Old and Middle Japanese
Summary and Keywords
Old and Middle Japanese are the pre-modern periods of the attested history of the Japanese language. Old Japanese (OJ) is largely the language of the 8th century, with a modest, but still significant number of written sources, most of which is poetry. Middle Japanese is divided into two distinct periods, Early Middle Japanese (EMJ, 800–1200) and Late Middle Japanese (LMJ, 1200–1600). EMJ saw most of the significant sound changes that took place in the language, as well as profound influence from Chinese, whereas most grammatical changes took place between the end of EMJ and the end of LMJ. By the end of LMJ, the Japanese language had reached a form that is not significantly different from present-day Japanese.
OJ phonology was simple, both in terms of phoneme inventory and syllable structure, with a total of only 88 different syllables. In EMJ, the language became quantity sensitive, with the introduction of a long versus short syllables. OJ and EMJ had obligatory verb inflection for a number of modal and syntactic categories (including an important distinction between a conclusive and an (ad)nominalizing form), whereas the expression of aspect and tense was optional. Through late EMJ and LMJ this system changed completely to one without nominalizing inflection, but obligatory inflection for tense.
The morphological pronominal system of OJ was lost in EMJ, which developed a range of lexical and lexically based terms of speaker and hearer reference. OJ had a two-way (speaker–nonspeaker) demonstrative system, which in EMJ was replaced by a three-way (proximal–mesial–distal) system.
OJ had a system of differential object marking, based on specificity, as well as a word order rule that placed accusative marked objects before most subjects; both of these features were lost in EMJ. OJ and EMJ had genitive subject marking in subordinate clauses and in focused, interrogative and exclamative main clauses, but no case marking of subjects in declarative, optative, or imperative main clauses and no nominative marker. Through LMJ genitive subject marking was gradually circumscribed and a nominative case particle was acquired which could mark subjects in all types of clauses.
OJ had a well-developed system of complex predicates, in which two verbs jointly formed the predicate of a single clause, which is the source of the LMJ and NJ (Modern Japanese) verb–verb compound complex predicates. OJ and EMJ also had mono-clausal focus constructions that functionally were similar to clefts in English; these constructions were lost in LMJ.
1. Linguistic Periods and Typological Overview
The attested history of the Japanese language is usually divided into the following periods:
Old Japanese (OJ; mainly 700–800 ad)
Early Middle Japanese (EMJ; 800–1200)
Late Middle Japanese period (LMJ; 1200–1600)
Modern Japanese (NJ; 1600–)
These periods largely coincide with and are based on main political periods in Japanese history, and although alternative periodizations have been proposed, this periodization does allow us to capture significant general facts about the changes that have taken place between OJ and NJ. Thus, most fundamental changes in phonology took place from OJ through EMJ, while most important changes in grammar in the main set in towards the end of the EMJ period and took place in the course of the LMJ period. Externally, a strong influence from Chinese is observable in particular during the EMJ period. Since the early NJ period, from around 1600, Japanese has changed relatively little, other than its large intake of loanwords and creation of loan translations, as well as other influence, from European languages, in particular English, from the last quarter of the 19th century onwards. Despite the fact that the capital and political center moved from Kyoto in the Kansai area in central Japan to Tokyo in the Kantō area in the eastern part of Japan from the beginning of the 17th century, and thus from the beginning of the NJ period, standard, or common, NJ is a fairly direct descendant of OJ, EMJ, and LMJ, as the Kyoto-based common language was the language of the educated, ruling classes who migrated to the new capital. Although it did receive some influence from other varieties, this variety came to form the overwhelmingly dominant part of the language of the educated classes there, which in turn served as the basis for the establishment and codification of a ‘standard’ Japanese around the beginning of the 20th century.
All stages of Japanese share a number of properties:
Japanese is verb final and left-branching, has extensive pro-drop, accusative alignment, marks (or can mark) grammatical relations by clitic case particles, and has a topic–comment construction distinct from subject–predicate. Relative, or adnominal, clauses precede their head noun directly without relative pronouns or complementizers. Japanese has morphological and other systematic expression of ‘honorific’ language, which also serves more general referent tracking and deictic functions.
The main lexical parts of speech are verbs, adjectives, nouns, pronouns, and adverbs. Verbs have extensive morphology, which is mainly agglutinating, but with some fusion. Adjectives almost always combine with a bound, inflecting copula. Other lexical parts of speech have no inflectional morphology, but some nominal suffixation is found.
There are three main types of grammatical clitics: Copulas (with defective or irregular inflection) follow nouns to form nominal predicates. Predicate extensions (with verbal or adjectival morphology) follow predicates to express some added deontic or epistemic modality, including evidentiality. Particles are non-inflecting. Those that follow nouns are mainly case, topic, and focus particles, while those that follow predicates are mainly conjunctional particles that form subordinate clauses, complementizers, and final particles that contribute to utterance modality.
In a segmentational model, the structure of Japanese verb forms may in general be described as follows, for any stage of Japanese, with three main morphemic layers, as shown in (1). This is illustrated in (2) with the verb sak- ‘come into bloom’ (appearing in different stem shapes), for NJ (saita “(it) bloomed”, sakasemashita “(I) made (it) bloom”) and OJ (saku “(it) comes into bloom”, sakyerikyeri “(it) was blooming”).
Flectives express the obligatory, paradigmatically opposed categories of which any verb form must express one and only one, e.g., NJ past tense or OJ conclusive. Auxiliaries are bound inflecting suffixes. (The term ‘auxiliary’ is unfortunate or at least unusual, as these morphemes are not auxiliary verbs by any stretch of the imagination, but it has become commonly used for Japanese and it is retained it here.) Auxiliaries are optional and express a variety of verbal categories (e.g., NJ causative, polite, OJ stative, modal past tense). Some auxiliaries can combine, whereas others are mutually exclusive; the last auxiliary in a chain carries the inflection of the whole verb form. Although this basic morphological structure has been the same throughout the attested history of Japanese, the range of obligatory inflectional categories and the inventories of auxiliaries, the categories they express and their combinatory possibilities changed profoundly between OJ and the end of LMJ (and NJ).
This article will describe salient features of Old Japanese (section 2), especially as it differs from NJ, and the main changes that took place between OJ and the end of the Late Middle Japanese period (section 3). References are kept to a minimum, and are generally only given when the treatment or description in Frellesvig (2011, which has further references) is insufficient.
2. Old Japanese
2.1 Sources, Writing, and Script
The bulk of sources for OJ are from the 8th century AD. The vast majority has been handed down in copies from later periods. There are a few epigraphical materials on stone and metal from the 8th century and earlier, but most are fragmentary and short; in addition, a large number of wooden slips with writing on them (mokkan 木簡) from the 7th and 8th century have been excavated over the past decades, valuable mainly for the study of the history of writing. Some texts were compiled in the 8th century from older, since lost texts or from oral tradition, with the earliest texts thought to date from the late 5th century. A small proportion of text represents Eastern OJ dialects, but overall the language represented is that of the central region of the capital Nara. Japan has no indigenous script and all OJ texts are written in Chinese characters (kanji), adapted to be used in principle in two ways: logographically (to represent words or morphemes) and phonographically (to represent sound, specifically syllables). In practice, writing in OJ is quite varied, with some texts written exclusively phonographically and some logographically, but most in a mixture of logographic and phonographic writing, and with many examples of playful or deliberately complex writing. The OJ texts are available online through the Oxford Corpus of Old Japanese (OCOJ).
OJ phonographic writing was syllabic, mainly with one syllable represented by one kanji (called ‘man’yōgana’ when used phonographically to write Japanese). There are several complicating factors: First, most syllables could be written with a range of different man’yōgana. Second, a number of man’yogana were used to represent different syllables. Third, almost all kanji used as man’yōgana were also used logographically, and to a large extent also vice versa. Some of the textual material is therefore difficult to decipher. Much of the logographically written text is open to some interpretation, and some of it can only be read for sense in a general way, including the Kojiki (古事記 “Record of ancient matters”) from 712, which is a large history and mythical account of the origin of the Yamato state. Apart from that, the volume of OJ text is around 110,000 words, out of which more than 90% is poetry; the remainder is highly ritualized prose, found in prayers and blessings (known as Norito 祝詞) and imperial edicts (Senmyō 宣命). The single largest source for OJ is the poetry anthology Man’yōshū (万葉集 “Collection of Myriad Leaves”, compiled sometime after 759) which consists of around 4,500 poems (approximately 90,000 words), spanning three centuries and a variety of genres and mix of types of writing, but with a large number of texts from the 8th century; although phonographically used kanji are used in the other OJ texts as well, they are called man’yō-gana after their extensive use in the Man’yōshū, which, however, also contains a large proportion of logographic writing.
OJ had the following phoneme inventory:
This system of five short vowels has remained unchanged to the present day in most varieties of Japanese, including the main descendants of OJ, EMJ, and LMJ: Kyoto and Tokyo NJ.
The lax obstruents and the liquid, /b, d, g, z, r/, were not used in initial position in lexical words of native word-stock or older, assimilated loans. The intake of Sino-Japanese vocabulary, which set in during the OJ period, overruled this phonotactic restriction, but even today word-initial /b, d, g, z, r/ is largely limited to Sino-Japanese and other borrowed vocabulary (as well as a few native forms that result from irregular sound change, e.g., loss of initial vowel, NJ das- ‘put out’ < OJ idas-, or vowel loss and contraction, de case particle < nite).
Tense /p, t, k, s/ were distinct from lax /b, d, g, z/. Voicing was not the phonemic distinctive feature, as the tense obstruents were voiced ([-b-, -d-, -g-, -z-]) intervocalically (or, word medially: medial voicing). Conversely, lax /b, d, g, z/ were prenasalized ([mb, nd, ŋg, nz]); this reflects the diachronically secondary status of the lax obstruents, which derive from contractions of nasal + (tense) obstruent, sometimes following vowel reduction and loss. In some cases, the etymological source is readily recoverable, e.g., yamadi ‘mountain path’ < yama-miti ‘mountain-path’. The origin of the lax obstruents makes their non-occurrence in word-initial position readily understandable. Sibilant (strident) /s, z/ were distinct from nonsibilant (mellow) /p, t, k, b, d, g/. Both sibilants and nonsibilants exhibited allophonic variation with respect to continuousness, such that sibilants had both fricative and affricative variants, e.g., /s/ [s, ts], and nonsibilants had both stop and fricative variants, e.g., /p/ [p, Φ]. Other features of allophonic variation include palatalization before /i, e, y/ and nasalization of vowels before nasals and (prenasalized) lax obstruents.
A morphophonological process applying to compounding, whereby an initial tense obstruent in the second member of a compound becomes lax (e.g., matu ‘pine’ + para ‘plain’ => matu-bara ‘plain with pines’) is known by the Japanese term ‘rendaku’. This term literally means ‘sequential muddying’, after the traditional Japanese term for syllables with initial lax obstruent, dakuon ‘muddy sounds’. (Rendaku is usually translated into English as ‘sequential voicing’, but that is misleading, as the phonemic category for obstruents is not voicing, but tenseness.) Etymologically, rendaku originated in reduction of nasal initial particles, in particular genitive no: *matu no para > matu-bara (with [-mb-]).
Rendaku is still found in the modern language and the following holds for all stages of Japanese: Rendaku is blocked if the second member of the compound already contains a lax obstruent (‘Lyman’s Law’), e.g., matu + kage ‘shade’ => matu-kage, not *matu-gage, and it is not an automatic phonological rule: There are many examples where rendaku could, but does not occur, e.g., kara ‘Korean, Chinese, foreign’ + koromo ‘clothes’ => kara-koromo ‘foreign robe’, as opposed to nare- ‘get/be used to’ + koromo => nare-goromo ‘well-worn and familiar clothes’.
2.2.2 Syllable Structure
OJ Syllable structure was very simple, with only open syllables and with only few complex syllable onsets (Cy-, Cw-, with restricted distribution). The total number of distinct syllables in early OJ was 88, (5). Vowel-initial syllables were only found in word initial position.
Cye, Cwi, and Cwo type syllables were lost and merged with Ce, Ci, and Co type syllables in the transition to and the early part of the EMJ period and they are not reflected or represented in the EMJ kana script (see section 3.1). The existence of the Cye, Cwi, and Cwo type syllables was only finally discovered in the 1910s and at the time they were described and thought about in terms of ‘extra’ syllables in OJ, paired with the syllables they later merged with and labeled in relation to them:
Some transcription systems use diacritics or subscript numbers or letters to distinguish between these syllable types, rather than a phonemic transcription. Until the 1970s it was commonly thought that the differences between these syllable types resided in differences in vowel quality, and you will sometimes still find references to an ‘eight-vowel’ system in OJ (positing unitary vowel phonemes for /-wi, -ye, -wo/).
2.3.1 Verb Morphology
OJ verbs obligatorily inflect for the following categories. Inflection is mainly for syntactic, modal, and conjunctional categories (as opposed to LMJ and NJ which inflects for tense and aspect). OJ has eight verbal conjugation classes, which fall in two major groups: consonant base verbs (approximately 75% of all verbs) and vowel base verbs (approximately 25%), each with a number of subclasses. The inflected forms are here exemplified with verbs from the two main verb classes, regular sak- ‘come into bloom’ and ake- ‘dawn’, irregular ar- ‘exist’ and se- ‘do’ (both of which have important grammatical functions).
Finite forms can conclude a main clause. In addition to primarily modal categories (exclamatory, imperative, negative conjectural, and optative), a main distinction is between the conclusive (“CONCL”) and adnominal (“ADN”). The conclusive was used to conclude declarative main clauses (9) (as well as being followed by modal extensions and final particles), whereas the adnominal was used as follows.
The distinction between conclusive and adnominal verb forms is functionally important in OJ and EMJ. However, it is important to bear in mind that while this distinction has morphological expression in most verb classes, and for most auxiliaries, there is syncretism between the conclusive and the adnominal in the largest verb class of all, the regular consonant base verbs which make up close to 75% of all OJ verbs.
The nonfinite, or converb, forms conclude nonfinal, mostly subordinating, clauses. They include a range of specific subordinations (continuative ‘while’, conditional ‘if’, provisional ‘as, when’, concessive ‘although’), as well as two general subordinating forms, infinitive and gerund. The infinitive also derived a verbal activity noun, e.g., kwopwi- ‘to love’ => kwopwi ‘loving’, which could be predicated by se- ‘to do’ and which in some cases were lexicalized as common nouns, kwopwi ‘love’). The nominal (“NMNL”) generally expresses abstract nominalizations (‘(the fact) that’), as in (12) which also exemplifies the provisional (“PROV”).
A number of verbal morphological categories, including voice, aspect, negation and tense, are optionally expressed by ‘auxiliaries’ (inflecting suffixes). Most auxiliaries belong to one of the two main verbal conjugation classes, but a few (negative, simple past, subjunctive) have highly irregular and/or suppletive paradigms.
Auxiliaries attach to the basic stem of vowel base verbs, and for consonant base verbs either to one of two derived stems (ending in /a/ or /i/) or to the basic stem. They are here shown in their basic stem form in the leftmost column, and in the conclusive form when attached to verbs (with the suppletive adnominal forms shown in brackets for the negative and the simple past).
As shown, some auxiliaries do not combine with the largest group of vowel base verbs, reflecting that that verb class is younger in the language and was only formed shortly before the OJ period; from EMJ onwards there were no such restrictions on combinations of auxiliaries with verbs from different morphophonological classes.
The stative auxiliary is used about both progressive and resultant states. Etymologically the stative incorporates the existential verb ar-, e.g., sak-yer- ‘be blooming’ < *saki ar-. This construction was morphologized prior to the formation of the main vowel base verb classes and therefore the stative was not used with these verb classes. It is supplemented by a productive periphrastic stative, formed by the gerund and an existential verb, e.g., iki-te ara-ba live-GER exist-COND “if you are alive”, which was used with all verb classes. Through EMJ the periphrastic stative replaced the stative auxiliary and itself morphologized to give a new stative auxiliary, -te ar- > -tar- (cf. section 2.3.5); for that reason -tar- is included in square brackets in (13), although it was not yet a morphologized auxiliary in OJ. The stative construction consisting of nonfinite verb + existential verb is pervasive throughout the history of Japanese and has gone through several cycles of morphologization followed by innovative analytic formation. It is also the way in which statives are formed in NJ, with the new animate existential verb i- (suwatte i- ‘be sitting’), and dialectally a number of variants are found.
The two perfectives are selected by different verbs, n- by unaccusatives (e.g., sak-, ake-) and -te- by unergatives (e.g., wem- ‘laugh’) and transitives (e.g., yak‑ ‘burn’, tuke- ‘attach’, which have been added to (13) above). Perfective auxiliary selection is the main expression of split intransitivity in OJ.
The modal past tense contains an element that can be characterized as ‘speaker commitment’, which is absent in the simple past. The modal past is thus paradigmatically opposed both to the simple past and to the conjectural, whereas simple past and conjectural could combine to give a past conjectural -kye-mu.
Generally, auxiliaries can combine between (but not within) the main groups (respect, voice, aspect/negation and tense/mood (apart from the combination of simple past and conjectural)) to form longer syntagms, e.g., apa-sa-zu meet-RESP-NEG “you do not meet (me)”, uwe-te-kyeri plant-PFV-MPST “(you) have planted”, pupum-yeri-ki bud-STAT-S[imple]PST “(the cherry trees) were budding”, asipuma-si-na-mu step-RESP[ect]-PFV-CONJ[ectural] “you will end up stepping”. Auxiliaries occur in a fixed order, respect being closest to the verb stem, tense/mood being rightmost, although far from all seemingly possible combinations are attested. The negative combined directly with the modal past and with some forms of the simple past, but not with the s- initial forms of the simple past, or with the conjectural or subjunctive; however, analytic constructions with the existential verb ar- allowed such combinations (see section 2.3.4).
Japanese adjectives are often regarded as an inflecting part of speech. However, they can be used on their own in exclamatory or adverbial function (paya ‘fast, quickly, early’), and the predicational flective with which they most frequently occur is best thought of as a bound copula which inflects for many of the categories that verbs inflect for, e.g., conclusive paya-si ‘(s/he) is fast’, adnominal paya-ki se (rapids of a river) ‘the fast rapids’, infinitive paya-ku ake-naba (dawn-PFV.COND) ‘if it dawns early’, conditional/provisional paya-kyeba ‘as/when/if it is fast’, etc. The adjectival copula could combine directly with only very few auxiliaries (conjectural and, vestigially, negative); in order to combine with the full range of auxiliaries, an analytic construction with the existential verb ar- was used (see section 2.3.4).
So-called ‘adjectival nouns’ (or, ‘nominal adjectives’) are a subclass of adjectives which are predicated by the regular copula (2.3.3) and not by the adjectival copula, e.g., sakari ‘be in full bloom; maximal’ in (19).
Nominal predications used a clitic copula, which has minimal, highly defective, and irregular inflection, reflecting an earlier, fuller morphological system.
The adnominal was used only in adnominal function (and not in the other functions of the adnominal of verbs and of the adjectival copula), e.g., nara no miyakwo “the capital, (which is) Nara”. The infinitives were on their own mainly used to adverbialize nouns, ‘as, being’ (namida kwosame ni puri tears fine.rain COP.INF fall “tears fall like fine rain”, pana yuki to puri flower snow COP.INF “flowers fall like snow”), in raising and absolute (see section 2.3.9) constructions, and with ‘become’ verbs (yupubye ni nareba evening COP.INF become.PROV “when it becomes evening”).
Full predication and inflection for the full range of categories of verbs and combination with auxiliaries took the form of analytic constructions with an existential verb, by far the most frequently ar-, following the copula infinitive:
2.3.4 Other Analytic ar- Predications with Stative Predicates
Extended analytic predications with the existential verb ar- were an important part of predicational morphology also for the adjectival copula and negative predicates, allowing combination with a range of auxiliaries. These extended predications also used the infinitive of the adjectival copula (“ACOP”) (waka-ku ari-ki young-ACOP.INF exist-SPST.CONCL ‘he was young’) and the negative auxiliary shown in (16).
(17), which includes the necessitive (“NEC”) predicate extension be-, which morphologically is an adjective, contains two analytic ar- predications.
2.3.5 Fused Forms and New Morphemes
Due to regular vowel deletion, the analytic ar- constructions formed with the infinitives of the copula, the adjectival copula, and the negative auxiliary, as well as the analytic stative -te ar-, had abbreviated, fused forms as in (18), illustrated by the copula in (19).
Through EMJ these forms morphologized and became important parts of EMJ/LMJ morphology, giving rise to secondary, but full and regular, conjugations of the negative auxiliary (-zar-) and the adjectival copula (-kar-), to a new stative auxiliary (-tar-), and to a fully inflected clitic copula verb (nar-). Except for negative -zar-, these forms are all reflected in NJ. However, the analytic constructions continued to exist alongside the morphologized forms, and also the analytic ar- constructions remain important parts of the morphology of stative predicates today.
2.3.6 Complex Verbal Predicates
In addition to predicates of some morphological complexity, a salient and important feature of predication in Japanese is the frequent use, also in OJ, of complex predicate constructions, in which two verbs in one clause contribute semantically to the predication, e.g., kwopwi- ‘long for’ and ki- ‘come’ in (20) (see Aoki & Frellesvig, forthcoming).
Various types, and degrees of grammaticalization, are found in these constructions, including vector verb constructions in which one verb performs a mostly grammatical function, as in (21) where watar-, which was and remains a lexical verb ‘cross’, is used to express ‘continuative’.
In OJ, the two verbs in a complex predicate seem not (always) to have formed morphological compounds, but they exhibit the same functions as and constitute the origin of the very frequent V-V compounds in NJ (e.g., kaki-tuzuke- write-continue ‘continue writing’).
2.3.7 Pronouns, Demonstratives, and Interrogatives
OJ had a set of personal pronouns and a morphological system of two-way demonstratives + interrogative.
The short forms are used in compounds and derivatives (e.g., ko-yopi ‘this night’, wa-dori ‘my bird’ (tori ‘bird’), i-ka ‘how’), or with a following particle, including by far the most frequently genitive particles, e.g., wa ga kokoro (I GEN heart) ‘my heart’. The long forms are used in isolation, for example as topics, and with various particles, but never with a genitive particle (except idure which has no corresponding free short form). Of the two genitive particles (see section 2.3.9), personal pronouns take ga and demonstratives take no. Nare, onore, and sore are very rare in OJ, but became full members of the system in EMJ.
Although most descriptions of OJ will project the three-way demonstrative system (proximal, mesial, distal) found in EMJ and onwards in Japanese (see section 3.4.1) on to OJ, there is no proper attestation of the third of these terms in OJ, and the system is in fact a two-way system: proximal referring to what is within the speaker's domain of direct sensory perception or experience, versus nonproximal referring to what is outside of the speaker’s domain of direct experience. The nonproximal forms were very often used anaphorically, whereas the proximal forms almost exclusively were used deictically.
2.3.8 Topic and Focus
In addition to the topic–comment constructions well known also from NJ, with a topic, often marked by one of the topic particles pa (> NJ wa) or mo ‘also, even’ occurring on the left edge (or sometimes right-dislocated, (9) above), OJ had a focus construction, which involved a focus particle marking the focused element and use of the adnominal or exclamatory form to conclude the sentence, rather than the conclusive form; this construction is traditionally known as kakari-musubi (‘hanging-tying’), meaning that one element is set (“hung”) up and the predication completed (“tied up”) by use of a particular verb form. The focus particles are so~zo (> EMJ zo), namo (> EMJ namu~nan) and interrogative ka, ya, with which the adnominal form is used to conclude the sentence, and koso with which the exclamatory (“EXCL”) is used.
These focus constructions have a semantic effect similar to it-clefts in English, but are mono-clausal. A focus construction establishes a copular, predicative relation between presupposition (kizo no ywo ime ni mi-ye-turu “(what) I saw in a dream last night”) and focus (e.g., wa ga kwopuru kimi so “you, my beloved lord”), like that between subject and nominal predicate, and that explains the use in the presupposed predicate of the adnominal form, which was used in nominalizations (see section 2.3.1); this probably also holds etymologically for the use of the exclamatory, which is often thought to go back to a nominalized form. These focus constructions are a conspicuous and prominent feature of both OJ and EMJ, but disappear in LMJ.
2.3.9 Case Particles, Grammatical Roles, and Word Order
The main OJ case particles were:
Unlike NJ, OJ or EMJ had no nominative case; it was only in late LMJ that the OJ/EMJ genitive ga changed to become a nominative case particle (see section 3.4.3).
The two genitives were used for possession, or adnominal modification, and for subject marking in certain clause types (see section 2.3.10), with no difference in function between them. The details of their distribution has yet to be understood fully, but overall they were selected by different nouns, determined to a large extent by animacy and close personal relation, with demonstratives (ko, so) and most common nouns taking no, but personal pronouns (wa, a, na, si, ta, ono) and a smaller set of nouns, some very frequent, mainly referring to close people taking ga (e.g., kimi ‘(my) lord’, imo ‘sister, my beloved’, sekwo ‘older brother; beloved (male)’. However, there seem to be exceptions with regards to people reference (e.g., pito ‘person’, opokimi ‘great lord’, tukapi ‘messenger’, or masurawo ‘gentleman’ taking no) and some nouns (e.g., kwo ‘child; dear’ or matu ‘pinetree’) select both ga and no. The distributional tendencies are clearer with genitive marked subjects than with genitive-marked possessors (or noun modifiers), partly, though not exclusively, because some modified head nouns always are preceded by ga, e.g., always X ga ne ‘sound’, never *X no ne.
Accusative wo was used to mark (specific) direct objects (see section 2.3.10) and for a range of for example traversal and durational arguments and adjuncts which cross-linguistically often are associated with accusative marking; wo also marked raised subjects and subjects in some absolute constructions (which as constructions probably originate in raising constructions), as in:
Ni was the general oblique case marker, used for indirect objects and a variety of other functions, including direction, location, time, agent (in passives), and causee. Other case particles include ablative (‘from, along’ etc.) ywori, ywo, yuri, yu, comitative (‘with, and, than’) to, as well as newly emerging ablative kara, allative pye and allative made.
There is a close historical relation between the copula (see section 2.3.3) and some of the case particles. OJ dative ni, genitive no and comitative to derive from and are homonymous with the main simple forms of the copula, infinitive, and adnominal. This still holds in NJ, which in addition has homonymy between copula gerund de and instrumental and locative case particle de (which derives from the copula). In OJ, as in NJ, dative or instrumental marking of loose adjuncts can sometimes be difficult to distinguish from adverbialization by copula ni or de, but the distinction of adnominal copula no from genitive no is clear, cf. copula nara no miyakwo “the capital, which is Nara; the capital, Nara”, as opposed to genitive kuni no miyakwo “the capital of the province/land”.
2.3.10 Subject and Object Marking
OJ had, like later stages of Japanese, accusative alignment. However, extensive pro-drop (probably reinforced by the poetic genre of the texts from the period), combined with widespread topicalization and extensive use of relative clauses from which arguments are extracted, means that the majority of clauses in the material do not overtly express all their arguments within the clause, and the details of clause-internal marking of subjects and objects have proven elusive. However, we can say the following:
Subjects of declarative, imperative, or optative main clauses were not case-marked. Subjects of subordinate, nominalized and interrogative clauses can be marked by the genitive particles ga, (16), (20), and no (30), subject to the distribution of ga and no described in 2.3.9, but the distribution of bare and genitive marked of subjects in these types of clauses has not yet been described fully.
Objects were accusative case marked in accordance with a system of differential object marking, in which only objects which were specific (definite or otherwise discourse linked) were accusative case marked, whereas nonspecific objects were bare (Frellesvig, Horn, & Yanagida, in press), such as here where context makes clear that (31) is specific and (32) nonspecific, as reflected in the translations.
This system of differential object marking was lost in EMJ where also nonspecific objects can be marked by wo. The OJ system is further complicated by the fact that some specific objects were bare, due to case drop, the precise conditions of which are not yet clear. Case drop continued in EMJ and subsequent stages of the language, so that also today not all objects are accusative case marked.
2.3.11 Word Order
NJ has persistent, basic S O V word order, subject to the possibility of scrambling. Again presumably because of the genres of the texts available to us, OJ exhibits a great deal of scrambling. However, it does appear that the basic word is S O V, but there is one consistent and noteworthy exception (Yanagida & Whitman, 2009): wo-marked objects usually appear to the left of a subject, and always to the left of ga-marked subjects (which themselves often are adjacent to the verb), as in (33), which has a nominalized predicate in the adnominal form in a focus construction and therefore allows genitive subject marking (cf. 2.3.10); (33) also has a stative construction with the continuative (“CONT”) followed by ar-:
This suggests first of all that wo-marked objects (which are specific) move into a higher position in structure, and second that the connection between a ga-marked subject and a subordinate and/or nominalized verb is close. This word order rule was lost in EMJ where ga-marked subjects regularly precede wo-marked objects.
3. Middle Japanese
3.1 Script, Writing, and Sources
The EMJ period saw the development of the kana script, with the variants hiragana and katakana, which arose through graphic simplification of the man’yōgana. The kana letter categories were few, less than 50, as they, as opposed to the man’yōgana, did not distinguish tense from lax obstruents, and they made for easy, direct written representation of the language. The development of the kana letters accompanied a surge in writing in Japanese, and we have a much greater volume of text from this period than from OJ, both poetry and a large body of important and well-known literature, including the Genji monogatari (‘The tale of Genji’) and the Makura no Sōshi (‘The pillow book’), both from around 1000 AD and both written almost exclusively in kana. Other writing, however, continued to include logographically used kanji. Up until towards the end of the EMJ period we have a large body of text that constitutes a fairly close record of the contemporary language of the capital (which in 794 had moved to Kyoto, like Nara also in the Kansai region). However, towards the end of the EMJ period, the written language starts fossilizing and a normative ‘classical Japanese’ language emerges which remained the main medium of writing in Japanese up until the end of the 19th century, and during that long period contemporary spoken language features are in most texts reflected only sporadically and incidentally. This makes it difficult to trace in detail several of the grammatical changes which took place through the LMJ period, but we have a good-sized valuable body of Romanized texts, dictionaries and grammars produced by Portuguese and Spanish missionaries in Japan, dating from the turning point between LMJ and NJ, around 1600, giving us an accurate picture of the language at the time. The main texts from EMJ and increasingly also texts from LMJ are being made available online through the Corpus of Historical Japanese (CHJ), published by the National Institute of Japanese Language and Linguistics, Tokyo.
3.2.1 Syllable Structure
A major structural phonological change took place in the transition between OJ and EMJ, namely the introduction of the distinction between metrically short (or light) and long (or heavy) syllables. This is the main structural sound change to have affected Japanese in its attested history. Thus, EMJ acquired syllables of the shape CVV and CVC, with complex nuclei. Through this change Japanese became quantity sensitive and the mora became a relevant unit in the language. The phonemes which occur in the new syllable position are mainly /-i, -u/, forming diphthongs or long vowels (/CVi, CVu/), and /-N-, -Q-/, obstruents which were distinguished only by nasality and which would copy other features from the immediately following consonant. /N/ later came to be found in word-final position and before vowels, there usually pronounced as an unreleased, most frequently uvular, nasal, sometimes described as an approximant.
The change in syllable structure came about in the main through a set of non-automatic sound changes (known as onbin ‘euphony’) in which a short syllable with a high vowel (/Ci, Cu/) was phonetically reduced and subsequently phonologically reinterpreted as a single segment incorporated into the preceding syllable, e.g., kagupasi ‘fragrant’ > kanbasi (/kaNbasi/), kaubasi; nipi(-)ta ‘new(-)field’ proper name > nitta (/niQta/), niuta; yomi-te ‘read-GER’ > yon-de (/yoN-de/), you-de; na-ki ‘not.exist-ACOP.ADN’ > na-i; paya-ku ‘fast-ACOP.INF’ > paya-u. In some cases both vocalic and consonantal forms resulted, and when the reduced syllable had initial (prenasalized) lax or nasal consonant, a following tense consonant became lax (prenasalized). As shown, both verbal inflected forms and forms of the adjectival copula were affected by this set of changes, which therefore also had an impact on morphophonology, in particular resulting in a new stem for consonant base verbs and changed forms of the adnominal and infinitive of the adjectival copula. The change in syllable structure also allowed easier phonological accommodation of the large intake of loanwords from Chinese that took place during the EMJ period (see section 3.3).
3.2.2 Segmental Sound Changes
Sound changes through the EMJ and LMJ periods had almost no effect on the inventory of onset consonants, with the exception of the split of /p/ into /p, f/ at some point, most likely early LMJ (and subsequent change of /f/ to /h/ in NJ, after 1700). It was only in the second half of the 19th century that the sudden large-scale intake of loanwords from European languages resulted in the addition of more onset consonants.
Most segmental sound changes during EMJ and LMJ fall in two overall groups, phonotactic changes, which resulted from conditioned loss, change and split, and changes to phonetic realization rules. The former affected the glides /w, y/ and /p/:
(a) Over several centuries OJ, from very late 8th through to the 13th century, /w, y/ were lost in many environments: /w/ as lost before /o, i, e/ (first in /CwV/ and later in syllable initial position), and similarly /y/ was lost before /e/ (first in /Cye/, later in syllable initial position).
(b) OJ /p/ lenited: Intervocalic /-p-/ changed to /-w-/ in the 10th century (and was subsequently lost before /u, o, i, e/, cf. immediately above); later, remaining /p/ changed to /f/ in most environments.
Some examples of changes in phonemic word shapes resulting from these two sets of changes are given in (34), showing the OJ and the current, present-day forms in bold, and indicating when the forms reached their present-day phonemic shapes.
These changes resulted in a reduction in the number of different short syllables, but also in an increase in vowel sequences. Of these, /Vi, Vu/ became long syllables within the new syllable structure (e.g., monosyllabic /koi/), whereas /Ve, Vo/ remained dissyllabic and were pronounced with an on-glide on the second vowel: /ma.e/ [maje], /ka.o/ [kawo]. This feature of pronunciation was only lost in the 19th century.
During LMJ /Vu/ diphthongs changed to long vowels, which remained metrically long and which therefore usually are transcribed as double vowels: /iu/ > /yuu/, /eu/ > /yoo/, /au/ > /ɔɔ/, /ou/ > /oo/. In NJ, /ɔɔ/ merged with /oo/, but until then the long vowels had a height distinction between back vowels which was absent among short vowels (long /ɔɔ/ ≠ /oo/ ≠ /uu/, but short /o/ ≠ /u/). This is often remarked upon as being a typologically unusual feature of LMJ phonology, but rather than having a typologically anomalous vowel system, it is perhaps more likely that LMJ and early NJ /ɔɔ/ was also synchronically a derived realization of underlying //au// (=> /ɔɔ/); this finds some support amongst others in the formation of the volitional (“VOL”) verb form (see section 3.4.2), which is best described as akyoo <= ake (basic stem) + u VOL and sakɔɔ <= saka (derived a-stem) + u.
The prominent realization rules of OJ were lost during LMJ (intervocalic voicing) and NJ (continuousness variation, prenasalization of lax obstruents); concurrently with the loss of continuousness variation in early NJ, /d, z/ merged as /z/ before /u/ and as /d/ before /i/ (Frellesvig, forthcoming). Both intervocalic voicing and prenasalization are found dialectally in NJ.
Well-assimilated loanwords from Chinese, and other continental languages, may be identified within OJ, which were acquired through direct spoken language contact. However, of far more profound impact was the influence from Chinese, which arose primarily from reading practices in Japan of Chinese texts. This originated from before the OJ period and certainly such influence is seen already in the OJ texts, but it is from EMJ that this influence manifests on a large scale. There are two kinds of influence. One reflects the dissemination into the common language of practices which arose in the translation of Chinese texts into (or ‘rendition’ in) Japanese (kanbun-kundoku (漢文訓読 ‘reading Chinese in Japanese’). The full extent of the lasting influence of kanbun-kundoku practices on Japanese remains to be charted and described, but it includes loan translation collocations such as ame-tuti ‘heaven and earth; the world’ and ko-no-yo ‘this world (as opposed to before- and after-life)’, the noun modifiers iwayuru ‘so-called’ and arayuru (lexicalized from the passives of iw- ‘say’ and ar- ‘exist’), as well as extensive use of sentence initial and modal adverbs, e.g., osoraku ‘likely, probably’ (from the nominal form of osori- ‘to fear’), sikasite ‘and, then’, or imada ‘(even) now’ (> NJ mada) in correlation with a negative verb form to mean ‘not yet’.
Far more conspicuous and manifest is the intake of a large number of direct loanwords from Chinese, again through the medium of Chinese texts. Sino-Japanese (“SJ”) loanwords are mostly nouns. Many early loans are court titles, Buddhist or philosophical terms, or words for technology, but there is also from early on common nouns such as niku ‘meat’, nikki ‘diary’, and net/neti (‘fever’. Also the Chinese numerals 1–10, 100, 1,000, and 10,000 were borrowed, and while the native numerals still are in use today, they are restricted and the SJ numerals are the main, unmarked numerals. Verbs were usually taken in as verbal nouns, predicated by se- ‘to do’, e.g., gu-se- ‘furnish, be furnished (with)’, si-se- ‘die’, rongi-se- ‘debate’.
On the whole, it is difficult to judge the dissemination of Sino-Japanese vocabulary among the general population in EMJ, or the extent to which SJ vocabulary primarily was a written form. It is clear, though, that a fair number of everyday SJ words were common, and the missionary materials from the very end of LMJ clearly show a large SJ vocabulary that was in widespread use and seems to have been fully assimilated at the time.
In the Tale of Genji from around 1000 AD, just under 5% of distinct words and 12.5% of words in running text are SJ, while on the other hand, SJ vocabulary was normatively excluded from poetry. This shows both that SJ vocabulary was in fairly widespread use among people at court (such as Murasaki Shikibu, the author of Genji) and also that SJ to a large extent remained identifiable. This is also reflected in the fact that SJ vocabulary in some genres (though not in the Tale of Genji, for example) often would be written with the kanji used to write them in Chinese.
A few phonological features were exclusive to SJ vocabulary (though obviously not shared by all SJ vocabulary), further contributing to the identifiability of some SJ vocabulary: syllable final /-t/ (lost in NJ), new syllables with complex onsets /kwa, gwa/ (lost in most varieties of NJ) and /Cya, Cyo, Cyu/ (which subsequently also arose in the native word-stock through regular sound change, and which are frequent in NJ), and use of word initial /b, d, g, z, r/ (also in NJ found in loans from other languages).
In NJ, a much larger proportion of the vocabulary is Sino-Japanese, both in the lexicon and in text. That is, however, due to the very large amount of vocabulary created in the course of the modernization of Japan in the late 19th and early 20th century as translations into Sino-Japanese of words from European languages, especially English.
3.4.1 Pronouns, Demonstratives, and Interrogatives
The simple system of personal pronouns in OJ (see section 2.3.7) was lost in the course of EMJ. Some of the forms continued to be used, e.g., ware ‘I’ and univerbated waga (‘mine’), but the system as such disappeared. Already in OJ the pronominal system was supplemented by a number of lexical terms of address, and in EMJ such usage grew to displace the pronominal system, with a number of terms of address and of speaker- and self-reference, including speaker: maro, mi (‘body’ < OJ mwi), onore-ga-mi/ono-ga-mi, ware, wa-ga-mi, midukara; hearer: nandi, kimi, omape (> omawe > omae), gozen. Kimi and omae are still used today for hearer reference; the NJ speaker reference watakusi (‘he/she’ in EMJ) and ore (hearer reference in EMJ) came to be used from LMJ. The personal interrogative pronoun tare ‘who’ remained in use.
From EMJ a three-way demonstrative + interrogative system develops and by the end of the LMJ period the system is close to that found in NJ:
The short forms of OJ lexicalized with genitive no to become modifiers. EMJ did not have forms for all the slots in the set, but over time forms arose which filled the empty slots to make up a full and consistent paradigm. The shape of the forms has shown some change which may be thought to be due to paradigmatic pressure: The distal forms were kano, kare, kasiko in EMJ, but eventually changed, with higher differentiation from proximal kono and so forth. More significantly, all the interrogative forms have undergone changes such as iduku > iduko > idoko > doko (which are all attested), acquiring the same vocalism and length as the proximal and mesial forms. The initial /d/ of the interrogative forms was in NJ extended to the interrogative personal pronoun which changed from tare to dare, and the interrogatives today form a neat morphophonological class which is the result of leveling over time.
3.4.2 Verb Morphology
LMJ saw large-sale changes in verbal morphology and by the end of the period the grammatical categories expressed by finite inflection and by auxiliaries had changed profoundly (whereas the system of nonfinite, converbal forms remained largely as before), and the overall system then was more or less as in NJ today.
These are the finite inflected verb forms from the end of LMJ.
The distinction between conclusive and adnominal forms was lost, as the two forms merged in the shape of the adnominal, probably before the middle of the LMJ period. The resulting system included a nonpast opposed to a past tense. This is a major morphological change, from a system with optional tense to a system with obligatory expression of tense as a part of core verb inflection. The past tense flective attaches to the stem of the consonant base verbs which arose in the course of the onbin sound changes (see section 3.2.1), for example sai-.
The only finite form from the OJ paradigm surviving in shape and function is the imperative. The exclamatory, optative and negative conjectural were lost, whereas volitional, intentional and past conjectural were acquired. The volitional remains a core form in NJ, but the intentional was lost during NJ and also the past conjectural is now no longer used in standard NJ.
Most of the OJ/EMJ auxiliaries were lost. However, two of them are sources of new inflected forms, stative -tar- and conjectural -m-: As the conclusive/adnominal distinction and also the exclamatory form were lost, these auxiliaries thereby lost their own main, finite inflection, and they were in turn reinterpreted as flectives expressing new inflectional categories. The EMJ stative -tar-, which in the course of MJ changed to become a perfect tense, is the source of the past tense, -ta (having lost the final /ru/ of its original adnominal form, -taru), e.g., aketa, which was re-interpreted as a past tense flective in opposition with the outcome of the merger of the conclusive and adnominal of verbs without auxiliaries, as in the case of akuru, which was reinterpreted as a nonpast. Conjectural -m- is reflected in the volitional: saka-mu (adnominal of the conjectural attached to sak-) > saka-u => sakɔɔ ; ake-mu > ake-u => akyoo. Past conjectural reflects EMJ stative + conjectural: -tara-mu > -tara-u > -tarɔɔ. Also the intentional reflects a construction with the conjectural.
All of OJ respect (-s-), perfective (-te- ~ -n-), stative (-yer-), simple past (-ki / -si ), modal past (-kyer-), and subjunctive (-masi / -mase-) were lost from the language with no trace. Thus, the category of perfective disappeared, but stative was retained as a category, expressed analytically (gerund -te + i- ‘exist’) (see section 2.3.1).
Late LMJ had the following auxiliaries. As shown, the categories expressed by auxiliaries are the same as in NJ, as is to a large extent also the morphological material.
The OJ causative and passive auxiliaries were reshaped and replaced during EMJ, but the categories remained, in the shape they still have today. As opposed to OJ, causative and passive became able to combine (-sase-rare- ‘be made to’) from EMJ, and the restrictions on combination with particular verb classes disappeared. The only auxiliary to survive directly as such was the OJ negative, whose inflectional paradigm, however, did undergo some change. (In NJ, the negative paradigm has been reformed significantly, but negation remains expressed by an auxiliary in NJ, -na-, which morphologically is an adjective: saka-na-i bloom-NEG-ACOP.N[on]P[a]ST “doesn’t bloom”). A potential auxiliary, -e-, arose, used first with consonant base verbs, but in NJ spreading to vowel base verbs, in the shape -re-. Desiderative and evidential arose, as did a politeness auxiliary.
Through the LMJ period a set of changes in syntax took place that are in some way related to or involve the loss of the conclusive/adnominal distinction in verb inflection. With the merger of those two forms (and the emergence of tense as an obligatory inflectional category), the expression by verbal morphology of the distinctive functions of the two forms (see (8)–(11) in section 2.3.1) was lost. The two main forms that reflect the conclusive and the adnominal, the nonpast and the past, combined the functions of the conclusive and adnominal forms and thus came to function as:
(39)–(42) are all from the translation into Japanese of Aesop’s fables, produced and printed in 1593 in Japan by Jesuit missionaries as Esopo no fabulas.
The past and nonpast forms of NJ are still used in main clauses and in adnominal clauses, but they no longer form direct nominalizations. In this function, and thus in examples corresponding to (40)–(42), NJ uses no as a nominalizer. This use of no, which only developed in the 17th century, in NJ, served to re-establish morphological expression of one of the most important functions of the OJ/EMJ adnominal. There is, however, several centuries’ time between the loss of the conclusive/adnominal distinction and the emergence of no as a nominalizer (see Wrona, 2012).
The kakari-musubi focus construction (see section 2.3.8) was lost in the course of LMJ. This is often said to be directly related to the loss of the conclusive/adnominal distinction, but the fact that the largest of all verb classes had syncretism between those two forms (see (7) in section 2.3.1) shows that the availability of a distinct adnominal form was not a prerequisite for the kakari-musubi construction, and so it is difficult to accept a causal link between those two changes. Of the focus particles, koso came to and continues to be used as an emphatic particle, ya now means ‘and, or, or the like’, and namu was lost completely; both ka and zo changed to final particles: zo was used in exclamatives and wh-questions and ka in yes/no questions, suggesting that one important function of the adnominal verb form, namely to mark sentences as exclamative or interrogative, was taken over after the conclusive/adnominal merger by two former focus particles, first of all zo, but in yes/no questions by ka.
In NJ this has changed, so that zo is exclamative and ka is used in both wh- and yes/no questions.
Finally, the use of the two genitive particles no and ga changed in the course of LMJ: The use of no to mark subjects in subordinate clauses was gradually curtailed. As seen in several of the LMJ examples above, no could mark subjects in adnominal and directly nominalized clauses, but that was no longer possible in other subordinate clauses (and in NJ no can mark subjects in only a subset of relative clauses). No continued to have possessive and other adnominal uses as a genitive case particle. Conversely, ga lost use as a (possessive or adnominal) genitive, while retaining and eventually expanding use as a subject marker to include use in declarative main clauses, as in (39). At the end of LMJ ga had completed the transition from a genitive to a nominative case particle, and no remained as the sole genitive particle.
Frellesvig, B. (2011). A history of the Japanese language. Paperback edition. Cambridge, U.K.: Cambridge University Press.Find this resource:
Frellesvig, B., & Whitman, J. (2008). Proto-Japanese: Issue and prospects. Amsterdam and Philadelphia: John Benjamins.Find this resource:
Martin, S. E. (1987). The Japanese language through time. New Haven and London: Yale University Press.Find this resource:
Miller, R. A. (1967). The Japanese language. Chicago: University of Chicago Press.Find this resource:
Seeley, C. (1991). A history of writing in Japan. Leiden: Brill.Find this resource:
Takeuchi, L. (1999). The structure and history of Japanese: From Yamatokotoba to Nihongo. London: Longman.Find this resource:
Vovin, A. (2003). A reference grammar of Classical Japanese prose. London: RoutledgeCurzon.Find this resource:
Vovin, A. (2005–2009). A descriptive and comparative grammar of Western Old Japanese, 2 volumes. Folkestone, U.K.:Global Oriental.Find this resource:
Aoki, H., & Frellesvig, B. (forthcoming). Verb Verb complex predicates in Old and Middle Japanese. In T. Kageyama, P. Hook, & P. Pardeshi (Eds.), Verb verb complexes in Asian languages . Oxford University Press.Find this resource:
CHJ (Corpus of Historical Japanese), published by the National Institute for Japanese Language and Linguistics. Retrieved from http://pj.ninjal.ac.jp/corpus_center/chj/overview-en.html.
Frellesvig, B. (2017). On feature ranking in Japanese onset obstruents. In A. Vovin & W. McClure (Eds.), Studies in Japanese and Korean historical and theoretical linguistics and beyond (pp. 24–33). Leiden: BrillFind this resource:
Frellesvig, B. (2011). A history of the Japanese language. Paperback edition. Cambridge, U.K.: Cambridge University Press.Find this resource:
Frellesvig, B., Horn, S. W., & Yanagida, Y. (in press). A diachronic perspective on differential object marking in pre-modern Japanese: Old Japanese and Early Middle Japanese. In I. A. Seržant & A. Witzlach-Makarevich (Eds.,) The diachronic typology of differential argument marking. Berlin: Language Science Press..Find this resource:
OCOJ (Oxford Corpus of Old Japanese). Retrieved from http://vsarpj.orinst.ox.ac.uk/corpus/.
Yanagida, Y., & Whitman, J. (2009). Alignment and word order in Old Japanese. Journal of East Asian Linguistics, 18, 101–144.Find this resource:
Wrona, J. (2012). The early history of no as a nominalizer. In B. Frellesvig, J. Kiaer, & J. Wrona (Eds.), Studies in Japanese and Korean linguistics (pp. 201–220). Munich: LINCOM.Find this resource: