Early Middle Japanese

Major phonological changes were characteristic of the period.

The most prominent difference was the loss of the Jōdai Tokushu Kanazukai, which distinguished two types of /i/, /e/ and /o/. While the start of the loss can already be seen at the end of Old Japanese, it was completely lost in Early Middle Japanese. The final distinction to be lost was /ko₁, go₁/ vs. /ko₂, go₂/.[2]

In the 10th century, /e/ and /je/ merged into /je/, and /o/ and /wo/ had merged into /wo/ by the 11th century.[3][4][5]

An increase in Chinese loanwords had a number of phonological effects:

• Palatal and labial consonant clusters such as /kw/ and /kj/
• The uvular nasal /ɴ/
• Length became a phonemic feature with the development of both long vowels and long consonants

The development of the uvular nasal and geminated consonants occurred late in the Heian period and brought about the introduction of closed syllables (CVC).[6]

• /a/: [a]
• /i/: [i]
• /u/: [u]
• /e/: [je][3][4][5]
• /o/: [wo]

Theories for the realization of /s, z/ include [s, z], [ts, dz], and [ʃ, ʒ]. It may have varied depending on the following vowel, like in Modern Japanese.

By the 11th century, /ɸ/ had merged with /w/ between vowels.[7]

Nouns occurred with postpositive case particles such as these:

• -ga and -no (genitive.)
• -wo (accusative). Optional.
• -ni (dative/locative).-ni had a wide range of functions ('to' or 'for' a person; 'by' an agent'; 'at' or 'to' a place; 'at' a time), and in some uses, especially when indicating time, it was optional.
• -yori (ablative).
• -made (terminative: 'until'; 'as far as').
• -to (comitative: 'with'; essive 'as').
• -fe (allative: 'to'). -fe w derived from the noun fe (邊 [辺] or 方) 'vicinity; direction', which wa occasionally found in the location noun structure Noun + -no + Location Noun to mean 'near', or in the noun-deriving suffix -be (< -no-fe) in such words as midube 'beside the water' or yamabe 'around the mountains'.

The nominative function was marked by the absence of a particle in main clauses and by the genitive particles in subordinate clauses. The dative/locative particle -ni was homophonous with the simple infinitive form of the copula -ni, with verbal suffixes supplies more complex case markers -ni-te ('at' a place) and -ni si-te or -ni-te ('by means of'). A number of particle + verb + -te sequences provided other case functions: -ni yori-te 'due to' (from yor- 'depend'), -ni tuki-te 'about, concerning' (from tuk- 'be attached'), and -to si-te 'as' (from se- 'do'). More complex structures were derived from genitive particle + Location Noun + appropriate case particle (typically locative -ni) and were used particularly to express spatial and temporal relations. Major location nouns were mafe 'front' (Noun-no mafe-ni 'in front of Noun'), ufe 'top' (Noun-no ufe-ni 'on top of Noun' ~ 'above Noun'), sita 'under' (Noun-no sita-ni 'under Noun), saki 'ahead' (Noun-no saki-ni 'ahead of Noun)', etc.

Early Middle Japanese verb inflection was agglutinative. All verbs were conjugated in a small number of 'stems' and could be combined with grammaticalized verbs to express tense, aspect, mood, voice, and polarity. Several of the grammaticalized verbs could combine in a string, and each component determined the choice of stem of the preceding component. A small number of other grammatical endings were not verbs but carried various co-ordinating or subordinating functions.

Early Middle Japanese inherited all eight verbal conjugations from Old Japanese and added one new one: Lower Monograde.

Traditionally, verbs were divided into five regular conjugations: quadrigrade (yodan 四段), upper monograde (kami ichidan 上一段), lower monograde (shimo ichidan 下一段), upper bigrade (kami nidan 上二段), lower bigrade (shimo nidan 下二段). There were also four 'irregular' conjugations: K-irregular (kahen カ変), S-irregular (sahen サ変), N-irregular (nahen ナ変), R-irregular (rahen ラ変). The conjugation of each is divided into six stems: irrealis (mizenkei 未然形), infinitive (ren'yōkei 連用形), conclusive (shūshikei 終止形), attributive (rentaikei 連体形), realis (izenkei 已然形), and imperative (meireikei 命令形). The English names for the irrealis and the realis differ from author to author, including negative and evidential, imperfective and perfective, or irrealis and realis.

The system of nine conjugation classes appears to be complex. However, all nine conjugations can be subsumed into variations of two groups: the consonant-root verbs, and the vowel-root verbs. Consonant-root verbs were quadrigrade, N-irregular and R-irregular verbs. The irregularity of N-irregular verbs occurred only in the conclusive and the attributive, and as there are no quadrigrade verbs with n-roots, quadrigrade and N-irregular verb patterns may be treated as being in complementary distribution.[8] Vowel-root verbs consist of bigrade verbs (the majority), a few monograde verbs (especially mi- 'see' and wi- 'sit'), the K-irregular verb ko- 'come', and the S-irregular verb se- 'do' (or -ze- in some compounds).[9] The difference between 'upper' and 'lower' bigrade or monograde verbs is whether the vowel at the end of the root was i or e. There was only one 'lower' monograde verb, kwe- 'kick', which was a 'lower' bigrade verb kuwe- in Old Japanese, changed pronunciation to ke- in early Late Middle Japanese (completed around 1300), and changed conjugation class again in later Japanese to become quadrigrade (Modern Japanese ker-). The difference between bigrade and monograde was whether in the conclusive, attributive and realis the initial u of the ending elided the vowel of the root or the vowel of the roots elides the initial u of the ending.

There are problems with that arrangement of stems.

• The irrealis did not occur by itself but always with another ending. There is good evidence that it actually was a fusion of the root of the verb with the a-sound that began the following ending: quadrigrade yom- 'recite' + -azu (negative suffix) has been later re-interpreted as a stem yoma- + -zu.[10]
• The infinitive had two functions, a linking function with another verb or with a verb ending and a nominal function as a verb-noun, and both functions were distinguished in having different pitch patterns. *The conclusive occurred as it is in the above table only at the end of a sentence (including before a quotative particle to), but before tomo 'even if', monograde verbs used the infinitive (quadrigrade yomu tomo 'even if ... recites', but monograde mi tomo 'even if ... sees'),[11] and before endings such as -besi 'have to' R-irregular verbs use the attributive instead (ari 'is' at the end of a sentence but aru-besi 'has to be'). Probably, the monograde verb form that was used before tomo is the earlier true conclusive form, and in Old Japanese, it was the only conclusive form attested (always before tomo). In Early Middle Japanese, the attributive form of monograde verbs came to be used also as a conclusive. With endings such as -besi, there is strong evidence that they were originally -ubesi and that a fusion of the root of the verb with the u-sound of the ending (quadrigrade yom- 'recite' + -ubesi) has been interpreted as conclusive yomu + -besi. That means that the apparently-anomalous u in aru-besi was part of the ending, not of the verb stem.[12] In Old Japanese, the root of the ending -ubesi, ube was attested as an independent noun. A more accurate representation of stems, therefore, would be:

Verb endings are attached agglutinatively to the various stems of verbs. They are divided into endings that behave like verbs themselves and endings that act as conjunctions.

Verb-like endings behaved like verbs in that they exhibited all or most of the stems that a lexical verb exhibits. The usual forms are listed below, and the way in which they were attached to the preceding verb follows the revised system of stems above. A verb could be followed by several such endings in a string.

Voice: 'passive' and 'causative':[13][14]

• Consonant-stem verbs + -aru, vowel-stem verbs + -raru (lower bigrade): passive voice; spontaneous voice (expressing lack of volitional control); honorific; potential ('can').
• Consonant-stem verbs + -ayu, vowel-stem verbs + -rayu (lower bigrade): passive voice; spontaneous voice (expressing lack of volitional control); honorific; potential ('can').
• Consonant-stem verbs + -asu, vowel-stem verbs + -sasu (lower bigrade): causative; honorific.
• Any verb + -asimu (lower bigrade): causative; honorific.

Tense/Aspect:[15][16]

• -eri (R-irregular): progressive or perfect aspect. Only attached to quadrigrade or S-irregular verbs.
• Infinitive + -tari (R-irregular): progressive or perfect aspect. Attached to any verbs.
• Infinitive + -nu (N-irregular): perfective aspect.
• Infinitive + -tu (lower bigrade): perfective aspect.
• Infinitive + -ki (unique conjugation): witnessed past tense. The morpheme has a suppletive conjugation: conclusive -ki, attributive -si, realis -sika, and no other stems.
• Infinitive + -keri (R-irregular): unwitnessed past tense, or emotive assertion.
• -amasi (unique conjugation): counterfactual ('would have ... ed'). The morpheme has conclusive -amasi, attributive -amasi or -amasiki, realis -amasika, and no other stems. The combination -amasika-ba expresses a counterfactual condition ('if ... had ... ed').

Other aspectual functions can be expressed with auxiliaries that also exist as independent verbs:

• Infinitive + -wori (R-irregular): progressive aspect. C.f. wori 'sit; live; be'.
• Infinitive + -oku (quadrigrade): preparative aspect, expressing an action performed in readiness for some future action. C.f. oku 'put'.
• Infinitive + -miru (monograde): speculative aspect, expressing an action performed experimentally, to 'see' what it is like. C.f. miru 'see'.

Other auxiliaries express direction, either literally or metaphorically.

Mood:[15][17]

• -amu (quadrigrade): tentative mood, expressing among other functions uncertainty ('maybe', 'shall I?'), intention ('I shall'), and hortative ('let's').
• -ubesi (siku-adjective): debitive mood, expressing 'can', 'should', or 'must'.
• -unari (R-irregular): hearsay mood.

Polarity:[18]

• -azu (unique conjugation): negative. The morpheme has conclusive and infinitive -azu, attributive -anu, and realis -ane.
• -azi (uninflected): negative of the tentative mood, functionally the equivalent of -amu + -azu, a combination that does not occur.
• -umazi (siku-adjective): negative of the debitive mood, an alternative of -ubekarazu (< -ubesi + -azu).

• Infinitive + -te: 'and (then/so), when, because'. It usually expressed a close sequential link between the verbs that it connects. The subjects of the two verbs connected by -te were usually the same.
• Realis + -ba: 'and (then/so), when, because'. It usually expressed a looser sequential link between the verbs that it connected. The subject of both verbs connected by -ba was usually different.
• -ade: negative 'and', 'without ... ing', 'rather than ... ', functionally an alternative to -azu-te.
• Realis + -do or -do-mo: 'although, but'. The longer form involved the addition of the particle -mo 'also, even' to the shorter form.
• -aba: 'if'.
• Infinitive + -tutu: 'while (at the same time)'.
• Various stems + -tomo: 'even if, even though'. Most verbs took the conclusive stem, bigrade verbs take the infinitive in earlier texts, and r-irregular verbs took the attributive stem.
• Infinitive + -nagara: 'while, while still' or 'despite'.

There were two types of adjectives: regular adjectives and adjectival nouns.

The regular adjective was subdivided into two types: those for which the adverbial form ended in -ku and those that ended in -siku. That created two different types of conjugations:

The class of siku-adjectives included a few adjectives that had z, rather than s: adverbial -ziku, conclusive -zi, attributive -ziki., e.g. imizi 'be terrible'. Onazi 'be the same' usually had -zi rather than -ziki, in its attributive form.

The -kar- and -sikar- forms were derived from the verb ar- "be, exists." The adverbial conjugation (-ku or -siku) was suffixed with ar-. The conjugation yielded to the r-irregular conjugation of ar-. The resulting -ua- elided into -a-.

The adjectival noun kept the original nar- conjugation and added a new tar-:

The nar- and tar- forms shared a common etymology. The nar- form was a contraction of the case particle ni and the r-irregular verb ar- "is, be": ni + ar- > nar-. The tar- form was a contraction of the case particle to and the r-irregular verb ar- "is, be": to + ar- > tar-. Both derived their conjugations from the verb ar-.