page 1 note 1 Yamada Yosio (hereafter, Japanese surnames will be given first), Naratyoo Buñpoosi ‘A Historical Grammar of the Nara Period’ (Tookyoo, 1913); Sir George Sansom, An Historical Grammar of Japanese (Oxford, 1928), referred to hereafter as Naratyoo and Grammar respectively.

page 1 note 2 Charles A. Hockett, A note on ‘structure’, IJAL 24.269-71 (1948).

page 1 note 3 For a fuller discussion of this subject, see § 3 The Method.

page 1 note 4 Hockett, Implications of Bloomfield's Algonquian Studies, Lang. 24.119 (1948).

page 2 note 5 The terms ‘early’ and ‘traditional’ will be used throughout this study to refer to those grammars which follow the traditional Japanese arrangement of verbs into conjugation classes called dañ ‘step’ or ‘grade’. The verbs are classified according to the vocalic endings of certain verb forms. For a detailed discussion of the traditional treatment, see § 7.

page 2 note 6 Naratyoo Buñpoosi (Tookyoo, 1913); Heiañtyoo Buñpoosi (Tookyoo, 1913).

page 2 note 7 Muromati Zidai Geñgo no Keñkyuu ‘A Study of the Language of the Muromati Period’ (Tookyoo, 1929); Tokugawa Zidai Ceñgo no Keñkyuu ‘A Study of the Language of the Tokugawa Period’ (Tookyoo, 1933).

page 2 note 8 Yamada, Nihoñ Buñpooroñ ‘Theory of Japanese Grammar’ 603, Naratyoo 366; Sansom, Grammar 273.

page 2 note 9 Yamada, Naratyoo 193; Sansom, Grammar 187.

page 2 note 10 Sansom describes it thus, see Grammar 273.

page 3 note 11 The possibility of regarding such morphs as tum- (tumu 'piles up') and sak- (saku 'blooms') as verb roots has occurred to Japanese grammarians, see Haga Yaiti, Buñpooroñ 'Theory of Grammar' 72 ff.; Yamada Yosio, Nihoñ Buñpoogaku Cairoñ ‘Introduction to the Study of Japanese Grammar’ 35. But it remains only a theoretical possibility to them.

page 3 note 12 A comparison of the method used here with the traditional treatment of inflected forms will be given in some detail in § 7.

page 4 note 1 The anthologies do not agree as to the total number of poems. The figure varies from 4,313 to 4,560.

page 4 note 2 It was intended to leave all poems from books 14 and 20 out of the main body of material, regardless of whether they showed dialectal differences or not, but several new inflectional forms which were what one might expect to find, deducing from the rest of the material, appeared in books 14 and 20. These have been incorporated into the analysis, but marked with an asterisk, thus *uwetekeru.

page 5 note 3 The translations used were J. L. Pierson's Manyôśû, vols. 1-6, and The Manyōshū, a selection of one thousand poems published by the Nippon Gakujutsu Shinkōkai (Nippoñ Gakuzyutu Siñkookai).

page 5 note 4 Syntactic environment is significant for this material in that it conditions the particular form of an inflected word. Some of the environments are absolute clause-final position, clause-final position with particles before or after, position within a clause with particles, and position before a noun or before another inflected form.

page 5 note 5 The books consulted were Origuti Sinobu's Mañyoosyuu, a complete translation of the poems into modern colloquial Japanese, 3 vols.; Takeda Yuukiti's Mañyoosyuu Siñkai ‘New Interpretation of the Mañyoosyuu’, a translation of selected poems into modern colloquial Japanese, 3 vols. (referred to hereafter as Siñkai); Sasaki Nobutuna's Mañyoo Ziteñ ‘A Dictionary of the Mañyoosyuu’, 1 vol. (referred to hereafter as Ziteñ); Masamune Atuo's Mañyoosyuu Daiziteñ ‘A Dictionary of the Mañyoosyuu’, only vol. 1, words beginning with ‘a’; and Sano Yasutaroo and Huzii Kañ's annotated edition of the Mañyoosyuu, 1 vol.

page 5 note 6 Examples of verbs and adjectives will be given in the indicative mood.

page 5 note 7 Only one occurrence has been noted, in poem 2877 (the poems are numbered according to the numbers in the Kokka Taikañ ‘Thesaurus of Japanese Poems’, which are regarded to be the standard), but it does not appear to affect the following verb form as na mo does in the prose material: Itu ha na mo kohizu ari to ha aranedo mo ‘though it is not that I do not love you at all times’. One would expect the verb aru rather than ari, i.e. the contingent mood rather than the indicative.

page 6 note 8 The use of Chinese characters as it developed in Japan made it possible for one character to represent several different (approximately synonymous) words. For a discussion of this subject, see ‘Introduction of Writing’, Sansom, Grammar 1-40.

page 6 note 9 Pierson, Manyôśû 1.15.

page 6 note 10 The katakana and the hiragana are two shapes of the same syllabary. One or the other is often used to indicate the proper reading of the Chinese characters. The term kana is used to designate both.

page 6 note 11 Mañyoosyuu Soosakuiñ, 4 vols. (Tookyoo, 1929), hereafter referred to as the Index.

page 7 note 12 Koohoñ Mañyoosyuu, 5 vols. (Tookyoo, 1924). The Kañei edition of 1643 is also used as the basic text for this work.

page 7 note 13 The writer has not been able to consult this work.

page 7 note 14 Actually no new forms were observed which could be attributed to a unique rendering of the Chinese characters by Masamune.

page 8 note 1 The following description is based on Dr. Bernard Bloch's lectures in his courses, Language Structure, spring term 1948, and Linguistic Analysis, spring term 1949, Yale University; and on readings in descriptive linguistics.

page 8 note 2 Although in the presentation of the analysis, morphophonemics precedes morphemics, in the actual analysis of the material, morphemics is the prerequisite of morphonemics.

page 8 note 3 Hockett, Problems of Morphemic Analysis, Lang. 23.322 ff. (1947).

Sometimes a meaning is found which cannot be assigned to any utterance fraction with overt form. In such cases, zero morphs are set up. However, no zero morph can be set up as the sole member of a morpheme; it must belong to a group of one or more morphs with overt phonemic shapes. A zero morph, then, is set up to account for a morpheme whose meaning is present in the utterance fraction, but for which no overt form is found. As a result, statements of tactics become less complicated.

In this study it has not been necessary to use other convenient devices such as empty morphs and portmanteau morphs. See ibid.

page 9 note 4 This description of the processes involved in morphemic analysis is based on Hockett's description, ibid. 341-2, except for this particular section on grouping requirements where a slightly different set of criteria is used. Hockett uses three criteria for grouping morphs, the first being similarity of meaning; the second, non-contrastive distribution which includes free variation as well as complementary distribution; and third, non-uniqueness in the total range of environments of a morpheme.

page 9 note 5 The latter constitutes the cases of automatic selection and non-automatic selection respectively (see §4.2.3, §4.2.4).

page 9 note 6 Because of this assumption, only the criterion of complementary distribution (see above) is used in the grouping of morphs, rather than Hockett's criterion of non-contrastive distribution, which also includes free variation.

page 10 note 7 A few instances have been found, however, of morphs which differ phonemically, but are apparently dialect variants with identical meaning. These have been considered, therefore, as variants of the same morpheme. They are the durative morphs er/ar (§4.3.4.7) after verb bases, the indicative and contingent morphs u/o (§4.3.5.1, §4.3.5.2) after the presumptive morph m, the directive morphs mu/mo (§4.3.5.9), and the precative morphs ne/ni (§4.3.5.10). In the Aduma dialect (see § 6, examples B D 1 ff.), the durative morph after verb bases is ar, and the indicative and contingent mood morphs after the presumptive morph m is o (see examples B P 1, B P 2). No example has been found in the Aduma dialect for the directive and the precative morphs mo and ni, but these are possible variant forms, judging from other correspondences between the two dialects.

page 12 note 1 For a diagram showing the order and distribution of the morphemes, see § 4.4.1.1 Diagram of Structural Type I.

page 12 note 2 A study of S. Yositake's The Phonetic System of Ancient Japanese (London, 1934) shows that there are possibly three extra vowel phonemes in ancient Japanese (up to 750 A.D.) as compared to those represented by the kana syllabary (invented about the 9th century, and which is assumed by this writer to be phonemic writing). This means that the phonemes /e/, /i/, and /o/ as used in this work, may possibly each represent either of two phonemes /e/ and /ε/, /i/ and /I/, and /o/ and /ɔ/ respectively. To transcribe the poems into what may possibly be the more accurate phonemic orthography, would require the prodigious task of studying each Chinese character, which is used phonetically to represent one of these three pairs of vowels sounds, to determine whether it is /e/ or /ε/, /i/ or /I/, etc. Since the purpose of this study is to analyze the morphological structure of the language, all that is relevant or necessary is to isolate and to identify the morphs in some manner. This can be done by using the kana orthography, as Japanese scholars have done who give the reading of the Chinese characters inevitably in kana. The problem here is not so much to give the correct phonemic shape to the morphs as to identify them. Strictly speaking /e/, /i/, and /o/, as used here, are not phonemes in the usual sense. A special name, perhaps, should be given to them. But to avoid adding a new term to a problem which is only discussed once here, the term ‘phoneme’ will be kept, but with the proviso that the three ‘phonemes’ /e/, /i/, and /o/ may possibly represent either of two phonemes.

Note Yositake's discussion of the phonetic value of the phoneme represented in this study by /h/, which he represents with the letter ƒ (ibid. 46-50).

From the extant material of this period, there is no way of arriving at accent, either of pitch or stress, which might indicate morphological distinctions.

page 13 note 3 In the transliteration of Japanese words, especially proper names and titles of books, other than those in the 8th-century material, the modern pronunciation of the kana is kept.

page 13 note 4 The mañyoogana makes some distinctions between u and wu, e and ye, i and yi, see Yositake, op. cit. 17-25, 27-9, 48, 54-61.

page 13 note 5 Bernard Bloch, Studies in Colloquial Japanese I, JAOS 66.98 fn. 3 (1946); referred to hereafter as SCJ.

page 14 note 6 A capital letter (s) usually represents a whole morpheme, including all the morphs belonging to it. In the case of the verb and adjective bases, however, the capital letter(s) represents a class of morphemes. When a particular member morph of a morpheme is meant, it is given following the symbol for the morpheme. In the case of the adjectival morpheme, however, the member morph given represents a set of morphs which occur with certain adjective bases (see §4.3.3 Adjective Classes).

page 14 note 7 In the symbolic representation of a given morpheme sequence, the letter B stands for the verb base as well as for the adjective base. The latter is distinguished from the former by the presence of the adjectival morpheme which occurs linked with it. The capital letters represent the premodal suffix morphemes: active A, stative S, negative N, perfective T, presumptive P, hypothetical H, and durative D. The Arabic numerals represent the modal suffix morphemes: indicative 1, contingent 2, exclamatory 3, nominal 4, concessive 5, conditional 6, provisional 7, optative 8, directive 9, precative 10, imperative 11, conjunctive 12, gerund 13, and continuative 14.

page 15 note 8 Represents verb classes (see §4.3.2).

page 15 note 9 The symbol P-H represents the linked morpheme pair presumptive-hypothetical. Other linked morpheme pairs are B-J (adjective base and adjectival morpheme), 8-9 (optative and directive), 12-13 (conjunctive and gerund), and 12-14 (conjunctive and continuative).

page 15 note 10 In the durative structural type, the active, negative, perfective, and the durative morphemes have a second position after the first durative morpheme (see §4.4.2.1). Morphemes in this position are marked with a prime when written in symbols.

page 16 note 11 It is possible to interpret this either as the variable initial vowel predominating, or the variable final vowel predominating. The first interpretation has been adopted to make the summary statement simpler.

page 16 note 12 These two verb classes have only one member each, the verbs su ‘does’, and ku ‘comes’ (see §4.3.2 Verb Classes).

page 17 note 13 Hereafter the symbol for the base will be omitted, whenever the class of verb base is given.

page 17 note 14 Verb classes II, IIIa, and IV also take the nominal morph uraku.

page 19 note 15 The Roman numerals after the adjective base indicate the adjective class to which it belongs, i.e. the particular set of adjectival morphs with which it occurs (see §4.3.3 Adjective Classes).

page 19 note 16 The numerals represent mood morphemes; the capital letters, the premodal morphemes, see fn. 7.

page 19 note 17 Class III adjectival morph zik does not occur before D (durative). The term ‘occurs’ as used in this study means occurrence as observed in the text.

Conversely, the non-occurrence of a form means that it was not observed in the texts. This may mean either that the non-occurrence is fortuitous, or that it is characteristic of the language. In some instance, it is possible to deduce from the distribution of certain morphemes which it is; in other instances, it is impossible to say.

page 21 note 18 Ramu usually follows verbs in the indicative mood: koh-u ramu B 1 ‘seems to love’. In addition, it follows class Ib verbs, and verbs and adjectives with the durative morpheme immediately preceding the mood morpheme, in the contingent mood: ar-u ramu Ib 2 ‘there probably is’; sak-er-u ramu B D 2 ‘is probably blooming’; tahuto-k-ar-u ramu B-J D 2 ‘is perhaps noble’. Ramu also follows class IIIb verbs in the conjunctive mood: m-i ramu IIIb 12 ‘will probably see’.

page 22 note 19 See fn. 7.

page 24 note 20 Besi usually follows verbs in the indicative mood: miy-u besi B 1 ‘should see’. In addition, it follows class Ib verbs, and adjectives with the durative morpheme immediately preceding the mood morpheme, in the contingent mood: ar-u besi Ib 2 ‘ought to have’; ure-sik-ar-u besi B-J D 2 ‘may be happy’. It also follows class IIIb verbs in the conjunctive mood: m-i besi.

page 24 note 21 Gotosi follows verbs and adjectives in the contingent mood and is sometimes preceded by the particle ga: tir-in-uru gotoki B S 2 ‘like the falling’; toho-k-i ga gotoki B-J 2 ‘as if far-distant’. It also follows a noun plus the particle no: titose no gotoki ‘like a thousand years’.

page 24 note 22 Rasi usually follows verbs in the indicative mood: id-u rasi ‘seems to leave’. It also follows class Ib verbs in the contingent mood: ar-u rasi ‘there seems to be’, and class IIIb verbs in the conjunctive mood: n-i rasi ‘seems to be boiling’.

page 25 note 23 If it were not for onazi, it would be possible to regard the z in these adjectives as a negative morph, but no negative meaning can be found for the very common adjective onazi.

page 25 note 24 Masizi occurs after verbs in the indicative mood: ahi-katu masizi ‘not likely to meet’. It does not occur after adjectives. It is found in both the prose and poetic works of this period. Mazi, a later form of masizi according to Japanese grammarians, occurs only in the prose works. The only other inflected form of masizi (mazi) found was masiziki (maziki).

page 27 note 25 The word(s) enclosed in brackets in the text are ‘pillow words’ which are conventionalized metaphors associated with a particular word (in the first example, haru) which they precede.

page 27 note 26 In both poems the active and stative morphs occur with the indicative mood morph u.

page 27 note 27 In the first half of both poems the stative morph in occurs with the indicative mood morph u before the adjective rasi, as haru tat-in-u rasi and aki-duk-in-u rasi. In both cases, it describes a state.

page 27 note 28 Both morphs occur in these two poems with the presumptive morph m and the contingent mood morph u. The presumptive morph has the meaning of ‘probability’ here; the contingent mood morpheme occurs before clause-final particles ka mo.

page 28 note 29 The Roman numerals refer to the standard number of the señmyoo, of which there are sixty-two extant.

page 28 note 30 Affirmation is expressed simply by the absence of a negative morpheme, or by the presence of the active or the stative morpheme. In actual distribution, as well as in meaning, the negative morpheme is mutually exclusive with the active and the stative morphemes (see §4.4.1.1 Diagram of Structural Type I).

page 28 note 31 There are two negative conjunctive forms zu-0 and n-i. The form n-i is much more limited in distribution than zu-0. It is found with only three verbs, siru ‘knows’, aku ‘tires’, and -gatu ‘withstands’ which forms compounds with other verbs. The form zu-0 occurs freely with many verbs, including the verbs siru and aku, but not with -gatu. The two forms, aside from the verb bases, share similar environments, occurring at the end of clauses, and immediately preceding the verb su ‘does’, but only n-i occurs before the particle to: kotahe wo ihi yaramu sube wo sira-n-i to tatite tumazuku ‘not knowing how to answer, thus standing, I falter’. The form n-i occurs chiefly with the verb siru (vocalic base alternant sira-) in such common phrases as semu sube siran-i ‘not knowing what to do’, or tazuki sira-n-i ‘not knowing the means of doing.‘ The traditional grammars explain n-i as an archaic form which survived until the Nara period, after which it became obsolete. (see Yamada, Naratyoo 189; Takeda, Siñkai 1.70; Pierson in The Manyôśû 3.234-6 discusses this form at some length).

page 29 note 32 This matter will be discussed more in detail in the section on durative constructions, (§4.4.2), where examples will be given, providing contrasts in meaning.

page 32 note 33 See §4.4.2 Structural Type II: Durative, for a fuller discussion.

page 32 note 34 Occurs chiefly after class Ia verb bases, though there are a few instances of the morph er occurring with classes IIb (su 'does'), IIIb (kiru 'wears') and IV (ku 'comes') verb bases.

page 32 note 35 The morphs er(a) and ar(a) both occur with the adjectival morpheme but not in free variation. The morph ar(a) occurs with J in all instances, but er(a) with J only if it is followed by a mood morph beginning with the vowel e, thus: tika-k-er-edo ‘although near’.

page 32 note 36 With the one verb motu ‘holds’ Ia, the durative morph ar occurs in the indicative mood, mot-ar-i ‘is holding’ (one instance, in poem 1790), and in the contingent mood, mot-ar-u aga ko ‘the child I am holding’ (eleven instances). But with other mood morphs, for example the concessive, the durative morph er occurs, thus: mot-er-edo ‘though holding’. Since in the Aduma dialect, the durative morph ar occurs regularly after the verb base, the variant forms given above may be Aduma dialect forms. The picture is further complicated by the fact that there is found one instance of the durative er occurring in the contingent mood with the verb motu in book 14 where poems in the Aduma dialect are found, thus: yoti wo zo mot-er-u ‘has a child about the same age’ (3440).

page 34 note 37 There is only one example of this morph occurring as a possible dialect variant of u after P m (3886).

page 34 note 38 Before the verb ramu, the adjectives besi and rasi, and the concessive particles to mo, verbs of this class occur in the conjunctive mood rather than in the indicative: m-i ramu instead of mi-ru ramu ‘seems to see’, m-i beki ‘should see’, etc.

page 34 note 39 The last three particles follow only adjectives.

page 34 note 40 Verbs of class Ib, for example ari, and inflected forms with a durative morph immediately preceding the mood morpheme, occur before ramu, besi, and rasi in the contingent mood rather than in the indicative (see fnn. 18, 20, and 22 for examples. No example has been found of these verbs occurring before the adjective masizi).

page 35 note 41 There are three examples of this morph occurring as possible dialect variant of u after P m (614, 1655, 4105).

page 35 note 42 Certain adjectives of Class II occur before the particle cluster ka mo with the adjectival morph s and the indicative mood morph i, instead of the adjectival morph sik and the contingent mood morph i, for example, kuya-s-i ka mo ‘oh, how regrettable’.

page 35 note 43 So far as it can be determined the meaning of the morpheme here does not differ from that of the indicative mood morpheme. The presence of one of the particles listed above calls for this particular mood morpheme.

page 35 note 44 Certain adjectives of classes I and III occur before nouns with the adjectival morphs s and z and the indicative mood morph i, instead of the adjectival morphs k and zik and the contingent mood morph i, for example, uma-s-i kuni ‘sweet land’, ona-z-i koto ‘the same thing’. Certain adjectives of class II occur before nouns with the adjectival morph s and the indicative mood morph i instead of the adjectival morph sik and the contingent mood morph i, for example, naguha-s-i Yosinu ‘the famous Yosinu’.

page 36 note 45 Bloch, SCJ II, Lang. 22.219 (1946).

page 37 note 46 The particles ya and ya mo occur also after the indicative mood morpheme in clause-final position (see §4.3.5.1). There, they turn a statement into a question without any rhetorical meaning. There are several instances of the question particle cluster ka mo occurring after the exclamatory morpheme in clause-final position with rhetorical meaning in books 14 and 20, but since there were no instances found in the main body of material, ka mo has not been included here.

page 37 note 47 The particle in parentheses indicates that it occurs somewhere earlier in the sentence.

page 38 note 48 No examples of the morph re occurring with class IIIb verb base, or of the morph ure with classes Ic, IIa, IIIa, and IV verb bases are found. These omissions are fortuitous. Judging from the conditional mood morphs which are homophonous with the exclamatory mood morphs, it can be safely assumed that the morphs re and ure would occur with the verb bases indicated.

page 38 note 49 Cf. contingent morpheme §4.3.5.2 occurrence number (5) of the inflected forms which appear with the same particles.

page 38 note 50 Not to be confused with a small group of adjectives derived from nouns, which in the conjunctive mood are homophonous with adjectives in this mood, for example: tahira-ke-ku-0 ‘is tranquil and’.

page 43 note 51 See fn. 31 for a discussion of the two negative conjunctive forms.

page 43 note 52 Occurs as a linked morpheme pair with the gerund and the continuative (12-13 and 12-14).

page 43 note 53 The particle ha follows the negative conjunctive form zu-0 in such expressions as ohokata ha kohitutu ara-zu-0 ha, sinamu masareri ‘without loving so indifferently, it would be better to die’, i.e. ‘rather than to love so indifferently, it would be better to die’. Ha also follows the conjunctive form of the adjective nasi ‘is without’ in clause-final position with much the same meaning.

page 43 note 54 Cf. §4.3.5.2 Contingent Morpheme and §4.3.5.4 Nominal Morpheme, both of which occur within a clause after the referent particle ha. The difference in meaning of these three morphemes with ha is difficult to distinguish, but may be found in the translations given. Contingent: yuku ha ‘the act of going’, ‘to go’, ‘those who go’; nominal: yukeku ha ‘the fact of going’, ‘that one goes’; conjunctive: yuki ha ‘the going’. Sansom in distinguishing between the contingent and the conjunctive morphemes, says that the latter ‘stands for a comprehensive or abstract idea’, while the former pictures the act or even sometimes the agent of the action (Grammar 138).

page 47 note 55 The analysis of morphemes by position, i.e. the study of their relative order of occurrence within a word, ultimately derives from Zellig S. Harris's diagram in which he shows substitutable morphemes arranged in columns (Structural Restatements: II, IJAL 13.175-86, 1947). Subsequently, at a meeting of the Yale Linguistic Club in the summer of 1948, Floyd G. Lounsbury presented a modification of this method, wherein relative order of occurrence of morphemes replaces substitutability as the prime criterion for classification. He uses this type of analysis in his study of Iroquian morphology (Comparative Iroquian Morphology, ms., Yale University Dissertation, 1949), where he describes the method at some length.

page 49 note 56 These morphemes occupy the same position within a word and are, therefore, mutually exclusive (see §4.4.1.1 Diagram).

page 59 note 57 An alternative analysis of the durative structural type is to regard the durative morpheme as a derivational suffix, forming complex bases, for example, B S T D, which are equivalent grammatically, but not semantically, to the simple base B. If this analysis is adopted, the second durative morpheme D╹, would likewise be a derivational suffix, forming complex bases such as B A D T D╹, equivalent grammatically to the simple base B. It is to be noted that the more complex such bases become, the more restricted they are in the suffixes which they take, the derived base B A D T D╹, for example, occurring only before M (see §4.4.2.1 Diagram).

The reason for not regarding the durative structural type as a derived form but a new inflectional type is that, insofar as it can be judged from a preliminary study of derivation, the formation of a derived base from an inflected form is an unusual derivational process in Japanese. Moreover, by treating the durative structural type as an inflectional type, its relation to the non-durative forms which it closely resembles in its constituent morphemes, can be shown and comparisons made.

page 69 note 1 ‘Josi’ are introductory verses similar to ‘pillow words’ but usually longer (more than five syllables in length). They are not as common as ‘pillow words’. One example is found in B S P-H 1 (1746).

page 83 note 1 A good part of book 20 consists of poems written by Yakamoti, one of the chief poets in the Mañyoosyuu whose poems are also found in books 17, 18, and 19. Several inflected forms found nowhere else in the Mañyoosyuu have been found in Yakamoti's poems. Since these are not Aduma forms, but forms which one might expect in the Yamato dialect, judging from the other Yamato forms, they have been incorporated into the main analysis, but marked with an asterisk.

page 83 note 2 In the examples hereafter, Yamato will be abbreviated to Y, Aduma to A.

page 84 note 3 The verb base yuk- is also found in the Aduma dialect.

page 85 note 4 Another possible Yamato equivalent is ko-n-u ni ‘as he does not come’ where the negative morph occurs in the contingent mood rather than in the conjunctive. This form actually occurs more frequently than ko-zu-0 ni, and moreover, there is a corresponding set Y n-u: A na in the example given above under B N 2 with the particles ka mo. The Yamato form ko-n-u ni has, however, an adversative meaning which the Aduma form kona ni does not have. In meaning, then, the latter is closer to the Yamato form ko-zu-0 ni ‘without coming’. Moreover, in the next Aduma example given nena na narinisi ‘it came to be that they no longer slept together’, the corresponding Yamato form would be ne-zu-0 ni narinisi, since the negative conjunctive is found before the verb naru ‘becomes’, with or without the particle ni. No case of the contingent negative has been observed before naru. For these reasons, it seems that ko-zu-0 ni is the more probable Yamato equivalent.

page 85 note 5 The usual explanation given of this form (Yamada, Naratyoo 515; Takeda, Siñkai 2.791; Sasaki, Ziteñ 309) is that the first na is the negative morph, the second is the optative morph. However, it is difficult to find an optative meaning in the six examples found; the meaning seems to be simply conjunctive, ‘without doing so and so, not doing so and so and’. Furthermore, there is no example of a Yamato form where the negative morph occurs with the optative morph. Since an example has already been found of an Aduma form na na corresponding to the Yamato form zu-0 ni, the Yamato form suggested here is hure- zu-0 ni, i.e. the negative morph in the conjunctive mood occurring with the particle ni, rather than the negative morph na occurring with the optative mood morph na. It might be further added in support of this interpretation that there are other examples of the particle ni occurring in the Aduma dialect as na.

page 86 note 6 The morpheme sequence B N P 3 does not occur in the Yamato dialect. The closest equivalent to this form would be B N D P 3 kake-z-ara-m-e ‘will not be hanging indeed’, a durative structural type. The negative morpheme offers the greatest variation in the Aduma dialect.

page 87 note 1 In this discussion, only the treatment of verbs will be compared, since what can be said of the verbs is essentially true of the adjectives as well.

page 87 note 2 The term ‘stem-forms’ will refer specifically to those verb forms which appear in the traditional arrangement of verb conjugations, and is not to be confused with the term ‘stem’ as used in descriptive linguistics.

page 88 note 3 This line represents the indicative forms of the verbs.

page 88 note 4 These correspond actually to suffixes, inflected words, and particles in modern descriptive terms. Grammarians after Motoori have redefined his tenioha as suffixes, inflected words, and particles, but their analysis differs in many respects from the one given in this study. In order to avoid confusion in terminology, the collective term tenioha as used by Motoori will be kept.

page 89 note 5 The tenioha ñ, keñ, and rañ correspond to the following in 8th-century Japanese: ñ to the presumptive and indicative suffixes m-u; keñ to the perfective, presumptive, and indicative suffixes ke-m-u; rañ to the class la verb ramu ‘seems, is likely’ which always follows another inflected form (see §4.3.2).

page 89 note 6 The inflected word meri ‘seems to be’ does not occur in 8th-century Japanese.

page 89 note 7 This verb does not occur in the 8th-century material.

page 89 note 8 In a previous discussion of this subject (see §1 The Problem) these same verb forms were chosen as examples, but a slightly different interpretation was given in order to limit the discussion at that time to simply illustrating the principle of segmentation. Here the description will take into account more facts concerning these verb forms.

page 90 note 9 Some suffixes are considered to inflect like verbs. This is discussed later.

page 90 note 10 Most frequently described as a suffix by the traditional grammarians, e.g. Yamada, Naratyoo 126. Sansom calls besi an ‘auxiliary adjective’ (Grammar 109-12), or ‘an inflected adjectival suffix’ (ibid. 127).

page 90 note 11 For convenience in description, the name most commonly used traditionally which is in small capitals here, has been adopted.

page 92 note 12 Sansom, Grammar 276.

page 92 note 13 Nihoñ Buñpoosi 40-1.

page 92 note 14 In segmenting the conjugation 2 verb, miru ‘sees’, into a base and a suffix morph, the fraction i is assigned to the base since it has a precedent in occurring with the base, as in mi-ba ‘if he sees’, mi-zu ‘he does not see’. The cut, therefore, is made mi-ru, with the suffix taking the vocalic base alternant. In the case of the conjunctive form mi, there is a choice of regarding it as a vocalic base miplus a zero morph (mi-0), or as a consonantal base m- plus the suffix i (m-i). Both the conjunctive morphs 0 and i occur elsewhere, but of the two, the latter is the more regular form, so the choice here is i rather than 0. This necessitates setting up a consonantal base alternant m-, but this is in conformity with the general practise of having both consonantal and vocalic alternants for verb bases.

page 93 note 15 Sansom, Grammar 127.