Sources of Old Japanese i2

J East Asian Linguist (2011) 20:219–228 DOI 10.1007/s10831-011-9075-2 On one more source of Old Japanese i2 Alexander Vovin Received: 4 April 2010 / Accepted: 8 March 2011 / Published online: 11 May 2011 Ó Springer Science+Business Media B.V. 2011 Abstract The goal of this article is to provide further and systematic evidence for an idea suggested only in passing that the Old Japanese vowel i2 derives diachronically not only from * y and *uy, as traditionally believed, but also from *oy. This solution proves to be a key to solving some puzzles in the phonological history of a number of common Old Japanese words. e Keywords Western Old Japanese  Eastern Old Japanese  Ryukyuan  Vocalism  Monophthongization 1 Introduction It is well known that among eight WOJ1 vowels: a, u, i1 ¼ [i], i2 ¼ [€ı], e1 ¼ [e], e2 ¼ [ y], o1 ¼ [o], o2 ¼ [ ],2 three vowels i2, e1, and e2 clearly have a diphthongal origin. The discussion of the origin of e1 and e2 falls outside the scope of the present article, as it deals exclusively with diachronic sources of i2. Traditionally it is believed e e This paper is dedicated to the memory of Samuel E. Martin. 1 The following abbreviations for languages are used in this article: EOJ—Eastern Old Japanese, MdJ—Modern Japanese, MJ—Middle Japanese, MK—Middle Korean, OJ—Old Japanese (both WOJ and EOJ), PA—Proto-Ainu, PJ—Proto-Japonic, PJN—Proto-Japanese, PR—Proto-Ryukyuan, WOJ—Western Old Japanese. 2 The sound values given here for WOJ vowels follow Miyake (2003). Note that there is no contradiction between PJ * y > WOJ i2 and WOJ e2 ¼ [ y], since PJ * y is a diachronic source for WOJ i2, and WOJ [ y] is a synchronic sound value for WOJ e2, which ultimately goes back to PJ *ay. e e e e A. Vovin (&) Department of East Asian Languages and Literature, University of Hawai’i at Mānoa, 382 Moore Hall, 1892 East-West Rd., Honoululu, HI 96822, USA e-mail: [email protected] 123 220 A. Vovin that i2 is derived from monophthongization3 of the diphthongs * y or *uy (Hattori 1978–1979), (Whitman 1985, pp. 41–42), (Martin 1987, p. 63), (Unger 1993, p. 26), and (Miyake 2003, pp. 80–81). The vowel i2 contrasts with i1 only in the position after labial consonants p, mb, m, and velar consonants k, Ng. After all coronals, the distinction between i1 and i2 is neutralized as i. Nevertheless, the particular original source, * y or *uy, can normally be deduced in most cases even after coronals due to morphophonological alternations that take place in OJ. The same is true for the cases when a form with i2 is not attested phonographically in OJ. These alternations mainly involve the alternations between free and bound forms for nouns and alternations between transitive and intransitive forms for verbs. Examples: e e * y Nouns Free form ki2 ‘tree’ (KK 47) e *yo2mi24‘Hades’ (M 9.1804) Verbs Intransitive oki2- ‘to rise’ (NK 83) opi2-‘to grow’ (KK 57) MJ5 ori- ‘to descend’ oti- ‘to fall’ (KK 82) pi2- ‘to get dry’ (M 5.798) *uy Nouns Free form kami2 ‘deity’ (KK 2) kuki2 ‘stalk, stem’ (KKHS 2) kuti ‘mouth’ (KKF 11) mi2 ‘body’ (KK 23) sati ‘bounty’ (KJK 1.52b) tuki2 ‘moon’ (KK 28) Bound form ko2-no2 pa ‘tree leaves’ (KK 20), ko2-ndat-i ‘grove’ (NK 105) yo2mo2-tu siko2 me1 ‘stupid woman of Hades’ (KJK 1.9b) Transitive oko2s- ‘to raise’ (M 19.4164) opo2s- ‘to cultivate’ (M 18.4113) oro2s- ‘to lower’ (M 15.3603) oto2s- ‘to drop’ (NR 2.4) pos- ‘to make dry’ (M 15.3712) Bound form kamu-kara ‘divine nature’ (M 17.4001) kuku-tat-i ‘group of stalks’ (M 14.3406)6 kutu-mbami1 ‘bit, bridle’ (NR 2.preface) mu-kuro2 ‘body’ (NSK 1.10) satu-yumi1 ‘hunting bow’ (M 5.804) tuku-yo1 ‘moonlit night, moonlight’ (M 20.4489) 3 One of the anonymous reviewers suggested that ‘monophthongization’ should be replaced by ‘contraction’. However, the important distinction between monophthongization and contraction was established by Russell (2003), and consequently accepted elsewhere (Vovin 2005, pp. 45–50). 4 WOJ *yo2mi2 is not attested phonographically in texts. yo2mi1 with i1 in man’yōgana script appears only in later Norito, and this is too late for the preservation of mi1 versus. mi2 contrast. The morphophonological alternations clearly point to *yo2mi2 with i2. 5 Ori- is not attested phonographically in OJ. 6 Although M 14.3406 is technically an Eastern poem (Azuma-uta), it has no typical Eastern features, so I treat it as a WOJ text. 123 Otso-rui i2 221 tuki2 ‘zelkova’ (KK 100) tuki2 ‘tribute’ (M 20.4360) tuku-yumi1 ‘zelkova-wood bow’ (NK 28) tuku-nop- ‘to compensate, to offer as a tribute’ (NR 2.25) To the same group we can also add the word tuti ‘hammer, mallet’ (NK 29), because the uncontracted form tutui ‘mallet’ appears four times in KK 10 and twice in NK 9, although the bound form *tutu- is not attested in OJ. This archaism is interesting, but the full discussion of it falls outside the scope of this paper. Verbs Intransitive ami2- ‘to bathe oneself’ (NR 1.8) puri- ‘to get old’ (M 17.3919) tuki2- ‘be exhausted’ (M 20.4458) Transitive amus- ‘to bathe someone’ (M 16.3824) purus- ‘to make old’ (M 7.1326)7 tukus- ‘to exhaust’ (M 18.4094) There is also a small group of words with i2 for which no alternations are attested. If any of these words has i2 after /k/, it is possible to trace the original proto-Japonic form depending on whether a cognate in Shuri has any palatalization. The Shuri palatalization k > č indicates original *uy, while the lack of the palatalization points to * y, cf. Shuri ciči ‘moon’ (OGJ 1998, p. 144) < PJ *tukuy, but Shuri kii ‘tree’ (OGJ 1998, p. 320) < PJ *k y. Thus, both ki2si ‘bank, stone edge’ (M 10.1818) and ki2ri ‘fog’ (M 15.3615) can be reconstructed as PJ *kuysi and PJ *kuyri, respectively, on the basis of Shuri evidence: Shuri čisi ‘bank’ (OGJ 1998, p. 164) and Shuri čiri ‘fog’ (OGJ 1998, p. 162). In addition, there are two OJ words, pasi ‘chopsticks’ (KJGC) and mi28 ‘winnow’ (HFK, Kamo-gun), (M 4.509) that should be reconstructed as PJ *pasuy and *muy on the basis of pre-OJ loans into Ainu: Ainu pasúy ‘chopsticks’ and múy ‘winnow’. e e 2 Another source for OJ i2 Martin (1987, p. 63) claimed that OJ i2 never alternates with OJ o1, but as the reader will see below this statement turns out to be incorrect. I will argue that, in fact, OJ i2 has one more source besides PJ * y and *uy, namely PJ *oy. Strictly speaking, this is not the first time that this idea is presented, but my predecessors always limited themselves either to one example without any discussion or to a mention in passing at best. Whitman (1985, p. 44) was the first to mention that MJ kuri ‘black mud in the water, black color’ (see #6 below) has to be derived from kuro1 ‘black’ + -i. The same example was apparently rediscovered independently by Majtczak (2008, e 7 One of the anonymous reviewers pointed out that WOJ PURUs-u ‘to make someone old’ is spelled mostly semantographically in M 7.1326 as . I have no access at this point to book seven of the Man’yōshū in the oldest manuscript that preserves this poem: Genryaku kōhon Man’yōshū, but in the is clearly glossed as purusu second oldest, Ruijū koshū, dating from the end of the Heian period, (RJKS 8: 21). 8 mi2 ‘winnow’ in OJ texts occurs only in semantographic writing, but we know that the word must have been phonetically [mi2], because the kun-yomi of the character ‘winnow’ was used for writing the syllable /mi2/, e.g. URAmi2 ‘circumference of a bay’ is written as (M 4.509). 123 222 A. Vovin p. 26), as Whitman (1985) is not listed in his bibliography. Frellesvig and Whitman (2008, p. 39) mention in passing the development PJ *oy > WOJ i2, again with just one example: iso1 ‘rock’~isi ‘stone’ (see #7 below). Finally, Serafim (2008, pp. 88– 89) also provides one more example, suNgo1s-~suNgi2- (see #4 below), mentioning again in passing the development PJ *oy > WOJ i2. I provide here, with extensive discussion, all the examples where, as I believe, one can safely reconstruct PJ *oy as a source of OJ i2. (1) Let me start with a case which also has clear philological support. Martin (1987, p. 405) believes that OJ pi2 ‘fire’ goes back to PJ *p -Ci. This belief is probably based on the form fo2-no2-fo1 ‘flame’, cited in Martin (1987, p. 64). Why Martin attributed a kō-otsu distinction to the MJ form fo-no fo ‘flame’ (lit. ‘top of fire’ or ‘grain-ear of fire’) that is attested phonographically9 for the first time only from the Heian period (KKS 1102), is unclear to me. It is even less clear to me what the basis is for assigning the o2 to the bound form po2- ‘fire’ in po2-no po1 ‘flame’. Quite to the contrary, Kojiki, the only text that differentiates between po1 and po2, has clearly po1-naka ‘inside the fire’, where po1 is spelled with (KJK 2.42a),10 the Kojiki man’yōgana sign for /po1/ (Mabuchi 1972, p. 63).11 Moreover, EOJ has pu ‘fire’, a hapax legomenon attested only in one poem from Munzasi (MdJ Musashi) province (M 20.4419). It is well known that OJ o2 and u do not alternate; what is more, the correspondence of WOJ o2 to Musashi EOJ /u/ is not otherwise attested. Thus, the development of *p y or *p to EOJ pu is irregular and puzzling. Meanwhile, Munzashi EOJ u can be easily understood as the result of the raising of PJ *o>u. With the overall EOJ tendency to have bound stems to be used for both WOJ free and bound stems as a result of a diphthong simplification process, e.g. WOJ tuki2~tuku- ‘moon’, but EOJ tuku ‘moon’ (M 14.3476), the correspondence of EOJ pu ‘fire’ to WOJ pi2 is expected.12 Consequently, we can surmise that WOJ pi2 and EOJ pu ‘fire’ underwent the following developments from PJ: e e e WOJ line of development: PJ *poy > WOJ pi2 (monophthongization of the diphthong *oy) EOJ line of development: PJ *poy > *puy > EOJ pu (raising of the diphthong *oy’s nucleus with subsequent deletion of -y). 9 PO-NO PO ‘flame’ is attested semantographically as in M 13.3344, therefore there is no indication that PO ‘fire’ should be read as /po1/ or /po2/. In addition, the contrast between /po1/ and /po2/ was already lost by the time of the Man’yōshū. 10 This spelling plus the change of o1+i>i2 was implicitly noticed by Shirafuji (1987, p. 98), but curiously enough he used it for the argument in favor of the loss of contrast between po1 and po2 in the Kojiki. 11 While Mabuchi (1957) theory that claimed that the man’yōgana script in the Kojiki differentiated po1 and po2 syllables found little support in Japan, it was accepted in the West (Martin 1987); (Whitman 1985); (Bentley 1997); (Miyake 2003); (Vovin 2005). 12 Other examples besides PJ *tukuy > EOJ tuku ‘moon’ exemplifying the development of PJ *uy or *oy > EOJ u: EOJ ko1pusi (M 14.3476) and kupusi (M 20.4345) ‘to be longed for’ < PJ *kopoysi, EOJ kandu ‘mulberry tree’ (M 14.3432) (cf. WOJ kandi) < PJ *kantuy or *kantoy, EOJ paru ‘needle’ (M 20.4420) (cf. WOJ pari) < PJ *paruy or *paroy, WOJ usiro1 ‘back’, siri back’, EOJ siru ‘back’ (M 20.4385) < PJ *siroy. 123 Otso-rui i2 223 (2) OJ ko1pi2- ‘to long for’ (M 5.819) has five different derived adjectival forms: WOJ ko1pi2si (M 17.3987) and ko1posi (M 5.875) ‘be longed for’ and EOJ ko1pi1si (M 20.4407), ko1pusi (M 14.3476), and kupusi (M 20.4345) ‘id.’ Martin (1987: 63) derived WOJ ko1posi from ko1po2-si(-), but this is impossible, since o1 and o2 can never occur together within the same morpheme (Ōno 1977, pp. 203–206). The raising of o1 to u is well attested in EOJ and in the history of Japanese in general (Hattori 1978–1979); (Hayata 1998); (Frellesvig and Whitman 2008), etc.; for EOJ cf. (1) above and also WOJ kumo1 ‘cloud’ (KK 1), but EOJ kumu id.’ (M 20.4403). But the reverse process, namely the lowering of u to o1 cannot be supported by any evidence (see also (4), below). Thus, I can rule out PJ *kopuy- as well. Therefore, the only possible solution is to reconstruct PJ *kopoy- > OJ ko1pi2- ‘to long for’. (3) Unfortunately, WOJ *ki2 ‘yellow’ is not attested phonographically in the texts, but we can safely conclude that it was *ki2, and not *ki1 on the basis of the bound form ku- found in WOJ ku-Ngane ‘gold’ (lit. ‘yellow metal’) (M 18.4094). Shuri či-iru ‘yellow’ (OGJ 1998, p. 149) with palatalization k- > č- indicates PJ *kuy, too. However, MJ ko-Ngane ‘gold’ (TM 2913) with ko- rather than ku- creates a puzzle if one goes along with the traditional point of view that recognized that OJ i2 can be derived only from PJ *uy or * y: MJ ko- points to PJ *k y, but *k- before * y does not palatalize in Shuri. In addition, u and o2 do not alternate. Therefore, the simplest solution is to propose that WOJ *ki2 ‘yellow’ < PJ *koy: this will account for the correspondence of WOJ u in ku-14 to MJ o in ko-, and for the palatalization in Shuri či-iru. We will see the same palatalization before *oy again in (4), below. (4) OJ suNgi2- ‘to pass (intr.)’ (M 5.816) has two different transitive counterparts: WOJ suNgus- ‘to pass (tr.)’ (M 5.804) and EOJ suNgo1s- ‘id.’ (M 14.3564). Martin believes that suNgus- > suNgo1s- and posits here the lowering of o1 to u (1987, p. 63). Although his assumption follows the Japanese tradition (Saeki and Mabuchi 1969, pp. 474, 476), it seems that there is no actual evidence supporting the lowering of o1 to u.15 Although we have seen the raising of o1 to u in EOJ in (2) above, we also know that EOJ sporadically preserves PJ *o (Thorpe 1983, p. 235). There are doublets in MJ: suNgus- (MS 86)16 and suNgos- (TM 31). The existence of lexical doublets always indicates that one form is inherited, and another is borrowed.17 Since we already saw e e e 13 Numbers after TM indicate the page number according to the NKBT edition of this text. 14 WOJ ku- is a result of the raising of PJ *o1 to WOJ u; see also (4) below. 15 This belief in Japanese tradition probably rests on the ad hoc assumption that all phenomena found in WOJ are archaisms, just because WOJ is the oldest attested standard form of Japanese. Thus, any later attestations, or even contemporary attestations in dialects are automatically viewed as deviations, and WOJ is essentially treated ‘‘as though it stands in the same relationship to all other known varieties of Japonic as Latin to all Romance languages’’ (Vovin 2010, p. 4). 16 Numbers after MS indicate the number of the dan according to the NKBT edition of this text. 17 An anonymous reviewer raised the issue that lexical doublets cannot be always explained as loans, and as an example s/he provided Mandarin de and zhi that are in his/her opinion both native to Mandarin. This is an excellent example that perfectly illustrates my point: Mandarin zhi is, of course, borrowed from Classical Chinese and is not ‘native’ to Mandarin. Borrowings are certainly not limited to borrowings from other language families: they can come from an earlier stage of the same language (as MdJ ware-ware ‘we’ is borrowed from bungo), from a different dialect (like MdJ dekkai ‘big’), or from a closely related language (like chanpuru from Okinawan). The same holds true for MJ suNgos- and suNgus(the latter borrowed from OJ suNgus-). 123 224 A. Vovin in (3) above that MJ is more conservative than WOJ as far as the preservation of o1 is concerned, it does not seem unreasonable to suggest that the form suNgos- is native to MJ, while the form suNgus- is borrowed from WOJ, even though the latter is more frequent than the former. Thus, the original nature of EOJ suNgo1s- is supported both by directionality of change and additional evidence from MJ. Consequently, OJ suNgi2- ‘to pass (intr.)’ can go back only to PJ *suNgoy-, but not *suNguy-. It is worth noting that the Shuri form is siji- ‘to pass (intr.)’ (OGJ 1998, p. 487), demonstrating the palatalization g > j. Note that in (3) above we also had the palatalization k > č. Compare also Shuri ciči ‘moon’ (OGJ 1998, p. 144) < PJ *tukuy, but Shuri kii ‘tree’ (OGJ 1998, p. 320) < PJ *k y. The examples in (3) and (4), as well as these two examples, show that both PJ velars *k and *Ng have the same development in Shuri involving palatalization before *uy or *oy in contrast to the position before PJ * y that does not palatalize. Thus the developments of PJ *kuy, *koy, and *k y in Shuri probably underwent the following paths: e e e PJ *kuy > PR *kwi > pre-Shuri *ki > Shuri či PJ *koy > PR *kwi > pre-Shuri *ki > Shuri či PJ * k y > PR *k€ı > pre-Shuri *k€ı > Shuri ki e (5) WOJ naNgi2- ‘to get calm’ (M 4.753) and corresponding adjective forms: OJ na go1-ya ‘calm’ (M 4.524, M 14.3499), and especially WOJ naNgo1 ‘id.’ (M 6.963) that is not derived, but represents the root and has -o1 in the final position, demonstrate that WOJ naNgi2- < PJ *naNgoy-, contrary to Martin’s claim that naNgo1(-ya-ka) is a secondary formation from *naNgu, as attested in naNgusam- ‘to become calm’ (1987, p. 63). Since, normally, PJ *o is preserved in WOJ as o1 only in the last syllable of a disyllabic nominal stem (Hayata 1998), the raising of *o1>u in naNgusam- is expected. (6) MJ kuri ‘black mud in the water, black color’ (SJ 10.32a), which is not attested phonographically in OJ texts, must be derived from WOJ kuro1 ‘black’ (KK 4). As I mentioned before, this was already noticed by Whitman (1985, p. 44) and Majtczak (2008, p. 26). The latter specifically notes that this derivation involves o1, and not o2. But if kuri is derived from kuro1, then it surely must go back to PJ *kuroy. (7) OJ isi 2.2 ‘stone’ (KK 10, M 14.3398) is likely to be etymologically connected with OJ iso1 2.1 ‘rock, rocky shore’ (KK 5, M 14.3563). The latter also has other specifically EOJ forms: osi (M 14.3359) and osu (M 14.3385). The relationship between OJ isi ‘stone’ and OJ iso1 ‘rock, rocky shore’ is further strengthened by the fact that the name of the shrine Iso1-no2 kami1 ‘Top of the rock’ is consistently spelled semantographically with the characters ‘stone top’ (Omodaka etal. 1967, p. 76). There is also a secondary version of the same name Isu-no2 kami1 (Omodaka et al. 1967, p. 76) that has already undergone raising o1>u. Consequently, if OJ isi ‘stone’ is a derivative of OJ iso1 ‘rock’, its reconstruction must be PJ *esoy.18 The -y in *esoy probably can be analyzed as a diminutive; thus, N 18 On the reconstruction of initial *e- rather than *i- for both PJ *esoy ‘stone’ and *eso1 ‘rock’ primarily based on EOJ data, see Vovin (2010, pp. 126–127). 123 Otso-rui i2 225 PJ *eso-y ‘stone’ is historically ‘little rock’.19 The words for ‘stone’ and ‘rock’ were traditionally considered to have a complex phonological history, particularly due to the correspondence of -i to -o1. With the introduction of *oy as another source of OJ i2, I hope that this complex history is essentially explained. (8) It is traditionally believed that the WOJ form corresponding to MdJ usiro ‘back’ was usiro2 with otsu-rui vowel o2 (Omodaka et al. 1967, p. 115). However, the basis for such a belief is actually rather meager. WOJ usiro2 ‘back’ occurs once in the Kojiki (KK 42), twice in two very late imperial edicts (SM 51 and SM 58) and once in a late Heian period man’yōgana gloss in Konkomyō saishō-ō kyō ongi (1079 AD). On the other hand, WOJ usiro1 ‘back’ appears once in the Nihonshoki (NK 120). As far as the phonetic accuracy of transcription is concerned, the Nihonshoki should have the precedence over the Kojiki, as it has a much better history of transmission. Apparently related to this word are WOJ siri ‘back’ (KK 22, 45, 46), (NK 37, 38), (M 18.4108), EOJ siri ‘id.’ (M 14.3431), and EOJ siru ‘id.’ (M 20.4385). Since EOJ u does not alternate with o2, the only possible source for WOJ and EOJ siri is PJ *siroy, which is further confirmed by WOJ usiro1. Thus, in this case, philology and historical phonology are remarkably concordant. I believe that in the eight cases listed above, PJ *oy can be reconstructed uncontroversially as the source of OJ i2. In addition, there are three other groups of examples : (A) the evidence for *oy is extremely compelling, but not altogether conclusive, (B), the evidence remains inconclusive: both PJ *oy and *uy are possible, and (C) there is no evidence for *oy, although superficially it may initially look like a possibility. (A) Compelling evidence for *oy: (9) WOJ niko1- ‘gentle, soft’ (M 20.4309) and EOJ niko1- ‘id.’ (M 14.3370), as well as the WOJ derivatives niko1-yo2-ka n-i ‘gently, softly’ (M 11.2762, M 20.4309) and niko1-ya-ka n-i ‘id.’ (NR II.27) are apparently somehow connected to WOJ niki1 ‘soft, gentle’ (NSK IX: 246.3). However, in this case we have an unexpected form niki1 with i1 instead of the expected *niki2 with *i2. Since the alternation of i1 with o1 is virtually unknown except in this case and in example (11) below, this turns out to be a puzzle. The case is reminiscent, though, of WOJ -ki1 ‘sake’ in WOJ mi1-ki1 ‘honorable sake’ (KK 39), (M 19.4262). As Murayama (1988, pp. 251–253) demonstrated, the vowel i1 in mi1-ki1 is a result of progressive assimilation under the influence of the vowel in the prefix mi1-, because the WOJ word is otherwise attested as -ki2 in KURO1 ki2 ‘black sake’ (SM 38) and SIRO1 ki2 ‘white sake’ (SM 39). Thus, the evidence for reconstructing WOJ niko1 -~niki1 ‘soft, gentle’ as PJ *nikoy is compelling, but, unfortunately not completely conclusive as, in contrast to WOJ mi1-ki1 ‘honorable sake’, we have no way to tell whether ni- goes back to pre-OJ *ni1 or *ni2. (10) There is a famous puzzling correspondence of WOJ ninzi ‘rainbow’ (NR I: 5) to EOJ no1nzi ‘id.’ from Kami1tuke2no1 province (M 14.3414). Both WOJ and EOJ words are hapax legomenoi, but there is no lack of attestations for MJ ninzi ‘id.’ and later corresponding forms. The most sensible solution that seems to 19 Note, however, that *-y in PJ *k y ‘tree’, *muy ‘body’, *tukuy ‘moon’ as well as in most other words ending in -*y in all probability belongs to the stem, as it cannot be explained away as a suffix. e 123 226 A. Vovin reconcile these two forms is PJN *noynsi, although, unfortunately, as in the case above, we have no way to tell whether WOJ ni- in ninzi goes back to pre-OJ *ni1 or *ni2. The case for original *oy, however, might be further strengthened by Ryukyuan data: Yuwan n€ı:dziN, Kumejima nidzi (Uchima and Arakaki 2000, p. 431), indicating PR *e (that can correspond to either PJ *e or to PJ sources of OJ i2), although many other Ryukyuan dialects show contradictory evidence, essentially supporting the EOJ form: Izena, Maejima no:dzi, Hateruma no:dz€ıN, etc. (Uchima and Arakaki 2000, p. 431). In light of these forms agreeing with EOJ in the vocalism of the first syllable, one might want to claim that WOJ ninzi underwent the regressive assimilation, but caution must be taken as they all exhibit long vowels that are secondary in the dialects cited above and might well have a diphthongal origin. (B) Inconclusive evidence for either *oy or *uy: (11) A very similar case to the one described in (9) above is presented by the alternation of WOJ nipi1 ‘new’ (KK 25, 100, 101), (M 17.4000) and EOJ nipi1 ‘id.’ (M 14.3350, 14.3506, 14.3537a) with EOJ nipu- ‘id.’ (M 14.3460). It is not possible to tell whether the PJN form was *nipuy or *nipoy, although we have evidence from Ryukyuan for the development of pre-OJ *ni1- > ni- in this case, since Ryukyuan languages show mii-: Shuri mii- (OGJ 1998, p. 371), Nase mii-, Ishigaki m€ı€ı-, etc. (Hirayama 1966, p. 425).20 (12) With WOJ kandi ‘mulberry tree’ (KMJ 24) and EOJ kandu ‘id.’ (M 14.3432), the evidence for either PJ *kantuy or *kantoy is inconclusive. (13) With WOJ pari ‘needle’ (M 18.4128) and EOJ paru ‘id.’ (M 20.4420), the evidence for either PJ *paruy or *paroy is inconclusive. (C) No evidence for *oy: (14) One of the anonymous reviewers brought up the personal communication with Leon Serafim, where the latter suggested that OJ kami2 ‘deity’ may be from *kamoy, and the name Kamo may be from *kamo1. The anonymous reviewer further pointed out that Ainu has kamúy and not *kamoy, and posited two questions: whether it is possible that Ainu kamúy was borrowed after raising and whether there are any traces of PJ *e and *o in Japonic loanwords in Ainu. I believe that there are several practically insurmountable difficulties to derive OJ kami2 from PJ *kamoy, as well as to identify it with the placename Kamo. First, there is indeed Ainu evidence for PJ *e and *o in early loanwords from Japonic (better for the latter than for the former) (Vovin 2010, pp. 34–35). Second, it is not really clear whether Ainu kamúy is a loanword from Japonic, or PJ *kamuy (> OJ kami2) is a loanword from Ainu. In any case, there is no evidence for PA *kamoy ‘deity’, and the word is clearly reconstructed as *kamúy (Vovin 1993, p. 99). Third, there is no evidence that the placename Kamo (whether a name of the shrine or the river) indeed goes back to kamo1: the earliest man’yōgana spellings that I am familiar with are kamo and gamo that include the character mo that does not indicate either mo1 or mo2. In addition, both spellings are attested only in a fragment of the Yamashiro Fudoki that is found itself in the late Kamakura text Shaku Nihongi, which makes the point of derivation even more moot. Fourth, the river name must precede the 20 As far as the initial consonant is concerned, Ryukyuan is more archaic than Old Japanese, since only *mi- can palatalize to *ni-, but not vice versa. 123 Otso-rui i2 227 name of the shrine, and not vice versa, as river names as a rule precede the names of settlements or edifices that are erected on their banks. Fifth, even if we opt for reading Kamo1, the reality might be much more prosaic: Kamo1-Ngapa might mean simply ‘duck river’, and not ‘divine river’. After all, there are plenty of ducks swimming in the Kamo river nowadays, and there is no reason to believe that they were not engaging in the same kind of activity 12 or 13 hundred years ago. 3 Conclusion As a conclusion, I would like to say that the establishment of the fact that OJ i2 has yet another source *oy, in addition to the traditionally recognized * y and *uy, not only allows us to solve some puzzles in PJ reconstruction and to achieve more precision in this reconstruction, but also allows us to better evaluate certain external etymological proposals for Japanese. Thus, for example, the reconstruction of the Japonic word for ‘fire’ as *poy rather than *p y completely rules out a comparison with MK púl [p€ır] ‘fire’, because PJ *o does not correspond to MK u [€ı].21 In a similar vein, a reconstruction of the Japonic word for ‘stone’ as *eso-y ‘little rock’ holds little hope for the comparison with MK tw olh ‘stone’, or other ‘Altaic’ words for ‘stone’, such as Old Turkic taš, Middle Mongolian čila’un, and Ewenki jolo. Similarly, a comparison of PJ *koy ‘yellow’ with Proto-Austronesian *kuniN ‘id.’ turns out to be an etymology based exclusively on the initial phoneme. e e Acknowledgments I wish to express my gratitude to Dr. Sven Osterkamp (Bochum University/Kyoto University) and to Prof. Tomasz Majtczak (Jagiellonian University/Bochum University), with whom the underlying idea for this article was discussed during my stay in Bochum University 2008-2009, and to two anonymous reviewers whose comments helped to make this article much better than it was in its original version. References Primary sources HFK KJK KJGC KK KKF KKHS KKWS KMJ M MS NK NR 21 Harima Fudoki, ca. 713 AD Kojiki, 712 AD Kōtai Jingū Gisiki chō, 804 AD Kojiki kayō, 712 AD Kinkafu, eighth century AD Kakyō hyōsiki, 775 AD Kokin waka shū, 921 AD Shōsōin komonjo, from 702 AD Man’yōshū, ca. 759 AD Makura-no sō shi, ca. 1000 AD Nihonshoki kayō, 720 AD Nihon ryō iki, 824 AD Being driven by the mistaken assumption that WOJ pi2 < PJ *p y, and not *poy, and, in spite of the fact that I am very critical of the suggested Japonic-Korean genetic relationship, I have recently claimed that this comparison can be acceptable as a potential cognate (Vovin 2010, p. 107), but the reconstruction of a different vowel in the PJ form completely rules it out. e 123 228 NSK RJKS SJ SM TM A. Vovin Nihonshoki, 720 AD Ruijū koshū, mid to late twelfth century AD Shinsen jikyō, ca. 898–901 AD Senmyō, seventh–eighth centuries AD Taketori monogatari, late nineth or early tenth century Secondary sources Bentley, John R. 1997. MO and PO in Old Japanese. MA thesis, University of Hawai’i at Mānoa, Honolulu. Frellesvig, Bjarke, and John Whitman. 2008. Evidence for seven vowels in proto-Japanese. In ProtoJapanese: Issues and prospects, ed. B. Frellesvig and J. Whitman, 15–41. Amsterdam/Philadelphia: John Benjamins. Hattori, Shirō. 1978–1979. Nihon sogo ni tsuite [On Proto-Japonic]. Gekkan Gengo 1–22. Hayata, Teruhiro. 1998. Jōdai Nihongo no onsetsu kōzō to o-retsu kō-otsu no betsu [The syllabic structure of Old Japanese and kō-otsu distinctions in the o row]. Onsei kenkyū 2(1): 25–33. Hirayama, Teruo, ed. 1966. Ryūkyū hōgen no sōgōteki kenkyū [A comprehensive study of the Ryukyuan dialects]. Tokyo: Meiji shoin. Mabuchi, Kazuo. 1957. Kojiki no SI, O, PO no kana [The Kana signs SI, O, PO in Kojiki]. Kokugogaku 31: 61–90. Mabuchi, Kazuo. 1972. Jōdai no kotoba [Old Japanese]. Tokyo: Shibundō. Majtczak, Tomasz. 2008. Japońskie klasy czasownikowe w perspektywe diachronicznej [Japanese verbal classes in a diachronic perspective]. KrakóW: WydaWnictwo Universytetu Jagiellońskiego. Martin, Samuel E. 1987. The Japanese Language through time. New Haven and London: Yale University Press. Miyake, Marc H. 2003. Old Japanese: A phonetic reconstruction. London: RoutledgeCurzon. Murayama, Shichirō . 1988. Nihongo no kigen to gogen [The origins and etymology of the Japanese language]. Tokyo: San’ichi shobō. OGJ. 1998. Okinawa go jiten [A dictionary of the Okinawan language]. Tokyo: ōkurashō. Omodaka, Hisataka et al., eds. 1967. Jidai betsu kokugo dai jiten. Jōdai hen [A big dictionary of Japanese by periods. Ancient period]. Tokyo: Sanseidō. Ōno, Susumu. 1977. On’in no hensen [Phonological changes]. In Iwanami Kōza Nihongo [Iwanami lectures on the Japanese Language], ed. S. Ōno and T. Shibata, Vol. 5, 148–219. Tokyo: Iwanami. Russell, Kerri. 2003. Contraction and Monophthongization in Old Japanese. In Nihongo keitōron no genzai [Perspectives on the Origins of the Japanese Language], ed. A. Vovin and T. Osada, 511–539. Kyoto: International Center for Japanese Studies. Saeki, Umetomo, and Kazuo Mabuchi. 1969. Kogo jiten [A dictionary of premodern Japanese]. Tokyo: Kōdansha. Serafim, Leon A. 2008. The uses of Ryukyuan in understanding Japanese language history. In ProtoJapanese: Issues and prospects, ed. B. Frellesvig and J. Whitman, 79–99. Amsterdam/Philadelphia: John Benjamins. Shirafuji, Noriyuki. 1987. Nara jidai no kokugo [The Japanese language in the Nara period]. Tokyo: Tōkyōdō. Thorpe, Maner L. 1983. Ryūkyūan language history. PhD diss., University of Southern California, San Diego. Uchima, Chokujin, and Arakaki, Kumiko. 2000. Okinawa hokubu nanbu hōgen no kijutsuteki kenkyū [A Descriptive Study of Northern and Southern Okinawan dialects]. Tokyo: Kazama shobō. Unger, J. Marshall. 1993[1977]. Studies in Early Japanese Morphophonemics. Bloomington: Indiana University Linguistics Club. Vovin, Alexander. 1993. A reconstruction of Proto-Ainu. Leiden & Köln: Brill. Vovin, Alexander. 2005. A descriptive and comparative grammar of Western Old Japanese. Part 1: Phonology, script, lexicon, and nominals. Folkestone: Global Oriental. Vovin, Alexander. 2010. Koreo-Japonica. A re-evaluation of a common genetic origin. Honolulu: University of Hawai‘i Press. Whitman, John B. 1985. The phonological basis for the comparison of Japanese and Korean. PhD diss., Harvard University, Cambridge. 123