Miller 1967

Linguistic Society of America
Old Japanese Phonology and the Korean-Japanese Relationship

Author(s): Roy Andrew Miller
Source: Language, Vol. 43, No. 1 (Mar., 1967), pp. 278-302
Published by: Linguistic Society of America
Stable URL: http://www.jstor.org/stable/411398
Accessed: 17-11-2015 12:19 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact [email protected].
Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access to Language.
http://www.jstor.org
This content downloaded from 138.73.1.36 on Tue, 17 Nov 2015 12:19:14 UTC
All use subject to JSTOR Terms and Conditions
OLD JAPANESE PHONOLOGY AND THE KOREAN-
JAPANESE RELATIONSHIP
ROY ANDREW MILLER

Yale University
Samuel E. Martin has recently, for the first time, unequivocally demonstrated
the Korean-Japanese relationship, but his sound correspondences are in the main
based on comparisons between Middle Korean and Modern Japanese cognates.
The Japanese forms which enter into his reconstructions are phonologically
'antiqued' on the basis of the traditional (and largely historical) orthography, but
they do not take full cognizance of our present knowledge of the phonology of
Old Japanese. The problem is particularly critical because of the nature of the
Old Japanese vocalism, which, in contrast to the modern five-vowel system of the
standard language of Tokyo, can be shown to have been an eight-vowel system.
The present paper collects the available Old Japanese evidence for those Japanese
forms which enter into Martin's comparisons, with the particular aim of seeing
what light Old Japanese phonology may cast upon his reconstructions for the
proto-Korean-Japanese unity.
Sometime in the second half of the 5th century, a Chinese or Korean scribe
composed an inscription for a sword which centuries later was to be found to-
gether with other artifacts in the Funayama Tumulus, a proto-historic burial-
cult site in modem Kumamoto Prefecture, on the main island of Japan. The in-
scription is entirely in Chinese except for one word, probably the name of the
sword-smith, recorded in Chinese characters used as phonograms; this appears
to read i-ta-[o? ka?],' and is perhaps to be understood as some derivative from the
Japanese verb i-ru 'to cast metal'. At about the same time another scribe was
composing an inscription for a metal mirror that eventually found its way into a
proto-historic tumulus in modern Wakayama Prefecture; in this text, which on
archeological grounds has been dated in the period 443-503, several Japanese
words recorded in Chinese characters used as phonograms have also been identi-
fied (Kayamoto, no. 9; Mizuno and Kobayashi, 520-1). A Chinese historical
text known as the Wei chih (Records of Wei), by Ch'en Shou (233-297), which is a
part of the San-kuo-chih (History of the Three Kingdoms), contains in its Chapter
30 an account of Japan as it was known to the Wei dynasty which ruled in China
during the period 221-265;2 this contains many transcriptions of Japanese words
in Chinese phonograms, of which a few may be identified with surviving Japanese
forms with some degree of confidence. Apart from this text, however, which is as
rich in philological and historical difficulties as it is in linguistic information,3
1 Inscription no. 8 in
Kayamoto (1955:255-62). The reading i-ta-ka is suggested in Fu-
kuyama 1934, according to Mizuno and Kobayashi (1959:869), which also see for additional
literature on this inscription.
I Translated with a commentary in Tsunoda and Goodrich
1951, where however the
readings for the OJ forms in this text and the etymologies proposed for them simply repro-
duce the views of traditional Japanese scholarship at the turn of the century, and are al-
most totally without scientific value.
a The philological and historical problems as well as some of the linguistic conundrums
posed by this text are treated in a definitive fashion in Hashimoto 1956; see especially his
278
OLD JAPANESE PHONOLOGY AND KOREAN-JAPANESE RELATIONSHIP 279
the two inscriptions on metal are the earliest textual evidence we have for the
language that in this paper is called Old Japanese (OJ). Its records were soon in-
creased by a sizable corpus of slightly later inscriptions on metal and stone, and
finally by the appearance of important historical and literary texts, notably the
Kojiki (Records of ancient matters, 712), the Nihon shoki or Nihongi (Records
of Japan, 720), the poetic anthology Man'yoshu (Collection of a myriad leaves,
mostly put together in the period 645-760), the extremely important inscription
known today as the Bussokuseki no uta (Songs on the stone [inscription preserving
the traces of the veritable] footprints of the Buddha, ca. 753), still in situ at the
Yakushi-ji in Nara,4 and certain portions of miscellaneous liturgical (norito),
documentary (senmyo), and gazetteer (fudoki) texts. Of these, only the Busso-
kuseki inscription is totally in Japanese and written entirely in phonograms. The
Kojiki has portions of phonetically written Japanese interspersed with passages
of Chinese; the Nihon shoki is by and large in a simple variety of literary Chinese
but preserves a rich corpus of phonetically transcribed OJ words and names. The
Man'yoshi is entirely in OJ, but since its text often employs a variety of compli-
cated systems of writing involving the use of the Chinese script over and above
its employment as phonograms, it is not always possible to cite its forms as evi-
dence for OJ. The liturgical, documentary, and gazetteer texts are by and large
in Chinese, but preserve occasional OJ forms in phonogram transcriptions. The
cut-off point for this OJ corpus is generally taken to be the shift of the national
capital from Nara to Kyoto in 794.5
Study of these written records in terms of our knowledge of Chinese historical
pp. 192-205on the official titles recorded in the Wei chih for early Japan, and their possible
interpretation. Hamada 1900 is a basic early linguistic study, now largely of historical
interest. Omori 1955a deals inter alia with the implications for the interpretation of OJ
written records of the fact that the orthographic devices by which the later texts distinguish
OJ i/i, e/e, and o/l can already be observed in part in the Wei chih transcriptions. Else-
where he attempts a comprehensive summary of just what the Wei chih transcriptions tell
us about Old Japanese (Omori 1955b:166-94); this still errs slightly on the side of incor-
porating too many traditional but linguistically indefensible 'readings' for these transcrip-
tions, but it is nevertheless a first-rate piece of modern scholarship which has done much
to clear away decades of learned obfuscation. Since the problems of the Wei chih text in-
volve the remote history and ultimate origins of the Japanese imperial house, its study
could not be conducted in a wholly scientific manner until Japan's defeat in World War II;
the various forces which influenced all types of scholarship on the subject until that time
are discussed in Young 1958, on which see also Hall 1960.
4 The translation in Mills 1960 is useful; but since he transcribes the document not in OJ
but as if it were in the modern language, and in addition garbles the text of poems 7 and
17, it is necessary to use and cite instead the text in the edition of Tsuchihashi Yutaka (in
Tsuchihashi and Konishi 1957:239-47). I have verified Tsuchihashi's readings, which are
accurate, with a rubbing of the inscription which I acquired at the Yakushi-ji in Nara in
April, 1965.
6 For general introduction in a western language to the periods in Japanese linguistic
history and to the major remains of the literary records, see Lewin 1959;and for a methodi-
cal treatment of virtually all the Japanese sources and secondary literature, chiefly con-
cerned with but by no means limited to phonetics and phonology, Wenck 1954-59.Wenck's
rich bibliographic information provides an extremely useful guide to the widely scattered
Japanese literature.
280 LANGUAGE,VOLUME43, NUMBER1 (1967)
phonology has made it possible to reach a significant number of fairly firm con-
clusions about the phonology of OJ; I have summarized the methods of such
studies and something of their history elsewhere.6These studies reveal, in briefest
summary, a language in which the morpheme regularly consisted of C + V
sequences; different interpretations of the data give different consonant inven-
tories, but a convenient one with which to begin the discussion is the following:
[k g t d ts dz s z r n F b m w y], with [F] the symbol for a voiceless bilabial
fricative going back to a [p] at a somewhat earlier stage of the language. On the
phonemic level different analyses are also possible, and result in different treat-
ments of the relationship between [s] and [ts] and between [z] and [dz];in one
system, [s] is found only before OJ i and e, with [ts] always elsewhere; so also with
[z] before [i] and [e] but [dz]elsewhere. There has been a considerable amount of
important new scholarship on these problems in recent years,7 but the issues do
not directly affect the discussion in the present paper.
In vocalism the written records show an eight-vowel system for OJ, so that to
the a i u e o known from most of the modern Japanese dialects, notably and typi-
cally from the modern Tokyo Standard Language (J), three additional vowels
must be added. These are generally written'i e (though for this last Japanese
scholars often prefer other more complicated symbols, including a, 3, and e).
There has been much speculation, largely inconclusive, on the phonetic qualities
of these three vowels, but fortunately this too need not be resolved for our present
purposes, since the written records (as well as the results of internal reconstruc-
tion and the comparative method) make it clear that in OJ we are dealing with
an eight-vowel system, regardless of precisely what some of the eight may have
sounded like.
6 See Miller 1964. Izui 1963 reached me too late for inclusion in the bibliography of that
article. Reviewing Izui's paper, Householder (1965:309) remarks that 'Izui's contribution
is largely news to most of us, since virtually none of the cited literature was published in
a western language'; this is of course true of the work upon which Izui reports, but this
was in large measure due only to Izui's own unfamiliarity with the literature on the sub-
ject available for three decades in western languages. The most important such work, and
one often overlooked, is the monograph by Yoshitake 1934, a thoroughgoing treatment of
OJ pronunciation from the point of view of the written records, particularly of the pro-
nunciation of the vowels. This monograph made easily available everything that had been
done on the problem up to that point both in Japan and the West, and had complete refer-
ences to the other earlier western literature. Yoshitake's work was soundly based upon
Japanese scholarship, but thanks to his remarkably original turn of mind he always did
more than simply repeat the findings of others. He attempted a fairly narrow phonetic
reconstruction of OJ pronunciation and unfortunately did not concern himself with pho-
nology, but his 'Examples of Ancient Japanese sounds' (pp. 62-71), actually a short lexicon
of OJ in his reconstruction, was not only a pioneering attempt, showing considerable cour-
age, but remains an extremely useful check-list for OJ. For the period between 1934 and
1959there is of course now also Wenck (see fn. 5 above), as well as Haguenauer 1956,which
in spite of its title deals largely with linguistic problems. With the availability of this
much secondary literature in English, French, and German, it is difficult to understand
why Izui's account should have come as a surprise. Cf. now also Miller 1967:172-80.
7 Much of this is the work of Mabuchi
Kazuo, who has recently challenged much long-
accepted doctrine in this field; see, for example, on the problem of the allophones of OJ
/s z/ etc., Mabuchi 1959, reprinted with some changes in Mabuchi 1963.
The distribution of the vowels with respect to the consonants in OJ is of far

greater interest and has much greater significance for the history of the language
than does their pronunciation. According to the most commonly accepted read-
ing and interpretation of the OJ written records, basically that resulting from the
studies of Arisaka Hideyo (1908-1952),8 the entire repertory of eight vowels
(a i u e o ; e 6) was found in the language only after k g m; 6 was missing after
F b, and i e were missing after t d n r. Only a u o 6 were found after t8 dz, and
only i e after s z. After initial # (zero, or possibly a syllable-initial glottal phoneme;
cf. Hattori 1955, 1961, Martin 1959:375-6), only a i u e o were found, a pattern
further reduced by the non-occurrence of homorganic sequences after y w, so
that w was followed only by a i e o and y only by a u e o-a nearly symmetrical
pattern upset only by the occurrence of the sequence y6.
This system has recently been challenged, notably by Mabuchi 1957, who has
brought together evidence for adding 6 following #, i following s z, and 6 follow-
ing F. Internal reconstruction makes it possible to add e following z, perhaps to
add i e followings, certainly to add e following t n r w, and perhaps 6 following F.
The comparative method makes it possible to add e and perhaps i following z,
e following r, i e following n, and 6 following b9 Putting all this together, and at
8 Most important are Arisaka 1944, 1955. In traditional Japanese terminology the OJ
vowels written here as i e o are koorui 'Type A', and i e 6 are oturui 'Type B'. Consideration
of the phonemic status of these Type B vowels along the lines which Martin (1959:380)
suggests seems to point to an analysis of the Type A i and e as /yi/ and /ye/. These con-
trast with the Type B i and e as /i e/, perhaps to be understood as [i e], an interpretation
which parallels some of the earlier attempts by Japanese scholars and others to make more
precise phonetic statements about the pronunciation of the Type B vowels. For a con-
venient summary of these attempts, from 1934 through 1955, see Wenck (1959:4.15, ?914).
On this subject Martin has now written me further under date of 16 May 1966: '[This] con-
clusion [was] also reached, apparently independently (though we may have discussed it, I
forget), by Hattori [Shir6]in Nihongo no keito (Tokyo, 1959) 285-6; but I suggest that the
'Type' labels are perhaps to be reversed for labials. This was, as I recall, based on an exam-
ination of Ryikyu material and the reconstructed OJ vocabulary, where the COMMON
type is what the philologists have called 'Type A' except after labials where it is what
they have called 'Type B'. I also suggested (loc. cit.) that the Type-A back vowel ("o")
was really /wo/ or /wa/; the Type-B back vowel ("o") was the common one, and I assume
that it was the simple /o/. Of course, this raises questions about syllable type and so on,
e.g. "Is there any evidence that would require us to reconstruct postconsonantal /ya yo
yu/ and /wi we wu/"-there may be, but I can't cite it at the moment. Leaving that aside,
we can still maintain the Cy/wV schema as a way of showing at least the phoneme COMPO-
NENTS, without worrying whether the y/w goes with C or V, and thus clarify the class
identification of the 'odd' member of each attested pair. Note that I imply that the 'neu-
tralized' version after certain consonant types at various periods was not phonetically
'in between' or indeterminate but rather is to be identified with the COMMONtype in the
polarized pairs and not be treated as an archiphoneme.' The philologists do not appear to
have had any good reasons for assigning the A-B type distinctions after different conso-
nants; apparently their labeling of each category was arbitrary, depending in each case
on whether the first word in the texts clearly identified as belonging to one or the other
category was written with Type A or Type B vowels. Hence the traditional labeling (which
we perpetuate in our transcription of OJ) is significant, it at all, of nothing more than the
order of the original philological discovery of individual OJ lexical items.
9The evidence from internal reconstruction and the comparative method is brought
together in Miller 1964, and in Miller 1967:181-9.
282 LANGUAGE, VOLUME 43, NUMBER 1 (1967)
the same time collapsing the s t8 and the z d' categories in terms of their compli-
mentary distribution, the scattered patterning of the traditional Arisaka version
of the phonology of the OJ written records is replaced by a virtually gap-free
system in which all eight vowels occur after a majority of the consonants, i.e.
after k g t dZn r F b m; the only gaps remaining are i e missing after #, i missing
after t d y w, and the symmetrical non-occurrence of the homorganic sequences
*yi *ye *wu *wo.
It is, however, both somewhat risky and somewhat inconvenient to attempt to
cite OJ forms exhibiting this virtually complete system, in which all but a few of
the vowels appear after each of the consonants. It is risky chiefly because some of
the 'filled-in gaps' ('filled-in' with reference to the rather more restricted system
of Arisaka, for example) depend solely upon what are still somewhat moot points
in the philological evidence. If the interpretation which now appears to fill in
these gaps in the system eventually wins general approval, well and good; but
if it is later disallowed, then any further linguistic extrapolation erected upon its
basis will also collapse. The 'filled-in gaps' resulting from internal reconstruction
or those based upon the comparative method are probably less risky, but it is
still better to avoid citing forms containing them for the time being, since these
two techniques of linguistic reconstruction are still unfamiliar to and hence
unaccepted by most Japanese scholars working on the history of Japanese. This
means that they have yet to gain general acceptance in the literature, and that to
introduce them into a further comparative discussion at this time would be
unnecessarily to complicate and confuse other issues which at best are compli-
cated enough. This is why, despite the availability of these other sources of in-
formation for the nature and structure of the OJ vocalism, it is still best in any
comparative work at present to limit our citations of eight-vowel OJ forms to
those which can be verified in the written records.'0
The possibility of establishing a historical relationship between Japanese and
Korean on the basis of regular sound correspondences between a large number of
lexical items in the two languages has recently been most strikingly demonstrated
by Martin 1966; on the basis of his comparisons he is able to reconstruct forms
for proto-Korean-Japanese (KJ) which express the correspondences he estab-
lishes between forms in the two languages. The possibility of demonstrating such
a relationship had been suggested over and over again during the past century,
but Martin's work for the first time removes the problem from the realms of
speculation and unsubstantiated hypotheses, placing it on a solid scientific
foundation. As far as the Korean lexical items are concerned, his KJ reaches a
10But it must be noted that few Japanese scholars are this cautious; at best they seldom
distinguish between results derived from the inspection of written records and results
coming from other types of historical study. Examples of forms resulting from the ensuing
confusion of levels are *w6sofor OJ woso (deprecatory morpheme), *6n6 for OJ ono 'self',
*F6so- for Foso- 'thin', narrow', *6Fo- for OJ oFo- 'great', *us6 for OJ uso 'lie', and kuso
for OJ kuso 'dung'; all these are to be found in Izui (1956:989-1020).There is nothing in-
trinsically wrong with forms like *w6s6,except for the fact that they symbolize a consider-
able body of unverbalized assumptions both about the processes of reconstruction which
underlie them and about the prehistory of OJ. Hence it is a risky business to cite them
without specific identifications of these factors.
considerable diachronic depth, since virtually every citation of Korean evidence

involves a verification with or exegesis of individual developments in terms of
Middle Korean (MK) phonology of the 15th century. With the Japanese evi-
dence, however, the situation is somewhat different. Martin cites a few forms
which he identifies as OJ, but virtually all his Japanese citations are instead J,
i.e. the modem standard language of Tokyo, somewhat 'antiqued' in its con-
sonantal orthography. In some respects (but not all) this 'antiquing' has taken
the forms back perhaps as far as the 12th century, but their time-depth is less
important than the fact that they are orthographic projections backward in time
rather than true earlier historical forms. The forms he identifies as OJ are obso-
lete or archaic lexical items not surviving in the modem language, but (apart from
their 'antique' consonant orthography) they too do not reflect any significant
features of the phonology or structure of OJ. The Japanese forms upon which the
bulk of Martin's comparisons is based, then, are the forms of modem standard
Japanese with minor orthographic adjustments along historical lines: syllable-
initial /p/ is written historically and on the basis of the traditional orthography
for modern /h/, i.e. for the initial element in each of the five sequences [ha hi 4u
he ho]; and an initial /y/ is regularly interpolated before /i/, breaking up J i
into two components for the purposes of comparison.1'In other words, Martin's
lexical comparisons are essentially comparisons between a partially 'antiqued'
J and MK, and the phonetic correspondences which result from these compari-
sons (and by implication his KJ reconstructions) are based on these two lan-
guages; but they do not involve our knowledge of OJ phonology. Naturally, we
will want to see what differences, if any, a consideration of this knowledge makes
in the earlier common language revealed by Martin's lexical comparisons; and
that is the primary purpose of the present paper.
n1I would prefer reconstructing initial /y-/ only in those cases where it is unmistakably
indicated by internal reconstruction (ik-u, doublet with yuk-u 'go'; i-i, doublet with yo-i
'be satisfactory', OJ yosino 'fair field' [place name], var. esino < yesino in Kojiki poem 97
etc.; other doublets in Wenck 1959:4.184, ?983.8), or by comparative evidence, tracing it
back in most cases to an original (Altaic?) dental initial (this initial can also sometimes be
recovered by internal reconstruction; e.g. yama-i 'illness', i-tam-u 'be painful'). A good
deal of evidence which indicates a correlation of OJ or pre-OJ initial *y- with initial den-
tals in related languages has been collected in Murayama 1962. Of course, as in the case of
the OJ eight-vowel system, the entire process of reconstruction of earlier forms becomes
more impressive when it can be shown that both the comparative method and internal
reconstruction lead to the same forms. One such case involving OJ *y- is that of OJ yo-
'four' and OJ ya- 'eight', which go together in the well-known o/a ablaut relationship, and
may be compared with Mongol dorben< *dor-ben,Manchu dujin < *do-gun,Evenki dygin,
proto-Altaic *do- 'four' (the initial of Old Turkish tort 'four' shows that it is an old loanword;
the normal development would have had initial j-); and with Evenki djapkun 'eight' and
other forms from proto-Altaic *da- 'eight'. (On these cf. Nomura 1955:583-4;Murayama
1962:108; Ramstedt 1952:62; Poppe 1955:244 and 1960:22, 110). Within Japanese itself the
same initial can be established by internal reconstruction: ya- 'eight', ta-ko 'octopus'. The
same process can be employed with OJ yo, var. yoro (Fudoki poem 3, Tsuchihashi and
Konishi 226), J yoru 'night', which Murayama, loc. cit., has comparedwith 'proto-Tungusic'
*dol-bo;this goes directly with Mongol iides (i.e. u-des) 'evening', where the final consonant
preserves an example of proto-Altaic *-12, for which the normal correspondence is Mongol,
Manchu, Tungus 1, Turkish s, but which here has been upset by borrowings.
Our first step will be to cite here, in the same order as Martin's original lexical
comparisons, all the forms involved in his reconstructions for which we have
textual evidence involving the occurrence of OJ o, e, and z. To insure that we err
on the side of caution, if err we must, only forms attested in the written records
have been taken into consideration (with exceptions involving internal recon-
struction to be noted immediately below); this means that we are dealing with
the most restricted system of distribution of the eight-vowel system, and also
that, as a result, there are relatively few forms for which the OJ vocalization is
significantly different from that of J. OJ forms have been drawn chiefly from
citations in DGK (= Otsuki Fumihiko, Daigenkai, 1932-37); but page refer-
ences have not been given to this generally available if now somewhat old-
fashioned reference work, except when the discussion there is of particular
interest, since anything that I have drawn from it can easily be verified in each
case sub verbo. From the DGK I have also drawn a considerable number of OJ
forms attested solely in WM (= Wamyoruijusho), a Heian lexical compilation
by Minamoto Shitago dating from ca. 93412 The OJ phonetic glosses in WM are
of course not always as reliable as they would be if they were attested in some
earlier document, and there is some evidence that its phonology confuses, for
example, OJ /e/ and /ye/ (Ono, Nakada, et al. 1961:109); but in the main its
glosses are based upon earlier sources, and by and large they preserve the original
OJ eight-vowel system. A few citations are also drawn, again via DGK where
they may be verified with full citations sub verbo, from JK (= Jikyo), another
somewhat later lexical compilation of unknown author and date, which was
probably put together shortly after the first part of the 12th century. By this
time the eight-vowel system of OJ had changed in the direction of the modem
five-vowel J, but the phonetic glosses of JK are mostly based on earlier sources
and reflect a phonology far earlier than the text itself. Most valuable of course
are citations directly from the OJ text corpus itself, especially those from the
Nihon shoki, here cited from Ono (1953a:229-75), where OJ materials from the
Nihon shoki are for the first time made conveniently available with glosses and
with meticulous attention to their vocalism as shown in the documents (cited
simply with Ono's name plus page numbers). Note that most of the DGK cita-
tions which here carry no other identification also go back to the OJ corpus,
often to the Man'yoshu or to other primary sources for the language; the sigils
WM and JK have been used here only to indicate the absence of such earlier
primary evidence, which in any individual case may or may not be significant.
In a few cases it has been useful to cite the Man'yoshu directly as M, with poem
numbers as in the great edition of Ono, Takagi, and Gomi (1957-62); Ono was
solely responsible for all the linguistic glosses in this edition, which in this and
in many other respects is a monument of contemporary Japanese linguistic and
literary scholarship. A few forms are also cited directly from BSS (= Bussokuseki
inscription; cf. fn. 4).
In addition to the details of the OJ eight-vowel system, another phonological
problem which arises from time to time in any confrontation of OJ forms with
12 For an introduction to
these sources see Karow 1951; less useful from the linguistic
point of view is Bailey 1960.
their J cognates is that of the sporadic occurrence of intervocalic voiced con-

sonants in certain later forms. The representation of voiced consonants in OJ
texts written phonetically in Chinese characters is now much better understood
than it was one or two generations ago, when some scholars seriously argued that
perhaps OJ had had no voiced consonants at all, or at most only sporadically
voiced consonants in free variation with their voiceless norms.l3 Now we know
that it is possible to trace voicing in the OJ records with a considerable amount of
confidence, but it is also clear that this does not always correspond to the presence
or absence of voicing in the cognate J forms. The circumstances under which this
voicing, which now appears to be completely sporadic, arose in each individual
form are rarely clear; probably shifts in the position of the pitch accent and dia-
lect mixture were the major factors in most developments, but the overall ex-
planation is still lacking. Nevertheless, poorly understood as their origins remain
for the present, it is important not to involve these later voicings with our KJ
reconstructions, since all the evidence points to their being secondary develop-
ments within Japanese. Several of these cases have been noted in the glosses
which follow on the individual items in Martin's lexical correspondences; on one
occasion at least (item 87 below), the OJ evidence corroborates that J is at best
only a convenient but never a methodologically desirable point of departure for
such comparisons.
Finally, some OJ forms whose vocalization has been recovered by internal
reconstruction, or by internal reconstruction in conjunction with the compara-
tive method, have been taken from Miller 1964 (hereafter abbreviated IR);
in most cases these reconstructions can also be verified by reference to the
written records cited s.v. in DGK.
The following, then, is the OJ evidence which may be added to Martin's lexical
correspondences for Japanese and Korean, to which reference is made below
with his original numbers and English tag glosses.14
13On this subject see Ono 1960.In transcribingOJ I have followedthroughoutthe
indicationson voicing/voicelessandkoo/otu-membership for the man'yogana graphsfound
in the tables in the Sekai gengo gaisetsu 2.266-9; another convenient and equally reliable
sourceforsuchdataarethe tablesin Ono,Takagi,andGomi(1957:1.32-5); botharethe work
of Ono.
14Verbbases whichwouldhave had OJ e or i in certainformsonly have generallynot
beennotedas partof the abovedata;thus, forexample,with 1 ACCUMULATE, wherethe form
whichMartincites as tur(e)-wouldhave had OJ -e- in thoseformswherethis vowelwas in
turn immediatelyfollowedby OJ -k-, -F-, or -m-,but OJ -e- in all others.Similarexamples
are, among others, 5 ARRIVE, 15 BLOCK IT, 40 1CHANGE, 56 CROWD, 60 DAMPEN, 68 DOWN
(with -i- in those forms with OJ -k-, -F-, and -m- immediately following), and 69 DROOP.
Indications of suprasegmental features have been omitted from all citations, partly because
they appear to be without significance for the discussion, and also because data for the
suprasegmental configuration of Japanese forms earlier than J are not conveniently at
hand, though a certain amount of work has been done on the historical development of the
Japanese pitch accent; see for example McCawley (1966:174-5) on Kindaichi Haruhiko.
Perhaps omitting such data here has been a rash decision. But no convincing formulation
is possible until we have more complete information on the earlier nature and location of
this feature in Japanese, and the pitch which Martin indicates on his J forms is that of the
modern standard language, which can hardly be taken over intact as a guide to the supra-
segmental configuration of OJ. Even so, there are several interesting indications of a
2. ACCUMULATE. OJ mor- 'heap them up' (Ono 270) OJ moromoro'all, both,

many, together' (BSS 18.245).
4. ALTHOUGH. OJ concessive morpheme -d6 'although it is so'; OJ toFoked6
'although [she] is far away' (M 3981).
7. BAMBOO. OJ take(IR 11, 15).
14. BIRD. OJ sagi 'a heron' (WM, JK).
18. BOAT.OJ Funr (IR 14).
19. BODY. OJ mi (Ono 246).
29. BRUSH. If OJ Fude is, as often suggested, a development from Fumite
(and that form is actually attested in WM), then it may be reconstructed with
considerable confidence as OJ Fude, since the final -te would then be OJ -te
'hand' (IR 15; cf. Yoshitake 1930).
37. CAGE. OJ wori, which cannot go with the Korean forms cited in com-
parison; hence it is probably necessary to abandon the proposed KJ *ori/*eri.
Fortunately (as Martin points out to me in his letter of 16 May 1966) K uli <
MK eli can be accounted for under other etymologies, uli under 44 CLAN, and
eli under 259 2wEAVE.
38. CARRY.OJ nor- 'ride, mount' (M)
43. CHICKEN. OJ tori 'bird'(M), but tori (WM).
45. CLAW. OJ tume (IR 11).
47. CLOUD. OJ kumo (Ono 242). For OJ 'black' the Nihon shoki has both kuro
and kuro, since the graph used for the second syllable is capable of both inter-
pretations (Ono 242), but internal reconstruction shows kuro (IR 15). OJ keburi
'smoke' is attested in a Nara-period gloss to the Chinese translation of the
Avatamhsakasitra, cited s.v. in DGK.
possiblecorrelationbetweenthe location of the pitch accent in J and the appearanceof

what must for the time being remaina 'sporadic'phenomenonof voicing of intervocalic
stops at different times in the history of the language. It would seem that this voicing ap-
peared in many cases when the pitch accent immediately followed the consonant in question,
while in words without this suprasegmental feature, or in words in which it preceded the
consonant, the consonant remains unvoiced. Examples of the voicing operating under these
circumstances are 87 FOOT, K pal, J hagi < Faki; 269 ARM, K phal, J hidi < Fiti; 136 MARK,
K al, el, J azd < asa; 119 JAW,K akali, akuli, J ag6 < aki; 183 REPAY, K mul-, J mod-s- <
moto. These are to be compared with cases in which the voicing did not occur in environ-
ments which are identical except for the absence of the pitch accent immediately following
the consonant: 12 BEE,K pOl,J hati; 113 HUSBAND,K pes < MK pet, J oto- < woto-; 143
MOON,K tal, J tuki < tuki; 150 MOUTH, K kut-, J kuti. Similarly the voicing did not occur
when the J pitch accent immediately precedes the consonant: 16 BLOW,K pul-, J hlvk-;40
kCHANGE, K pakku-, J bdk(e)-; 90 FULL, K michi-, J mit(i)-; 146 1MOUNTAIN, K thek, J tdka-;
156 OAK, K kal, J kdsi; and others could be cited. Comparisons of the type of 141 MENACE,
K ili-, J odor6k- < otrok-, and 151 MUDDY,K tel'w/p-etc., J dor6, make one wonder if the
same process did not also operate at a distance one-syllable removed in cases where the
intervening syllable consisted of -r- plus vowel. Also extremely suggestive are such com-
parisons as 69 DROOP,K suk-, J sdg(e)- 'hang it down', sagdr- 'sag, fall', J sakd '(down)
slope of a hill'. These show that while the overall outlines of the historical process operative
here can still be discerned through the traces which it has left in J, other changes in the
history of the language have at times operated to obscure it, with the result that a complete
formulation must await more information about the historical location and nature of the
pitch accent.
53. CORRECT. OJ mato 'target' (WM).

55. CRANE. OJ suffix -me, common in naming nonhuman animates, particu-
larly birds; cf. suzume 'sparrow', tubakurame 'swallow,' and kamome 'sea-gull';
all have OJ -me in WM.
61. DEEP. OJ kiFa 'extreme'; OJ kewasi- 'be steep, precipitous' (JK).
64. DIRECTION. OJ yoko (Ono 273), though WM has OJ yoko- already par-
tially levelled out in yokoFuye 'horizontal flute'.
65. DIRT. OJ kita-na- 'be dirty' (Ono 240).
66. DIRTY. OJ nigor/s- 'get, make dirty' (M). Martin remarks, 'left unex-
plained is the relationship, if any, [between J yogor(e)- "get dirty", yogos- "make
it dirty", and] the synonymous Japanese etymon in nigor- "get dirty" and nigos-
"make it dirty".'
The relationship here can be traced through different Japanese reflexes for a proto-Altaic
initial *~i-,which gives n- in Japanese before the modern front vowels, but zero (alternating
with y- in some way that cannot yet be described) before the modern back vowels. Thus
J arata 'newly' and nii- < OJ niFi- 'new' go together historically and can both be traced
either to the form eventually underlying Mongol nirai < *-iarai 'neugeboren,frisch', Buriat
harai 'neugeboren, frisch, jung (Pflanzen)', Manchu iarxun, written nijarxun, 'neu, frisch,
griin', Old Turkish jaz < *jar2 < *r2 'Friihling' (these forms are difficult to keep separate
from Mongol and Middle Mongol nilqa 'jung, klein, frisch, unvernminftig'and others going
with proto-Altaic *'alawun 'roh, unangenehm nach rohem Fisch schmeckend, widerlich'
and *Aalugar'schleimig, glitschig'; Poppe 1960:38-9). The forms with original -i show up in
Mongol nogoyan < *nigoyan 'griin, das Griin, Gras, Gemiise', Manchu 1owaygdn, written
niowaygijan 'das Griin, griin', and ?ioro-,written nioro-, 'griinen, grin werden' (Poppe
1960:38). With the forms involved in this comparison we may further bring into the con-
sideration J -aka in teaka 'dirt rubbing off on and soiling an article handled often' (with
te- 'hand') as an example of the back vowel development in Japanese with zero initial; the
sporadic y- development shows up in yogor(e)- and yogos- which go together with nigor-
and nigos- and eventually with Mongol nijasun < *1vyasun < *4agasun 'Eiterkliimpchen
in den Augen', Buriat rlagahay 'id.', Manchu 1aki, written nijaki, 'Rotz', Evenki niakse
'Eiter', Lamut fawulan 'Eiterung' (Poppe, ibid.) 1 Martin, 95 GREENS,comparesK namul <
MK nomalh and J na (KJ *na), which must also be considered together with these Altaic
forms.
75. ENJOY.OJ y6orokbosi 'delightful', OJ yoro-si 'be acceptable' (Ono 273).
76. ENTER. OJ yor- 'approach (etc.)' (Ono 273).
77. FAST. The problem of the s- in MK spalo- 'be swift, be pointed', cognate
with OJ Faya- 'be fast, swift, early', must surely be related to the mysterious OJ
tiFayaburu, in Japanese poetry a makurakotoba 'pillow word' (cf. Dickins 1908),
i.e. a conventionalized poetic epithet for gods, men, and the place Uji; thus often
in the Man'yosha, e.g. M 4402.
Traditionally the ti- element in tiFayaburu has been variously interpreted, including
morpheme identifications with ti 'thousand', ti 'blood', and ti 'road, way', but ultimately
these etymologies do little but preserve some of the fantasies of the traditional orthography.
The gloss of MK spola- 'be swift, be pointed' is also reminiscent both of the evidence for
phallic worship in early Japanese religion (Kat6 1924) and of the poetic epithet tamahiko
no 'jewel-stalked' for roads, through association with their phallic-shaped markers (e.g. M
3995; cf. Ono's gloss in M 1.339).16This is not the only example which may be cited of a
16This correspondenceis treated in greater detail in Miller 1967:73-5.
16Gusinde and Sano (1960:448,plate d) illustrate megalithic remains which
give a certain
makurakobotawhich apparently has its ultimate origins, if not as far back as the KJ unity,
at least in a period remote enough for it to have been possible for the OJ poets efficiently to
fashion one of their extremely sophisticated literary devices on the basis of elements drawn
from a common Korean-Japanese cultural as well as linguistic heritage. Another such
makurakotobais OJ komaturuki 'a sword of Kogury6 [one of the ancient Korean 'Three
Kingdoms', founded 37 B.C.]'. (The vocalism of OJ koma in this sense is attested in WM;
corresponding to J turugi 'sword [obsolete]' OJ has both turuki (Ono 253) and turugi (Ono
241); cf. Wenck (1954:2.62, ?491b). In M 109 the phrase komaturukiis a makurakotobafor
wazami ga Fara 'the plain of Wazami', in M 2983 it is a makurakotobafor na ga kokoro'my
heart'. The traditional explanation for these locutions, cited DGK 2.358d,goes back to the
monk Keichf (var. Kaichf) (1640-1701),the great early Tokugawa Man'yosha exegete, and
attempts to find their rationale in a pun between OJ wa- 'I' and OJ wa 'a ring, circle, loop',
the semantic link being provided by the fact that an early variety of sword from well-known
Tumulus cult sites in Japan (and presumably of continental origin) has a circular (wa)
hand-guard corresponding to the modern tuba 'sword-guard' (with which, however, it is
not to be connected etymologically-tuba is a late contracted form based on a late OJ
tumiFa, in which the etymology of the initial tumi- is not clear but where the final -Fa must
surely be Fa 'blade'; internal voicing resulting from the contraction and disappearance of
the sequence -mi- is a well-attested phenomenon in late OJ and from that point on through-
out the history of the language). Thus komaturukias a makurakotobawas employed by the
poets to elicit the wa of wazami and the na of na ga kokoro.But Ono's gloss in M 3.284 to
M 2985 suggests much more plausibly that the connecting link is instead to be found in
OJ na 'myself' and at the same time in K nal 'blade'-cf. J katana '(a later variety of)
sword, the modern Japanese sword' (though it is difficult to follow Ono, loc. cit., in his
attempt to identify kata- here with J kata- 'single, one-sided'; a much more satisfactory
etymology is to involve J kata-i 'be hard, firm, unyielding', OJ kata 'land alternately hidden
or revealed by changes in the tide'-cf. Mongol qada 'Felsen' and qata- 'austrocknen, hart
werden'; after all, the two most important characteristics which distinguished the newer
and more satisfactory katana from the old and ornamental but highly inefficient turuki >
turugi were the harder, stronger metal used in its blade and its double-edge). In 100HATCHET
Martin compares K nas 'sickle' and J nata 'hatchet, machete', relating them by means of
KJ *nas(a), and asks 'is K nal < MK nolh 'blade' related?' The use of komaturukias a
makurakotobaeliciting OJ na provides additional evidence for the reality of the relation-
ship.
78. FERTILE. OJ koy-(e)- 'get fat, flavorful' (JK).

79. FIELD. OJ Fatake (WM).
81. FILTER IT. OJ kos- 'pass (it) over, through' (Ono 243-if the form belongs
here).
82. FIRE. OJ Fi (Ono 262); the word contrasts with OJ Fi 'sun, day'.17
84. FLOOR. OJ muro 'plastered sleeping-place in house; protected, tightly
enclosed shed or house; cave' (Ono 236), but also OJ muro 'id.' (WM). The early
Japanese bath was always a 'Turkish' steam-bath (J musiburo) rather than the
modern soak-bath, and there is an etymological connection between this root
amount of substance to these speculations; their conclusions are inadequate partly because
of the nature of their subject, and partly because they ignore linguistic data which might
bear on the solution of their problem. Even so, it is significant that they tend to the view
that the remains which they have investigated are not, as commonly held, the work of the
Ainu but rather of 'Tungusic invaders'. The linguistic implications of the Ainu folklore
belief which associates these stone remains with the Koropukguru (cf. OJ kara 'the conti-
nent, Korea, China'), the pit-dwelling peoples of Ainu legend, urgently need investigation.
17 On this contrast, cf. Izui (1956:1008) and Ono
(in Ono, Takagi, and Gomi 1.176), in
his gloss to poem M 336.
and the original form underlying modern o-huro 'bath'. The connecting semantic
link must have been the fact that the muro was the only room in the building con-
structed tightly enough to permit it to be used as a steam room. Perhaps origi-
nally 'air-tight room' and 'bath' were kept apart by the o/6 distinction, but even
if this were so, it is impossible at this point to say which vowel went with which
meaning.
85. FOG. OJ sagiri, with prefixed sa- and secondary voicing (Ono 245).
87. FOOT. 'OJ Fagi 'shin' is often compared directly to K pal; if it were *paki
we could reconstruct KJ *palgyi'; but fortunately this is one of the words in
which the modern intervocalic voicing is of late origin, and OJ has exactly the
form in -k- to be expected from Martin's comparison with Korean, OJ Faki
'shin' (WM, JK).
90. FULL. OJ mit(i)- 'get full', mitas- 'fill it' (M).
96. GULL. OJ kamome 'seagull' (WM).
100. HATCHET. See under 77 FAST above.
106. HOLD. OJ tor-, tor- 'hold in the hand' (Ono 254). See the discussion below.
113. HUSBAND (or MAN = MALE). OJ wotoko 'young man' (Ono 275).
114. HOUSE.OJ iFe (Ono 233, IR 14). Cf. the verbal formation seen in OJ
iFor- 'dwell in a hut; inhabit temporary lodgings', as in Fudoki poem 3 (Tsuchi-
hashi and Konishi 226): tukuFane ni iForite tuma nasi ni 'spending the night
without a bedfellow on Mt. Tsukuba'.
119. JAW.OJ aki is the form to be compared directly with MK akwi 'mouth',
not J ago, which however goes back at least to the form ago registered as 'vulgar'
in the Wa-Kan sansai zue of 1715 (correct Wenck 1959:4.174, ?980; citation
s.v. in DGK).
The Ryukyf dialects are of no help in determining the history of the Japanese forms, which
they do not appear to have inherited; instead Shuri has ?utugee (cognate with J otogai
'chin'), which the Okinawagojiten (Iwabuchi 1963) glosses with J ago-unless it is possible
to detect some distant echo of aki > ago in the last element of the Shuri form Putugaku
glossed loc. cit. 570 as 'a vulgar term for ?utugee'.The modern lexicographers have generated
a ghost word *agi, which the texts do not substantiate, apparently as a convenient midpoint
between OJ aki and J ago. They also engage in extremely involved semantic gymnastics in
attempting to relate all these forms for 'jaw' to OJ akito, J agito 'gill of a fish' (e.g. Matsuoka
1962:6). Most of this seems as pointless as it is unlikely, particularly in view of the vocaliza-
tion of the second syllable of OJ akito, attested in the same text, WM, that registers OJ
aki 'jaw'.
122. LATE. OJ noti 'later on' (M, BSS 10.243).
123. LESS. OJ otor- 'be inferior' (BSS 13.244). The modern form otootois from
late OJ otiut6 (WM), and derives from oto 'younger brother' (Ono 236) and
Fit6 'person'.
127. LIQUOR. OJ sake (IR 14).
128. LITTLE. OJ sukosi 'a little' (JK); both 'little' and 'child' were OJ ko
(Ono 243, WM), which contrasted with the deictic k6-.
136. MARK. OJ asa; the voicing which eventually resulted in J aza appears
from the sources to have taken place prior to the time of the Ruijumyogisha,
i.e. sometime in the first part of the 12th century.
141. MENACE. OJ ot6ork- 'take fright' (JK). This is another etymology in
which the identification of the late, secondary character of the -d- in J odorok-
'id.' is of importance.
142. MILDEW. No OJ form appears to be citable, but one wonders if OJ kabi
'bud, ear of rice' (WM following a gloss in the Kojiki) is related.
143. MOON. OJ tuki (M 3586).
146. 1MOUNTAIN. OJ take 'peak' (Ono 250).
151. MUDDY. On J doro 'mud' Martin remarks, 'I assume the Japanese initial
voicing is either expressive or dialect in origin', and compares it with Korean
forms in initial t- going eventually with KJ *ter(e)-. Shuri has duru 'id.' The
DGK 3.605d cites the Japanese form from only three texts-a commentary on
the Mencius dated in the period 1469-86, and two historical romances which can
be dated only roughly, one probably in the early 12th century, the other some-
time in the mid-13th century; apparently no earlier citations are possible. But
whether the DGK's citations accurately reproduce the orthography of the
manuscripts is a moot question-and moreover even if the scribes did write
(without the voicing diacritics) toro, this might have been intended either for
toro or for doro! So in a critical case like this it is impossible to trust second-hand
kana citations in the lexical sources without further information on the actual
text and manuscript tradition. Nor is it possible to cite J *toro 'mud', but the
form prior to the expressive or dialect voicing of its initial is not difficult to
reconstruct, since it clearly goes with a sizable group of forms, the most im-
portant of which are tororo 'grated yam', torori 'melting, fluid, oozing', torotoro
'dozing, drowsy; slow, simmeringly; bum slowly'. The earth in dorois in a liquid
suspension, just as the grated yam is in tororo.It also seems more appropriate to
accommodate here J tiri 'dust; dirt', euphemistic for 'feces' in tirigami 'toilet
paper', as going together with *toro, doro rather than with Martin, under 71
EARTH, as a deverbal noun from tir- 'scatter'.
155. NIGHT. OJ yo, contrasting with both OJly6 'world, age' and with OJ
2y6- 'be good, be satisfactory' (Ono 272). Cf. fn. 11.
156. OAK. The custom of wrapping certain kinds of cakes in oak leaves, for
which this etymology provides old linguistic evidence, is represented on the
Japanese side by J kasiwamoti 'glutinous rice-cakes (or today more usually other
confections) wrapped in oak leaves'; the ultimate origins of this object are
extremely involved, and as this etymology shows, extend at least as far back as
the Korean-Japanese unity.
Completely to unravel all the strands here would require treatment at monograph length,
but the principal elements are probably to be related ultimately to the root common to
OJ kasiko-m- 'obey reverentially', surviving in J kasikomarimasita'I have understood and
will comply; yes (servant to superior)', and OJ kasiko-si 'august, awe-inspiring, sagacious
(an attribute of indigenous Japanese deities)'. The kasiwamoti cakes become involved here
as religious offerings to the gods (*'august, reverent rice-cakes' > 'oak leaf-wrapped rice-
cakes') by a semantic process closely parallel to that seen in the development of J kasiwade
'clapping the hands together at the beginning of Shinto worship' (cf. J te 'hand'), which
last clearly has no connection (except in the traditional script) with kasiwa 'oak tree'. A
further level of etymological complication resulted when KJ *kasyi as represented in OJ
kasi was early specialized in the sense of the evergreen Quercusmyrsinae-folia, while kasiwa
(where the suffix is probably OJ Fa 'leaf'; cf. MK kalap, for which Martin notes 'probably a
variant of kal-(n)iph "oak leaf" ') was specialized for the deciduous Q. dentata, the leaves
OLD JAPANESEPHONOLOGYAND KOREAN-JAPANESERELATIONSHIP 291
of whichareused for wrappingthe cakes.To be kept totally separatehereis non-standard

J kasiwa 'chicken'.
157. ONE. OJ Fito-tu (Ono 263).

160. OYSTER.OJ kaki (in Kojiki; see DGK).
163. PERSIMMON.If, as Martin suggests, J kaki does include -ki 'tree', then
this goes back to OJ -ki (IR 14). But the etymology is difficult on other scores
as well; Japanese scholars have tried to find in it an apheretic allomorph of aka
'be red'.
166. PIGWEED.OJ Fiyu (WM).
167. laPLACE.OJ toko 'raised alcove' (Ono 255).
169. 2PLACE.OJ toki 'time when' (Ono 254).
171. 1PLENTIFUL.OJ oFo-ki(-) 'be big' (Ono 236).
174. POLISHIT. OJ tog- (M).
175. PUT ASIDE. OJ nokor- 'leave, leave behind' (Ono 260, BSS 10.243). The
problem of the alternation of J n- and initial zero in roots from original *n- has
already been noted above under 66; if the solution suggested there is to be in-
voked here (though even then the reasons for the particular alternation here
would remain obscure), we might use OJ n6kor- as the basis for postulating OJ
ok- 'put', agreeing exactly with the Kojiki form Mabuchi attempts to reconstruct
on the basis of the written records. Still another doublet which might be cited
here is J yok- 'avoid, shun (it)', going with nok(e)- 'leave, finish', to which is also
to be compared the vocalization in the Korean cognates cited and the recon-
structed form for J nigas- 'release it' in 181 below.
179. REDBEAN. Again, J aduki may eventually involve OJ -ki 'tree'. Cf.
PERSIMMON.
180. 1RELEASE. OJ nig(e)- 'escape' (Ono 258).
181. 2RELEASE.See 175 above.
183. REPAY. OJ motiFor- 'revert' (Ono 269). The etymology here involves
OJ motW'basis, root', and the -d- in J modos-, modor-, etc., is secondary.
184. RICEPLANT. Martin compares K pye < MK pye 'riceplant; kernel of
rice' and OJ Fo 'ear of grain' to reconstruct KJ *pYe. But OJ Fo appears in the
earliest texts (Kojiki poem 30 = Nihon shoki poem 22; Kojiki poem 41 =
Nihon shoki poem 34) in contexts which suggest that its use in the meaning 'ear
of grain' is only a secondary or even tertiary semantic development from an
earlier and far more general sense of anything projecting or jutting out-specifi-
cally, in the texts cited, a peak of land or a promontory-and this does not argue
very effectively in favor of the comparison. Nor do Mabuchi's suggestions
(1957:88) about a possible philological basis for distinguishing Fo and Fo in the
early texts add much on the phonological level, since in Kojiki 30 = Nihon
shoki 22 the word is written with a graph which he would read as Fo, but in
Kojiki 41 = Nihon shoki 34 with a graph which he would read as Fo.
It is, however, still possible to add something to Martin's comparison from our knowledge
of OJ phonology. Martin remarks, 'the Japanese word for 'riceplant' is ine; if evidence can
be found that this is an aberrant development from Proto-Japanese *wine, rather than the
yine implied by the traditional orthography, it can be directly compared with K pye by
assuming pye < pgy (the common northern form) and ... [KJ] *vyinye. If the initial high
vowel is prothetic, however, we could compareJ ine to K ipssal < MK ni-psol 'rice (grains)'
..., [KJ] *nye.' The traditional orthography makes no provision for the sequence yi-, and
does not necessarily imply its existence in this form; but nevertheless a /y/ initial here is
well authenticated by internal reconstruction (Miller 1964:14), and the first-syllable vocal-
ization of the relevant o-grade ablaut form OJ yone is attested in WM. The Nihon shoki
poem 34 has the combining form ina- in inagara 'stalk of a riceplant', which is evidence for
reconstructing a final -ne in all these forms, one of the cases where internal
reconstruction shows earlier traces of a more complete distribution of -e than that revealed
in the written records. But OJ y6ne < yone and yine < yine both appear to have originated
as conscious taboo-avoidance remodelings of an original J sine < sine 'id.', which must in
its turn somehow be related to the old doublet tane, sane 'seed' (in the modem language tane
survives as the normal term for 'seed', and sane as an extreme vulgarism for 'clitoris'). OJ
sine appears as a combining form of *yine in a few compounds registered in early texts, e.g.
migisine 'fine rice', arasine 'rough rice'. The remodeling, which avoided homophony with
sin- 'to die', is paralleled in yosi 'reed, bulrush' (a plant of considerable economic signifi-
cance in early Japan) remodeled from asi 'id.' to avoid homophony with asi- 'be evil'. The
combining form sine probably also survives in uruti 'nonglutinous rice' which may be inter-
preted as a somewhat irregular but by no means unprecedented contraction < *urusine
(again reminiscent of the t-, s- variation in the tane, sane doublet). Martin also comments
usefully upon Otsuki's highly speculative attempt (DGK 1.333d) to derive ine by contrac-
tion from a form which Otsuki cites as iFine (Martin's ipi-ne); this form is an important one
for the entire comparison and deserves a somewhat more elaborate treatment. It is a hapax
legomenon attested (according to DGK loc. cit., and DGK 2.11a, the only two rubrics under
which I can locate it) only in the early Japanese herbal generally cited as the Honzo Wamyo,
a text which is probably to be dated ca. 918; this document knows it only as a term meaning
'feed for birds'. Otsuki's speculation hinges upon associating the iFi- of this iFine with OJ
iFi 'steamed or boiled grain'. There are serious problems in all this (presumably the seman-
tic link between 'feed for birds' and 'steamed or boiled grain' is to be sought in something
like 'mash'), and it is difficult indeed to make much of the historical significance of a hapax,
particularly one appearing in such a specialized context. But be that as it may, and perhaps
improperly setting aside the semantic problems, the form iFine itself presents no particular
obstacle to postulating a prior *iwine as its phonological predecessor-in fact, such a postu-
lation has substantial parallels in the development of other well-attested words at the same
time in the history of the language. Most impressive is the parallel presented by the suc-
cessive developments of early OJ motiwir- 'use, employ', a compound consisting of mot-i-,
an enlargement of mot- 'have, hold', and wir-, a doublet-form with wor- 'be, exist; be in a
state of doing something'. This etymology unequivocally establishes the priority of the
forms with intervocalic -w-. But in the last part of the 10th and the early part of the llth
century, forms in which this -w- had already shifted to -F- are attested. Wenck (1959:4.89,
?943.4) cites motiFir- from 1020; and these secondary shifted forms become the point of
departure for the later generalized literary forms based upon motiF-. Wenck also is able
(loc. cit.) to cite examples of word-play between forms based upon motiF-i- and motiFi
(> J moti) 'glutinous rice' in the second official poetic anthology compiled in 951. Other
forms which he cites showing developments parallel to those postulated here for iFine <
*iwine are OJ awi 'indigo', written aFi in a gloss to the Saddharmapundarikadated in
correspondenceas 1136, and OJ kurawi 'degree, station, status', written kuraFi in the WM.
For the loss of the initial i- in *iwine (which thus actually does become prothetic, as Martin
speculated, but for somewhat different reasons!) there are abundant parallels: imada 'not
yet' alongside later mada 'id.'; imokasa 'smallpox' alongside mokasa; ibari 'urine' alongside
bari; idas-, ide-r- 'put out, come out' alongside das-, de-r-; iduko > idoko 'where?' alongside
doko; ina 'nay, how much [more] ...' alongside na; and still others (many involving dialect
or philological difficulties) in Wenck (1959:4.176, ?980a). One way in which this initial i-
came to fall and disappear is shown by the cases which Wenck cites (loc. cit.) from 11th
century texts of word-play between ima ya 'now, indeed' and uma ya 'stable', where the
two forms are to be related by an intermediate stage in which the initial vowels were first
nasalized and then completely assimilated to the following consonant: [im-, fim-] > [m-,
m-]. To sum up, the form *wine, which goes well with the Korean evidence and supports
the reconstruction KJ *vyinye, is if anything more easily substantiated from our knowledge
of OJ than the form *yine. The earliest form which can be recovered from the Japanese
materials is *iwine, which survives though substantially altered both phonetically and
semantically in the early 10th century hapax citation iFine. Both KJ *vyinye and KJ *nye
can be traced in the Japanese evidence; and both OJ *yine and yone are to be discarded as
having little if any comparative significance, though in other respects they are valuable
forms.
185. RIDE.Are the Japanese cognates here eventually to be connected with the
root appearing in OJ nobor- 'climb up upon' (Ono 260)? Martin reminds me
(letter of 16 May 1966) that J nobor-is usually taken to go with K noph 'be high'
and also sometimes associated with Ainu nupuri 'hill'.
190. SEA.The etymology involving OJ wata 'sea' with MK palol can be some-
what regularized by putting together instead MK palol and OJ watar- 'to cross
over, go over'; then OJ wata can go instead directly with K pata (though this
still leaves the problem of the apparently aberrant initials). On watar-/watas-
see also Martin's 177 RECEIVE.
191. SEABREAM. OJ taFi.
196. SHEPHERD'S-PURSE. OJ natuna (WM), naduna (JK). Note the secondary
voicing sometime between these two sources.
198. 1SIDE.If J soba 'side, vicinity' actually involves the deictic J so- 'that' +
-ba 'place', then it goes back to OJ so.
200. SIMILAR.The voicing in J -goto- is secondary; the form goes back to OJ
koto (Ono 244). Ono's gloss on M26 (in Ono, Takagi and Gomi 1957:25) has an
important note on the Altaic cognates of this form.
203. SLOW.OJ oso- 'be slow'. Once again the problem is that of the reflexes of
an original *n- (cf. 66, 175, and 181 above).
204. SMELL.It is unwise to involve J kag- < OJ kak- 'smell it' in this etymol-
ogy; they should both probably rather be taken as loan-words from Middle
Chinese xiav 'incense, fragrance' (Karlgren 1940, ?717a).
205. SOAK. OJ s6m(e)- 'dye it' is traditionally associated with sim-u 'steep,
soak up', which item actually agrees better with the proposed Korean cognates
both in sense and in form.
207. SOFT.OJ moro- 'be fragile, brittle' is apparently registered in a lexical
source of the period 1124-25, s.v. J moro-si- in DGK 4.642c.
209. SOUND. OJ otW (Ono 236).
212. SPADE. OJ saFi.
216. SPITTLE.
OJ tuFaki (WM, JK), which contrasts with OJ tuFaki 'camellia'
(WM, JK).
222. STINKING.OJ kuso 'dung' (WM).
227. STUPID.The same late lexical source cited in 207 above has OJ oroka.
It is not clear if J oroka 'trifling' is to be kept distinct from this word; this was
also OJ oroka.
231. SURPLUS.OJ nokor- (Ono 260); cf. under 175 above. OJ noke-r- in BSS
5.241 is an aberrant development in that text, with -e- as a result of a contraction
of -i- + -a-.
233. SWELL.OJ Futo- 'fat' (Ono 264).
294 LANGUAGE,VOLUME43, NUMBER 1 (1967)
235. SWIM.The semantics of the proposed etymology involving Korean he- <
hey- 'swim' and J kog- 'row' are difficult; the Japanese form goes back to OJ
kog-am- 'id.', but presents many problems quite apart from Martin's comparison.
It has generally been treated as a secondarily voiced counterpart of kak- 'scratch;
scrape back and forth', but this is probably little more than a phonetically so-
phisticated folk etymology. Also somehow involved here is OJ kayi 'rudder'
which eventually is to be traced to an old loan from some form cognate with
Turkish qayiq 'boat', Oirot qayiq 'oar' (cf. Sinor 1961:163ff.); the proto-Korean-
Japanese were probably far more at home on the water than they were in it. On
reconstructing OJ kayi rather than kai, Murayama (107) remarks, 'ein kai hatte
lautgesetzlich kd [i.e. OJ ke] ergeben miissen'.
237. THAT. OJ so(-re) 'that (near you)'. Cf. 198 above.
239. THING. OJ koto (Ono 234). The OJ 6/a ablaut variation, 'the most com-
mon in the language' (Ono 1953b:49), makes possible the association of OJ
k6t6 and OJ katar- 'tell'.
240. THIS. OJ kd(-re) 'this' (Ono 244).
242. 1TIME. OJ toki 'time when' (Ono 254).
244. TORTOISE. OJ kame (Ono 240).
246. TWIST. OJ yor- (Ono 273).
247. TWO. The vocalization of OJ ture 'companion' can be recovered on the
basis of the membership of the verb tur-u 'to accompany' in that class of OJ
verbs with -e- in certain forms (cf. fn. 14).
250. VAST. J obi-tadasi- 'be extreme(ly many)', attested as early as texts of the
second half of the 10th century, is the result of an irregular process of develop-
ment but nevertheless somehow related to the morpheme otherwise generally
appearing as J oo- 'be many' < OJ oFo- 'great'; in obi- the OJ intervocalic -F-
has been voiced rather than, as regularly happened, disappearing in the course
of the development of the long vowels of Middle Japanese.
251. WAIST. OJ k6si (Ono 244), contrasting with OJ kos-i- 'pass it through
(Ono 243).
253. 2WALK. OJ kayoF- 'go to and fro'. But J aruk- cannot be directly used in
comparisons, since it is a later changed form with secondary -u- vocalization;
OJ had arik-.
256. WASHTUB. OJ taraFi.
257. WATER. OJ had both mitu and midu; but OJ moi 'drinking water'
should be corrected to OJ moFi.
261. WIDE. OJ n6b(i)- 'stretch, grow'.
265. YEAR. OJ tosi.
Martin's 'Supplementary list' (his 266 through 320) includes a number of
etymologies which are for one reason or another to be treated with greater than
normal caution. To avoid piling questions up on questions, OJ data involving this
group of comparisons have not been introduced into the discussion of sound
correspondences which concludes this paper; however, since this 'Supplementary
list' contains a number of important words, I complete these remarks by giving
here the relevant OJ information for these also.
266. ABOVE. OJ uFe (IR 14; Miller 1967:186, 188, 201).

269. ARM. OJ Fiti 'elbow' (WM, JK). The voicing in J hidi is secondary.
271. BASKET. OJ Fako (WM).
273. BOTTOM. OJ moto 'base, root, origin' (Ono 269); cf. 183 above, which
seems to be a better place for the Japanese cognate.
274. CLEAN. OJ kiyo- 'clear, clean, transparent' (M).
276. DANDRUFF. Does this involve OJ ke 'hair, fur' (SM)?
277. DEEP INSIDE.OJ soCk'bottom, depths' (Ono 249).
282. FILBERT. OJ Fasi-bami (WM).
285. FLOUR. OJ ko (WM).
291. HEART. OJ kokoro (Ono 243).18
293. JUmip.OJ t6b- 'jump, fly' (Ono 255).
295. LIVER. OJ kimo (WM).
302. PAN.OJ nab6 (WM).
305. 3PLACE. OJ kI-ko 'here', s6-kd there' (Ono 243, 249); OJ koro 'approxi-
mate time' (M).
307. PLATE. Japanese sara 'plate' is usually taken as a loan from some Indic
form going with Skt. sardva 'shallow cup, dish, plate'; but perhaps the loan was
not directly into Japanese but into proto-KJ.
309. REAR. OJ ato (BSS 1.240 and passim). Ono 230 has OJ ato, but this is in
a passage (Nihon shoki poem 96) with tor- for t6r-; cf. below.
310. RICECAKE. For the vocalization of the OJ form, both OJ sitoki (JK) and
sitoki (WM) may be cited, but either makes it clear that the -g- of J sitogi is
secondary.
315. SNAKE. A variety of taboo-avoidance and other folklore factors have
intervened to complicate the history of this word. Japanese scholars generally
consider the first member of the doublet hemi, hebi to be the older historically;
BSS 19.246 has Fimi; WM registers both OJ Febi and OJ Femi; and OJ mi 'tail;
serpent (in the animal cycle)' is also probably involved in the tangle of forms and
meanings.
318. VACANT. OJ Fima 'opening, (spare) time' (WM).
319. WHEAT. The -g- of J mugi is secondary; OJ has muki (WM) but the form
(or perha'ps, though less likely, the proto-KJ form) is difficult to disassociate
from Middle Chinese mwek 'wheat' (Karlgren ?932a-c).
320. YAWN. OJ akubi (JK).
The principal implication of this OJ data for Martin's reconstruction of KJ
is the fact that it now becomes possible to 'fill out' the patterns of distribution
which Martin reconstructs for combinations of the KJ vowels with KJ *y and
18 This is one of the few OJ wordsfor which there has been a serious
attempt in the
literatureto take accounton the comparativelevel of the contrastbetweenOJ o and o.
Hattori (1948:131,fn. 57) noted that 'ShichiroMuragamiattemptedthe followingcom-
parisonat a lecture...: Mongoliank6ok- "breast",Turkic kioki-z,Japanesekoko-ro[i.e.
OJ koko-ro] "heart"; Mong. (ene)ku, Japanese ko [i.e. OJ ko-] "this".' The comparisons of
Murayama(108),in whichhe attemptsto correlateOJ o with proto-Tungusic*o, but OJ o
with proto-Tungusic *o, are a further refinement of this same general attempt, and one is
tempted to speculate that Hattori's 'Muragami'is simply a translator's slip for 'Murayama'.
*w. The following combinations occur in Martin's system:

yi
ye yo
ya yo
wa wo
The OJ data make it possible to postulate, in addition, the KJ combinations
*wi, *we, and *wo, which virtually complete the possibilities of occurrence in the
patterns immediately above. These new combinations account for the following
correspondences involving OJ:
1. KJ *...wi > K ...i, OJ ...z, as in BIRD, Korean sd < say 'bird', OJ sagi
'heron', KJ *sagwi (Martin reconstructs *sagi). This contrasts with KJ *...i >
K ...i, OJ ...i, as in SEABREAM, K tomi, OJ taFi, KJ *twombi. KJ *i regularly
gives K i, OJ i, as in PIGWEED *pirfm, DEEP *kipxa-, LEG (OJ 'shin') *palgyi,
and FULL *micx-. KJ *...yi also regularly gives OJ i, as in OYSTER *kwalgyi,
SPADE *salpyi, and TIME *ce/okyi. When OJ i is found in this environment, it
represents instead KJ *wi, so that MOON,K tal < tol, OJ tuki, must be recon-
structed *talgwi,not *tolgyi. OJ i also occurs in the reflexes of some but not all of
those words for which Martin reconstructs KJ *ya > K i, J i; thus, FIRE, K
pul < pil, OJ Fi; and FOG,K hili- 'be cloudy', OJ -giri 'fog' with non-etymolog-
ical -g- for original -k-. Apparently giving a different reflex as OJ i is WATER,
K mul < mil, OJ mitu, midu. Here the problem is complicated by difficulties
with the history of the vocalization of the Korean form, on which see Martin,
loc. cit.; perhaps these problems also mean that it is unwise at present to take
the difference here between OJ i and i too seriously. If the contrast is a valid one
after all, we can account for it by reconstructing FIRE as *pywal and FOG as
*xywari, with KJ *ywa giving K i, OJ i, but WATER as *myaldu with Martin,
with KJ *ya giving K i, OJ i.19
2. KJ *e > K e, OJ 6 was the usual and most ordinary development, attested
in many sets of cognate words, including ENJOY, l^PLACE, THING, 1TIME, WAIST,
LESS, ACCUMULATE, SURPLUS, SWIM, THAT, and WIDE. But we must also recon-
struct the sequence KJ *we, which gave K e, OJ o, in order to account for NIGHT,
K ce-nyek 'evening', OJ yo 'night', KJ *jwe;O2FERTILE, K keIlm 'fertilizer' <
19 If such sequences as KJ *pyw- appear 'strange' or 'unlikely' in shape, it is not perhaps
completely irrelevant to recall that among the Middle Chinese syllables (as reconstructed
by Karlgren) used in the man'yogana to transcribe the sequences OJ Fi, Fi, bi, and bi the
following are found: *pjwi, *b'jwij, *pjwig, *p'jw6i, *pjwei, and *b'jwj; among those used
to transcribe the sequence OJ mi are found *mjwi and *mjwig; and among those used for
OJ mi, Middle Chinese *mjwwiwas the most common. (These materials may now be exam-
ined in great detail in Wenck's Band II, Die Phonetik der Manyogana, or in the far less
complicated but extremely useful conspectus in Yoshitake 1934:22 ff., from which I have
selected the Middle Chinese forms cited here.) The proposed KJ reconstructions are of
course completely independent of this Chinese data, and their sequential configurations
were developed quite apart from Karlgren's reconstructions, but it is certainly interesting
and perhaps even significant to note how closely Karlgren's phonetic reconstructions of the
Middle Chinese morphemes used in the man'yogana transcriptions of OJ correspond to our
phonemic formulations for the KJ unity out of which OJ grew.
20 Since the second morpheme in K ce-nyek 'evening' is from nyekh 'direction', it is im-
possible not to speculate on some connection between this formation and J yuugata 'even-
kel- 'be fertile', OJ koy- 'get fat, flavorful', KJ *kwer-; and FILTER IT, K keli-,
OJ kos-, KJ *kweS-.KJ *ye regularly gave OJ e, whether in the environment
(Martin's 16a) where the Korean reflex is i or in the environment (Martin's
16b) where the Korean reflex is zero; BAMBOO is an example of the former, and
for the latter BRUSH, CLAW, FIELD, LIQUOR, TORTOISE, and MOUNTAIN may be
cited. But KJ *ywe is necessary in order to account for the reflex as OJ e in HOUSE,
K cip, OJ iFe < *yiFe, KJ *jipywe; and FRONT, Kaph < alph, OJ maFe, KJ
*alpxy we.
3. KJ *o > K o, OJ 6 was the usual and most ordinary development, attested
in many sets of cognate words, including ALTHOUGH, 2TIME, 3PLACE, DIRECTION,
PUT ASIDE, THAT, THIS, TWIST, and YEAR. But we must also reconstruct the
sequence KJ *wo, which gave K o, OJ o, in order to account for FAT, MK pulo-
'be inflated', OJ Futo- 'be fat', KJ *pulwo-. Another pattern of regular develop-
ment is probably indicated by the correspondence KJ *yo > K o, OJ i, in BODY,
K mom, OJ mi, although it is dangerous to generalize on the basis of this single
example. Also apparently following a regular pattern of development is the
combination KJ *yo in the case of DIRT, K hilk, dial. halk < MK holk 'dirt', OJ
kita-, 'be dirty', KJ *xyola; and also in CLEAN, K ka < MK kay- (?) < *koy-
'clear up', OJ kiyo- 'be clean', KJ *kyo- (this last is however from Martin's
'Supplementary list').
4. KJ *5 regularly developed into OJ 6; the Korean reflexes are extremely
complicated, chiefly K a < o, e < o, and i(/a) < o. For OJ 6 from KJ *6 we may
cite SIMILAR, RIDE, 2PLACE, ENJOY, LATE, ONE, SOAK, and THIS. But we must also
reconstruct the sequence KJ *wo in order to account for the development as OJ
o in HOLD, K til-, OJ tor- 'hold, take in the hand'; and also for such relatively
late forms as OJ tori 'bird', in WM, going with K talk < MK tolk, KJ *tw6rkyi
'chicken', leaving Martin's original construction KJ *t6rkyi to account for the
otherwise generally attested and relatively earlier OJ tori 'id.', as for example
throughout M and the Nihon shoki.
In this paper OJ has been treated as if it were a perfectly uniform language,
but of course this was not the case. The last two sets of comparisons raise a num-
ber of interesting problems that involve the relationship of the standard OJ
ing' < yuFu + kata, where the final element is kata 'direction'. Unraveling all
the etymological connections here poses several major unsolved problems, but what seems
to have happened, in general terms, is that an early *nok 'direction' developed in two differ-
ent ways, either in two different languages or in two different dialects, and probably at two
different times, giving in the first set of changes KJ *yok going with K ccok 'direction',
(c)c?/,,k'in a row', and OJ yoko 'side, horizontal' (Martin's 64 DIRECTION);in the second set
of changes the original *n- initial developed differently before different vocalization, result-
ing ultimately in K nyekh 'direction' etc. The forms in the set represented by K ccok etc.
also, in still a third set of developments, underwent metathesis which resulted in K kyeth <
kySth,kyech 'side', alongside K keth, kecuk < MK kech 'face, surface, outside', cognate with
J kata 'direction', all going back to *k(Y)atx(a), *k(Y)acx(a) (Martin's 199 "SIDE). The
differences in vocalization here can also be traced in Japanese; with J kata compare J
koti-ra 'this side, direction; here', which probably incorporates the deictic OJ ko- 'this',
and goes back to the same etymon but in the shape *k(w)etx(i). Perhaps the same semantic
process, 'night' + 'direction' > 'evening', is to be traced in J yuube, non-standard Tokyo
yumbe 'last night', by identifying the -be element with OJ Fe 'direction, neighborhood'.
dialect, in particular that of the Nara area of government and cultural prestige,
with other non-standard OJ dialects. Japanese scholars generally refer to the
non-standard varieties of OJ as the Azuma dialects, with the term azuma < OJ
aduma 'east'. A major bundle of ancient dialect isoglosses separated a 'Western'
variety of OJ, the standard language, from this 'Eastern' variety, following
quite closely the modern dividing line between Shizuoka and Nagano Pre-
fectures on the one hand and Aichi and Gifu Prefectures on the other. Our
principal sources for those non-standard Azuma or 'Eastern' dialects are books
14 and 20 of the Man'yosha, which record a considerable number of poems in the
local OJ dialects originally found to the east of this line. A good deal of modern
Japanese scholarship has been devoted to the analysis of these written records,
and as a result we probably know more about some of the dialect distinctions in
8th century OJ than we do about many similar ones in some of the surviving
modern non-standard dialects.
For the verb which is the OJ cognate of K til- 'hold in the hand', the texts
show forms based on two different vocalizations of the verb base, OJ tor- and
tor-. In the Man'yoshi the form tor-i-, for example, occurs seven times and the
form t6r-i- twenty-six times; and tor-u does not appear to be registered in this
text at all, while t6r-u is found there eight times (Masamune 1931:714-7). In
the poems of the Nihon shoki the following forms occur:2'
with vocalization tor- with vocalization tor-
tori (42 [=50]) tori (35 [cf. 43])
itoramu (43 [ = 51]) t6re (51)
tori (96, twice) torase (51, twice)
-dori (96, twice) torasane (60)
isanatori (68)
torame (108)
torasu (108)
The non-standard Azuma or 'Eastern' dialect correspondences of OJ o and
OJ 6 have been studied in considerable detail by Fukuda,22and his findings help
21 The numbersin
parenthesesfollowingeach formare those of the Nihon shokipoems
as cited, for example,in Ono, and also in Tsuchihashiand Konishi 124-214.The textual
context and originalman'yoganatranscriptionfor each citation may be easily verifiedin
that edition. Glosses have been omitted since only the forms, not their meanings,are
relevantto the discussion,andbecausean attemptto translatethe citationsout of context
wouldin most casesinvolveconsiderablephilologicalexegesis.Numbersin squarebrackets
identify Nihon shokipoemsappearingalso in the Kojiki, with their numbersin that text
(ibid.,34-122);the equalsign is usedfor poemsappearingthereverbatim,and 'cf.' for close
textual parallels.Ono'sremarkson tor-/tor-in M 2.430are also important.
22 Most recently and in great detail by Fukuda (1965:273ff.,305-6). On the OJ dialects in
general see also Fukuda 1962.He is the 'Y. Hukuda' who is the author of the paper in English
(Fukuda1954),which attemptsto show that muchlate OJ sound-changewas a result of
contact betweenthe Azumaand the standarddialects; I know this work only from its
unsympatheticnotice in Haguenauer(640). Yokoyama (1950:83-6)discusses the mor-
phologyof the non-standardOJ texts in the Man'yosha;but since she transcribedall her
materialswith the five-vowelsystem of J, it is difficultto correlateher analysiswith the
languageof the texts.
us to understand something of the historical and dialect background of this

situation, which on the descriptive level could otherwise only be written off as
free variation. Fukuda shows that in the texts, standard OJ o corresponds to
Azuma 6 in five cases and to Azuma i in seven cases, and that while there is
virtually no confusion of standard OJ o and o in the texts, in those poems which
reflect the Azuma dialects 12 of 64 cases of words which would have OJ o in the
standard language turn up with 6 in the Azuma texts, while of 278 cases of words
which would have 6 in the standard language, only 19 turn up witho in the Azuma
dialects. In other words, as Fukuda puts it, approximately twenty percent of all
cases of standard OJ o are confused with 6 in Azuma, as against only two per-
cent of the cases of standard OJ o confused with o. On the basis of his exhaustive
investigation of the sources, he concludes that 6 was the vowel that 'charac-
terized' the Azuma dialects, which tended to level both standard OJ o and 6
under their 6. To quote Fukuda (1965:274), this Azuma version of 6, whatever
it may have been in precise articulatory terms (a subject upon which he wisely
refrains from speculating), was the 'vowel which formed the basis of the vocaliza-
tion of the Azuma dialects (azuma hoogenno boin sosiki o koosei suru boin de atta)'.
All this does much to explain the variation in our texts between the tor- and
tor- forms of the root 'hold in the hand', a variation which in historical terms we
may now treat in terms of different forms representing separate stages of histori-
cal development within proto-Korean-Japanese, original KJ *twor- and secon-
dary KJ *t6r-. When later lexical sources such as the WM show OJ tori 'bird'
for earlier tdri, they simply reflect the beginning of the eventual collapse of the
older eight-vowel system, and do not point in the direction of an earlier *wJ
alongside (or instead of) an earlier *5. But with the verb 'hold in the hand' the
situation is different. Those portions of the Man'y6shii composed in the Azuma
dialect have only forms with tor- (M 3420, 3459, 3485). The preponderance of
forms in -6- against forms in -o- in the balance of this text shows the intrusion of
Azuma forms into the standard OJ language of the other portions of the Man'-
yoshu already well underway; the same is also true of the forms cited above from
the Nihon shoki. The Azuma forms built upon OJ tor- go back to KJ *t5r-,while
the standard OJ forms built upon tor- go back to KJ *twor-. Somewhere in
certain areas of pre-OJ the sequence *tw6-was simplified into the sequence *to-,
the earlier and more complex one ultimately producing the standard language
forms, the later and less complex one resulting in the Azuma dialect forms.
Such simplifications appear to have played a role of some importance in the
development of OJ, and other evidence may be cited which makes it clear that
changes of this sort underlie many of the important developments in the early
history of the language. These changes are also probably related in some way to
the KJ phoneme which Martin reconstructs as *Y, representing the corre-
spondence between K y... and J #... following a consonant in both languages; of
this *Y, Martin says that it 'may merely represent sporadic palatalization
(through dialect mixture?)', and cites such doublets as K myek, mok 'neck'.
The attempt by Mabuchi 1957 to bring together evidence for postulating the
OJ i/i distinction following OJ s and the OJ 0/6 distinction following OJ F
and # (or initial?) in the phonetically written OJ portions of the Kojiki has
already been noted. Unfortunately a large proportion of the words for which he
is able to cite convincing philological evidence for earlier OJ /si/ are place-
names, personal names, or god names, and hence they are virtually impossible to
verify or even to consider on the comparative level. For a number of other words
he finds a good deal of uncertainty on the part of the Kojiki scribes, who tend to
waver between /si/ and /si/. Among the cases where the records show scribal
indecision between i and z are the forms sita 'lower side', si-ga 'self', isi 'rock',
asi 'foot', and the verbal suffixal sequences -r-as-i 'seems to be' and -m-as-i
'if [it] were so [but it is unlikely]', which are most often written with /si/; sir6
'age, generation, period', kasi 'oak', siFo 'salt', and sise- 'die', evenly divided be-
tween i and i;23and sisi 'wild boar', which is written twice as sisi, twice as sisi.
Finally, sid-u- 'hanging down', siba 'brushwood, firewood', and sima 'island',
are written sid-u-, siba, and sima more often than they are written sid-u-, siba,
and sima.24For BRUSHWOOD Martin constructs KJ *syibxa on the basis of K
seph < MK sip and J siba. For ISLANDhe reconstructs KJ *sYyima, with the
'sporadic palatalization' phoneme *Y, on the basis of K sim < MK syem and J
sima. The forms OJ siba and sima can be accounted for by the reconstructions
*swibxa and *sYwima, and it is easy to see how subsequent changes, parallel to
those postulated above for *tw6rkyi'bird', led to the development of the forms
siba and sima. An original *swibxa, which underlies the form siba of the Kojiki,
first underwent assimilation of its initial sequence to become *syibxa, the form
underlying the siba of the other OJ written records. In the word for ISLANDthe
changes were parallel, but complicated by the presence of the 'sporadic' *Y,
whose 'sporadicity' probably resulted from its own assimilation to the following
semi-vowel; *sYwima > OJ sima and MK syem, but after assimilation giving
*sYyima > OJ sima, and then after further assimilation, in which *Yy simpli-
fied to *y, becoming *syima which regularly produced K sem. It is interesting to
note that two other words in which Martin reconstructs this same *Y appear
following KJ *p- which developed into OJ F- -in other words, in one of the other
environments for which from the Kojiki materials Mabuchi attempts to recon-
struct the OJ o/6 distinction; but unfortunately none of these three words, STAR,
BONE, and RICEPLANT, is attested in sufficiently antique glosses to permit involv-
ing the OJ written records on the level of its lexical comparison with Korean.
REFERENCES
ARISAKAHIDEYO.
1944. Kokugoon'inshino kenkyu. Tokyo, Sanseid6.(Rev. ed., 1957.)
- . 1955.Jodai on'inko.Tokyo, Sanseido.
BAILEY,DONC. 1960. Early Japaneselexicography.MonumentaNipponica16.1-52.
DICKINS,F. VICTOR.1908. The makura-kotoba of primitiveJapaneseverse. Transactions
of the Asiatic Society of Japan 35.4:1-113.
FIUKDAYOSMSUKE.1954. A study of the formationof the Easterndialect in the Nara
'8 Mabuchi 1957:76; I reproduce the data from his tabular summary on the top of this
page, omitting proper names and god names, and also the forms kitasi and ofo-gitasi (sic),
for which I am unable to supply a translation gloss.
24Ibid. On OJ sima < sima, see also Mabuchi, 80.
era, with special referenceto its phonetic phenomena.Bulletin of the Faculty of

Literature,KyushuUniversity,2.1-19.
- . 1962. Kodai Nihon no hogenken.Kokubungakukaishakuto kansh6 1962:2.25-35.
. 1965. Nara jidai Azumahogenno kenkyu.Tokyo, Fukan shobo.
FuKnYAMA TOSHIO. 1934. Eta hakkutsudaito oyobi Suda Hachimanjinja kagami no
seisakunendai ni tsuite. K6kogakuzasshi 24.1.
GUSINDE, MARTIN,and CHIYESANO.1960. Stone circlesin northernJapan. Anthropos
55.441-55.
HAGUENAUER, CHARLES. 1956. Origines de la civilisation japonaise, Introduction a l'6tude
de la prehistoire du Japon, premiere partie. Paris, Imprimerie Nationale.
HALL,JOHN.1960. Review of The location of Yamatai: a case study in Japanese histori-
ography,720-1945, by John Young. JAOS80.281-2.
HAMADA ATSUSHI. 1900. Gishiwajindennado ni shokenno kokugogoi ni kansurufutatsu
mittsu no mondai.Jinbunkenkyu3:8.
HASHIMOTO ZOKICHI. 1956. Toyoshij6yori mitaru Nihon jokoshi kenkyi. (Toyo bunko
ronso,dai-38.)Tokyo, TSoy bunko.
HATTORI SHIRO.1948. The relationshipof Japaneseto the Ryukyu, Koreanand Altaic
languages.Transactionsof the Asiatic Society of Japan, 3rd series, 1.101-33.
--. 1955. Nihongo.II. On'in.Sekai gengogaisetsu,ed. by IchikawaSanki and Hattori
Shiro, 2.153-60. Tokyo, Kenkyusha.
--. 1961. Prosodeme,syllablestructureand laryngealphonemes.Studiesin descriptive
and appliedlinguistics1.1-27.
HOUSEHOLDER, FREDW. 1965. Review of Trendsin modernlinguistics,ed. by Christine
Mohrmann,F. Norman,and Alf Sommerfelt.Lg. 41.308-12.
[IWABUCHIETSUTARO.]1963. Okinawago jiten. (Kokuritsu kokugo kenkyujo shiryoshii, 5.)
Tokyo, Okurash6insatsukyoku.
Izui HISANOSUKE. 1956.Jodai Nihongoni okeruboin soshikito boin k6tai. Kyotodaigaku
bungakubugojushunenkinen ronshu989-1020.
----. 1963. Recent trends in Japaneselinguistics.Trends in modernlinguistics,ed. by
Christine Mohrmann,F. Norman, and Alf Sommerfelt,pp. 38-55. Utrecht and
Antwerp,Spectrum.
KARLGREN, BERNHARD. 1940. Grammataserica: Script and phonetics in Chinese and
Sino-Japanese.Stockholm,Museumof Far EasternAntiquities.
KAROW,OTTO. 1951. Die W6rterbuicherder Heianzeitund ihre Bedeutungftir die japa-
nischeSprachgeschichte.MN 7.156-97.
KATO GENCHI. 1924. A study of the development of religious ideas among the Japanese
people as illustratedby Japanese phallicism.TASJ, 2nd series, supplementto v.
1.1-70.
KAYAMOTOMORITO. 1955. Kofun jidai ni okeru kinsekibun. (Nihon k6kogaku koza, 5.)
Tokyo.
LEWIN, BRUNO.1959. Abrissder japanischenGrammatik,auf der Grundlageder klassi-
schen Schriftsprache.Wiesbaden,Otto Harrassowitz.
MABUCHI KAzUO.1957. Kojiki no shi, o, ho no kana. Kokugogaku31.61-90.
- . 1959.Jodai,chukoni okerusagyotoon no onka. Kokugoto kokubungaku36:1.60-7.
-- . 1962-63. Nihon ongakushino kenkyi. 2 vols. Tokyo, Nihon gakujutsushinkokai.
MARTIN, SAMUEL E. 1959. Review of Japanische Phonetik, by Guinther Wenck. Lg.
35.370-82.
-- . 1966. LexicalevidencerelatingKoreanto Japanese.Lg. 43.185-251.
MASAMUNEATSUO.1931. Man'yoshf sosakuin, tangohen. Tokyo, Hakusuisha.
MATSUOKA SmZUO.1962. Nihon kogo daijiten. Tokyo, Toko shoin. (Rev. ed., Nihon
kogo jiten.)
MCCAWLEY, JAMES D. 1966. Review of Papers of the CIC Far EasternLanguageInsti-
tute, The Universityof Michigan,ed. by Joseph K. Yamagiwa.Lg. 42.170-5.
MILLER,ROY ANDREW.1964. The reconstruction of the Japanese vowels. Papers of the

CIC Far Eastern Language Institute, Indiana University, 5-18.
- . 1967. The Japanese language. Chicago, University of Chicago Press.
MILLS,DONALDE. 1960. The Buddha's footprint stone poems. JAOS 80.229-42.
MIzuNo SEIICHIand KOBAYASHI YUKIO.1959. Zukai kokogaku jiten. Tokyo, Sogensha.
MURAYAMA SHICHIR6. 1962. Etymologie des altjapanischen Wortes ir6 'Farbe, Gesichts-
farbe, Gesicht'. Ural-Altaische Jahrbiicher 34.107-12.
NOMURAMASAYOSHI.1955. Mokogo, ?4, Sushi gokon no hikaku. Sekai gengo gaisetsu,
ed. by Ichikawa Sanki and Hattori Shirs, 2.583.4. Tokyo, Kenkyusha.
OMORISHIRO.1955a. Jodai kanazukai kara mita Yamataikoku no suitei. (Tokyo joshi
daigaku ronshf, 5:2.)
-- . 1955b. Gishi wajinden no kenkyu. Tokyo, Hobunkan.
ONOSUSUMU.1953a. Jodai kanazukai no kenkyu. Tokyo, Iwanami shoten.
-- . 1953b. Nihongo no doshi no katsuyokei no kigenni tsuite. Kokug to kokuobungaku
350.47-56.
- . 1960. Nara jidai no seion, dakuon ni tsuite. Man'yoshf (Nihon koten bungaku
taikei, 3), pp. 42-6. Tokyo, Iwanami shoten.
- , NAKADANORIO,Doi TADAO,YOSHIDASUMIO,and MATSUMURA AKIRA. 1961.
Nihongo no rekishi. 6th ed. Tokyo, Shibundo.
, TAKAGIICHINOSUKE, and GOMITOMOHIDE. 1957-62. Man'yoshiu. (Nihon koten
bungaku taikei, 4-7). Tokyo, Iwanami shoten.
OTSUKIFUMIHIKO. 1932-37. Daigenkai. Vols. 1-5. Tokyo, Fuzanbo.
POPPE,NIKOLAUS. 1955. Introduction to Mongolian comparative studies. (Memoires de
la Societ6 Finno-Ougrienne, 110.) Helsinki, Suomalais-Ugrilainen Seura.
-- . 1960. Vergleichende Grammatik der altaischen Sprachen, Teil 1, Vergleichende
Lautlehre. (Porta linguarum orientalium, neue Serie, 4.) Wiesbaden, Otto Harrasso-
witz.
RAMSTEDT, G. J. 1952. Einfiihrung in die altaische Sprachwissenschaft, II, Formenlehre,
bearbeitet und herausgegeben von Pentti Aalto. (M6moires de la Soci6et Finno-
Ougrienne, 104:2.) Helsinki, Suomalais-Ugrilainen Seura.
SINOR,DENIS. 1961. On water-transport in Central Eurasia. Ural-Altaische Jahrbiicher
33.163-79.
TSUCHIHASHI YUTAKAand KONISHIJIN'ICHI. 1957. Kodai kayoshu. (Nihon bungaku
koten taikei, 3.) Tokyo, Iwanami shoten.
TSUNODARYNSAKU,tsl., and L. CARRINGTON GOODRICH, ed. 1951. Japan in the Chinese
dynastic histories. (Perkins Asiatic monographs, 2.) South Pasadena, P.D. and Ione
Perkins.
WENCK,GUNTHER.1954-59. Japanische Phonetik, Band I, Die Lautlehre des modernen
Japanischen; Die Geschichte des Lautbewusstseins und der Lautforschung in Japan;
Die Quellen der japanischen Lautgeschichte; Band II, Die Phonetik der Manyogana;
Band III, Die Phonetik des Sinojapanischen; Band IV, Erscheinungen und Probleme
des Japanischen Lautwandels. Wiesbaden, Otto Harrassowitz.
YOKOYAMA, MASAKO.1950. The inflections of 8th century Japanese. (Language disserta-
tions, 45.) Baltimore, Linguistic Society of America.
YOSHITAKE, S. 1930. Etymology of the Japanese word fude. BSO[A]S 6:1.45-53.
-- . 1934. The phonetic system of ancient Japanese. (James G. Furlong Fund, vol. 12.)
London, The Royal Asiatic Society.
YOUNG,JOHN.1958. The location of Yamatai: a case study in Japanese historiography,
720-1945. (Johns Hopkins University, studies in historical and political science, ser.
XXV, 2.) Baltimore.

Miller 1967

Uploaded by

Copyright:

Available Formats

Miller 1967

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Miller 1967

Uploaded by

Copyright:

Available Formats

Linguistic Society of America

Old Japanese Phonology and the Korean-Japanese Relationship

ROY ANDREW MILLER

The distribution of the vowels with respect to the consonants in OJ is of far

considerable diachronic depth, since virtually every citation of Korean evidence

their J cognates is that of the sporadic occurrence of intervocalic voiced con-

2. ACCUMULATE. OJ mor- 'heap them up' (Ono 270) OJ moromoro'all, both,

possiblecorrelationbetweenthe location of the pitch accent in J and the appearanceof

53. CORRECT. OJ mato 'target' (WM).

78. FERTILE. OJ koy-(e)- 'get fat, flavorful' (JK).

of whichareused for wrappingthe cakes.To be kept totally separatehereis non-standard

157. ONE. OJ Fito-tu (Ono 263).

266. ABOVE. OJ uFe (IR 14; Miller 1967:186, 188, 201).

*w. The following combinations occur in Martin's system:

us to understand something of the historical and dialect background of this

era, with special referenceto its phonetic phenomena.Bulletin of the Faculty of

MILLER,ROY ANDREW.1964. The reconstruction of the Japanese vowels. Papers of the

You might also like