¹ The first paper in the series, dealing with inflection, appeared in JAOS 66.97-109 (1946); the third, dealing with the derivation of inflected words, will appear in JAOS, Vol. 66, No. 4 (1946). These two papers will be referred to as SCJ I and SCJ III respectively.

My thanks are due to the American Council of Learned Societies, which granted me a fellowship in the Intensive Language Program from February to August 1942, and thereby made it possible for me to begin the study of Japanese. I am grateful also to the many Japanese informants with whom I have worked, in particular to Lincolna McKinnon, Mikiso Hane, Toshio Kono, Kentaro Ikeda, Masako Yokoyama, and Miyo Okada. Mr. Kono kindly checked the phrasing and accentuation of the examples used in the present paper.

In arriving at the final form of this article I have been greatly assisted by discussions with Rulon S. Wells, Henry M. Hoenigswald, and Elizabeth F. Gardner, and above all by the careful and illuminating criticism of Leonard Bloomfield. The more general debt which I owe to Bloomfield is obvious on every page.

² The only approach to such an attempt that I know of is Henry Saburoo Tatsumi's Simplified grammar table of the spoken Japanese (Tokyo, 1936). Tatsumi tries to exhibit the over-all structure of Japanese sentences on a single chart (approximately 31 by 40 inches); but his arrangement is impossibly cumbersome, and his analysis both unscientific and incomplete. The first sentence under the heading Syntax indicates the general approach: ‘The usual word order in a Japanese sentence is as follows: (A) Subject, (B) Time, (C) Means of transportation, (D) Place, (E) Indirect Object, (F) Instrument, Means, Material or Agent, (G) Direct Object, (H) Complements, (I) Quotations, (J) Predicate verb or adjective, (K) Auxiliary suffixes.‘

³ But the period will be omitted when the Japanese expression is followed by a gloss.— The transcription used here is a modification of the official Japanese romanization (Kokutei Rōmazi), as set forth in SCJ I, fn. 4. The chief points in the phonemic analysis are repeated here.

A syllable contains a syllabic, alone or with one non-syllabic preceding. Syllablcs are of two kinds, vocalic and consonantal. Vocalic syllablcs are a, o, u, e and ya, yo, yu, ye; the second group palatalize a preceding non-syllabic, or, when forming a syllable alone, begin with a palatal glide. But ye is [i], and will be written i for simplicity. Consonantal syllables, never initial after a pause and never preceded by a non-syllabic, are ñ and q. The former is [m] before a labial, [n] before a dental, [ŋ] before a velar, elsewhere [n] or [ŋ] in free variation. The syllabic q is voiceless and homorganic with the following non-syllabic; it occurs only in the sequences qp, qk, qt, qs, phonetically [p·,k·, t·, s·], here written pp, kk, it, ss respectively. (The writing of pp for qp, etc., conforms to traditional romanization. In SCJ I, the writing with q was used because it makes morphophonemic statements easier.)

Thirteen consonants function as non-syllables : h, p, k, t, s, b, g, z, m, n, r occur before all the vocalic syllables; d occurs only before a, o, e; and w occurs only before a. Between voiceless consonants, u and i (ye) are voiceless; so also is u final after s in rapid conversation, esp. in the forms -másu and désu. Note that ty is phonetically [č], sy is [š], zy is [dž] (rarely [ž]), tu is [tsu], hu is [fu] or [φu] or [hu].

On intonations and pauses see §§1.1-2 in the text; on the word accent see §1.4 below. Capital letters are used in the transcription according to the rules of English orthography. In the glosses, round parentheses enclose explanatory matter, literal translations, and equivalents of Japanese words not required in the English sentence; square brackets enclose English words not corresponding to any words in the Japanese expression.

⁴ There is a minimal pause-group ma meaning ‘space, room, interval’, and another minimal pause-group ti meaning ‘blood’; but neither of these is to be regarded as part of the word matí ‘town’. The semantic criterion by which these three pause-groups are kept separate is the same as the one that is used to identify morphemes; cf. Leonard Bloomfield, Lang. 2.155 (1926) and Language 161 (New York, 1933).

⁵ Accordingly, if the accented syllable is the last or the only syllable of a word, the high pitch is perceptible only when the word is followed without pause by another word with a low-pitched first syllable. When such a word is not so followed, it loses its accent. We infer from this that no accented syllable is ever followed by another accented syllable; i.e. no word has an accent on its last or only syllable when the following word has one on its initial syllable.

In citing words with an accent on the last or only syllable, and pause-groups ending with such a word, we shall nevertheless indicate the place of the accent by an acute. Our transcription in this one respect is thus morphophonemic.

^5a Unless one of the preceding syllables is q, i.e. the first part of a long voiceless consonant (fn. 3). Such a syllable is always neutral with respect to pitch: it is neither high nor low. Thus in the word massúgu ‘ straight’ (more strictly maqsúgu), the syllable ma is low-pitched, q is neutral, sú is high-pitched, and gu is low-pitched ; in the word mizuppói ‘ is watery’ (more strictly mizuqpói), the neutral syllable q stands between the two high-pitched syllables zu and pó.

⁶ The division dó notaté- is ruled out. The syllable ta is low-pitched, but would be high-pitched if it were the second syllable of a word with accent on the third.

⁷ The accent does not rule out the possibility that the last syllable (-i) might be a separate word; but morphologic considerations lead us to regard it as a suffix. Knowing tábeta to be a separate word, we might attempt to analyze tabetá rasíi as a phrase, explaining the shift of accent (tábeta > tabetá) as a function of the environment. But since rasíi has high pitch on both the first and second syllables, it cannot be a separate word.

In emphatic speech, some formations with -rasíi have an accent on the first syllable instead of the second-last; e.g. tábetarasii, emphatic variant of tabetarasíi. But other formations have no such variant; e.g. musumerasíi ‘seems to be [his] daughter’. In discussing these formations below (§1.5 and §3.12, and fnn. 21, 22) we shall disregard the emphatic variants with initial accent.

⁸ Cf. Bloomfield, Language 183: ‘ In languages which use bound forms, the word has great structural importance because the constructions in which free forms appear in phrases differ very decidedly from the constructions in which free or bound forms appear in words. Accordingly, the grammar of these languages consists of two parts, called syntax, and morphology.‘

⁹ I owe the formulation of this principle to Rulon S. Wells.

¹⁰ It is convenient to make an exception to this rule in the case of synonymous words which differ only by the presence or absence of the phoneme q before p, t, k, s, or of the phoneme ñ before m, n. Thus totemo and tottemo ‘quite, very’, sírokute and sírokutte ‘being white’, amari and an̄mari ‘too much’ are alternants of the same word respectively.

^10a Cf. Zellig S. Harris, Morpheme alternants in linguistic analysis, Lang. 18.169-80 (1942).

¹¹ However, some non-final clauses occur at the ends of utterances as minor sentences; see §2.3, §8.2.

¹² Elinor M. Clark has counted sentence types in a large sample of recorded dialog, chosen as an extreme type of rapid, elliptical conversation. She found the following proportions: major sentences 58%, exclamations 15%, other fragments 27%. (On the distinction between exclamations and other fragments see §8.) The relatively high proportion of exclamations is probably due to the presence of a drunken man among the speakers. It is doubtful whether any sample of conversational Japanese would show a lower proportion of major sentences.

¹³ Cf. Bloomfield, Language 171 ff.

¹⁴ For the morphology of inflected words, and for terms relating to inflectional categories, reference is here made once and for all to SCJ I, JAOS 66.97-109.

¹⁵ Some sentence particles and clause particles are homonymous with particles of other types, occurring in different environments. In accordance with the convention stated in §1.6 (No. 3), we regard such homonyms as different words.

¹⁶ These classes will be named and defined in SCJ III.

¹⁷ In addition to the three types of verb phrases enumerated here, there is a combination consisting of the verb áru ‘is, there is’ or its equivalent preceded by the gerund of a copula phrase; e.g. heitai de áru ‘is a soldier’. Combinations of this kind are more conveniently classified as copular expressions, and will therefore be treated in §3.13.

¹⁸ The immediate constituents of this last expression are shown by the following arrangement: [〈(isóide) arúite〉káette] kuru.

¹⁹ Nouns will be defined in §5.1.

²⁰ Noun expressions will be defined in §5.1 and §5.3-6. The noun expressions used as examples in this paragraph are limited to the simplest type : a single noun, with or without one modifier preceding. Relational phrases will be defined in §4.3.

²¹ The scope of -rasíi resembles the scope of the English possessive suffix; cf. Bloomfield, Lang. 2.156. In the phrase the King of England's daughter, the immediate constituents are the King of England's and daughter; the immediate constituents of the prior expression are the King of England and -s: the suffix, though attached to the noun England by a morphologic construction, has a syntactic scope which includes the whole preceding phrase. (Note that the possessive suffix sometimes appears at the end of a phrase whose last word is not a noun: the man that went by just now's daughter; the man I saw's daughter.) Japanese and English are thus alike in that each has one bound form that combines with phrases in a syntactic construction.

²² We treat -rasíi in this way because we wish to keep our description of the syntax as free as possible from exceptions and special cases. By recognizing that -rasíi is unique among bound forms in serving as a constituent of syntactic constructions, we escape the necessity of describing anomalous combinations—e.g. a noun modifier (§5.2) preceding an adjective, as tomodati no ‘of a friend’ in the expression tomodati no, musumerasíi, or kono ‘this’ in the expression kono kadorasíi ‘seems to be this corner’.

The semantics of the construction agree with our analysis. An adjective group ending with -rasíi means ‘seems to be or do x, looks like x, is x-like’; the x is the meaning of the whole noun expression or clause preceding -rasíi within the adjective group. When someone says of a girl, Tomodati no> musumerasíi, he means not merely that she seems to be a daughter, but that she seems to be a friend's daughter; when he says of her, Tamanégi o, tabetarasíi, he means not merely that she seems to have eaten, but that she seems to have eaten onions: the meaning of tomodati no or of tamanégi o is part of the x. However, the semantics of the construction are not enough to make the analysis here adopted necessary or even particularly desirable. Similar semantic relations obtain in expressions that we analyze differently. Thus the expression tamanégi o, tabetái ‘I want to eat onions’ (where the suffix -tai attached to the infinitive of the verb tabéru carries the desiderative meaning) might be described as having the immediate constituents tamanégi o, tábe ‘eating onions’ and the bound form -tai ; but we find it simpler in the long run to say that the phrase tamanégi o here is not part of the predicate, but rather a clause attribute preceding the predicate tabetái ‘I want to eat’.

²³ In this expression the combination de wa is optionally replaced by zya—a morphophonemic alternation found also in other contexts. Cf. SCJ I, §8.3 (JAOS 66.109).

²⁴ Not all questions end with ka. Some questions are distinguished from statements only by rising intonation; e.g. Moo, kité irá n̄ da? ‘Has he come already?'. When such a question has a presumptive copula phrase as its final predicate, the meaning of the question is often rhetorical; e.g. Nakanaka, zyoozú daroo? ‘He's pretty skilful, isn't he?'.

²⁵ The theoretical combination dá ka ne (where dá is the plain non-past indicative of the copula) is usually replaced by one of two equivalent combinations : by ka ne, with dé omitted (i.e. replaced by a zero alternant); or by da ne, with ka omitted. The question Nihoñzín̄ ka ne? cited in the text occurs also as Nihoñzín̄ da net—In rapid speech, the sequences ka ne and da ne are replaced by the conglomerates kai and dai respectively ; thus Nihoñzín̄ kai? and Nihoñzín̄ dai?

²⁶ In these examples, square brackets enclose the English equivalent of a possible final clause added after the clause particle to complete the sense.

²⁷ This flavor of courtesy is of course entirely conventional and often largely fictitious. The use of polite predicates does not necessarily imply, and is not necessarily understood to imply, any genuine respect on the speaker's part. The difference between plain and polite predicates may be compared to the difference between French tu and vous, German du and Sie.

²⁸ Even in very polite speech, adjectives are almost never in the polite state outside a final predicate and certain types of non-final predicates. The provisional form of the copula (nára, náraba) has no polite correspondent.

²⁹ For the details of the morphologic structure see SCJ III, §3.5.

³⁰ What is here said of adjectives applies equally to adjectives constituting adjectival expressions alone and to adjectives serving as nuclei of longer expressions.

³¹ A predicate consisting of a copula plus a particle (with no other words preceding the nucleus within the predicate) is never preceded by a clause attribute.

³² The difference is due to the difference between the particles which follow the noun hón̄ in the two sentences. When a noun is followed by the particle ga, o, wa, or mo, a quantity noun (§5.1) referring to it may precede or follow as an adverbial phrase. When a noun is followed by any other particle, such an adverbial phrase always precedes.

^32a The adverb áruiwa is initial in the clause (and is therefore classified as a connective adverb) except in discontinuous noun expressions of the type discussed in §5.51 below.

³³ Relational phrases may be symbolized as follows. In the simplest type, a relatum— say a noun n—is followed by a particle p. When the relatum is itself a relational phrase, we have (n p) p. When the relatum is a relational phrase whose own relatum is a still smaller relational phrase, we have [(n p) p] p. Note that in each case the immediate constituents are the final particle on the one hand and everything that precedes it on the other.

³⁴ The grammatical subject of a Japanese predicate is the word or the expression most intimately connected with it. It is often but not always the ‘actor’.

³⁵ Note that the predicate in this sentence is the negative adjective inai. In the synonymous sentence cited above (Áni dake, iru), the predicate is the underlying verb. A relational phrase with daké followed by an affirmative predicate is equivalent to a relational phrase with sika followed by a negative predicate.

³⁶ The noun before ka is always an interrogative, defined in §5.1.

³⁷ The relatum in a relational phrase with wa is unemphatic, in the sense that it does not constitute the novel or important part of a statement. Compare Heitai wa, kureta with Heitai ga, kureta above (both sentences translated ‘The soldier gave it to me') : the former might be an answer to the question Heitai wa, náni o, sitá ka ‘What did the soldier do?'; the latter might be an answer to the question Dáre ga, agetá ka ‘Who gave it to you?'

³⁸ Hypostatized elements in Japanese are regularly treated as prior constituents of quotational phrases. On hypostasis see Bloomfield, Language 148.

³⁹ Note that we regard this word as different from the connective adverb to ‘so saying’ (§4.2), the referent particle to ‘with’ (§4.3), and the clause particle to ‘whenever’ (§3.22). There is also a conjunctive particle to ‘and’ (§5.41). Cf. §1.6, No. 3, and fn. 15.

⁴⁰ This sentence and the preceding each contain two quotational phrases. In the second sentence, the phrase dóo ka contains the zero alternant of the copula dá; cf. §3.13 end, and fn. 25.

⁴¹ The difference in use between the series ití—zyúu and hitótu—tóo is one of the details which fall outside the scope of this paper.

⁴² This classification of nouns into syntactic subdivisions is independent of the morphologic classification into such groups as complex and compound. Only participial nouns are characterized as a class by a particular morphologic structure (see §6.21). The derivation of nouns will be treated in a later article.

If it should seem strange to apply the term ‘noun’ to such words as yori ‘more than’, iroiro ‘various’, dekisóo ‘looking as if it were possible’, and arukinágara ‘while walking’, our justification is that these words fit the definition of nouns that we have adopted. Grammatical terms must be defined anew for every language; and the choice of particular terms (for instance ‘noun‘) is at best a matter of convenience.

⁴³ The last three words are homonymous with three words belonging to other word classes : the noun onazi ‘the same’, and the verbs káku ‘ writes’ and áru ‘is, there is’. In the expression onazi desu ‘it is the same’, onazi is a noun; in the nearly synonymous expression onazi kotó desu ‘it is the same thing’, onazi is a demonstrative.

⁴⁴ As these examples show, a clausal modifier often corresponds to an English relative clause, either restrictive or non-restrictive.

⁴⁵ Since no, the alternant of the copula, is homonymous with the referent particle no ‘of, some expressions are ambiguous. Isya no, ozi means ‘my uncle, who is a doctor’ if no is the copula, but ‘ the doctor's uncle’ if no is the particle. Since the head of each expression is the noun ozi ‘uncle’, there is no way of distinguishing them except by meaning. To analyze a sequence A no B (where A and B are nouns), we apply the following semantic test: if the statement A da ‘someone or something is A’ provides a description of B, then no is the copula; if not, it is the particle.

⁴⁶ This is the meaning of the sentence if to is the conjunctive particle ‘and’. If to is the referent particle ‘with’, the sentence means ‘A soldier was talking with a sailor’. This ambiguity is resolved most directly by a semantic criterion: to is a conjunctive particle only if the sentence is equivalent in meaning to the two statements Súihei ga, hanásite ita ‘A sailor was talking’ and Heitai ga, hanásite ita ‘ A soldier was talking’, taken together. That is, if A and B are nouns, and p is a conjunctive particle, and x is the element that follows the series (a referent particle, the copula, or the predicate of the clause), then the sequence A p B p x or A p B x is equivalent to the two sequences A x and B x taken together, or to either A x or B x taken separately. For this semantic criterion it might be possible to substitute a rather complicated distributional criterion based on the selection of the nouns before and after the particle in question. It is not possible to base the identification of conjunct nouns simply on the selection of the conjunctive particle, since three of these (to, ni, and ka) are homonymous with referent particles.

⁴⁷ An epigram expressive of the impermanence of perfection, quoted from H. G. Henderson, Handbook of Japanese grammar 202 (Boston etc., 1943). On the sentence type see §8.1, fn. 51.

⁴⁸ If it were not for the parallelism, we could regard the pause-groups hitóri and hutarí in the former sentence, and the pause-groups nán̄nin̄ and nán̄nin̄ in the latter, as constituting two separate adverbial phrases. The analysis given in the text is preferable because it recognizes the semantic and structural similarity of the sentences with and without to.

⁴⁹ There may be other bound morphemes deriving participial nouns, but only these two are known to me. The morpheme -nagara appears also in (non-participial) nouns derived from nouns, e.g. bin̄boonagara ‘although [someone] is poor’ from bín̄boo ‘pauper’; -sidai appears both in nouns derived from nouns and as a noun by itself, with the meaning ‘ order, precedence’. But these uses of the morphemes in question we regard as different from their use in deriving participial nouns.

⁵⁰ Some referent particles occur also after other elements (adverbs, demonstratives, pseudo-clauses, etc.); but it is the occurrence after nouns that defines the class.

⁵¹ A third type, collocation, appears chiefly in proverbs and epigrams. This consists of two parallel parts, each containing one or more adverbial or relational phrases. Examples: Íma no, hitóhari; noti no, tóhari ‘One stitch now, ten stitches later’ (i.e. A stitch in time saves nine) ; Kane no, kiremé ga; éñ no, kiremé ‘The end of money, the end of love (lit. affinity)‘. Cf. also the epigram cited in §5.41 (fn. 47).

⁵² Cf. the list in J. K. Yamagiwa, Modern conversational Japanese 16-7 (New York and London, 1942), from which some of these examples are taken.

⁵³ Arthur Rose-Innes, Kana 65-85 (= Japanese reading for beginners, Vol. 1; Yokohama etc., [1923]).