Jump to content

Old Japanese

From Wikipedia, the free encyclopedia

Old Japanese
Rubbing of a stone with eleven columns of standard Chinese characters
Rubbing of Bussokuseki-kahi poems carved c. 752, recording Old Japanese using Chinese characters
Era8th century
  • Old Japanese
Early form
Language codes
ISO 639-3ojp
ojp [a]
This article contains IPA phonetic symbols. Without proper rendering support, you may see question marks, boxes, or other symbols instead of Unicode characters. For an introductory guide on IPA symbols, see Help:IPA.

Old Japanese (上代日本語, Jōdai Nihon-go) is the oldest attested stage of the Japanese language, recorded in documents from the Nara period (8th century). It became Early Middle Japanese in the succeeding Heian period, but the precise delimitation of the stages is controversial. Old Japanese was an early member of the Japonic language family. No genetic links to other language families have been proven.

Old Japanese was written using man'yōgana, using Chinese characters as syllabograms or (occasionally) logograms. It featured a few phonemic differences from later forms, such as a simpler syllable structure and distinctions between several pairs of syllables that have been pronounced identically since Early Middle Japanese. The phonetic realization of these distinctions is uncertain. Internal reconstruction points to a pre-Old Japanese phase with fewer consonants and vowels.

As is typical of Japonic languages, Old Japanese was primarily an agglutinative language with a subject–object–verb word order, adjectives and adverbs preceding the nouns and verbs they modified and auxiliary verbs and particles appended to the main verb. Unlike in later periods, Old Japanese adjectives could be used uninflected to modify following nouns. Old Japanese verbs had a rich system of tense and aspect suffixes.

Sources and dating[edit]

Two pages of a manuscript, with the main text in standard characters and annotations in a cursive style
11th-century annotated manuscript of the Man'yōshū

Old Japanese is usually defined as the language of the Nara period (710–794), when the capital was Heijō-kyō (now Nara).[1][2] That is the period of the earliest connected texts in Japanese, the 112 songs included in the Kojiki (712). The other major literary sources of the period are the 128 songs included in the Nihon Shoki (720) and the Man'yōshū (c. 759), a compilation of over 4,500 poems.[3][4] Shorter samples are 25 poems in the Fudoki (720) and the 21 poems of the Bussokuseki-kahi (c. 752). The latter has the virtue of being an original inscription, whereas the oldest surviving manuscripts of all the other texts are the results of centuries of copying, with the attendant risk of scribal errors.[5] Prose texts are more limited but are thought to reflect the syntax of Old Japanese more accurately than verse texts do. The most important are the 27 Norito ('liturgies') recorded in the Engishiki (compiled in 927) and the 62 Senmyō (literally 'announced order', meaning imperial edicts) recorded in the Shoku Nihongi (797).[4][6]

A limited number of Japanese words, mostly personal names and place names, are recorded phonetically in ancient Chinese texts, such as the "Wei Zhi" portion of the Records of the Three Kingdoms (3rd century AD), but the transcriptions by Chinese scholars are unreliable.[7] The oldest surviving inscriptions from Japan, dating from the 5th or early 6th centuries, include those on the Suda Hachiman Shrine Mirror, the Inariyama Sword, and the Eta Funayama Sword. Those inscriptions are written in Classical Chinese but contain several Japanese names that were transcribed phonetically using Chinese characters.[8][9] Such inscriptions became more common from the Suiko period (592–628).[10] Those fragments are usually considered a form of Old Japanese.[11]

Of the 10,000 paper records kept at Shōsōin, only two, dating from about 762, are in Old Japanese.[12] Over 150,000 wooden tablets (mokkan) dating from the late 7th and early 8th century have been unearthed. The tablets bear short texts, often in Old Japanese of a more colloquial style than the polished poems and liturgies of the primary corpus.[13]

Writing system[edit]

Artifacts inscribed with Chinese characters dated as early as the 1st century AD have been found in Japan, but detailed knowledge of the script seems not to have reached the islands until the early 5th century. According to the Kojiki and Nihon Shoki, the script was brought by scholars from Baekje (southwestern Korea).[14] The earliest texts found in Japan were written in Classical Chinese, probably by immigrant scribes. Later "hybrid" texts show the influence of Japanese grammar, such as the word order (for example, the verb being placed after the object).[15]

Chinese and Koreans had long used Chinese characters to write non-Chinese terms and proper names phonetically by selecting characters for Chinese words that sounded similar to each syllable. Koreans also used the characters phonetically to write Korean particles and inflections that were added to Chinese texts to allow them to be read as Korean (Idu script). In Japan, the practice was developed into man'yōgana, a complete script for the language that used Chinese characters phonetically, which was the ancestor of modern kana syllabaries.[16] This system was already in use in the verse parts of the Kojiki (712) and the Nihon Shoki (720).[17][18]

For example, the first line of the first poem in the Kojiki was written with five characters:[19][20]

Middle Chinese[b] yae kjuw maw ta tu
Old Japanese ya-kumo1 tatu
eight-cloud rise.ADN
'many clouds rising'

This method of writing Japanese syllables by using characters for their Chinese sounds (ongana) was supplemented with indirect methods in the complex mixed script of the Man'yōshū (c. 759).[21][22][23]


In man'yōgana, each Old Japanese syllable was represented by a Chinese character. Although any of several characters could be used for a given syllable, a careful analysis reveals that 88 syllables were distinguished in early Old Japanese, typified by the Kojiki songs:[24][25]

Syllables in early Old Japanese, with common man'yōgana[c]
a ka 加,迦 ga sa za ta da na pa ba ma ya ra wa
i ki1 gi1 si 斯,志 zi ti di ni 爾,迩 pi1 bi1 mi1 ri wi
ki2 gi2 pi2 bi2 mi2
u ku gu su zu tu du nu pu bu mu yu ru
e ke1 ge1 se 勢,世 ze te de ne pe1 be1 me1 ye re we
ke2 ge2 pe2 be2 me2
o 淤,意 ko1 go1 胡,呉 so1 zo1 俗,蘇 to1 do1 no1 po 富,本 bo mo1 yo1 ro1 漏,路 wo 袁,遠
ko2 go2 so2 zo2 to2 do2 no2 mo2 yo2 余,與 ro2

As in later forms of Japanese, the system has gaps where yi and wu might be expected. Shinkichi Hashimoto discovered in 1917 that many syllables that have a modern i, e or o occurred in two forms, termed types A (, ) and B (, otsu).[24][26] These are denoted by subscripts 1 and 2 respectively in the above table. The syllables mo1 and mo2 are not distinguished in the slightly later Nihon Shoki and Man'yōshū, reducing the syllable count to 87.[27][28] Some authors also believe that two forms of po were distinguished in the Kojiki.[29] All of these pairs had merged in the Early Middle Japanese of the Heian period.[30][31]

The consonants g, z, d, b and r did not occur at the start of a word.[32] Conversely, syllables consisting of a single vowel were restricted to word-initial position, with a few exceptions such as kai 'oar', ko2i 'to lie down', kui 'to regret' (with conclusive kuyu), oi 'to age' and uuru, the adnominal form of the verb uwe 'to plant'.[33][34] Alexander Vovin argues that the non-initial syllables i and u in these cases should be read as Old Japanese syllables yi and wu.[35]

Frequencies of Old Japanese syllables in the Man'yōshū[36]
- k- g- s- z- t- d- n- p- b- m- y- r- w-
-a 4612 7616 3358 3473 255 5212 734 5891 6450 2195 6018 3184 4213 2581
-i1 3679 5771 762 8070 350 2195 335 7101 3489 585 5818 3901 270
-i2 690 404 756 140 589
-u 1556 4855 444 2507 904 4417 1065 1449 2905 389 2692 2190 3656
-e1 45 1145 13 1220 210 2831 727 1425 1101 203 318 644 2598 342
-e2 1011 489 959 287 1406
-o1 2441 1995 138 536 8 485 269 583 1870 75 7788 871 215 3657
-o2 3407 436 1206 122 5848 882 9618 1312 1177

The rare vowel i2 almost always occurred at the end of a morpheme. Most occurrences of e1, e2 and o1 were also at the end of a morpheme.[37]

The mokkan typically did not distinguish voiced from voiceless consonants, and wrote some syllables with characters that had fewer strokes and were based on older Chinese pronunciations imported via the Korean peninsula. For example,

  • ki1 was written with the character , pronounced *kje in Old Chinese and tsye in Middle Chinese, and
  • to2 was written with the character , pronounced *tjəʔ in Old Chinese and tsyi in Middle Chinese.[38]


Several different notations for the type A/B distinction are found in the literature, including:[39][40][41]

Common notations for the type A/B distinction
index notation i1 i2 e1 e2 o1 o2
Kindaichi, Miller, Tōdō i ï e ë o ö
Vovin i ï e ɛ o ə
modified Mathias–Miller î ï ê ë ô ö
Yale (Martin) yi iy ye ey wo
Unger, Frellesvig and Whitman i wi ye e wo o


There is no consensus on the pronunciation of the syllables distinguished by man'yōgana.[42] One difficulty is that the Middle Chinese pronunciations of the characters used are also disputed, and since the reconstruction of their phonetic values is partly based on later Sino-Japanese pronunciations, there is a danger of circular reasoning.[43] Additional evidence has been drawn from phonological typology, subsequent developments in the Japanese pronunciation, and the comparative study of the Ryukyuan languages.[44]


Miyake reconstructed the following consonant inventory:[45]

Old Japanese consonants
Labial Coronal Palatal Velar
Obstruent voiceless *p *t *s *k
voiced prenasalized *ᵐb *ⁿd *ⁿz *ᵑɡ
Nasal *m *n
Liquid *r
Approximant *w *j

The voiceless obstruents /p, t, s, k/ had voiced prenasalized counterparts /ᵐb, ⁿd, ⁿz, ᵑɡ/.[45] Prenasalization was still present in the late 17th century (according to the Korean textbook Ch'ŏphae Sinŏ) and is found in some Modern Japanese and Ryukyuan dialects, but it has disappeared in modern Japanese except for the intervocalic nasal stop allophone [ŋ] of /ɡ/.[46] The sibilants /s/ and /ⁿz/ may have been palatalized before e and i.[47]

Comparative evidence from Ryukyuan languages suggests that Old Japanese p reflected an earlier voiceless bilabial stop *p.[48] There is general agreement that word-initial p had become a voiceless bilabial fricative [ɸ] by Early Modern Japanese, as suggested by its transcription as f in later Portuguese works and as ph or hw in the Korean textbook Ch'ŏphae Sinŏ. In Modern Standard Japanese, it is romanized as h and has different allophones before various vowels. In medial position, it became [w] in Early Middle Japanese and has since disappeared except before a.[49] Many scholars argue that p had already lenited to [ɸ] by the Old Japanese period, but Miyake argues that it was still a stop.[50]


The Chinese characters chosen to write syllables with the Old Japanese vowel a suggest that it was an open unrounded vowel /a/.[51] The vowel u was a close back rounded vowel /u/, unlike the unrounded /ɯ/ of Modern Standard Japanese.[52]

Several hypotheses have been advanced to explain the A/B distinctions made in man'yōgana. The issue is hotly debated, and there is no consensus.[39] The traditional view, first advanced by Kyōsuke Kindaichi in 1938, is that there were eight pure vowels, with the type B vowels being more central than their type A counterparts.[53] Others, beginning in the 1930s but more commonly since the work of Roland Lange in 1968, have attributed the type A/B distinction to medial or final glides /j/ and /w/.[54][40] The diphthong proposals are often connected to hypotheses about pre-Old Japanese, but all exhibit an uneven distribution of glides.[40]

Examples of reconstructions of type A/B distinctions[39]
i1 i2 e1 e2 o1 o2 Author
i wi e we wo o Kikusawa (1935)
i ï e ë o ö Kindaichi (1938), Miller (1967)
i ïj e əj o ə Arisaka (1955)
ji i je e o ɵ Hattori (1958)
ji i je e wo o Lange (1968, 1973)
i wi je e wo o Unger (1977), Frellesvig and Whitman (2008)
i ï e ɛ o ɵ Ōno (1982)
i ɨ e əj o ə Miyake (2003)

The distinction between mo1 and mo2 was seen only in Kojiki and vanished afterwards. The distribution of syllables suggests that there may have once been *po1, *po2, *bo1 and *bo2.[28] If that was true, a distinction was made between Co1 and Co2 for all consonants C except for w. Some take that as evidence that Co1 may have represented Cwo.[citation needed]


Although modern Japanese dialects have pitch accent systems, they were usually not shown in man'yōgana. However, in one part of the Nihon Shoki, the Chinese characters appeared to have been chosen to represent a pitch pattern similar to that recorded in the Ruiju Myōgishō, a dictionary that was compiled in the late 11th century. In that section, a low-pitch syllable was represented by a character with the Middle Chinese level tone, and a high pitch was represented by a character with one of the other three Middle Chinese tones. (A similar division was used in the tone patterns of Chinese poetry, which were emulated by Japanese poets in the late Asuka period.) Thus, it appears that the Old Japanese accent system was similar to that of Early Middle Japanese.[55]


Old Japanese words consisted of one or more open syllables of the form (C)V, subject to additional restrictions:

  • Words did not begin with r or the voiced obstruents b, d, z, and g, with the exception of a few loanwords.[56]
  • A bare vowel did not occur except for word-initially: vowel sequences were not permitted.[33]

In 1934, Arisaka Hideyo proposed a set of phonological restrictions permitted in a single morpheme. Arisaka's Law states that -o2 was generally not found in the same morpheme as -a, -o1 or -u. Some scholars have interpreted that as a vestige of earlier vowel harmony, but it is very different from patterns that are observed in, for example, the Turkic languages.[57]


Two adjacent vowels fused to form a new vowel when a consonant was lost within a morpheme, or a compound was lexicalized as a single morpheme. The following fusions occurred:

i1 + ae1
  • saki1 'bloom' + ari 'exist' → sake1ri 'be blooming'[58][59]
  • ki1 'wear' + aru 'be.ATT' → ke1ru 'wear.FIN'[60]
Further examples are provided by verbs ending with the retrospective auxiliary -ki1 and the verbal suffixes amu 'conjecture' or ari 'exist':[61]
  • tir-i-ki1 'fall, scatter.INF.RET' + am-utirike1mu '(it) has surely fallen'[62]
  • ari-ki1 'exist.INF.RET' + ar-iarike1ri 'it existed'[62]
i1 + o2e1
  • utusi 'real' + oyomi1 'person' → utuse1mi1 'living person'[63][d]
a + ie2
  • naga 'long' + iki1 'breath' → nage2ki1 'sigh'[58]
  • taka 'high' + iti 'market' → take2ti (place name)[58][64]
o2 + ie2
  • tono 'palace' + iri 'enter' → toneri 'attendant'[58][60]
o2 + ii2
  • opo 'big' + isi 'rock' → opi2si 'big rock'[58][60]
u + ii2
  • waku 'young' + iratuko1 'term of veneration (male)' → waki2ratuko1 (title)[58][60]
u + ao1
  • kazu 'number' + ape2 'to join' → kazo1pe2 'to count'[59][62]
u + oo1
  • situ 'ancient type of native weaving' + ori 'weaving' → sito1ri 'native weaving'[58][59]

Adjacent vowels belonging to different morphemes, or pairs of vowels for which none of the above fusions applied, were reduced by deleting one or other of the vowels.[65] Most often, the first of the adjacent vowels was deleted:[66][67]

  • to2ko2 'eternal' + ipa 'rock' → to2ki1pa 'eternal rock; everlasting'[68][69]
  • ama 'heaven' + ori 'descend' → amori 'descend from heaven'[70]

The exception to this rule occurred when the first of the adjacent vowels was the sole vowel of a monosyllabic morpheme (usually a clitic), in which case the other vowel was deleted:[66][67]

  • mi1 (honorific) + uma 'horse' → mi1ma 'honourable horse'[71]
  • ko1 'child, egg' + umu 'birth' → ko1mu 'give birth, lay an egg'[72]

Cases where both outcomes are found are attributed to different analyses of morpheme boundaries:[66][69]

  • waga 'my' + ipe1 'house' → wagi1pe1 'my house'
  • wa 'I' + ga GEN + ipe1 'house' → wagape1 'my house'

Pre-Old Japanese[edit]

Internal reconstruction suggests that the stage preceding Old Japanese had fewer consonants and vowels.[73]


Internal reconstruction suggests that the Old Japanese voiced obstruents, which always occurred in medial position, arose from the weakening of earlier nasal syllables before voiceless obstruents:[74][75]

  • b /ᵐb/ < *-mVp-, *-nVp-: e.g. ami1 'net' + pi1ki1 'pull' → abi1ki1 'trawling'
  • d /ⁿd/ < *-mVt-, *-nVt-: e.g. yama 'mountain' + mi1ti 'path' → yamadi 'mountain path'
  • z /ⁿz/ < *-mVs-, *-nVs-: e.g. mura 'village' + nusi 'master' → murazi (title)
  • g /ᵑɡ/ < *-mVk-, *-nVk-

In some cases, such as tubu 'grain', kadi 'rudder' and pi1za 'knee', there is no evidence for a preceding vowel, which leads some scholars to posit final nasals at the earlier stage.[56]

Some linguists suggest that Old Japanese w and y derive, respectively, from *b and *d at some point before the oldest inscriptions in the 6th century.[76] Southern Ryukyuan varieties such as Miyako, Yaeyama and Yonaguni have /b/ corresponding to Old Japanese w, but only Yonaguni (at the far end of the chain) has /d/ where Old Japanese has y:[77]

  • ba 'I' and bata 'stomach' corresponding to Old Japanese wa and wata
  • Yonaguni da 'house', du 'hot water' and dama 'mountain' corresponding to Old Japanese ya, yu and yama

However, many linguists, especially in Japan, argue that the Southern Ryukyuan voiced stops are local innovations,[78] adducing a variety of reasons.[79]

Some supporters of *b and *d also add *z and *g, which both disappeared in Old Japanese, for reasons of symmetry.[80] However, there is very little Japonic evidence for them.[56][81]


As seen in § Morphophonemics, many occurrences of the rare vowels i2, e1, e2 and o1 arise from fusion of more common vowels. Similarly, many nouns having independent forms ending in -i2 or -e2 also have bound forms ending in a different vowel, which are believed to be older.[82] For example, sake2 'rice wine' has the form saka- in compounds such as sakaduki 'sake cup'.[82][83] The following alternations are the most common:

  • i2/u-: kami2/kamu- 'god, spirit',[82][83] mi2/mu- 'body',[82][84] nagi2/nagu- 'a calm'.[84] tuki2/tuku- 'moon', kuki2/kuku- 'stalk'.[85]
  • i2/o2-: ki2/ko2- 'tree',[82][83] yomi2/yomo2- 'Hades',[82]
  • e2/a-: me2/ma- 'eye',[82] ame2/ama- 'heaven', ame2/ama- 'rain', kage2/kaga- 'shade',[86] ke2/ka- 'day, sun', tume2/tuma- 'nail, hoof', take2/taka- 'bamboo'.[85]

The widely accepted analysis of this situation is that the most common Old Japanese vowels a, u, i1 and o2 reflect earlier *a, *u, *i and *ə respectively, and the other vowels reflect fusions of these vowels:[87]

  • i2 < *ui, *əi
  • e1 < *ia, *iə
  • e2 < *ai
  • o1 < *ua, *uə

Thus the above independent forms of nouns can be derived from the bound form and a suffix *-i.[82][83] The origin of this suffix is debated, with one proposal being the ancestor of the obsolescent particle i (whose function is also uncertain), and another being a weakened consonant (suggested by proposed Korean cognates).[88]

There are also alternations suggesting e2 < *əi, such as se2/so2- 'back' and me2/mo2- 'bud'.[82] Some authors believe that they belong to an earlier layer than i2 < *əi, but others reconstruct two central vowels *ə and *ɨ, which merged everywhere except before *i.[63][89] Other authors attribute the variation to different reflexes in different dialects and note that *əi yields e in Ryukyuan languages.[90]

Some instances of word-final e1 and o1 are difficult to analyse as fusions, and some authors postulate *e and *o to account for such cases.[91] A few alternations, as well as comparisons with Eastern Old Japanese and Ryukyuan languages, suggest that *e and *o also occurred in non-word-final positions at an earlier stage but were raised in such positions to i1 and u, respectively, in central Old Japanese.[92][93] The mid vowels are also found in some early mokkan and in some modern Japanese dialects.[94]


As in later forms of Japanese, Old Japanese word order was predominantly subject–object–verb, with adjectives and adverbs preceding the nouns and verbs they modify and auxiliary verbs and particles consistently appended to the main verb.[95] Japanese at all stages has used prefixes with both nouns and verbs, but Old Japanese also used prefixes for grammatical functions later expressed using suffixes.[96] This is atypical of SOV languages, and may suggest that the language was in the final stage of a transition from a SVO typology.[96][97]



Many Old Japanese pronouns had both a short form and a longer form with attached -re of uncertain etymology. If the pronoun occurred in isolation, the longer form was used. The short form was used with genitive particles or in nominal compounds, but in other situations either form was possible.[98]

Personal pronouns were distinguished by taking the genitive marker ga, in contrast to the marker no2 used with demonstratives and nouns.[99]

  • The first-person pronouns were a(re) and wa(re), were used for the singular and plural respectively, though with some overlap. The wa- forms were also used reflexively, which suggests that wa was originally an indefinite pronoun and gradually replaced a.[99]
  • The second-person pronoun was na(re).[100]
  • The third-person pronoun si was much less commonly used than the non-proximal demonstrative so2 from which it was derived.[101]
  • There were also an interrogative pronoun ta(re) and a reflexive pronoun ono2.[100]

Demonstratives often distinguished proximal (to the speaker) and non-proximal forms marked with ko2- and so2- respectively. Many forms had corresponding interrogative forms i(du)-.[102]

Old Japanese demonstratives[103]
Proximal Non-proximal Interrogative
Nominal ko2(re) so2 idu(re)
Location ko2ko2 so2ko2 iduku
Direction ko2ti so2ti iduti
Degree ko2kV- so2kV- iku-
Manner ka sate
kaku sika ika
Time itu

In Early Middle Japanese, the non-proximal so- forms were reinterpreted as hearer-based (medial), and the speaker-based forms were divided into proximal ko- forms and distal ka-/a- forms, yielding the three-way distinction that is still found in Modern Japanese.[104]


In later texts, such as the Man'yōshū, numerals were sometimes written using Chinese logographs, which give no indication of pronunciation.[105] The following numerals are attested phonographically:[106]

Phonographically attested Old Japanese cardinal numerals
1 pi1to2 10 to2wo 100 mo1mo1 1000 ti 10,000 yo2ro2du
2 puta 20 pata
3 mi1 30 mi1so1
4 yo2 40 yo2so1
5 itu 50 (iso1) 500 ipo
6 mu
7 nana 70 (nanaso1)
8 ya 80 yaso1 800 yapo
9 ko2ko2no

The forms for 50 and 70 are known only from Heian texts.[107]

There is a single example of a phonographically recorded compound number, in Bussokuseki 2:[108]











mi1so-ti amar-i puta-tu no2 katati

thirty-CL exceed-INF two-CL GEN mark

'thirty-two marks'

This example uses the classifiers -ti (used with tens and hundreds) and -tu (used with digits and hundreds).[109]

The only attested ordinal numeral is patu 'first'.[110] In Classical Japanese, the other ordinal numerals had the same form as cardinals, and this may also have been the case for Old Japanese, though there are no textual occurrences to settle the question.[111]


Old Japanese nominals had suffixes or particles to mark diminutives, plural number and case. When multiple suffixes occurred, case markers came last.[112]

Unmarked nouns (but not pronouns) were neutral as to number.[113] The main plural markers were the general-purpose ra and two markers restricted to animate nouns, do2mo2 (limited to five words) and tati.[114]

The main case particles were[115]


Old Japanese had a richer system of verbal suffixes than later forms of Japanese.[118] Old Japanese verbs used inflection for modal and conjunctional purposes.[119] Other categories, such as voice, tense, aspect and mood, were expressed by using optional suffixed auxiliaries, which were also inflected.[120]

Inflected forms[edit]

As in later forms of Japanese, Old Japanese verbs had a large number of inflected forms. In traditional Japanese grammar, they are represented by six forms (katsuyōkei, 活用形) from which all the others may be derived in a similar fashion to the principal parts used for Latin and other languages:[121]

Mizenkei (irrealis)
This form never occurs in isolation but only as a stem to which several particles and auxiliaries are attached.[122] This stem originated from resegmentation of an initial *a of several suffixes (auxiliary verbs) as part of the stem.[123][124]
Ren'yōkei (adverbial, infinitive)
This form was used as the infinitive.[125] It also served as a stem for auxiliaries expressing tense and aspect.[126]
Shūshikei (conclusive, predicative)
This form was used as the main verb concluding a declarative sentence.[119] It was used also before modal extensions, final particles, and some conjunctional particles.[127] The conclusive form merged with the attributive form by about 1600, but the distinction is preserved in the Ryukyuan languages and the Hachijōjima dialects.[128]
Rentaikei (attributive, adnominal)
This form was used as the verb in a nominalized clause or a clause modifying a noun.[129] It was also used before most conjunctional particles.[130]
Izenkei (realis, exclamatory, subjunctive)
This form was used as the main verb in an exclamatory sentence or as the verb in an adverbial clause.[131] It also served as a stem for the particles ba (provisional) and do (concessive).[132]
Meireikei (imperative)
This form expressed the imperative mood.[131]

This system has been criticized because the six forms are not equivalent, with one being solely a combinatory stem, three solely word forms, and two being both.[133] It also fails to capture some inflected forms.[134] However, five of the forms are basic inflected verb forms, and the system also describes almost all extended forms consistently.[135]

Conjugation classes[edit]

Old Japanese verbs are classified into eight conjugation classes that were originally defined for the classical Japanese of the late Heian period. In each class, the inflected forms showed a different pattern of rows of a kana table. These rows correspond to the five vowels of later Japanese, but the discovery of the A/B distinction in Old Japanese showed a more refined picture.[136]

Three of the classes are grouped as consonant bases:[137]

Yodan (quadrigrade)
This class of regular consonant-base verbs includes approximately 75% of verbs.[137] The class is so named because the inflections in later forms of Japanese span four rows of a kana table, corresponding to four vowels. However, the discovery of the A/B distinction revealed that this class actually involved five different vowels in Old Japanese, with distinct vowels e1 and e2 in the exclamatory and imperative forms respectively.[136] The bases are almost all of the form (C)VC-, with the final consonant being p, t, k, b, g, m, s or r.[138]
Na-hen (n-irregular)
The three n-base verbs form a class of their own: sin- 'die', -in- 'depart' and the auxiliary -(i)n- expressing completion of an action. They are often described as a "hybrid" conjugation because the adnominal and exclamatory forms followed a similar pattern to vowel-base verbs.[139]
Ra-hen (r-irregular)
The irregular r-base verbs were ar- 'be, exist' and other verbs that incorporated it, as well as wor- 'be sitting', which became the existential verb or- in later forms of Japanese.[140]
Conjugation of consonant-base verbs[141]
Verb class Irrealis Infinitive Conclusive Adnominal Exclamatory Imperative Gloss
quadrigrade kaka- kaki1 kaku kaku kake2 kake1 'write'
n-irregular sina- sini sinu sinuru sinure sine 'die'
r-irregular ara- ari ari aru are are 'be, exist'

The distinctions between i1 and i2 and between e1 and e2 were eliminated after s, z, t, d, n, y, r and w.

There were five vowel-base conjugation classes:

Shimo nidan (lower bigrade or e-bigrade)
The largest regular vowel-base class ended in e2 and included approximately 20% of verbs.[138]
Kami nidan (upper bigrade or i-bigrade)
This class of bases ended in i2 and included about 30 verbs.[138]
Kami ichidan (upper monograde or i-monograde)
This class contains about 10 verbs of the form (C)i1-. Some monosyllabic i-bigrade verbs had already shifted to this class by Old Japanese, and the rest followed in Early Middle Japanese.[142]
Ka-hen (k-irregular)
This class consists of the single verb ko2- 'come'.[143]
Sa-hen (s-irregular)
This class consists of the single verb se- 'do'.[143]

Early Middle Japanese also had a Shimo ichidan (lower monograde or e-monograde) category, consisting of a single verb kwe- 'kick', which reflected the Old Japanese lower bigrade verb kuwe-.[144][145][146][147]

Conjugation of vowel-base verbs[141]
Verb class Irrealis Infinitive Conclusive Adnominal Exclamatory Imperative Gloss
e-bigrade ake2- ake2 aku akuru akure ake2(yo2) 'open'
i-bigrade oki2- oki2 oku okuru okure oki2(yo2) 'arise'
monograde mi1- mi1 mi1ru mi1ru mi1re mi1(yo2) 'see'
k-irregular ko2- ki1 ku kuru kure ko2 'come'
s-irregular se- si su suru sure se(yo2) 'do'

The bigrade verbs seem to belong to a later layer than the consonant-base verbs.[148] Many e-bigrade verbs are transitive or intransitive counterparts of consonant-base verbs.[149] In contrast, i-bigrade verbs tend to be intransitive.[150] Some bigrade bases also appear to reflect pre-Old-Japanese adjectives with vowel stems combined with an inchoative *-i suffix:[151][152][153]

  • *-a-i > -e2, e.g. ake2- 'redden, lighten' vs aka 'red'.
  • *-u-i > -i2, e.g. sabi2- 'get desolate, fade' vs sabu- 'lonely'.
  • *-ə-i > -i2, e.g. opi2- 'get big, grow' vs opo- 'big'.

Verbal auxiliaries[edit]

Old Japanese had a rich system of auxiliary elements that could be suffixed to verb stems and were themselves inflected, usually following the regular consonant-stem or vowel-stem paradigms, but never including the full range of stems found with full verbs.[154]

Tense and aspect were indicated by suffixes attached to the infinitive.[126] The tense suffixes were:

  • the simple past -ki (conclusive), -si (adnominal), -sika (exclamatory).[155][156] The variation may indicate an origin in multiple forms.[157]
  • the modal past or retrospective -ke1r-, a fusion of the simple past with ar- 'exist'.[158][159]
  • the past conjectural -ke1m-, a fusion of the simple past with the conjectural suffix -am-.[160][161]

The perfective suffixes were -n- and -te-.[162][163] The tense and aspect suffixes were replaced with a single past-tense suffix -ta, derived from -te + ar- > -tar-, during the Late Middle Japanese period.[118][164]

Other auxiliaries were attached to the irrealis stem:


Old Japanese adjectives were originally nominals and, unlike in later periods, could be used uninflected to modify following nouns.[175][176] They could also be conjugated as stative verbs and were divided into two classes:[177]

Conjugation of stative verbs[178][179]
Class Stem Infinitive Conclusive Adnominal Exclamatory Gloss
-ku kata kataku katasi kataki1 katasa 'hard'
-siku kusi kusiku kusi kusiki1 kusisa 'precious'

The second class had stems ending in -si, which differed only in the conclusive form, whose suffix -si was dropped by haplology.[180] Adjectives of this class tended to express more subjective qualities.[181] Many of them were formed from a verbal stem by the addition of a suffix -si, of uncertain origin.[182]

Towards the end of the Old Japanese period, a more expressive conjugation emerged by adding the verb ar- 'be' to the infinitive, with the sequence -ua- reducing to -a-:[177]

Innovative conjugation of stative verbs[183]
Irrealis Infinitive Adnominal Gloss
katakara- katakari katakaru 'hard'

Many adjectival nouns of Early Middle Japanese were based on Old Japanese adjectives that were formed with suffixes -ka, -raka or -yaka.[184][185]


The capital (Nara) and the eastern provinces (hatched) in the 8th century

Although most Old Japanese writing represents the language of the Nara court in central Japan, some sources come from eastern Japan:[186][187][188][189]

  • 230 azuma uta 'eastern songs', making up volume 14 of the Man'yōshū,
  • 93 (101 according to some authors) sakimori uta 'borderguard songs' in volume 20 of the Man'yōshū, and
  • 9 songs in the Hitachi Fudoki (recorded 714–718, but the oldest extant manuscripts date from the late 17th century and show significant corruption[190]).

They record Eastern Old Japanese dialects,[190] with several differences from central Old Japanese (also known as Western Old Japanese):

  • There is no type A/B distinction on front vowels i and e, but o1 and o2 are distinguished.[191]
  • Pre-Old Japanese *ia yielded a in the east, where central Old Japanese has e1.[191]
  • The adnominal form of consonant-base verbs ended in -o1, but central Old Japanese ended in -u for both the adnominal and the conclusive forms.[192] A similar difference is preserved in Ryukyuan languages, suggesting that central Old Japanese had innovated by merging those endings.[193]
  • The imperative form of vowel-base verbs attached -ro2, instead of the -yo2 used in central Old Japanese.[194] This difference has persisted into modern eastern and western dialects.[195]
  • There was a group of distinctive negative auxiliaries, but they do not seem to be the source of the different negatives in the modern eastern and western Japanese dialects.[196]
  • A significant number of words borrowed from Ainu.[197][198]

See also[edit]


  1. ^ Described as "The ancestor of modern Japanese. 7th–10th centuries AD." The more usual date for the boundary between Old Japanese to Middle Japanese is c. 800 (end of the Nara era).
  2. ^ Readings are given in Baxter's transcription for Middle Chinese, omitting marking of tones, which are not relevant here.
  3. ^ These are the characters most used in the Kojiki songs, except for go1 and zo1 (which do not occur in the Kojiki) from the Man'yōshū.
  4. ^ An alternative form, utuso2mi1, obtained by deleting the vowel -i, is also attested.[63]


  1. ^ Shibatani 1990, p. 119.
  2. ^ Miyake 2003, p. 1.
  3. ^ Miyake 2003, p. 17.
  4. ^ a b Frellesvig 2010, p. 24.
  5. ^ Miyake 2003, pp. 19–20.
  6. ^ Bentley 2001, p. 6.
  7. ^ Miyake 2003, pp. 5–8.
  8. ^ Miyake 2003, p. 10.
  9. ^ Seeley 1991, pp. 16–25.
  10. ^ Miyake 2003, p. 12.
  11. ^ Miyake 2003, p. 66.
  12. ^ Seeley 1991, pp. 55–56.
  13. ^ Frellesvig 2010, p. 22.
  14. ^ Miyake 2003, pp. 8–9.
  15. ^ Seeley 1991, pp. 25–31.
  16. ^ Shibatani 1990, p. 126.
  17. ^ Seeley 1991, pp. 41–49.
  18. ^ Miyake 2003, pp. 18–20, 28–40.
  19. ^ Miyake 2003, pp. 1, 18, 22.
  20. ^ Frellesvig 2010, p. 19.
  21. ^ Seeley 1991, pp. 49–53.
  22. ^ Miyake 2003, pp. 20, 24–27.
  23. ^ Frellesvig 2010, pp. 17–20.
  24. ^ a b Miyake 2003, pp. 49–51.
  25. ^ Frellesvig 2010, pp. 26–27.
  26. ^ Frellesvig 2010, pp. 28–29.
  27. ^ Miyake 2003, p. 51.
  28. ^ a b Frellesvig 2010, p. 30.
  29. ^ Miyake 2003, p. 61.
  30. ^ Miyake 2003, p. 84.
  31. ^ Frellesvig 2010, p. 26.
  32. ^ Frellesvig 2010, p. 34.
  33. ^ a b Frellesvig 2010, p. 39.
  34. ^ Vovin 2020, pp. 48–50.
  35. ^ Vovin 2020, pp. 49–54.
  36. ^ Frellesvig & Horn 2023.
  37. ^ Frellesvig 2010, p. 49.
  38. ^ Inukai 2024, pp. 466–468.
  39. ^ a b c Miyake 2003, p. 62.
  40. ^ a b c Frellesvig 2010, p. 32.
  41. ^ Vovin 2020, p. 45.
  42. ^ Miyake 2003, p. 2.
  43. ^ Miyake 2003, pp. 54–55, 63–64.
  44. ^ Miyake 2003, pp. 64–65.
  45. ^ a b Miyake 2003, p. 196.
  46. ^ Miyake 2003, pp. 75–76.
  47. ^ Miyake 2003, pp. 183, 186.
  48. ^ Shibatani 1990, p. 194.
  49. ^ Miyake 2003, p. 74.
  50. ^ Miyake 2003, pp. 71, 164–166.
  51. ^ Miyake 2003, pp. 198–203.
  52. ^ Miyake 2003, pp. 207–211.
  53. ^ Miyake 2003, p. 55.
  54. ^ Miyake 2003, pp. 55–57.
  55. ^ Miyake 2003, pp. 37–39.
  56. ^ a b c Frellesvig 2010, p. 43.
  57. ^ Frellesvig 2010, p. 44.
  58. ^ a b c d e f g Russell 2003, p. 532.
  59. ^ a b c Frellesvig 2010, p. 48.
  60. ^ a b c d Miyake 2003, p. 81.
  61. ^ Russell 2003, pp. 521–522.
  62. ^ a b c Russell 2003, p. 531.
  63. ^ a b c Erickson 2003, p. 499.
  64. ^ Frellesvig 2010, p. 46.
  65. ^ Russell 2003, p. 523.
  66. ^ a b c Unger 2000, p. 662.
  67. ^ a b Vovin 2020, p. 55.
  68. ^ Russell 2003, p. 528.
  69. ^ a b Frellesvig 2010, p. 40.
  70. ^ Russell 2003, p. 525.
  71. ^ Russell 2003, p. 530.
  72. ^ Russell 2003, p. 529.
  73. ^ Miyake 2003, pp. 73, 80–81.
  74. ^ Miyake 2003, p. 73.
  75. ^ Frellesvig 2010, pp. 42–43.
  76. ^ Miyake 2003, pp. 71–73.
  77. ^ Shibatani 1990, p. 195.
  78. ^ Vovin 2010, pp. 36–44.
  79. ^ Pellard 2024, pp. 45–49.
  80. ^ Unger 2000, p. 666.
  81. ^ Miyake 2003, pp. 68–71.
  82. ^ a b c d e f g h i Frellesvig 2010, p. 45.
  83. ^ a b c d Miyake 2003, p. 80.
  84. ^ a b Shibatani 1990, p. 134.
  85. ^ a b Vovin 2020, p. 85.
  86. ^ Shibatani 1990, p. 133.
  87. ^ Frellesvig 2010, pp. 44, 48.
  88. ^ Frellesvig 2010, pp. 45, 131–132.
  89. ^ Frellesvig 2010, pp. 45–47.
  90. ^ Unger 2000, p. 661.
  91. ^ Frellesvig 2010, pp. 47–48.
  92. ^ Frellesvig 2010, pp. 47–48, 153.
  93. ^ Vovin 2010, pp. 32–36.
  94. ^ Osterkamp 2017, pp. 46–48.
  95. ^ Shibatani 1990, pp. 122–123.
  96. ^ a b Frellesvig 2010, p. 79.
  97. ^ Vovin 2020, p. 530.
  98. ^ Frellesvig 2010, pp. 136–137.
  99. ^ a b Frellesvig 2010, p. 138.
  100. ^ a b Frellesvig 2010, p. 136.
  101. ^ Frellesvig 2010, pp. 138–139.
  102. ^ Frellesvig 2010, pp. 139–140.
  103. ^ Frellesvig 2010, p. 141.
  104. ^ Frellesvig 2010, pp. 140, 246–247.
  105. ^ Frellesvig 2010, p. 285.
  106. ^ Vovin 2020, pp. 328–352.
  107. ^ Vovin 2020, pp. 344–345.
  108. ^ Vovin 2020, p. 344.
  109. ^ Vovin 2020, p. 354.
  110. ^ Vovin 2020, pp. 352–353.
  111. ^ Vovin 2020, p. 352.
  112. ^ Vovin 2020, p. 109.
  113. ^ Vovin 2020, p. 110.
  114. ^ Bentley 2012, p. 195.
  115. ^ a b Frellesvig 2010, p. 126.
  116. ^ Frellesvig 2010, pp. 126–127.
  117. ^ Frellesvig 2010, p. 134.
  118. ^ a b Shibatani 1990, p. 123.
  119. ^ a b Frellesvig 2010, p. 53.
  120. ^ Frellesvig 2010, p. 59.
  121. ^ Frellesvig 2010, pp. 114–118.
  122. ^ Frellesvig 2010, pp. 111–112.
  123. ^ Frellesvig 2010, pp. 112, 120–121.
  124. ^ Unger 2000, p. 664.
  125. ^ Frellesvig 2010, pp. 56–57.
  126. ^ a b Frellesvig 2010, pp. 109–111.
  127. ^ Frellesvig 2010, pp. 123–124, 133.
  128. ^ Shibatani 1990, pp. 195, 207, 223–224.
  129. ^ Frellesvig 2010, pp. 54–55.
  130. ^ Frellesvig 2010, p. 133.
  131. ^ a b Frellesvig 2010, p. 55.
  132. ^ Frellesvig 2010, pp. 112–113.
  133. ^ Frellesvig 2010, p. 117.
  134. ^ Frellesvig 2010, pp. 113, 117.
  135. ^ Frellesvig 2010, pp. 116–118.
  136. ^ a b Frellesvig 2010, p. 115.
  137. ^ a b Frellesvig 2010, p. 96.
  138. ^ a b c Frellesvig 2010, p. 97.
  139. ^ Frellesvig 2010, p. 105.
  140. ^ Frellesvig 2010, pp. 101–103.
  141. ^ a b Frellesvig 2010, pp. 54, 114.
  142. ^ Frellesvig 2010, p. 106.
  143. ^ a b Frellesvig 2010, pp. 107–108.
  144. ^ Frellesvig 2010, pp. 227–228.
  145. ^ Yamaguchi et al. 1997, p. 18.
  146. ^ Kondō, Tsukimoto & Sugiura 2005, p. 41.
  147. ^ Omodaka 1967, pp. 37–38.
  148. ^ Frellesvig 2010, p. 120.
  149. ^ Frellesvig 2010, pp. 118–119.
  150. ^ Whitman 2008, p. 164.
  151. ^ Unger 2000, p. 665.
  152. ^ Frellesvig 2010, p. 119.
  153. ^ Whitman 2008, p. 165.
  154. ^ Frellesvig 2010, pp. 58–62.
  155. ^ Frellesvig 2010, p. 62.
  156. ^ Vovin 2020, pp. 827–830.
  157. ^ Frellesvig 2010, p. 121.
  158. ^ Frellesvig 2010, pp. 48, 121.
  159. ^ Vovin 2020, pp. 879–881.
  160. ^ Frellesvig 2010, pp. 48, 122.
  161. ^ Vovin 2020, pp. 833–834.
  162. ^ Frellesvig 2010, pp. 66–68, 123.
  163. ^ Vovin 2020, pp. 844–866.
  164. ^ Frellesvig 2010, p. 69.
  165. ^ Frellesvig 2010, pp. 69–72, 121.
  166. ^ Vovin 2020, pp. 701–713.
  167. ^ a b Frellesvig 2010, pp. 63–64.
  168. ^ Vovin 2020, pp. 745–758.
  169. ^ Vovin 2020, pp. 771–776.
  170. ^ Frellesvig 2010, pp. 62–63.
  171. ^ Vovin 2020, pp. 758–771.
  172. ^ a b Frellesvig 2010, p. 78.
  173. ^ Vovin 2020, pp. 713–730.
  174. ^ Vovin 2020, pp. 606–610.
  175. ^ Vovin 2020, pp. 373–383.
  176. ^ Frellesvig 2010, pp. 79–80.
  177. ^ a b Bentley 2012, pp. 197–198.
  178. ^ Bentley 2012, p. 198.
  179. ^ Frellesvig 2010, p. 82.
  180. ^ Bentley 2001, p. 104.
  181. ^ Frellesvig 2010, p. 90.
  182. ^ Frellesvig 2010, p. 91.
  183. ^ Bentley 2001, p. 138.
  184. ^ Frellesvig 2010, p. 235.
  185. ^ Vovin 2020, pp. 386–389.
  186. ^ Vovin 2020, pp. 11–14.
  187. ^ Miyake 2003, p. 159.
  188. ^ Frellesvig 2010, pp. 23–24, 151.
  189. ^ Vovin & Ishisaki-Vovin 2022, pp. 1–16.
  190. ^ a b Kupchik 2011, p. 1.
  191. ^ a b Frellesvig 2010, p. 152.
  192. ^ Frellesvig 2010, pp. 152–153.
  193. ^ Bentley 2012, p. 189.
  194. ^ Frellesvig 2010, p. 154.
  195. ^ Frellesvig 2010, p. 402.
  196. ^ Frellesvig 2010, p. 153.
  197. ^ Vovin 2014, pp. 13–15.
  198. ^ Vovin & Ishisaki-Vovin 2022, pp. 35–38.

Works cited[edit]

  • Bentley, John R. (2001). A Descriptive Grammar of Early Old Japanese Prose. BRILL. ISBN 978-90-04-12308-3.
  • ——— (2012). "Old Japanese". In Tranter, Nicolas (ed.). The Languages of Japan and Korea. Routledge. pp. 189–211. ISBN 978-1-136-44658-0.
  • Erickson, Blaine (2003). "Old Japanese and Proto-Japonic word structure". In Vovin, Alexander; Osada, Toshiki (eds.). Nihongo keitōron no genzai 日本語系統論の現在 [Perspectives on the Origins of the Japanese Language]. Kyoto: International Research Center for Japanese Studies. pp. 493–510. doi:10.15055/00005265. ISBN 978-4-901558-17-4.
  • Frellesvig, Bjarke (2010). A History of the Japanese Language. Cambridge: Cambridge University Press. ISBN 978-0-521-65320-6.
  • Frellesvig, Bjarke; Horn, Stephen Wright; et al., eds. (2023). "The texts". Oxford-NINJAL Corpus of Old Japanese (ONCOJ). Retrieved 5 April 2024.
  • Inukai, Takashi (2024). "What mokkan (wooden documents) can tell us about ancient Japanese language". In Frellesvig, Bjarke; Kinsui, Satoshi (eds.). Handbook of Japanese Historical Linguistics. Berlin: De Gruyter Mouton. pp. 463–472. ISBN 978-1-61451-285-1.
  • Kondō, Yasuhiro; Tsukimoto, Masayuki; Sugiura, Katsumi (2005). Nihongo no Rekishi 日本語の歴史 [A history of the Japanese language] (in Japanese). Tokyo: Hōsō Daigaku Kyōiku Shinkōkai. ISBN 978-4-595-30547-4.
  • Kupchik, John E. (2011). A grammar of the Eastern Old Japanese dialects (PhD thesis). University of Hawai'i. hdl:10125/101739.
  • Miyake, Marc Hideo (2003). Old Japanese: A Phonetic Reconstruction. London; New York: RoutledgeCurzon. ISBN 978-0-415-30575-4.
  • Omodaka, Hisataka, ed. (1967). Jidaibetsu Kokugo Daijiten: Jōdaihen 時代別国語大辞典 上代編 [Comprehensive dictionary of Japanese by historical period: ancient edition] (in Japanese). Tokyo: Sanseidō. ISBN 978-4-385-13237-2.
  • Osterkamp, Sven (2017). "A mokkan Perspective on Some Issues in Japanese Historical Phonology". In Vovin, Alexander; McClure, William (eds.). Studies in Japanese and Korean Historical and Theoretical Linguistics and Beyond. Languages of Asia. Vol. 16. Brill. pp. 45–55. doi:10.1163/9789004351134_006. ISBN 978-90-04-35085-4.
  • Pellard, Thomas (2024). "Ryukyuan and the reconstruction of proto-Japanese-Ryukyuan". In Frellesvig, Bjarke; Kinsui, Satoshi (eds.). Handbook of Japanese Historical Linguistics. Berlin: De Gruyter Mouton. pp. 39–68. ISBN 978-1-61451-285-1.
  • Russell, Kerri (2003). "Contraction and Monophthongization in Old Japanese". In Vovin, Alexander; Osada, Toshiki (eds.). Perspectives on the Origins of the Japanese Language. Vol. 31. Kyoto: International Center for Japanese Studies. pp. 511–539. doi:10.15055/00005266.
  • Seeley, Christopher (1991). A History of Writing in Japan. Leiden: BRILL. ISBN 978-90-04-09081-1.
  • Shibatani, Masayoshi (1990). The Languages of Japan. Cambridge: Cambridge University Press. ISBN 978-0-521-36918-3.
  • Unger, J. Marshall (2000). "Reconciling Comparative and Internal Reconstruction: The Case of Old Japanese /ti, ri, ni/". Language. 76 (3): 655–681. doi:10.2307/417138. JSTOR 417138.
  • Vovin, Alexander (2010). Korea-Japonica: A Re-evaluation of a Common Genetic Origin. University of Hawaii Press. ISBN 978-0-8248-3278-0.
  • ——— (2014). "Man'yōshū" (Book 20). Leiden: Brill. ISBN 978-90-04-26198-3.
  • ——— (2020). A Descriptive and Comparative Grammar of Western Old Japanese (2nd ed.). Leiden, Boston: Brill. ISBN 978-90-04-42211-7.
  • Vovin, Alexander; Ishisaki-Vovin, Sambi (2022). The Eastern Old Japanese Corpus and Dictionary. Leiden, Boston: Brill. ISBN 978-90-04-47119-1.
  • Whitman, John (2008). "The source of the bigrade conjugation and stem shape in pre-Old-Japanese". In Frellesvig, Bjarke; Whitman, John (eds.). Proto-Japanese: Issues and Prospects. John Benjamins. pp. 160–173. ISBN 978-90-272-4809-1.
  • Yamaguchi, Akiho; Suzuki, Hideo; Sakanashi, Ryūzō; Tsukimoto, Masayuki (1997). Nihongo no Rekishi 日本語の歴史 [A history of the Japanese language] (in Japanese). Tokyo: University of Tokyo Press. ISBN 978-4-13-082004-2.

Further reading[edit]

External links[edit]