which is othcruise basedon a tablc found in Shimabukuro'sdissertationthat presentsthe reconstructionof two (a) and (b) for acccntclasses2.3-2.5(Shirnabukuro2002:369diffcrent subclasses 70).'' One can clearly scethat words bclongingto subclass(a) havevowel length in their first syllables.supportingMartin's hypothesisabout the origin of thc low registerfrom vowel lcngth. But it can be supportedonly partially,since thcre are no tracesof vowel length in the u'ordsbelongingto subclass(b). However,it does not make sensethat some words itt p.l have initial low register.and others have r - I t o o k t h c l i b c r r y o 1 ' n t a k i n gs o r r c c h a n g e st o S h i r r r a b u k u r o ' sp J a n d p R r e c o n s t r u c t i o n st h a t , 1 b r *o and xc. which underwcnl cxample. rcflcct the cument u'isdonr on presen'ation of prir.nary pJ r a r s i n g i n ( ' e n t r a l J a p a n c s ct o , ' u / a n d , , i i . t s e s i d c s* o r d s b c l o n g i n g t o 2 . 3 a . I h a v e a l s o t n a d c s o t r t e o t h e r a d d i t i o n sr v h e r c S h i m a b L r k u r o ' so n s i n a l c x a r n r r l e sd i d n o t s e e r nt o b e s u f l ] c i e n t . wal.c 2.ib s h a n k2 . l b b r c a t h2 . 4 a ntoftar 2.4a c h o p s t i c k s2 . 4 a n e e d l e2 . 4 a board 2..{b lbotprinr 2..{b s l t o u l d c l1 . 4 6 drcg: ).4b n u n r h c rl . 4 b s ki r r 1 . . { b h u e k e rl . 5 a b r i d c g r o o n r2 . 5 a r oicc 2.5u :pidel l.5u shrdou l.5a s i rc a t 2 . 5 b r a i r r2 . 5 b : p r r n gl . 5 h c r a n c2 . 5 b origin in pre-pJ scgment sorneu'ords belongingto tl secr.nsto be the only conc initial low registerdevelo seglnents.Therearesimply a e c o u n lo f -w o r d s b e l o n g i words belonging to subc PROTO-.IAPANEB S E Y O N DT H E A C C E N TS Y S T E M )nceof words o significant ptions to the whereinitial 2002"204): initial vowel length. with low registcr and vou,el lcngth later rncrging as low r e g i s t e r l.t i s r n o s t l i k e l y t h a t t h e i n i t i a l l o w r e g i s t e ri n s u b c l a s s( b ) a l s o h a s a n Gloss pJ * ka:may* ooO * o : k ix . . o * ka:rre* ooo *ntu:tulx ooO) * phana* OOI * vanra* OOI x khuntu* OOI * rranrcy+ OO * rlari * OO * ntoto* OO * pana* 0O * * 1,a1ta OO * kuntwo+ OO x ma:may* oo(.) * nra:ri* ooO * nro:tos oo() * pana* OO * yalra * Oo * kunto* (X) * ?cnu* OOI * narni* OOI * s u n i* O O I s'?i:ki)*ooo) *'?u:si)*ooo) *pha:si)*ooO) *pha:ri)*ooO) x ?ital * ool * 'latul * OO'l * irru* OO * natli + OO * s u n c ' *O O *ilkyi *o[o *uf su *O[O *pafsiy*O[O *pafriy *O[O *i[ta *o[o *a[to *O[O + c'u + OC) * nanti* OC) * s u n c+ ( X ) * kharal* (x)-l * khasul* ool * khazu-l* Ool *,-phada-l * OOI *wo:khe)*ooO) *kalta *o[O *kafsu*O[o *kalnsu*O[o *mukwo-l*OOl *kq\\,cyl*OOl *kurnwo-l*(X) l *kagcyT*OO-l *ascl *ool *amcy-l*ool *mo:kol *ooO-l spring2.5b *rro:kho)*ooO) * k h u : c )* o o O ) *kho:bo)*ooO) *kha:ge)*ooO) * '/asil* ool * 'lanrcl* ool * pharuT* Ool *parul*ool *paruT*OOl cranc2.5b * thurul * ool *turuml *ool *turunrl*ool llou'er2.-3b nlountain2.3b cloud2.3b ran raised by ; dissertation, lruct separate gth for accent Shimabukuro one example 369)), their sial examples se basedon a uction of two sro 2002'.369: vowel length gin of the low iince there are wever, it does rd others have lructionsthat, for whichunderwent : alsomadesomc ent. pR *kha:rnc)*- ooO) * ' ? u : k i*) o o O ) pMJ * kamey* OO * oki *_OO * karnex C)C) tortoisc2.3a o l k h o r c2 . 3 a jar 2.3a bean2.3a ball 2.3a root2.3a lst have been Shimabukuro )welsin these Shimabukuro Tespondences t, for the sakc e three rvords t53 dog 2.3b u,ave2.3b s h a n k2 . 3 b breath2.4a tnorlar2.4a chopsticks 2.4a nccdle2.4a boarcl2.4b lborprinl2.4b shoulcliir 2.4b dregs2.4b numbcr2.4b skin2.zlb buckct2.5a bridcgroon2.5a v o i c e2 . 5 a spidcr2.5a shadow2.5a sweat2.5b rain 2.-5b xkha:rnc)* ooO) *nra:rlc)*_ooO) *nra:i)* ooO) *pafnra*O[O *wokeyT*Oo-l *'/i:[ki *oo[o *'/u:[su*oo[o *pa:[suy*oo[O *pa:[ruy*oo[{) *i[ta *o[c) *a[ro *o[C) *kafra*O[o *kalsu *o[o *kafnsu*o[o *pafnta*O[t) *wo:kayl *ooO-l * k o : r v a y*l o o O l *ko:nrfplo i +ooo-l *ka:nkay-l *ooOl *ascl *(x)l *amayl*t)Ol Table 5 origin in pre-pJ segrnentalphonology. My carlier hypothesisthat the initials in sotre words belongingto the low registerclassesgo back to voiced segntentsthus seelnsto be the only concretesolution basedon internalreconstruction,because initial low registerdcvelopseither from initial vowel lcngth or from initial voiccd segments.There are sirnplyno other possiblesources,and vowel lengthonly takcs accountof words belongingto subclass(a). I thereforcproposeto rcconstructp.l w o r d s b e l o n g i n gt o s u b c l a s s e 2s . 3 b - 2 . 5 bw i t h i n i t i a l v o i c e d o n s c t s ,i n c l u d i n g 154 A L E X A N D E RV O V I N voiced obstruents,nasal sonorants,and smooth vocalic ingress.All these words have short vowels in the first syllable. All words belonging to subclasscs2.3a2.5a are reconstructedwith long vowels in the first syllable.Theoretically,they could have voiced onscts as well, but we sirnply have no evidence lbr that. I tentatively reconstrucl voiceless onsets for words beginning with obstruents. vowcl-initial words are reconstructedwith initial glottal stop *?-, which also behavesas a voicelessonset.Neveftheless,I reconstructwords with initial nasal sonorantsbelonging to subclasses2.3a-2.5aas having voiced onset and initial vowel lcngth. Thc reasonfbr this is cornbinatorialand will becomeclear when I dcal with reconstructionof onsetsof high registerwords below. As I n-rentionedabove, there is a possible contradiction betwcen this rcconstructionand a traditionalreconstructionof pJ that positspJ *b- and *d- on the basis of the Sakishimab- and Yonaguni r/- evidence.The discussionof why the Sakishimaand Yonaguni evidenceis not admissiblefalls outsidethe scopcof this chapter,but I bclieve that we should reconstructjust pJ *w- and *y-, and I addrcssthis issue in greaterdetail in a forthcorningbook (Vovin, fbrthcorning). Thus, the major problem (2), rncntionedabove will be taken care o1, since therc will not be two diflerent scriesof initial voiced obstruentsreconstructedfor pJ. Thus, Table 5 abovecan bc re-writtenas in Table 6, elirninatinglow initial rcgisterfor pJ altogetherand replacingit with voiccd onsetsfor thc subclasses 2.3b-2.5band vowel lengthin first syllablcfor thc subclasses 2.3a-2.5a.tE A s t h e m a j o r p r o b l e r n s( l ) a n d ( 2 ) h a v c b e e n s o l v e d a b o v e . i t i s n o w possible to address problcm (3): that is, what kind of onsets should be reconstructcdfor the words belongrngto high-registerclasses.Since initial low registerhas been eliminatedfiorn thc reconstruction,it does not make senseto preserve high register either. I think that high rcgister words should all be reconstructedas words with voicelessonsetsand with shorl vowels in their first syllables.Thus, the following reconstructions are suggestedfor the three diffbrent types of onsets(l usethe syllableslal , ltal, and lnal as exarnples): The reason for reconstructing nasal sonorant onsets belonging to subclasses2.3a-2.5aas voiced with vowel lcngth is purely combinatorial,as I mentionedabove.Voiced nasalsoccur cross-linguisticallyrnuch more fiequerrtly in the languagesthat have both voiced nasalsand voicelessnasals(pre-glottalized nasalsfunction as voiceless).Sincc I recorrstruct*?n- and *'?m- fbr words with nasal sonorant initials belonging to traditional accent classes 2.1-2.2, rcconstructingprcglottalizednasalsfor subclasses 2.3a-2.5awould producemore preglottalizednasal sonorantsthan voiced nasal sonorantsin the proto-language, w h i c h w o u l d b e t y p o l o g i c a l l yu n e x p e c t e d . r " '" Notc t h a t w c c a n n o t c o r n p l e t e l y c l i m i n a t c n o n - i n i t r a l p i t c h : w , c h a v e n o s e g m e n t a lc x p l a n a t i o n l b r h i g h p i t c h o n t h c s e c o n d m o r a o f 2 . . l . F o r 2 . 5 . a s i t r v a s s u g g e s t e da b o v e , t h c l a l l i n g p i t c h o n t h c s e c o l t dm o r a c a n b c e x p l a i n c d a s t h c r c s u l t o f t h c l o s s o f f i n a l * - n t . b u t o n l y o n c w o r d i n t h e t e h l c h a s a d i r c c t e r i d c r r c el b l i t . " I t i s n e c e s s a r yt o k e c p i n r n i n c lt h a t a r e c o n s t r u c t i o no f p r e - g l o t t a l i z e d n a s a l s* ' l n - a n d * ? m - m a y a c t u a l l y b e r e w r i t t s n a s a r c c o n s t r u c l i o no f v o i c c l c s s n a s a l sx h n - a n d * h n t - . PROTO.JA Gloss t o r t o i s e2 . 3 a o l l . s h o r e2 . 3 a . 1 a2 r .3a bcan 2.3a ball 2.3a root2.3a flowcr 2.3b n r o u n t a i n2 . - l b c l o u d2 . 3 b dog 2.3b wavc l.3b s h a n k2 . 3 b brcath 2..1a nrofiar2.4a c h o p s t i c k s2 . 4 a nccdlc 2.'la board 2.,{b lootprint 2.4b s h o u l d c r2 . 4 b d r e g s2 . 4 b ttrrnbcr2..1b skin 2.:lb b u c k e t2 . 5 a b r i d c g r o o n r2 . 5 a voicc f .-5a s p i d e r2 . 5 a s h a c l o u2 . 5 a su cat 2.-5b rain 2.-5b spring 2.-sb cranc l.-5b T)'pe of onse vowcl zr0bstruentldnasal na- P R O T O - J A P A N E SBIE] Y O N DT H I l A C C E N TS Y S T E M )rds .3ahey rt. I )ntS. also asal itial enl PMJ * kanrey+ Of) * oki * oo pJ *ka:ruay ofllhorc 2.3a *kha:me) * ooC)\ * ' / u : k i )* o o O ) . j a r2 . - 1 a b c a n2 . 3 a *kha:nrc) * ooO) * m a : r n e ' )* o o O ) + kanrc* OO + nramey* OC) *ka:tnc ball 2.3a *ma:i) * ooC)) + n r u : t u )* o o O ) * nrari x 0O * moto * OO * pana * OO *rna:ri * yarla * C)O *_kunr*'o x OO * inu * OO *yanra * nanri + OO *naml * sunc* OO *i[t<i,i*o[o *ulsLr*O[o *zunc *'li:[ki *oolc) *'lu:fsu*otfo +pha:si)*ooO) *pha:ri)xoool *pafsiy *O[o *palriy xO[o *pa:lrLry*,ro[{) *pa:[ruy*oof{) * ?iral * ool * ool * '?aruJ x khatal* ool *i[ta*o[9 *,,ftq*o[t) *kafta*o[o *ilra*o[O *altq*o[o *gafta*O[tl *kafsu*OfO *kafnsu*o[o *galsu*O[O skin 2.zlb * khasul* OO'l *_khazul* ool * phadal* OOI *palnta*o[O *balnta*o[O b u c k c t2 . 5 a *ri,o:khc)*oool b r i d e g r o o n r2 . 5 a smo:kho)*ooo) *u,okcyl*Ool *nruku'ol*OO'l *wo:kayl *ooOT *nro:kol*ooC)l Gloss torloisc 2.3a root 2.-la 1 ' l o u , e2r . J b this -on mountain 2.3b why eof ndl dog 2.3b cloud2.3b r . v a cr l . J b s h a n k2 . 3 b ,'6t' brcath2.4a nere rnortar 2.'{a ..' c h o p s t i c k s2 . 4 a ltlal necdlc2.'la board 2..1b fbotprint 2.4b now be low ,e to lbe first )rent s h o u l d c r2 . 4 b drcgs2.4b n u n r b c r2 . ' 1 b pR *1hana * OOI * yama* OOI * khurnu* OOI * 'icnu* OOI * n a m i* O O I * suni* tJ()l *'?i:ki) *ooO) *,iu:sil *ooo) *o:ki *tna:may *t.nq:tq *bana +gut.no tcrtu *galnsu*o[o voicc2.5a *khu:c)*ooo) *kgr'n'cyl *oO1 *ko:wayl *oool s p i d c r2 . 5 a *kho:bo)*ooo) *kha:gc)*ooo) *kunnvol*(X)l *kagcy-l*ool *ko:mIp]ol*ooOl *oool *ka:nkay-l *asel*ool *amey l *oo l *paru'l*uol *t.,runrl*ool * a t c l* o o l *amav l *ool *ool *barLr-l *ool *durunr-l Ito shadou 2.5a ASI s \ \ ' c a t2 . - 5 b rntly lized with rain2.5b s p r i n g2 . 5 b c r a n c2 . 5 b * * * * '/asil* ool '?amcl l oo l * ooT pharu-l thurul* ool 1.\ rore rage, Table 6 2.1-2.2 *'1a- 2.3a-2.5a 2.3b-2.5b vorvcl a- +'/a:- *a- obstruent 1rr- *ta- Etit:- *da n a s a ll a - *'/na- +n a : - *na Tvpc of onset tatron :ch on in the r55 ,- may Table 7 r56 A L E X A N D E RV O V I N 4. Conclusion If rny reconstruction above is correct, it rnight have far-reaching consequences for the study of the extcnralrelationshipsof the Japaneselanguage family. A contrastbetwecn srnoothvocalic ingressand an initial glottal stop on one hand, and voiced and voicelessnasalson the other. has no parallels in the 'Altaic' world. Instcad,these fcaturesare widely found throughout South-East Asia. lf the Japaneseproto-languageindccd arrived in Japan togetherwith the Yayoi culture,it should not be surprisingto find some South-EastAsian elunents in Japanese, althoughat the presentstagcofour knowledgethe exact naturcofthe connectionbetween Japancseand any of the major languagefarnilies found in South-East Asia, if any, relnains at best hypothetical. The search for these possibleconncctionsis a legitimatc cnterprise,but it should bc carried out in a caLltlousmanner. RECONSTR PROTO-JAPANESE AND PROSPECTS ISSUE,S EditedbY BJARKEFRELLESVIG Universityof Oxford 6 Universityof Oslo JOHN WHITMAN CornellUniversitY COMPANY JOHN BENJAMINSPUBLISHING A M S T E R D A M / P H I L A D E L P HAI