proto-japanese beyond the accent system
TRANSCRIPT
CHAPTER 7
PROTO-JAPANESE BEYOND THE ACCENT SYSTEM
ALEXANDER VOVINUn ivers i t.t, o/' Huu'ui' i at lllc1nou
l. Introduc'tionThis chapter deals with tu'o problems: the controversy over the archaic
nature of thc 2.5 accent class of norninals lbund only in Kansai Japanesc ($2) andthe possiblc correlation bctween low register and vowel lcngth (\3). Although atfirst glancc these two problcms look unrelatcd, I wil l attempt to show bclow thatthey are both relevant lbr reconstructing ccfiain segr.r.rental features of pJ thateither have bccn considered controversial unti l now, or havc passed altogetherunnot iced.
The traditional reconstruction of pJ accent fbr binroraic nouns is presentcdin thc fbllowing table. Thc follor,l ' ing sources are used: the Ruiju rr.r,rigi.rir i (RM,l0ll l ; a dictionary which rnarks pitch accent), and also dialect data frorn tnodernKyolo, Tokyo, and Kagoshima. Segmcntal fonr-rs of nouns are given in theirMiddle Japanese (M.f ) form.
Accent c lass and reconstruct ion RM Kvoto Tokl 'o Kagoshima2. | /a.ii
'cdgc' p J * l l t l H H H H . H H - H L I I . L H - H I l L . L t l - L2 .2 l i r i i
' b r i dgc ' p . l * t l l HL I l L . HL -L L t l . LH -L HL . L I I - L2.3 /Lnt t
'1 lou'cr ' p. l +LL L I - I I L . H L - L L I I . L H - L L I I . L L - H1 . . 1 / . r i
' r hops t i c ks ' pJ *L I J LH L I I . LL -H I l L . I l L -L t -H . LL - l l2.5 / i r t r ' sp r ing ' p . l *LF LF iLHT LF , LH- t - I I L . I I L -L LH . LL -T I
Table I
This reconstruct ion, which can be t raced to Kindaichi (1975), and is a lsoadopted in iLTT, presents the pJ accent system as practically identical witl i theone found in Lhe futi ju nt.t,ogishct.lt is also quitc close to the rnodem Kyoto systcrnHowevcr, not everyone agrces: this systern is considered controversial by somelinguists (Tokugawa 1972, Ramsey 1979,1980), who proposed that the p.l sysrern
Pitch shapcs in ntodem dia lects arc g ivcn both in isolat ion and wi th a fo l louing paf i ic leI Depending on part icular manuscr ipts.
142 ALIJXANDER VOVIN
was closer to the rnodern Tokyo systetr. Nevcrtheless, the dctarled argumentsprcsented by Marrin in favor c.l1'the traditional systern (.ILTT: 162-1 5) havc beenlcti virtually uncontested unti l today. Twclvc years ago I also provided a detailedrefutation of thc Tokugawa-Ramsey hypothesis (Vovin 199-5). based on internal,phi lo logical , and external cv idence, therefore I wi l l not d iscuss the wholceontro\ crsy again herc.
2. The segmental source ofoccent class 2.5However, there sti l l rcmains one controversial point, which is in some
scnsc peripheral to the controversy betwccn traditional and Tokugawa-Rarnseyrcconstructions, but sti l l deserves our attention and needs to be resolved one wayor another. This point concerns the archaic nature of the accent class 2.5, which istbund only in modern Karrsai dialects. The philological evidence fbr its existenccis rrrore controversial, since not all manuscripts o1- the Rlrflr nt.togisho havc the'eastern dot' (to-ten), believed to ntark the distinctive fall ing pitch on the sccondr.nora of nouns belonging to the accent class 2.-5. Naturally, supporters o1'thetraditional rcconstruction, who are 'Kyoto-oriented'. bclieve in its archaic naturc,u,hile Tokugawa and Rarnscy, who are 'Tokyo-oriented', insist that this accentclass represents a local innovation in Kansai dialects. The r.nost recent challengeto the archaic nature ol- thc 2.5 class was presentcd by Unger who writes(2000a:20-2 I ) :
Thc Russiat t l inguist Pol ivanov spcculated that l inal la l l inu pi tch is a t racc o1'a" t r unca t cd " nasa l phonemc . Ma r t i n ( 1987 .361 n . I t o Chap ts r . { . s \ l , l ) t h rnks hcgot thc idca by contpar ing Korean rrc lo l r 'morning' rv i th. lapanese r- r .srr id. .which is 2.5 and coulc l bc taken back to an ear l ier *asa.r l . Howevcr. Mart infbund that a handful of Ryukyu * 'ords u, i th f lnal nasals unrcf lected in nrain-is lands cognatcs arc scat tcred randornly across reconstructcd accent c lasscs. andHirata ancl I havc huntcd in vain lbr addi t ional Koreirn ntatches rv i th f inal nasalslbr . lapanese 2.5 nasals. Poi l ivauov's explanat ion for c lass 2.5 is thcrcfbre-dubious. ( ln any case. he suggcsts no l ink bctwccn c lasses 1.2 and 2.-5. and apar lf ronr thc la l l ing pi tch in Kyoto- typc dia lccts. therc sccms to be nonc) . . . the f -actthat c lass 2.5 is rs lat ivc ly sntal l and dist inct ivc only in one dia lcct groupsuggcsls that i ts or ig ins should bc sought i r . r rc lat ivc ly recent dcrc loptnentswi th in that group.
Hcrc I assunlc that Unger confuses two different issucs that have noconnection bctween tl.rern whatsoever; f lrst, the correspondence of 2.5 class nounsto nouns with final -lr in Korean, and second, the conespondencc of f inal -ly',found in Haterulna, to -O in other Japanese dialects. The archaic nature ofHaterurna -N has been persuasively refutcd in the l iteraturc (JLTT:74, Oyler l997),and it wil l not conccnt us here. The flrst problern, however, remains, and I wil ld iscuss i t be low in deta i l .
First. thc idea that nouns in accent class 2.5 havc parallets in MiddleKorean (MK) words rvith flnal -nr indeed belongs to Polivanov (1924), who,however, givcs not just onc exarnple but two. Second, Unger and Hirata areapparently unaware of the othcr l i terature on the subject whcrc additional parallels
PROTO-JApr
reflecting a correlation betrMK words with final -m ahave been in vain. The foltsources where they were fir
Old Japanese'r
r r .va 2.-5 'ntoming
l l l l 2.5 'crane'
M. l l i l r 2.5 'p igweed'
l t un r 2 .5 ' sp r i ng '
. s l l x r r ' r vh i t c '
ad. jcc l ivc accent: B
r loun acccnt :2.5 < +siru 'o-m
Az rn r r r ' b l ack '
adjcct ivc accent: B
nor- ln acccnt : 2.5 < *kurwo-m
i ' r lz 'dark '
at i jcct ivc acccnt : A
noun acccnt : 2.5 < *kura-m
uka 'rctl'
ad. jcct tvc accent: A
lroun i lcccnt : 2.5 < *aka-m
rrx o 'b luc/grecn'
ad.jcctiVc accent: fl
noun acccnt : 2.-5 < *awo-m
The table above rerJapanesc nouns. However. i ,distribution of these words iKorean cxternal ev idence is
I Acccnt notat ions arc g ivcn accacccntual data on OJ arc not suffica The compar ison ol - . lapanese and( 1966:23t i ) . but u, i thout rhe discus2.5 in Japanesc.
tentsbeenr i ledmal.hole
,omenseywaych is
r thecondI theture,)centenge'ntes
PROTO-. IAPANESE BEYOND TI IE ACCENT SYSTEM t43
rcflccting a correlation between Japanese nouns belonging to accent class 2.5 andMK words rvith final -m are provided, other-wise their hunt would probably nothavc been in vain. The following table summarizes these known parallels and thesourccs where they were flrst suggested:
Old Japanese'r I \ l iddle Korean sourcc
zr.srr 2.5 'rrorning
/ r / /1 / 2 .5 'c ra r lc '
r ) r ' h r i r r ' t no rn i ng '
/l ri/rlrirl-i 'crirncPo l i r ano r l 9 l - l : l - 51
M.l l i l i r 2.5 'p igrvccd'
pun r 2 .5 ' sp r i ng '7r i f t i r i
'p igvn eed'r
7 ; l r i r i ' sp r i ng '
\ \ ' h i t n ren l 9 t l 5 : l 0 l
. r l r l r . r 'u 'h i tc '
adlcct ivc acccnt : B
noun accent: 2.5 . ' - +sinvo-m
lrrnrz 'b lack '
adjoct ivc acccrt t : B
l ro l ln i lcccnl : 2.5 ' - *kunvo-rrr
i r r i a 'dark '
ad. jcct ivc acccnt : A
noun acccnl : 2.-5 < *kura-r .n
rr la ' rcd '
ad- jcct ivc acccnt : A
noun accent: 2.5 < saka-nr
a r rn ' b l ue /g recn '
adjcct ivc acccnt : B
noun acccnt : 2.5 < +arvo-m
ch r i ' dancc ' ( r ' . )
t h - v t i - n t ' d a n c c ' ( n . )
. t l u 7 r - ' h c l p ' ( r , . )
tv t ) l l ' - t ' r -nt 'help ' (n. )
. r ' l r . r - ' laugh' ( r . ' . )
r t i r : - t t t i - t r t ' laugh' (n. )
. x r l - ' c r y ' ( r , . )
v i t l -v ' r i -nt 'cr1, ' (n. )
l r i l - 'change" ( r ' . )
k i l - i - n t ' changc ' ( n . )
Vov in 1994:150(pre-p.l nominalizcr *-r.n for
ad.jcct ivcs dcnoting colors)
e n o,ounsI -1/,re of9e1),wi l l
iddlewho,a areallels
Tablc 2
The table above represents external evidence in favor of *-m in some.lapanese nouns. However. it does not irnr-nediately tell us of the possible age anddistribution of these words in Japanese itself. Thus, we sti l l have to see wlietherKorean external evidence is applicable to pJ as a whole or is localized in Kansai
' Accenl nolalions are givcn according to EMJ as rcf'lcctcd in thc RrrlTl nt.vitgish6. sincc thc
accentual data on OJ are not suf f lc icnt and. /or arc controvcrs ia l .'The conrpar ison of . lapanesc and Korcan rvords for 'p ig ivccd'was proposed or ig inal ly by Mart in(1966:2313). but wi thout thc discussion o1'corrc lat ion between t inal -nr in Korean and accent c lass
2 .5 i n Janancsc .
t44 ALEXANDER VOVIN
alone. Traditionally all thc comparisons above, including the four lexicalcomparisons and the correspondencc of accent class 2.5 in nouns designatingcolcrrs to the norninalizer -m in MK were considered to be examples suppofiingthe genetic rclationship between .lapanese and Korean. In order to confim ordisprove this position. I wil l offer a detailed analysis of all etyrrologies below.
( l ) OJ usu 2 .5 ' r r o rn ing ' -MK dch6n r ' i d . ' . A l t hough th i s con rpa r i son i susr"rally used to support thc theory of genetic relationship betwecn Japancse andKorean. (cf. Martin 1966:236, Whitrnan 1985:244), therc are problelns with thisetyrrology. First. as whitrnan notcs, it is diff lcult 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 to pJ *asa 2 5 < pre-pJ*asam. Second, Japancse r.r.ra 'moming' has a Imited distribution in Japancse: it isfbund in Ryukyuan only in Shur i and Shi to i d ia lccts as ?a.sa (OGJ l99l :123.Hirayarr.ra 1966284). lt is significant that the word does not present itself in thesouthern Ryukyus. where the rnainland Japanese influence was nrinimal. Thus. thclirnited attestation exclusively in Shuri and Shitoi suggests that in thcse twodialects it is a loanword fiom rnainland Japanese, especially givcn that anotherword, pR *tutor.nuti 'rnoming' (a cognate of M.l tutomete 'carly rnoming'discussed below), is attcstod throughout the Ryukyus, and is also present in Shuri( l { i rayarna 1966:2t34, Hi rayarna 1967 202, OGJ l99l :123) . In addi t ion, the wordn.ra'nro-rning' is also attcsted in u:trrrtu poems, but rnost of the poems in which itappears' do not have very distinctivc EOJ features.(' I suspcct that EOJ nsn rnay bealso a loan front WOJ. Givcn all this and the fact that it is irnpossible to connectpJ *asa < pre-pJ *asam with pK *acoKorn or *aKocom, I think that OJ a.sa is aloan liorn some lbrm of old Korean like *acharn that already had an aspirated+-ch- (and not a sequencc *-Koc- or *-coK-) that would be expected to beborrowcd as o.f or pJ *-s-. Even though oJ a.sa and MK ach6rn are not geneticallyrelated, the accent pattern 2.5 is sti l l confirmed by the Korean word as derivingfiorn *-n, but since the Korean loanword is wcll attested only in WOJ, it canconfirrr the archaic nature of *-m only fbr central Japanesc (cJ),/but not for allo f Japanesc.
(2) O.f tunr 2.5'cranc'- MK rr.r,zilx,t int-i ' id. '. Thc correspondences seemto bc irnpeccable, and there are no problems that would point to a loanwordscenario, unlikc the etymology for 'rnorning' discussed above. The -i in thc MKforrn is l ikely to be a dirrunitive suffix, cf. tv'ulem without this suffix in thedivergcnt ceycwu dia lect (choy 1987:878). In addi r ion, a l though the word is notattested in EoJ, it docs not suff-er frorn l imited distribution in Ryukyuan:Kikaij ima (Amarni), Sani (Arnarni) tsuru; Tokunoshima (Arnami) Lwruntui;
' Scc cranrples convcnicnt ly pror , idcd in Mizushima (19134:646).
" 1 'hc only real cxccpt ion is MYS 3-502. which has sornc rypical EOJ vcrbal morphology.- 'C'entral
.lapanese' rcI-ers to a branch of Japancsc which cornprises WOJ, MJ, and rhe ntajority ofntodern rnain land c l ia lccts. but i t does not inc ludc I rO. l and the nrodcm Hachi jc l - j ima dia lect .
PROTO-J,,
Yoron. Izena. Kumej imtt.|ilru: Shirobe (Miyako),Takebu (Yaeyama) tfiruEven if the Japanese worprohably l iom the periorthis etymology fully supfrom a lost f inal *-m.
(3) MJ ./ i.vu 2.5 ,1
attestation on the JapanesH,,n:1|y4r . r . r i 19181. so i ti ls OJ lbnn was *p iy t wi/-wi/ goes back to pJ *qidiff i cult. However, sincecountelpart /-wi/, thereNcvcrtheless, there is atnamely the correspondenwork wc must first assumof -1- to -1-, which is possassurre that MJ /i.vu < ysupportcd, but it is entirYonaguni dialect which irnedial -.t- corresponds Iternporal and geographicelsewhere (JLTT:20). Thrthis chapter, but there areon Yonaguni r/- (Whitmalproto-fom of the Japanecorrespondence of pK *-t.MK) or a correspondenceand both require a greatetymology conditionallyetymology. and therefore,Japanese accent class 2.5 t,
(4) OJ paru 2.5 ,s1
grave phonetic problerns. I/ i/. Second, even if this wxaCu corresponding to Mmerely coincidental reseml
t Marlin proposed to reconsrrucrnowadays it seerns that this prop''
l dcdicatc a substanlial chaprer
nA
(
e)tl :
i ;
of
PROTO-JAPANESE BEYOND TIIE ACC'[]NT SYSTEM 145
Yoron, Izena, Kuntejirna (Okinawa), Ishigaki (Yaeyama) tsuru, Hirata (Miyako)
frrrr; Shirobe (Miyako), Kawahira (Yaeyarna) l^sirrr; Kuroshima (Yaeyama) .[r4:Takebu (Yaeyanta) t.f iru; (Nakamatsu 1987:4t3, 1 17, 165, 201, 248, 281. 326)'
Even if the Japanese word is a loanword from Korean, it rnust be a very early onc,
probably fiom thc period of rnutual cocxistence on the Korean peninsula. Thus,
this etyr-nology fully supports the interpretation of accent class 2.5 as rcsulting
from a lost f inal *-nt.
(3) MJ /i1,u 2.5 'pigweed' - MK pilt im ' id.'. First, there is a problem of
attestation on the Japanese side. The word appears fbr the first time in the herbary
Honzou,amt'ri (9ltt), so it is MJ, not OJ. This leaves room for unceftainty whether
its OJ fbrm was *piyu with ko-ruil- i l or *pwiyu with r.rl^stt-rui l-wil . Since olsu-rai
/-wi/ goes back to pJ *oi or *ui, this would make a comparison with Korean rather
il i ff icult. However, since k6-ruil- i l appears more fiequently in OJ than its tt lsu-nri
counterpart /-wi/, there is a strong possibil i ty that it was actually OJ *piyu.
Neverthcless. thcre is another possiblc phonetic problcrr.r in this comparison.
namely the correspondcnce of MJ --t,- to MK -/-.n In order for this conrparison to
work wc must f lrst assurle that MK pil i lm < pK *pitum, with well-known lenition
of -l- to -1-, whrch is possible, but again cannot be proven. Second, we also must
assume that MJ /i.t 'u < pJ *pidu. The hypothesis that O.l . l ' < pJ *d is widely
supported, but it is entirely based, I am afraid, on a mistaken theory that the
Yonagur.ri dialect which indeed has init ial z/- corresponding to QJ 1' (though oJ
medial -.1'- corresponds to Yonaguni --) ', not -zl-) preserves, alone anlong all
ternporal and geographical varieties of Japar.rese. a pJ *d which shifted to }
elsewhere (JLTT:20). The full discussion of this problem is outside the scope of
this chapter, but there are some good prelirninary argulnents against pJ *d based
on yonaguni r / - (whi tman 1985:18) . ' 'Wi th pJ *p idu being re jected as a possib le
proto-fbrn-r of the Japanese word, we are faced with two possibil i t ies: either
corespondence of pK *-t- to pJ *-y- (assuming that a lenition took place prior to
MK) or a correspondence of pK *-l- to pJ *-y-. Neither is an enticing solution.
and both rcqr.rire a great deal of explanation, but I ar.n wil l ing to accept this
etymology conditionally for t ire time bcing. Sti l l , i t is not a very reliable
etymology, and therelbre, it cannot provide full support for the correspondcnce of
Japanesc accent class 2.5 to MK -rl.
(4) Oi paru 2.5 'spr ing ' - MK pw6m ' id . ' . Here. I bel ieve. we have two
grave phonetic problcrns. First, OJ -/1 never corresponds to MK -O- excepl befbre
/i/. Second. even if this was to be confinned, there are no other instances of pJ*aCu corresponding to MK u,o. Thus, I think that thcse two words represcnt a
mcrely coincidental resetnblance.
' Mart in prgposcd to rcconstruct pJ K +i on thc basis of th is corrcspondcncc ( Mart in | 966:2 I I ) . but
nowadays it scct.t.ts that this proposal is no longer supported.
" l dcdicate a substant ia l chapter to th is issuc in a fbr lhcomtrrg book (Vovin. fbnhconl ing).
146 ALEXANDER VOVIN
(5 ) . lap i lnese acccnt c lass 2.5 for nominal forms of adjcc l ives l " designal ingcolors - MK verbal norrinalizer -rr. Phonctically, the conespondence seetns to bcimpeccablc. However, thcrc are trvo problcms. First. this particular shift from anoriginal accent class to accent class 2.5 when a word is uscd as a nominal is
l irnited in Japanese to adjcctives only. Second, we cannot verify that the sal"neprocess was applicable to other varieties simply because accent 2.5 class is not
attcsted in other varieties of Japancse outside MJ and the Kansai dialccts ofr-nodcrn Central Japanese (CJ). Thc l imited attestation within Japanese in this case(quite sirnilar to the case of the word for 'n-rorning' discussed above), I bclieve.pcrints rnore to thc borrowing scenario lrom some varicty of Old Korean to CJthan to colnmon genetic inheritance. This is even rlorc l ikely in the l ight of thelact that we arc dealing here with derivational and not irrf lectional rnorplrology;and thc fbnner is nrorc casily borrowed than the latter. Thcrcfore, I believe, thiscorespondence confims the archaic naturc of accent class 2.5 within CJ, but not
rvithin Japanese itsell ' .
Thercfbre, an attempt to rcsolve this controversy with the help of externaldata leaves us with only a partial solution. lt could be confinned that accent class2.5 has an archaic nature in CJ. so Tokugawa and Ramsey's point of view that thisaccent class originated in the post-MJ era can be rcliably refuted. However, it isirnpossiblc to confirrn the existence of accent class 2.5 in pJ. bccause out of thefive exarnples discussed above only (2) has a potential for being explained eitheras a comlron inhcritance or as a very old borrowing.
Neverlhcless, external evidence should ncver take precedcnce in
cxplaining internal data, otherwise one runs the very high risk of fall ing into a
Moscou'Nostratic trap. where intcrnal etymologies arc disregarded and externaletyrnologies overridc the intemal data. I bclieve that there is substantial internalJapanese evidencc supporting the archaic nature of acccnt class 2.5 not only onthc level of CJ, but on the pJ level. This evidence is bascd rnainly on dcrivationalmorphology. Consider the fbllowing examples:
a. O.l x,rlsn 2.5 'cldcr'
cf. O.f v'o.iarlc- 'to rule' < *wosarnaCi- < *wosatn-(r)a-Ci-.
b. OJ togu 2.5 'b lamc, of fence'cf . C)J toguntc- 'to blarne, to reproach' < *togamaCi- < *togarn-(r)a-Ci-.
r" So-cal lcd adjcct i res in MJ and other latcr var iet ies ofJapancsc are in {act qual i ty vcrbs. and not"ad. ject i r ,es" in thc scnsc of thc tcrnr assigned to adject ivcs ofEuropean languages. Thus. e.g. . MJ
l i i r t r - docs not lnean 'dark ' but rathcr ' to bc dark ' .
PROTO-JI
c. OI tutwo(_ni)cf . MJ tutome'to stnve, tounclear.
d. OI pa.t,a 2.5 '<
Examples (a-d) alsome denominal verbs uulLtldnlL't 'u "to wann up'kl 'omeru 'to purify' (< tnagame- 'to prolong vot,analyzes these verbs as cr(ILTT 792). At first gtanthat intransitive verbs crwhich Martin reconstruclalalarnlru'to bc warmedare lhrce argurncnts again
( l ) * -ma- i s a su funnecessary entity. By Cwith the nominalizer *-m
(2) Mart in 's analyrproduces yet one more supafi as norninalizer *-m-
rcceive qui te a symmetr ic i
transitives: adjecinrransi t ives: adjec
(3) In Mart in 's anrbr.rllz nouns and adjecttvetrcated as a final phonemrafter adjectives.
One rnight argue iand acccnt class 2.5, be,or -ment bclong to the 2.rnake go to thc end', frrpai
" No t l i s t cd i n Zdb . Ohno (199only obl iquc cvidence fbr tnat tthc basis ofthc fact that the vo,,as u 'c l l as possib ly on the basrs rr r Vcrbs l ikc homaru'praiscs ' ot
n g" D
beanis
me1otof
tse
CJthegv;hisnot
nalASS
hist r stheher
lno analmalon
rnal
PROTO-.IAPANESE BEYOND THE ACCENT SYSTEM 147
c. OJ tutw,o(-n i ; r r 2 .5 'ear ly in the rnorn ing,cf. MJ rulotnere'id.', believed to be a derivation from the oJ tutv,ome_'to strivc, to tnake effbrt ', although thc sernantic connection rernainsunclear .
d. OI pa.tu 2.5 'quickly' , cf . MJ / 'o.t,unte_
,Io quicken..
Exanrples (a-d) allow us to reanalyzc the derivation ofall deadjectival andsorle denominal verbs in -me- such as aralotl lent .to renew' (< arala-.ncw,).atatemeru " to warm up' (< orataku- 'wanr ' , i lumeru. to hur t ' (< i tu_.painfu l ' ) ,k i .vomeru ' to pur i fy ' (< k i ) ,o- 'pure ' ) , sebameru ' to narrow' (< semlba- .narrow') ,nagume- ' to pro long voice when rec i t ing poetry ' (< naga- ' long ' ) , e tc . r r Mar l inanalyzes thesc verbs as consisting of stem+suffix *-rna- + transitivity f l ipper *_Ci_(ILTT 792) At f irst glance, his point of view scerns to be supported by the facrthat intransitive verbs comespon<1ing to trarrsit ives in -meru all cnd in -ntut.tt.whiclr Martin reconstructs as *-ma-ra- (19g7:792): uratamartL .to be renewed'.atulutnunt 'to be warmcd up', sehamunt 'ro become natrow', etc. However. thereare thrce arguments against Martin's analysis:
( l) *-rra- is a suffix with an unclear function, thus it represenrs an extraunneccssary entity. By Ockham's razor ir is, therefore, pref-erable to replace itwith the norninalizer *-m to which verbalizer *-(r)a_ is added.
(2) Martin's analysis of intransitiv es in -ntaru as having a stcm in *-ma-ra-produces yet one more suffix *-ra- with an unclcar function. I icanalyze tl.re -ntarpan as nominalizer *-m * ysl6slizer *-(r)a- + passivizing suffix *_r_. Thus, wcrecelve quite a synlmetrical picture of how transitives and intransitives arc derived:
transitives: adjectival stern + *-nr * *-(r)a- + *-Ci-intransitives: adjectival stern l *-1r * *-(r)a- + x-1-
(3) In Mafiin's analysis it is unclear why his verbalizer *-ma- ls added tobolh nouns and adjcctives. This strangc fact can be easily explained if *-rn istreated as a final phonerne of the nominal root, but as a nominalizing suffix *_rnaf tcr adject ives.
one rnight argue against the connection between the final r-m rn nounsand accent class 2.5, because not all nouns with denominal vcrbs in _nrcruror -tnant be long to the 2.5 class. Thus, oJ kipa 2.3 'brink', cf. oI kipa,t?e- .tomake go to the end' , k ipamar- , k iparn- ' to reach the end' , oJ.s iu, '2 .1 .wr ink le ' .
" Not l is ted in Zdb. Ohno (1990:t t t t4) t ranscr ibcs i t as tutv 'o rv i th type A / -u,o/ , a l though thcre isonly obl ique o ' idence for that (s ince thc word i tsc l f is nor at tested phonographical ly in O. l ) onthe basis of thc tact thal the vowels /u/ and 1oi do not normal ly .u,r . rb in. rog. tn.r wi th in 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'
ctr hu:irnLertr 'bcgins'. of coursc, do n.t bclonq hcre.
ci-
d not., MJ
l4u n LITXANDER VOVIN
OJ .s lnnnl - l r ' to wr ink le ' (Modem. lapanese s iv 'onte-) . The second case looks l ikea real exception. but I rvil l try to show later that there is evidencc for OJ kipa 2.3'br ink ' having an or ig inal 2.5 accent .
Therc are also sorne bird narnes belonging to accent class 2.5 besides 1lo"l' c rane 'a l ready rnen t i oned above : . s rg i ' sn ipe ' , OJ t c t t l u ' c ranc ' , l r . r b l ' k i t e ' ,tokilnki 'rbis'. lukulzrrl lr 'owl'. Cf. sufflx -me in other bird names: stt:ume'sparrow' . kumome'seagul l ' . . r in lc 'har t ' f inch ' . MJ lubukuronte 'swal low' . I t isinteresting that that none of thc bird narles found in acccnt class 2.5 has thissufflx. Thus. it is possible to speculate that accent class 2.5 in these wordsoriginatcd as a truncation of thc sr,rff lx -rie.
The evidence citcd above directly points to *-m as a source ofaccent class2.5. ln addi t ion. there is in tcrnal ev idence that ind icates that some k ind of nasal(no d i rect cv idence for * -m) undcr l ies accent c lass 2.5:
a. OJ lr.s[l lr.r 2.5 'a breath cxhaled through a narrou'cd rrouth', cf. OJusfir)obuk-lu,;lu' lontuk- 'to exhale a breath making one's ntouth narrowcr'< *usu'onpuka- < ?*uswom-puk-(r)a-.
b. J nusu 2.-5 'cggplant ' , c f . MJ nusubi 3.5b ' id . ' , wi th unclear suf f lxat ion <*nasunpi < ?*nasumpi. which can be an altcrnation of an earlier *nasutn-
l .
Finally, thcrc are some exafirples that constitute additional externalevidence (whether these are to be taken as cognates or as early loanwords):
a. OJ pemi 2.5 'snakc' has *-m preserved intact due to the fact that it isfollowed by obsolctc sufflx -r (dirninutive'?). Marrin believes IhaI OJ pemi2.5 < *paCi imi 3.6 (JLTT:404). Cf . MK p61'dm'snakc ' .
b. OJ nlrsl 2.5 'rnaster' "- ' OK *rrVlim ' lord' (> MK '.nint). The vocalism inthc first syllable may bc a problernl the vocalism of the first syllable in OKis unclear.
c . OJ mosu(-n i ) 2 .5 ' t ru ly ' , possib ly f rorn mu- ' t rue '* - .so ' ' ! ' , which can becompared with MK c'hom 'truly, really'.
d . OJ .saru 2.5 'monkcy ' - WM sannay' t ' in ' id . ' .c. OJ u.sagi 3.6 'harc'. cf. development of 3.6 to 2.5 in OJ pemi 'snake'
above. Cf. Koguryo *osigam 'hare'.
The last cxarnple brings us to an intcrcsting internal issuc. On the basis ofKoguryo *osigam we would expect that the p.l xusagi would bclong to accent
'Zdb l is ts.srurr i l - as a consonanlal vcrb. Thc cramples. both in logographic scr ipt in OJ and in
phonet ic in MJ. prol idc no basis lbr conclusion thal .v iualr - used to be a consonant rather than a
vou,e I r,crb in O.l. The earliest knovn'n attcstation of this "'erb
is in thc fom silrrbi/arrlit'r'ir 'has
*,rinkfed' (Konjuku nutnoguturi 26.2) that shows that it is either a consonantal or upper-bigradc
vou'c l vcrb (Ohno 1990:695).' ' I anr indebted to Bla inc Er ickson lbr noint ins th is out to l Ic .
PROTO-JI
c lass 3.7 (LHL-L) that caccent c lass 3.6 (LHH_Ihave a number of casesaccentuatton could makealthough the bulk of thr'sa lmon'
2.512.3 ' l or abo,typical for accent classesnrade above regarding thsuggests a hypothesis tlattrit ion: a number of wcunder 2.4, 2.3, and 3.6. r5lirrrhcr supporl rhis aflrit ir
OJ asa 'hemp' 2
Kyoto and Shuri is ineguaccentual class, although l
Now we can retur2. -5. OJ k ipami , l imi t ' ,
drKyoto has irregular accenand not 3.6, suggesting, hr
Final ly , I am norntonotroralc accent classwords. I would rather fol( lJH-L) and 2.2b (HL-L) .the fall of pitch that Ungethc fall within Ihe last mc
./it I I ov, i n g case pa rtic le, buThe isolation forrn for mo,to the fact that Kyoto autoalso ncccssary to keep in rregister class. Thus, one stbelonging to accent class I
Therelorc, I believcsolne extent external, to l irf inal consonant *-rn. It rs tto1'final *-r.r.r for every nourno surprise, because not erfiom these nouns. It is notexplanation is even smalleris that accent c lass 2.5 rsrightfully belongs ro pJ. Hbecause its specific shape
r5 Mart in notecl that . .a nunrber o lsh i f i cd t o 2 .1 " ( JLTT : I 77 ) .
case looks l ike-or OI kipct 2.3
.5 besides /lo' lr: ' , l r . r b l ' k i t e ' ,alnes" sllzltme;wal low' . I t is;s 2.5 has th isr these words
ofaccent c lassr kind of nasal
routh', cf. OJruth narrower'
r suffixation <arlier *nasum-
ional externalvords):
fact that it iss that OI pemi
re vocalism insyllable in OK
which can be
pemi 'snake'
)n the basis ofong to accent
ript in OJ and inrant rather than avabiturikeru 'has
or upper-bigrade
PROTO-JAPANESE BEYOND THE ACCENT SYSTEM I4g
class 3.7 (LHL-L) that conerates wi th accent crass 2.5 (LH_L), but r t belongs toaccent crass 3.6 (LHH-H), which corrcrates wi th 2.4 (LH-H). F.ur thennore, wehave a number of cases where either a lack of accentual dutu o. rrregurarity inaccentuation courd make a choice between 2.5 and 2.4 0r 2..5 0, i : qucrtionabre,a l though the burk of the ev idence norma'y points to crass z. i (c f . , e .g. , .sare'salmon' 2.512.3'r or above-rnentioned u,,^ra .elder,
2.512.4,r\. The sa'e prcture rstypical for accent crasses 3.6 and 3.7. This, in combinat ion wi th the observat ionmadc above regarding the potentiar Koguryo cognatc *osigar.r for pJ *usagi 3.6,suggests a hypothesis that the srnall Jasses 2.5 and 3.7"a/b becarne srrall byattrit ion: a number of words that used to bclong to those classes are now fbundunder 2.4, 2.3. and 3.6. r 's There is one addrt ronar extemal compa.son that courdlurther suppor-t this attrit ion hypothesis:OJ asu 'hernp'
2.3 _ MK sam ,id., . The accentuation in both modcrnKyoto and Shurr is irregular (JLTT:3g4), possibry suggesting a shift frorr anotheracccntual class. although Kyoto and Shuri contradict each other.Now we can return to the issue of why oI k ipa,br i 'k , is 2.3 rather than2 5. oJ kipctni 'r imit ',
derived frorn kipatn- .to reach the end,, is 3.6 (arthoughKyoto has irregurar accent). the 3-mora accent class that correrates with 2.3 is 3.4,and not 3.6, suggesting, however, that the original crass of tr4r,,rruy rruu" becn 2.-5.Finally, I arn, not. srrre that Unger is right when t "
lrr",-t, that the'onol rorarc accent c lass 1.2 (H-L) corre lates wi th accent crass 2.5 lbr b imoraicwords' I wourd rathcr follow the traditionar position that it corrcrates rvith 2.2a(llH-L) and 2.2b (HL-L). The principar diff 'erence between 1.2 and 2.5, regardingthe fall of pircrr that Unger rnentions, is thai in Kyoto-type diarects class 2.5 hasthc fall v'ithin rhe last mora ,in i.vtratetr /brnt, and r4ie,:ihe last mora he/itre tha./bllov'ing t 'ase parric/e, but crass 1.2 in hoth.a.rer has the fall ct/ ierrhefirst rnora.The isolation forrr fbr modem Kyoto is HL, the same as to, u...nt class 2.2, dueto the fact that Kyoto automatically rengthens one-mora words to two_moras. rt isa lso necessary ro keep in min. tnai z .s i , u to*- regis ter c lass, whi le I .2 is a h igh_register class. Thus. one should not expcct to find any final *-,n unJc.tyrng wordsbelonging to accent c lass 1.2,
Therefbrc. I berievc that there is suflrcient evidence, rnainly internal ancr tosorne extent extcrnal. to l ink the origins of accent class 2.5 in pl wrtn a loss of thcfinal consonant *-rn. rt is true that it- is not fossible to dernonsirate the probabirityof f inal *-rn for every noun belonging to accent crass 2.5, but that shourd come asno surp'se, because not every noun in this crass has verbal .oun*,pu., ' dcrivcdfrom these nouns' It is not surprising that the number of words with an cxternalexplanation is even smailer. Thus, th*e concrusion for the trrst pu.t ortnr,, chapteris that accent crass 2.5 is not a recent innovat ion in Kansai d iarects, but that i tr ightfully belongs to pJ. However, within the pJ accent systenr i1 is secon<lary,because its specific shape (the fall of pitch wrthin the last rnora) rs due to a
:;Y,::il i: j:i j l1l;X#i:r., of words in crass 2.1/5 with nnaropcrr vower in ord.rapancsc harc
1 5 0 ALI]XANDER VOVIN
sinplif ication of segmcntal phonetics, nanlcly the loss of the final consonant *-tn.
This should come as no sulprise. becausc pitch acccnts. l ikc lones. do not
represent an original characteristic of a language; they rnust colne from
somcwhere. This 'sorncwhere' usually tulxs out to bc certain segmental f 'catures
that cither merged w'ith other features, or have been lost altogethcr, leaving their
traces in the shape of tones or pitch accent. This brings us to the second problcnl
to be discussed in this chapter: the conelation bctween low register and vowel
length.
3. The segmental sources oJ low registerMartin suggested that pj low register probably reflects init ial vowel length
in pre-p.f (ILTT:249-53). I later argucd that this solution is l ikely to be only
partially plausible, because thcre are morc words with init ial low registcr than
words with high registcr, and this creates a typologically unjustif icd situation lvith
long vowels occurring more frcquently in init ial syllables than short vowels(Vovin 1995:131). I proposed to remedy th is by posi t ing two sources for words
with low register: pre-pJ vowcl length. as suggestcd by Martin, plus prc-pJ init ial
voiced consonants. and offcred somc external cvidence fbr the origin of low
register from init ial voiccd consonants (Vovin 1995 124-31). This solution,
although typologically more clegant, suf-fers from three rnaior problems:
(1) Until recently it was impossible to tell on thc basis of the internal
Japanese evidence which low register words would retlcct init ial vowel length and
which u'ould reflect init ial voiced consonantsl
(2) t was sti l l adhering at this point to thc traditional reconstruction that
posits pJ *b- and *d- on the basis of the Sakishima Ryukyuan evidence. This
crcated two different reconstructed series of voiced *b- and *d-: one on the basis
of the Sakishima evidence. and the other on the basis of the lou'register;
(3) It was unclear what to do with high register words that start with
vowels or nasal sonorants *m- and *n-: surely these are also voiced. Thus. on rny
earlier cxplanation, they should be all expected to be found in low register classes,
which ccrtainly also have the ir share of vowel and nasal sonorant inlt ial words.
My tentative solution was to posit an init ial glottal stop or *H- for high register
words starting with vowels and nasal sonorants versus smooth vocalic ingress and
pure nasal sonorants for low register words (Vovin 1995:131-32) . As I wi l l t ry to
demonstrate below, this solution may sttl l be partially true. although at that t ime
without a solution to ( I ) it was lrlere speculation.
Regarding the major problern ( I ) above, Hattori was the first l inguist to
suggest that Ryukyuan preservcs vowel length in some words belonging to accent
c lasses 2.3,2.4, and 2.5 (NSNT 21 ,22) . He a lso added a couple of exarnples of
words with vowel length from accent class 2.2. Hattori 's proposal was largely
PROTO-JA]
bascd on correspondencerand Onna, as vowel lenlHatror i labeled 2.1-2.5 nosarne classes without vow,2.3b-2.5b below). The apparc a number of examples(NSNT 2 l : 103 -05 ) , t he expresentatrons are very scaout. l lattori 's correspondtsignil icant refinement of Fby Moriyo Shirnabukuro,Shirnabukuro, this volumernostly corclate with thernost u,ords belonging tootherwise secondary (SShirnabukuro managed treconstruction of vowel ldialects rather- than just ondata are not consistent. INakij in dialec regularly cr2.3a-2.5a, whi le there is nclcngth in Shuri (subclasses
Gloss &
rnortar 2.
brcath 2.,
needle 2. ,
shadow 2
bucket 2.
bridegroc
flower 2..
nountaln
cloud 2.31
shoulder I
board 2.4i
ra in 2.5b
sweat 2.51
Shimabukuro recons2.5a in th is table and in the
' t ' Fn. I in Shimabukuro's chapter
PROTO- jAPANESE BEYOND TI IE ACCI INT SYSTEM l5 l
based on correspondences within several Okinawan dialects, in particular Shuri
and Onna, as vowel length does not seem to be prcserved outside Okinawa.
Hat tor i labeled 2.3-2.5 nouns wi th vowel length as subclass ( l ) and nouns in thc
same classcs without vowel length as subclass (2) (these are labeled 2.3a-2.5a and
2.3b-2.5b bclow). Thc apparcnt weakness of Hattori 's proposal is that while there
are a nurnber of examplcs with and without vowel lcngth fbr classes 2.4 and 2.5
(NSNT 2l :103-05) . the examples wi th vowel length fbr c lasses 2.3 and 2.2 in h is
presentations are vcry scarce (NSNT 22'.100-0 l). In addition, as Martin pointed
out . Hat tor i 's concspondences somet imes are not consistent (JLTT:252-53) . A
significant refinerncnt of Hattori 's proposal has been made in a reccnt dissertation
by Moriyo Shimabukuro. who demonstrated two impo(ant facts (2002: see also
Shirnabukuro, this volutne). First, words with l irst syllable vo\\ 'el length in Shuri
rrostly corrclate with the low pitch classes 2.3-2.5, and init ial vowel length in
most words bclonging to classes 2.1-2.2 is eithcr found in colnpounds or is
othenvise sccondary (Shimabukuro 2002:293). Second, most lmportant,
Shirnabukuro rnanagcd to find a strong intcrnal colTclation supporti lrg a
rcconstruction of vowel lcngth in Ryukyuan on the basis of two Okir.rawan
dialects rather than just onc, thus taking carc of Martin's objection that Hattori 's
data arc not consistent. Shirnabukuro demonstrated that init ial accent in thc
Nakij in dialcct regularly corresponds to vowcl length in Shuri for thc subclasses
2.3a-2.5a, rvhile thcre is no such colrcspondcnce in the case of absencc of vowcl
lcngth in Shur i (subclasses 2.3b-2.5b) (Shi rnabukuro 2002:203):" '
Gloss & accent class Naki i in Shuri
nrortar 2..la
brcath 2.4a
necdlc 2., la
shadou' 2.5a
bucket 2.5a
bridegroonr 2.5a
l lorver 2.3b
nrountain 2.3b
c loud 2 . lb
shoulder 2.4b
board 2.4b
rarn 2 .5b
su'cat 2.5b
Table 3
Shimabukuro reconstructs vowel length for the words belonging to 2.4a-
2.5a in this table and in the absence of any other cogent explanation I would l ike
' " Fn. I in Shirnabukuro's chaptcr rn th is book explains the syrnbols used in Tables 3-6
' lulsi(:) -
'?u:si
i lc i( :) : '1i : .^ i
p la l i ha : i
ha lg i ( : ) ka :g i
n u l t n i ( : ) u : k i
mu lhu( : ) ,n ru .ku
_phirna: hana
yarrlil: Jarlla
KUIIU: KLl l l lu
hata: kata
hica: '1 ica
_'?arli '/anri
hasi : '?asi
;t)
oIt)fi]
1 5 2 ALI ]XANDER VOVIN
to agree with him. What rernains to be explained, though, is the absence of wordsbelonging to 2.3a in this table; in this regard we achieved no significantirnprovement ovcr Hattori. Shirnabukuro also presents tl.rree exceptions to the
correspondence o1'long vowel in Shuri to init ial acccnt in Nakrjin, whcrc init ial
acccnt in Nakij in corresponds to short vowel in Shuri (Shimabukuro 2002:204):
Gloss & accen l c l ass Nak i i i n
bonc 2.- la
boat 2.4a
sca 2.4a
phLrlni( :)
phu-lni(: )' iulnri( :)
hun i
huni
?umi
Table 4
Shirnabukuro argues that 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 suppofis this point with cxarnples of long vowels in thesethrec words found in thc Kamishiro. Onna, and Matsuda dialects (Shimabukuro
2002:204), but this is sti l l reminiscent of Hattori 's irregular correspondencescrit icized by Martin. Although Shimabukuro may be ultimately right. for the sakeof kecping the corrcspondences as tight as possible, I wil l omit these three wordsfrom tufiher consideratton.
Therefore. although the issuc of regularity within Okinawan raised byMartin has bcen" I believe. rcsolved answered in Shimabukuro's dissertation.there renrains a problem as to whether it is possible to reconstruct separatesubclasscs (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 Shimabukurcrfailed to providc a decisive answer to this problem (both give iust one cxampleeach for subclass 2.3a (NSNT 22:100. Shirnabukuro 2002:204, 369)). theirintuit it,c solution is in fact correct, because there are uncontroversial examplesbelonging to subclass 2.3a that I add in Table 5 which is othcruise based on atablc found in Shimabukuro's dissertation that presents the reconstruction of two
diffcrent subclasses (a) and (b) for acccnt classes 2.3-2.5 (Shirnabukuro 2002:369-
7 0 ) . ' 'One can clearly sce that words bclonging to subclass (a) have vowel length
in their f irst syllables. supporting Martin's hypothesis about the origin of thc low
register from vowel lcngth. But it can be supported only partially, since thcre are
no traces of vowel length in the u'ords belonging to subclass (b). However, it doesnot make sense that some words itt p.l have init ial low register. and others have
r - I took thc l ibcrry o1'ntaking sorrc changes to Shirr rabukuro's pJ and pR reconstruct ions that ,1br
cxample. rcflcct the cument u'isdonr on presen'ation of prir.nary pJ *o and xc. which underwcnl
rars ing in ( 'entra l Japancsc to, 'u/ and, , i i . tsesidcs *ords bclonging to 2.3a. I have also tnadc sotr te
other addi t ions rvherc ShimabLrkuro 's onsinal cxarnrr les d id not seern to be suf l ]c ient .
PROTO-JAP,
init ial vowel lcngth, withregistcr. It is rnost l ikely I
Glosstor lo ise 2.3a
ol fshore 2.3a
ja r 2 .3a
bcan 2.3a
ba l l 2 . 3a
root 2.3a
l lowcr f .Jb
ntountain 2.3b
c loud 2 .3b
dog 2 .3b
wal .c 2. ib
shank 2 . l b
brcath 2.4a
ntoftar 2.4a
chopst icks 2.4a
needle 2.4a
board 2..{b
lbotprinr 2..{b
s l t ou ldc l 1 . 46
drcg: ) .4b
nun rhc r l . 4b
sk i r r 1. . {b
hueke r l . 5a
br idcgroonr 2.5a
r o i cc 2 .5u
:p i de l l . 5u
sh rdou l . 5a
s i r ca t 2 .5b
ra i r r 2 . 5b
:p r rng l . 5h
c ranc 2 .5b
origin in pre-pJ scgmentalsorne u'ords belonging to tlsecr.ns to be the only conctinit ial low register developseglnents. There are simplyaecounl of - words belonginwords belonging to subcl
PROTO-. IAPANESE BEYOND THE ACCENT SYSTEM
init ial vowel length. with low registcr and vou,el lcngth later rncrging as lowregister . l t is rnost l ike ly that the in i t ia l low register in subclass (b) a lso has an
Gloss pR pMJ pJ
tortoisc 2.3a *kha:rnc) *- ooO) * kamey * OO * ka:may * ooOolkhorc 2 .3a * '?u :k i ) * ooO) * ok i *_OO * o :k i x . .ojar 2.3a xkha:rnc) * ooO) * karne x C)C) * ka:rre * ooobean 2.3a *nra:r lc) *_ooO) * rranrcy + OO x ma:may * oo(.)bal l 2.3a *nra:i) * ooO) * r lar i * OO * nra:r i * ooOroot 2.3a *ntu:tul x ooO) * ntoto * OO * nro:to s oo()l lou'er 2.-3b * phana * OOI * pana * 0O * pana * OOnlountain 2.3b * vanra * OOI * 1,a1ta * OO * yalra * Oocloud 2.3b x khuntu * OOI * kuntwo + OO * kunto * (X)
dog 2.3b * ?cnu * OOI * irru * OO + c'u + OC)u,ave 2.3b * narni * OOI * natl i + OO * nanti * OC)shank 2 .3b * sun i * OOI * sunc ' * OO * sunc + (X)
breath 2.4a s'?i:ki) *ooo) * i lkyi *o[o * ' / i : [ki *oo[o
tnorlar 2.4a * '?u:si) *ooo) *uf su *O[O *' /u:[su *oo[o
chopsticks 2.4a *pha:si) *ooO) *pafsiy *O[O *pa:[suy *oo[O
nccdle 2.4a *pha:r i) *ooO) *pafr iy *O[O *pa:[ruy *oo[{)
boarcl 2.4b x ?ital * ool * i [ ta *o[o *i [ ta *o[c)
lborprinl 2.4b * 'latul * OO'l *a[to *O[O *a[ro *o[C)
shoulcl i i r 2.4b * kharal * (x)- l *kalta *o[O *kafra *O[o
dregs 2.4b * khasul * ool *kafsu *O[o *kalsu *o[o
numbcr 2.4b * khazu-l * Ool *kalnsu *O[o *kafnsu *o[o
skin 2.zlb *,-phada-l * OOI *pafnra *O[O *pafnta *O[t)
buckct 2.5a *wo:khe) *ooO) *wokeyT*Oo-l *wo:kayl *ooO-l
bridcgroon 2.5a *rro:kho) *ooO) *mukwo-l *OOl *mo:kol *ooO-l
vo ice 2 .5a *khu:c ) *ooO) *kq \ \ ,cy l *OOl *ko : rvay l *ooOl
spidcr 2.5a *kho:bo) *ooO) *kurnwo-l *(X) l *ko:nrf plo i +ooo-l
shadow 2.5a *kha:ge) *ooO) *kagcyT *OO-l *ka:nkay-l *ooOl
sweat 2.5b * ' /asi l* ool *ascl *ool *ascl *(x) l
rain 2.-5b * ' lanrcl * ool *amcy-l *ool *amayl*t)Ol
spring 2.5b * pharuT * Ool *parul*ool *paruT *OOl
cranc 2.5b * thurul * ool *turuml *ool *turunrl *ool
Table 5
origin in pre-pJ segrnental phonology. My carlier hypothesis that the init ials insotre words belonging to the low register classes go back to voiced segntents thusseelns to be the only concrete solution based on internal reconstruction, becauseinitial low register dcvelops either from init ial vowel lcngth or from init ial voiccdsegments. There are sirnply no other possible sources, and vowel length only takcsaccount of words belonging to subclass (a). I thereforc propose to rcconstruct p.lwords belonging to subclasses 2.3b-2.5b wi th in i t ia l vo iced onscts, inc luding
t 53
)nce of wordso significantptions to thewhere init ial
2002"204):
lst have beenShimabukuro)wels in theseShimabukuroTespondencest, for the sakce three rvords
ran raised by; dissertation,lruct separategth for accentShimabukuroone example369)), their
sial examplesse based on auction of twosro 2002'.369-
: vowel lengthgin of the lowiince there arewever, it doesrd others have
lructions that, forwhich underwent: also made somcent.
1 5 4 ALEXANDER VOVIN
voiced obstruents, nasal sonorants, and smooth vocalic ingress. All these wordshave short vowels in the first syllable. All words belonging to subclasscs 2.3a-2.5a are reconstructed with long vowels in the first syllable. Theoretically, theycould have voiced onscts as well, but we sirnply have no evidence lbr that. Itentatively reconstrucl voiceless onsets for words beginning with obstruents.vowcl-init ial words are reconstructed with init ial glottal stop *?-, which alsobehaves as a voiceless onset. Neveftheless, I reconstruct words with init ial nasalsonorants belonging to subclasses 2.3a-2.5a as having voiced onset and init ialvowel lcngth. Thc reason fbr this is cornbinatorial and wil l become clear when Idcal with reconstruction of onsets of high register words below.
As I n-rentioned above, there is a possible contradiction betwcen thisrcconstruction and a traditional reconstruction of pJ that posits pJ *b- and *d- onthe basis of the Sakishima b- and Yonaguni r/- evidence. The discussion of whythe Sakishima and Yonaguni evidence is not admissible falls outside the scopc ofthis chapter, but I bclieve that we should reconstruct just pJ *w- and *y-, and Iaddrcss this issue in greater detail in a forthcorning book (Vovin, fbrthcorning).Thus, the major problem (2), rncntioned above wil l be taken care o1, since thercwill not be two diflerent scries of init ial voiced obstruents reconstructed for pJ.
Thus, Table 5 above can bc re-written as in Table 6, elirninating low init ialrcgister for pJ altogether and replacing it with voiccd onsets for thc subclasses2.3b-2.5b and vowel length in first syllablc for thc subclasses 2.3a-2.5a.tE
As the major problerns ( l ) and (2) havc been solved above. i t is nowpossible to address problcm (3): that is, what kind of onsets should bereconstructcd for the words belongrng to high-register classes. Since init ial lowregister has been eliminated fiorn thc reconstruction, it does not make sense topreserve high register either. I think that high rcgister words should all bereconstructed as words with voiceless onsets and with shorl vowels in their f irstsyllables. Thus, the following reconstructions are suggested for the three diffbrenttypes of onsets (l use the syllables lal , l tal , and lnal as exarnples):
The reason for reconstructing nasal sonorant onsets belonging tosubclasses 2.3a-2.5a as voiced with vowel lcngth is purely combinatorial, as Imentioned above. Voiced nasals occur cross-linguistically rnuch more fiequerrtlyin the languages that have both voiced nasals and voiceless nasals (pre-glottalizednasals function as voiceless). Sincc I recorrstruct *?n- and *'?m- fbr words withnasal sonorant in i t ia ls belonging to t radi t ional accent c lasses 2.1-2.2,rcconstructing prcglottalized nasals for subclasses 2.3a-2.5a would produce morepreglottalized nasal sonorants than voiced nasal sonorants in the proto-language,which would be typological ly unexpected.r"
' " Notc that wc cannot cornplete ly c l iminatc non- in i t ra l p i tch: w,c have no segmental cxplanat ionlbr h igh pi tch on thc second mora of 2. . l . For 2.5. as i t rvas suggested above, thc la l l ing pi tch onthc secol td mora can bc expla incd as thc rcsul t of thc loss of f inal * -nt . but only onc word in thetehlc has a di rcct er idcrrce lb l i t ." I t is necessary to kecp in rn incl that a reconstruct ion ofpre-glot ta l ized nasals * ' ln- and *?m- mayactual ly be rewr i t tsn as a rcconstrucl ion ofvoicclcss nasals xhn- and *hnt- .
PROTO.JAI
Gloss
tor to ise 2.3a
ol l .shore 2.3a
.1a r 2 .3a
bcan 2.3a
ba l l 2 . 3a
roo t 2 .3a
f lowcr 2.3b
nrountain 2.- lb
c l oud 2 .3b
dog 2 .3b
wavc l . 3b
shank 2 .3b
brcath 2..1a
nrof iar 2.4a
chopst icks 2.4a
nccdlc 2. ' la
board 2.,{b
lootpr int 2.4b
shouldcr 2.4b
dregs 2.4b
t t r rnbcr 2. .1b
skin 2. : lb
bucke t 2 .5a
br idcgroonr 2.5a
voicc f . -5a
sp ide r 2 .5a
shaclou 2.5a
su cat 2.-5b
rain 2.-5b
spr ing 2.-sb
cranc l . -5b
T) 'pe of onse
vowcl zr-
0bstruent ld-
nasal na-
PROTO-JAPANESI ] BEYOND THI l ACCENT SYSTEM r 55
)rds.3a-heyrt. I)ntS.alsoasalit iale n l
Gloss pR PMJ pJ
tor lo isc 2.3a
of l lhorc 2.3a
. jar 2.-1a
bcan 2 .3a
ba l l 2 . 3a
root 2.- la
1 ' lou,er 2.Jb
mountain 2.3b
c loud 2 .3b
dog 2 .3b
r.var c l .Jb
shank 2 .3b
brcath 2.4a
rnortar 2.'{a
chopst icks 2.4a
necdlc 2. ' la
board 2..1b
fbotpr int 2.4b
shouldcr 2.4b
drcgs 2.4b
nunrbcr 2. '1b
skin 2.z lb
buckc t 2 .5a
br idegroonr 2.5a
vo i cc 2 .5a
sp idc r 2 .5a
shadou 2.5a
s\ \ 'cat 2.-5b
ra i n 2 .5b
sp r i ng 2 .5b
c ranc 2 .5b
*1hana * OOI* yama * OOI* khurnu * OOI* ' icnu * OOI* nami * OOI
* pana * OO *bana
* yarla * C)O *yanra
*_kunr* 'o x OO +gut.no
* inu * OO tcr tu
* nanr i + OO *naml
*kha:me) * ooC)\ * kanrey + Of) *ka:ruay
* ' / u : k i ) * ooO) * ok i * oo *o : k i
*kha:nrc) * ooO) + kanrc * OO *ka: tnc
*ma:rne' ) * ooO) + nramey * OC) * tna:may
*ma: i ) * ooC)) * nrar i x 0O *rna:r i
+nru: tu) * ooO) * moto * OO *t .nq: tq
this- o n
whye o fn d l, ' 6 t '
nere
. . 'l t la l
nowbe
low,e tol b efirst)rent
I t oA S I
rntlyl izedwith
1 . \
rorerage,
tatron
:ch on
in the
,- may
* suni * tJ() l * sunc * OO*'?i:ki) *ooO) *i [ t<i, i *o[o
*, i u:si l *ooo) *ulsLr *O[o
* ?iral * ool *i[ta *o[9* '?aruJ * ool *,,ftq *o[t)x khatal * ool *kafta *o[o
+pha:si) *ooO) *paf siy *O[o *pa:lrLry *,ro[{)
*pha:r i) xoool *palr iy xO[o *pa:[ruy *oof{)
*zunc
*' l i : [ki *oolc)
* ' lu:fsu *otfo
*i lra *o[O*altq *o[o*gafta *O[tl
* khasul * OO'l *kafsu *OfO *galsu *O[O
*_khazul * ool *kafnsu *o[o *galnsu *o[o
* phadal * OOI *palnta *o[O *balnta *o[O
*ri ,o:khc) *oool *u,okcyl*Oo l *wo:kay l *ooOT
smo:kho) *ooo) *nruku'ol *OO'l *nro:kol *ooC)l
*khu:c) *ooo) *kgr'n'cyl *oO1 *ko:wayl *oool
*kho:bo)*ooo) *kunnvo l * (X) l *ko :mIp ]o l *ooOl
*kha:gc) *ooo) *kagcy-l *ool *ka:nkay-l *oool
* '/asil * ool *asel *ool *atc l *ool* '?amcl l oo l *amey l *oo l *amav l *ool* pharu-l * ooT *paru'l *uol *barLr-l *ool* thurul * ool *t.,runrl *ool *durunr-l *ool
Table 6
Tvpc of onset 2.1-2.2 2.3a-2.5a 2.3b-2.5b
vorvcl a-
obstruent 1rr-
nasal la-
* '1a-
* ta-
* ' /na-
+ ' / a : -
E t i t : -
+ n a : -
* a -
*da
* n a
Table 7
r 56 ALEXANDER VOVIN
4. ConclusionIf rny reconstruction above is correct, it rnight have far-reaching
consequences for the study of the extcnral relationships of the Japanese languagefamily. A contrast betwecn srnooth vocalic ingress and an init ial glottal stop onone hand, and voiced and voiceless nasals on the other. has no parallels in the'Altaic' world. Instcad, these fcatures are widely found throughout South-EastAsia. lf the Japanese proto-language indccd arrived in Japan together with theYayoi culture, it should not be surprising to find some South-East Asian elunentsin Japanese, although at the present stagc ofour knowledge the exact naturc oftheconnection between Japancse and any of the major language farnil ies found inSouth-East Asia, if any, relnains at best hypothetical. The search for thesepossible conncctions is a legitimatc cnterprise, but it should bc carried out in acaLltlous manner.
RECONSTRI.