Hideki Kishimoto and Satoshi Uehara
2 Lexical categories
1 Introduction
Lexical items play diﬀerent grammatical roles depending on the categories they
belong to. Along with the distinction of the vocabulary strata – Native Japanese
(NS), Sino-Japanese (SJ), Foreign (F), and Mimetic (M) – the proper classiﬁcation of
lexical categories (word classes) is essential to discovering any grammatical rules
and principles in Japanese grammar. In the domestic grammatical tradition, consideration
of word classes has a long history starting as early as in the late 1600s, and
TŌJŌ Gimon (東条義門, 1786–1843) is generally credited with establishing the
fundamental bipartition of word classes into yōgen (inﬂecting class) and taigen
(non-inﬂecting class) that persists today. This bipartition is based on the morphological
criterion of the presence or absence of inﬂectional endings, with yōgen (inﬂecting
class) comprising verbs and adjectives, and taigen (non-inﬂecting class) nouns.
Such a polar opposition, however, will be called into question if there are categories
that fall in between the two poles. In fact, contemporary Japanese oﬀers at least two
categories that appear to be located between the two, namely Verbal Nouns (VN)
and Adjectival Nouns (AN). The present chapter will survey the characteristics of
Japanese lexical categories from generative and cognitive perspectives, with due
attention to the proper treatment of VNs and ANs in the system of Japanese word
classes.
School grammar (Japanese grammar taught at high schools) represents perhaps
the most widely spread classiﬁcation of Japanese lexical categories, which are
divided into two major groups: jiritsugo or “independent categories” in Table 1 and
fuzokugo or “dependent categories” in Table 2.
Table 1: Independent-form categories in school grammar
verb
(dōshi)
yomu ‘read’
agaru ‘climb’
kuru ‘come’
suru ‘do’
adjective
(keiyōshi)
utukusii ‘beautiful’
kanasii ‘sad’
yoi ‘good’
adjectival noun
(keiyōdōshi)
sizuka da ‘quiet’
kirei da ‘pretty’
derikeeto da ‘delicate’
barabara da ‘separate’
noun
(meishi)
hito ‘person’
anata ‘you’
terebi ‘TV’
wanwan ‘doggie’
adverb
(fukushi)
hakkiri ‘clearly’
sukosi ‘a little’
prenominal modiﬁer
(rentaishi)
arayuru ‘every’
aru ‘a certain’,
ano ‘that’
ookina ‘big’
conjunction
(setsuzokushi)
sikasi ‘but’
mata ‘or’
interjection
(kantōshi)
hai ‘yes’
aa ‘oh’
Table 2: Dependent-form categories in school grammar
auxiliary ( jodōshi) particle ( joshi)
(s)ase ‘causative’, -(r)are ‘passive’, -nai/ nu
‘negative’, -(y)oo ‘hortative’, -tai ‘desiderative’,
-masu ‘polite’, -ta/ da ‘past, perfect’, soo da
‘reportive’, -mai ‘negative expectative’, yoo da
‘modality (uncertainty)’, -rasii ‘modality
(hearsay)’, -da/ desu ‘copula’, etc.
1. case: ga (NOM), no (GEN), o (ACC), to (COM),
kara (ABL)’, de (INST), etc.
2. conjunctive: keredo ‘though’, ga ‘but’, ba ‘if’,
node ‘since’, etc.
3. adverbial: wa (TOP), mo ‘also’, sae ‘even’,
sika ‘only’, etc.
4. sentence-ﬁnal: na(a) ‘exclamatory’,
zo ‘emphatic’, yo ‘assertive’, etc.
In regard to vocabulary strata, verbs and adjectives are conﬁned to Native Japanese
words. Nouns could be Native Japanese (kuruma ‘car’), Sino-Japanese (gakkoo
‘school’), foreign (terebi ‘TV’), or mimetic (wanwan ‘doggie’) (and verbal nouns,
which are not included in the list of lexical categories in school grammar, come
from all the lexical strata, as in dokusyo(-suru) ‘reading (SJ)’, otetudai(-suru) ‘help
(NJ)’, tesuto(-suru) ‘test (F)’, tekuteku(-suru) ‘walk (M)’). Adjectival nouns are also
found in all the lexical strata, as in sizuka da/na ‘quiet (NJ)’, ganko da/na ‘stubborn
(SJ)’, derikeeto da/na ‘delicate (F)’, barabara da/na ‘separate (M)’.
The classiﬁcations in Tables 1 and 2 are far from deﬁnitive, and many diﬀerent
views are available in the literature; e.g. in modern Japanese linguistics (couched in
the Western linguistic paradigms), verbal nouns are often seen as constituting a
major lexical category, alongside verbs, nouns, adjectives, adjectival nouns (see
Martin 1975; Shibatani 1990), but they are included in the category of nouns in
traditional Japanese grammar. It is not an easy task to sort out Japanese lexical
categories, because, in Japanese, inﬂection/conjugation is exhibited in limited categories,
and in many cases, no clear boundaries can be drawn between free and bound
elements. Accordingly, a number of issues arise with regard to the question of how
lexical categories are distinguished and how many categories should be recognized.
This chapter reviews a number of theoretical and descriptive issues on lexical
categories. Section 2 brieﬂy surveys the classiﬁcation and terms of lexical categories
discussed in traditional grammar. Section 3 provides a discussion from the generative
perspective, addressing the question of how “adjectival nouns” and “verbal
nouns” – neither of which ﬁt into the classical cross-categorial feature system positing
only two features – can be deﬁned. It is also shown that certain adjectivallyinﬂecting
elements traditionally classiﬁed as dependent auxiliaries count as lexical
adjectives, despite their morphological status. Section 4 examines the nature of
Japanese lexical categories from the cognitive-typological perspective. The structural
organization of the overall lexical categorization of the language and its major
lexical categories (Noun, Verb, Adjective, Verbal Noun, and Adjectival Noun) are
characterized in terms of cross-linguistic markedness patterns and their functional
motivations are identiﬁed.
52 Hideki Kishimoto and Satoshi Uehara
2 Parts of speech in traditional Japanese grammar
This section provides a concise survey of how parts of speech were analyzed by some
pioneers of Japanese grammar in the domestic tradition of Kokugogaku philology.
Parts of speech were mentioned in grammar books written by European missionaries
who visited Japan (e.g. Rodriguez 1604–08), and in the pre-modern era (17th to 19th
centuries), Kokugaku scholars (FUJITANI Nariakira 富士谷成章, MOTOORI Haruniwa
本居春庭, and others) made various attempts to classify inﬂecting words. Among
them, Fujitani (1778) proposed a sophisticated “parts of speech” system of dissecting
sentences into four major parts of na (noun), yosohi (predicate), ayuhi (particle,
auxiliary), and kazashi (adverb and others), while TŌJŌ Gimon (1841), mentioned in
Section 1, established the distinction of inﬂecting and non-inﬂecting word classes,
i.e. yōgen and taigen. Later, in the late 19th century, ŌTSUKI Fumihiko 大槻文彦
(1889) proposed a system of eight parts of speech in line with the Western tradition
of modern linguistics in another important work investigating the notion of “word”
and “parts of speech”. Subsequently, in the 20th century, the early giants of Japanese
grammar (YAMADA Yoshio 山田孝雄, MATSUSHITA Daizaburō 松下大三郎,
HASHIMOTO Shinkichi 橋本進吉, and TOKIEDA Motoki 時枝誠記) also proposed
diﬀerent views on parts of speech, which are still inﬂuential and debated today.
Among others, Hashimoto’s ideas about lexical categories were incorporated in
school grammar (Monbushō 1947) and are taught even today. As noted in the Introduction,
the inventory of lexical categories includes dōshi (verb), keiyōshi (adjective),
keiyōdōshi (adjectival noun), meishi (noun), fukushi (adverb), rentaishi (prenominal
modiﬁer), setsuzokushi (conjunction), kantōshi (interjection), jodōshi (auxiliary) and
joshi (particle). The listed categories are categorized by two distinct sets of criteria,
morphological and distributional (or syntactic), as represented in the taxonomy
structure of (1), where M stands for morphological criteria and D for distributional
criteria.1
1 Categories are distinguished by making use of a feature system here, although this is not a com
mon practice in school grammar. In generative linguistic research, and also in traditional Japanese
grammar, lexical/grammatical categories (word classes, parts of speech) are distinguished on the
basis of their shared morphological and syntactic properties. Semantic criteria are not commonly
used for this purpose because, in many, if not all, cases, they are not considered to demarcate
categories in a reliable manner. For instance, the semantic criterion taking noun and verb to denote
an entity and an event, respectively, does not distinguish nagare ‘ﬂow (n.)’ as in mizu no nagare
‘water ﬂow’ from nagareru ‘ﬂow (v.)’, both of which denote events in one way or another; nagare
refers to an event, and yet functions as a noun.
Lexical categories 53
(1)
In (1), words are ﬁrst divided into two major groups, depending on whether they are
dependent or independent ([±dependent]). Both categories are further partitioned
into two classes, according to whether they inﬂect or not ([±inﬂected]). All the
inﬂecting categories are identiﬁed as predicates [+predicate], and non-inﬂecting
categories are divided into two sub-types, depending on whether or not they can
function as subjects ([±subject]). Non-inﬂecting categories that cannot act as subjects
are further divided into two classes, depending on whether they can serve as
modiﬁers [±modiﬁer].
The predicative group marked with [+predicate], which comprises verbs, adjectives,
and adjectival nouns, is traditionally labeled yōgen ‘predicative’, and nouns
are the sole member of the class labeled taigen ‘nominal’. Yōgen is divided into two
types, as in (2a), which are distinguished according to whether or not they have
imperative inﬂections (represented as [±imperative]).
(2) a. b.
Modiﬁers are divided into fukushi (adverbial) and rentaishi (prenominal modiﬁer),
which are distinguished by the kind of element modiﬁed. Setsuzokushi is used for
connecting clauses, and kantōshi for forming ‘exclamatives’, which may or may not
be single-word sentences.
54 Hideki Kishimoto and Satoshi Uehara
Of the categories listed above, keiyōdōshi (adjectival noun), rentaishi (prenominal
modiﬁer), jodōshi (auxiliary), and joshi (particle) call for special mention, as they are
not commonly found in European languages or are diﬀerent from their European
counterparts. Keiyōdōshi, referred to in this chapter as “adjectival noun” (a term
coined by Martin 1975), is chosen for the sake of consistency over competing English
translations, and the expression consists of a stem and an inﬂectional ending.
The inﬂectional endings of keiyōdōshi (e.g. adnominal na and conclusive da) are
morphologically distinct from those of adjectives (e.g. adnominal and conclusive
-i). As explained below, the category of keiyōdōshi poses a perennial question for
both Kokugogaku and Japanese linguistics in line with the Western tradition.
Rentaishi (‘prenominal modiﬁer’), which is often given the alternative term ‘prenominal
adjective’, include not only demonstratives like ano ‘that’, but also such adnominal
expressions as aru ‘a certain’, iwayuru ‘so-called’, taisita ‘considerable’, tiisana ‘tiny’,
and ookina ‘big’ that occur only in prenominal position. Jodōshi (‘auxiliary’), perhaps
the most undeveloped, problematic category in traditional grammar, is actually a
conglomeration of heterogeneous dependent elements that are agglutinated to the
preceding predicative elements, such as suﬃxes of passive, causative, honoriﬁcation,
spontaneity, and potential, as well as functional endings of tense, negation,
modality (see Chapter 13 [Takezawa, this volume] for inﬂection). Finally, various
particles labeled as joshi are dependent elements that follow noun phrases, clauses,
and the like (but are not agglutinated to them).
Although the classiﬁcations in school grammar are said to be largely based on
work by Hashimoto (1934, 1948), Hashimoto (1934) actually presented a ﬁner-grained
classiﬁcation including not only jiritsugo (independent category) but also fuzokugo
(dependent category), as in yōgen (verb, adjective), taigen (noun, pronoun, numeral),
fukushi (adverb), rentaishi (prenominal modiﬁer), setsuzokushi (conjunction), kantōshi
(interjection), jodōshi (auxiliary), as well as joshi (particle), a category that includes
various types of particles such as adverbial, nominal, conjunctive, coordinate, quasiadverb,
case, kakari, sentence-ﬁnal, and interjective particles. (In school grammar,
keiyōdōshi is treated as an established lexical category (cf. Hashimoto 1948), but in
his earlier work (Hashimoto 1934), Hashimoto eschews including keiyōdōshi in the
list of yōgen as an independent lexical category in modern Japanese, reserving the
judgment on the issue as to whether it should be analyzed as a simple word or a
complex expression, although he argues explicitly for its unique status there.)
The question of how words are classiﬁed involves a fundamental and yet highly
subtle issue, one that remains far from settled even today. The debates in the
Kokugogaku tradition largely center around the two somewhat related questions:
(i) whether it is plausible to posit keiyōdōshi (adjectival noun) as an independent
category, and (ii) how many parts of speech need to be recognized. Answers to these
questions may vary, depending not only on how much emphasis one places on what
kinds of criteria (morphological or distributional), but also on what kind of language
one analyzes (classical or modern). In traditional grammar, it was customary to
Lexical categories 55
consider category classiﬁcations based on classical Japanese or a mixture of classical
and modern Japanese. The ﬁrst question concerning keiyōdōshi is rooted in the structure
of classical Japanese. The name keiyōdōshi (lit. ‘adjectival verb’) derives from its
characteristic forms in classical Japanese, X-nari and X-tari, which exhibited verbal
inﬂections and inﬂected even for the imperative. In modern Japanese, where the
classical nari and tari have disappeared, however, keiyōdōshi has lost such erstwhile
verbal behavior. Hashimoto (1948) claimed that keiyōdōshi constitutes a lexical category
standing all on its own (although this view was only implied in Hashimoto
(1934)). On the other hand, Yamada (1908, 1936) considered this categorization
unnecessary on the grounds that the classical nari and tari originate historically
from ni ari [LOC be] and to ari [COMP be], where ari ‘be’ is a verb of existence. On
Yamada’s view, keiyōdōshi is thus composed of what he calls fukuyōgen ‘subpredicate’
(words that are neither inﬂected as predicates nor function as grammatical
arguments) followed by predicates of existence (sonzaishi ‘existential’ in
Yamada’s terminology).
In modern Japanese, keiyōdōshi is “inﬂected” with the endings na in adnominal
form and da in conclusive form. The ending na comes from the classical combined
form ni + ari [LOC + be], and the da ending is identical in form to the copula da as
in Nihon no syuto wa Tookyoo da [Japan GEN capital TOP Tokyo COP] ‘The capital
of Japan is Tokyo.’ This and other considerations led Tokieda (1950) to analyze
keiyōdōshi into a special type of nominal followed by the copula. The view that the
inﬂectional ending of keiyōdōshi is nothing but the copula is shared by Bloch (1946)
and Okutsu (1978). On the other hand, Mikami (1953), inspired by Sakuma (1951),
holds that adjectives and adjectival nouns make up a larger class of adjectivals,
the former represented as i-keiyōshi (i-adjective) and the latter as na-keiyōshi (naadjective).
The terms “i-keiyōshi” and “na-keiyōshi” are standardly used today in
textbooks for teaching Japanese as a foreign language, whereas “keiyōshi (adjective)”
and “keiyōdōshi (adjectival noun)” are taught in domestic school grammar.
The nature of the two adjectival categories will be discussed in great detail in the
subsequent sections of this chapter (see also Chapter 14 [Kageyama, this volume]
for the morphological status of the na/da ending in keiyōdōshi).
The writings of such great grammarians as Hashimoto, Yamada, Matsushita, and
Tokieda present diverse answers to the second question of how many parts of speech
should be distinguished. Although there is general agreement on the usefulness of
the fundamental distinction between taigen (nominal) and yōgen (predicative), these
grammarians disagree on the classiﬁcation of major categories and functional categories.
Yamada (1908, 1936) identiﬁes four categories consisting of taigen, yōgen,
fukushi (adverb), and joshi (particle) but does not regard auxiliaries as an independent
class. Matsushita (1924, 1930) does not identify auxiliaries and particles as
categories, because they are dependent, thus upholding the classiﬁcation of meishi
(noun), dōshi (verb), fukutaishi (adjective), fukushi (adverb), and kandōshi (exclama-
56 Hideki Kishimoto and Satoshi Uehara
tive). Tokieda (1950) includes taigen, yōgen, daimeishi (pronoun), rentaishi (prenominal
modiﬁer), and fukushi (adverb) as “core” lexical categories.
The various classiﬁcations that have been proposed in Kokugogaku are not without
problems from the perspective of modern linguistics. One characteristic feature
of traditional Japanese grammar is that lexical categories are ﬁrst divided into two
major classes according to whether they count as free or bound forms in morphological
terms. In fact, in all the classiﬁcations noted above, morphological criteria take
precedence over distributional criteria, but it is not entirely clear whether this
should be considered the optimal way of characterizing lexical categories.
In particular, the agglutinative character of Japanese makes it diﬃcult to tell
whether dependent elements appearing after main predicates should be analyzed
as auxiliaries or inﬂectional suﬃxes. A number of issues arise from this fact. One
such issue concerns the question of whether the element da combined with adjectival
noun stems should be regarded as a copula or an inﬂectional suﬃx (see Section
4). Copulas are dependent elements, but often regarded as non-aﬃxal, unlike other
auxiliaries, leading to the issue over whether adjectival nouns constitute single or
complex words. Another question is how the past (or perfective) tense marker
should be classiﬁed. Traditionally, it is often classiﬁed as an auxiliary, distinct from
the present tense element, which is taken to be part of an inﬂectional form, while
Bloch (1946) identiﬁes both present and past tense markers as inﬂectional suﬃxes.
The typological character of Japanese as an agglutinative language also raises an
issue over whether there should be a tight correlation between the distinction of
“agglutinated” versus “non-agglutinated (free)” forms and the “lexical” versus
“functional” distinction (see Section 3).
3 Lexical categories in the generative perspective
We are now in a position to discuss how “adjectival nouns” and “verbal nouns”
can be described from the generative perspective. It is shown in this section that
adjectivally-inﬂecting auxiliaries sometimes display syntactic behavior shared with
lexical adjectives, and that in Japanese, the categorical distinction of “lexical” and
“functional” categories is not tightly correlated with the morphological status of
words.
3.1 Criteria for classifying major categories
In generative grammar, mainly two kinds of distributional criteria – morphological
and syntactic – are used to identify categories (see, e.g. Carnie 2007). The morphological
criteria are based on inﬂections and derivations. Diﬀerent inﬂectional forms
Lexical categories 57
are usually found depending on their categories: in Japanese, verbs have inﬂectional
forms distinct from those found in adjectives or adjectival nouns, and nouns do not
inﬂect for person and number. The derivational criteria work more or less in a similar
way. In Japanese, adjectives as well as adjectival nouns can be identiﬁed as falling
into the adjective class by looking at whether the suﬃx –sa can be attached, since it
derives a noun from an adjectival expression. The syntactic (or distributional) criteria
are also usable for categorizing words, because diﬀerent categories have diﬀerent
syntactic distributions. A noun can appear in the frame ga nai [ NOM NEG]
‘there is no ’, but the same slot cannot be ﬁlled by an adjective, an adjectival
noun, a verb, or a postposition; e.g., hon ‘book’ is identiﬁed as a noun, because it
can appear in this frame.
Lexical categories consist of a number of word classes, which include the
major categories of noun (N), verb (V), adjective (A), preposition/postposition (P)
(i.e. content words), as well as several other minor categories (i.e. function words).
In Chomsky (1970), the lexical categories are not conceived of as primitives, but are
deﬁned by implementing cross-categorial features, i.e. the four major categories of
V, N, A, and P in English are characterized in terms of the features [±V, ±N].
(3) a. V: [+V, −N] b. N: [−V, +N] c. A: [+V, +N] d. P: [−V, −N]
One obvious advantage of the cross-categorial feature system, compared with positing
just distinct categories (or having simple binary classiﬁcations), is that it allows
us to capture commonalities across categories. In Chomsky’s system, the maximal
number of categories that can be deﬁned with the features is four. On the empirical
level, however, it is obvious that the cross-categorial feature system used to deﬁne
the categories in English often cannot be straightforwardly carried over to other
languages, including Japanese. (Even in English, no consensus has been reached
about the classiﬁcation of major lexical categories (content words), and in particular,
the exact categorical status of ‘preposition’ is often called into question;
Emonds (1985) and Baker (2003) suggest that prepositions fall into the class of
functional categories).
Japanese is often seen as possessing verbal nouns and adjectival nouns as major
lexical categories, alongside verbs, nouns, and adjectives (see Martin 1975, Shibatani
1990). Apparently, Japanese has more categories than can be deﬁned by the features,
so the feature system has raised an issue over how lexical categories should be
deﬁned. Kageyama (1982) suggests that the lexical categories in Japanese be deﬁned
with reference to the feature [±A], in addition to [±V, ±N] (see also Kageyama 1993).
(4) a. verb: [+V, −N, −A] b. noun: [−V, +N, −A]
c. adjective: [−V, −N, +A] d. adjectival noun: [−V, +N, +A]
e. verbal noun: [+V, +N, −A]
58 Hideki Kishimoto and Satoshi Uehara
Note that in (4), adjectives are characterized as [−N, −V], unlike Chomsky’s feature
characterization [+V, +N]. They are characterized negatively by the respective features
because [+A] is added. In Kageyama (1982), adjectival nouns are claimed to constitute
an independent lexical category, on the grounds that they display dual behavior as
nouns and adjectives. To be concrete, adjectival nouns pattern with adjectives, in
undergoing nominalization with the addition of the suﬃx –sa, which derives nouns
from adjectives.
(5) a. sizuka-sa b. utukusi-sa c. *tabe-sa d. *enpitu-sa
quiet-NMLZ beautiful-NMLZ eat-NMLZ pencil-NMLZ
‘quietness’ ‘beauty’ ‘eat-ness’ ‘pencil-ness’
On the other hand, adjectival nouns diﬀer from adjectives, in that the former, but not
the latter, allow their stems to be combine with the aﬃx –rasii. Adjectival nouns
pattern with nouns, since both of their stems can be combined with -rasii, as in (6).
(6) a. sizuka-rasii b. *utukusi-rasii c. gakkoo-rasii
quiet-looking beautiful-looking school-looking
‘look quiet’ ‘look beautiful’ ‘look like a school’
Neither nouns and adjectives show exactly the same behavior as adjectival nouns, so
adjectival nouns are characterized as [−V, +N, +A] in Kageyama’s system.
In Kageyama (1982), verbal nouns are regarded as constituting a category distinct
from the categories of noun and adjective, on the basis that verbal nouns serve
not only as arguments occurring with case particles, but also as predicates when
combined with suru ‘do’ (and the latter is not possible with nouns; e.g. *enpitu-suru
‘pencil-do’). (Note that verbal nouns can behave as predicates without suru when
they are combined with the suﬃx –tyuu, -sai, -ori and the like (see Kageyama 1993)).
(7) a. soodan-suru b. soodan o motikakeru
consultation-do consultation ACC bring.up
‘consult’ ‘ask for consultation’
Accordingly, verbal nouns are assigned the feature [+V, +N, −A], which is distinct
from nouns speciﬁed as [−V, +N, −A]. In his analysis, the commonalities shared by
nouns and verbal nouns are captured by the feature [+N], whereas the properties
which verbal nouns share with verbs are characterized by the feature [+V] (and verbal
nouns and verbs are assigned [−A], as they do not possess properties shared with
adjectives).
While Kageyama (1982) postulates the feature [±A], which is not implemented in
Chomsky (1970), Miyagawa (1987) advances an alternative analysis in an attempt to
Lexical categories 59
dispense with this extra adjectival feature and proposes an alternative characterization
of the lexical categories.
(8) a. verb: [+V, −N] b. noun, verbal noun: [−V, +N]
c. adjective: [+V] d. adjectival noun: [+V, +N]
In Miyagawa’s analysis, adjectives are conceived of as possessing the feature [+V],
while the feature [±N] being neutralized. In contrast, adjectival nouns are associated
with both [+V] and [+N]. One indication that adjectival nouns, but not adjectives, are
associated with [+N] is found in the adjectival nouns’ ability to occur with copula, in
an analogous way with nouns, as in (9).2
(9) a. kirei da b. sensei da c. *utukusi da
pretty COP teacher COP beautiful COP
‘be pretty’ ‘be a teacher’ ‘be beautiful’
Further, on the basis that adjectival nouns can be combined with soo ‘looking,
likely’, alongside verbs and adjectives, Miyagawa defends the position that adjectival
nouns are assigned the feature [+V].
(10) a. sizuka-soo b. utukusi-soo c. tabe-soo d. *gakkoo-soo
quiet-looking beautiful-looking eat-looking school-looking
‘likely quiet’ ‘likely beautiful’ ‘likely to eat’ ‘likely school’
Miyagawa claims that the generalization that adjectival nouns, adjectives, and verbs
can combine with soo ‘looking, likely’ can be optimally characterized by assuming
that these three categories have the feature [+V] in common.
In Miyagawa’s analysis, adjectival nouns are characterized as having [+N, +V],
whereas adjectives, which are marked by the feature [+V], are neutral with respect
to the feature [±N]. One problem with the proposed characterization is that the
nominalizing aﬃx -sa attaches to both adjectival nouns and adjectives, to the exclusion
of verbs and nouns, as we have seen in (5). Miyagawa claims that this distribution
can be captured on the assumption that -sa can be attached to a category with
[+V] that does not assign any case (i.e. verbs belong to the category that assigns
case, but adjectival nouns and adjectives do not). Under this proposal, adjectives
2 The adnominal form of the copula da is no when the preceding element is a noun, but the copula
has the form na when it is preceded by an adjectival noun. In Miyagawa’s analysis, the diﬀerence in
morphological inﬂection between nouns and adjectival nouns is taken care of by adjustment rules.
Miyagawa provides two more arguments which he claim can be used to distinguish adjectival nouns
from adjectives, i.e. the attachment of mitai ‘seem’ and reba ‘if’. Ohkado (1991) points out some
problems in using them as diagnostics, however.
60 Hideki Kishimoto and Satoshi Uehara
and adjectival nouns can be classed together by virtue of the case requirement
imposed on the categories to which -sa attaches, without recourse to the additional
feature [±A].
Another claim made by Miyagawa is that verbal nouns are equated with nouns
in terms of categorical features, the diﬀerence being reduced to the question of theta
role assignment. According to Miyagawa, the fact that suru ‘do’ directly attaches to
verbal nouns, but not nouns like enpitu (*enpitu-suru [pencil-do]), comes from the
condition that suru needs to inherit a theta role from a combined theta role assigner,
which he assumes is imposed independently of the category of the theta role
assigner. If a noun does not have any theta role to assign, as in enpitu, the resulting
N-suru is not capable of acting as a predicate, and hence is unacceptable. This
analysis allows us to account for the fact that verbal nouns (equipped with theta
roles to be assigned), but not simple nouns, can occur with the verb suru, without
any need to set up the extra lexical category of verbal noun.
Both Kageyama and Miyagawa attempt to capture generalizations across the
major lexical categories (content words) in terms of cross-categorial features. Even
though the issue concerns the status of major lexical categories, there are also minor
categories (function words), which are often identiﬁed with the categorical feature
[+F] (or [−L]) instead of [+L] assigned to a major lexical category (Abney 1987; Fukui
1986). Needless to say, just as English has minor functional categories such as
pronouns, auxiliary verbs, etc., so too does Japanese, which leads to another issue.
We will turn to this discussion in the next section.
3.2 From major to minor categories
In Japanese, auxiliaries, which are inﬂecting suﬃxal/agglutinative elements attaching
to predicates, are dependent categories, and many of them are likely to be
categorized as functional categories. Nevertheless, it is also true that some, if
not all, have an obvious link to major lexical categories, partly because the former
have often been derived from the latter via grammaticalization (see Hopper and
Traugott 1993), as can be inferred from the fact that auxiliaries often retain their
morphological forms of the original categories from which they are derived; some
display adjectival properties (or inﬂections) (e.g. the desiderative ta(i) ‘want’, the
hearsay rasi(i) ‘likely’), and others display verbal behavior (e.g. the passive rare(ru),
the causative sase(ru)).
The fact that auxiliaries can be divided into verbal and adjectival types can also
be discerned conspicuously by looking at what supportive verb is inserted when
tense is separated from them, as in (11).
Lexical categories 61
(11) a. {yomi/yom-are/yom-ase} mo su-ru
{read/read-PASS/read-CAUSE} also do-PRS
(lit.) ‘do also {read/be read/cause to read}’
b. {utukusiku/nomi-taku} mo ar-u
{beautiful/drink-want} also be-PRS
(lit.) ‘be also {beautiful/want to drink}’
The supportive verb suru ‘do’ is inserted to the left of tense for morphological support
when an adverbial particle separates the tense from the passive rare and the
causative sase, just like the main verb yomu ‘read’. This fact suggests that the auxiliaries
rare and sase belong to the verbal class (cf. Kishimoto 2013). In contrast,
when tense is separated from an adjective like utukusii ‘beautiful’ or the desiderative
auxiliary tai, the supportive verb aru appears to the left of the tense marker. Since
both elements pattern together, this fact shows that the auxiliary tai belongs to the
adjectival type.
In traditional Japanese linguistics, inﬂecting elements that are dependent, i.e.
those that cannot appear in isolation, are categorized as auxiliaries (by deﬁnition).
Given that major lexical categories are independent forms, it is tempting to think
that auxiliaries are assimilated to functional categories. Nevertheless, the fact of
the matter is that auxiliaries constitute a heterogeneous class, and it will be shown
below that some agglutinated auxiliaries showing adjectival inﬂection may be
categorized as lexical adjectives. By appealing to distributional/syntactic criteria,
it is argued that the morphological distinction of “agglutinated” versus “nonagglutinated
(free)” forms does not directly correlate with the distinction between
“lexical” and “functional” categories, and that bound morphemes can be lexical (or
content) morphemes, which should fall into the major lexical class.
To be concrete, tai ‘want’, which is categorized as an auxiliary in traditional
Japanese linguistics (Konoshima 1973; Kitahara 1981; and many others), is similar in
meaning to the adjective hosii ‘want’ (or the verb hossuru ‘want’), the most obvious
diﬀerence between the two being that the former is an agglutinative element [+Aggl],
and the latter a free form [−Aggl]. The auxiliary tai needs to combine with a verb, but
hosii stands as a syntactically independent element, as in (12).
(12) a. Watasi wa okane ga hosi-i.
1.sg TOP money NOM want PRS
‘I want money.’
b. Watasi wa osake ga nomi-ta-i.
1.sg TOP sake NOM drink-want-PRS
‘I want to drink sake.’
62 Hideki Kishimoto and Satoshi Uehara
The auxiliary tai can be separated from the preceding verb with an appropriate
morphological adjustment, i.e. with the addition of dummy verb suru to the left
of tai (e.g. kangae-tai [think-want] ‘want to think’ → kangae wa si-tai [think TOP
do-want]), and diﬀers from a derivational aﬃx like -sa (as in kata-sa [solid-ness]),
which can never be separated from its host. Both tai and hosii, despite the diﬀerence
in their morphological status, function as lexical adjectives, i.e. predicates speciﬁed
for [+Adj, +Pred], as conﬁrmed by the fact that they can appear as predicates
embedded under the verb omou ‘think’, which takes a small clause complement.
(13) a. Watasi wa [sono okane o amari hosiku] omowa-nakat-ta.
1.sg TOP that money ACC much want think-NEG PST
‘I did not want that money very much.’
b. Watasi wa [osake o nomi-taku] omo-u.
1.sg TOP sake ACC drink-want think-PRS
‘I think that I want to drink sake.’
Note that the small clause complement selected by omou can have an adjective (and
also an adjectival noun) as its predicate (Kishimoto 2007, 2008).
(14) Ken wa [sono kodomo o {kawaiku/sinsetu-ni}] omot-ta.
Ken TOP that child ACC {cute/kind} think-PST
‘Ken thought that child {cute/kind}.’
Example (13b) is acceptable because nomi-tai ‘want to drink’ is an adjectival predicate,
with the structure complement verb + lexical adjective (tai), i.e. the auxiliary tai
in (13b) functions as a fully lexical adjective with the feature [+Adj, +Pred, +Aggl],
despite the fact that it is a morphologically dependent (agglutinative) element.
By contrast, (15) is excluded because the embedded predicate has the sequence of
verb + function word (nai), i.e. the verb is negated by the function word nai.
(15) *Ken wa [sono kuruma o ure-naku] omot-ta.
Ken TOP that car ACC sell.can-NEG think-PST
‘Ken thought that car unlikely to be sold.’
In (15), the negative nai, which inﬂects like an adjective, is a functional category,
which can be labeled as [−Pred, +Aggl] (i.e. [−Pred] → [−Adj, −Pred]), and thus, it is
not allowed to occur in the small-clause complement when it is combined with a
verb.
The diﬀerence in the grammatical status of tai and nai can also be conﬁrmed by
embedding them under hosii ‘want’. When nomi-tai and noma-nai are embedded
under hosii, a diﬀerence in acceptability emerges, as in (16).
Lexical categories 63
(16) a. *Watasi wa [Ken ga osake ga nomi-taku-te] hosi-i.
1.sg TOP Ken NOM sake NOM drink-want-GER want PRS
‘I want Ken to want to drink sake.’
b. Watasi wa [Ken ga osake o noma-nai-de] hosi-i.
1.sg TOP Ken NOM sake ACC drink-NEG GER want PRS
‘I want Ken not to drink sake.’
Note that when a negated verb is embedded under hosii, the nai-de form instead of
the naku-te form must be used. The verbal te-form nai-de can be formed only when it
is preceded by a verb, but the naku-te form is available for both negated verbs and
adjectives (see e.g. Kuno 1973). The contrast in acceptability between (16a) and (16b)
comes from the requirement that the complement clause selected by hosii should be
verbal, but not adjectival. This is conﬁrmed by (17).
(17) a. Watasi wa [sono kuruma ga ure-te] hosi-i.
1.sg TOP that car NOM sell.can-GER want PRS
‘I want that car to be sold.’
b. *Watasi wa [kodomo ga itumo kawaiku-te] hosi-i.
1.sg TOP child NOM always cute-GER want-PRS
‘I want the child to be cute at all times.’
Thus, the diﬀerence in acceptability between (16a) and (16b) suggests that nomi-tai
has the constituent structure verb + adjective, and noma-nai, the structure verb +
function word. The data show that certain adjectivally-inﬂecting auxiliaries may be
construed as lexical adjectives, even though they are morphologically dependent.
Notably, predicates identiﬁed as morphologically free can be either lexical or
functional. In traditional grammar, the existential/possessive nai, which is the negative
form of the verb aru, is claimed to be an “adjective” mainly for morphological
reasons, but the examples in (18) illustrate that it does not function as a lexical
adjective.
(18) a. *Ken wa [soko ni hon o naku] omot-ta.
Ken TOP there LOC book ACC NEG think-PST
‘Ken thought the book not to be there’
b. *Watasi wa [soko ni hon ga {naku-te/nai-de}] hosi-i.
1.sg TOP there LOC book NOM {NEG GER/NEG GER} want PRS
‘I want the book to not to be there.’
The existential/possessive nai is not allowed to appear in the context where an
adjective can appear, nor can it occur in the syntactic context where a verb is
64 Hideki Kishimoto and Satoshi Uehara
allowed. This suggests that the existential/possessive nai should belong to the functional
class, despite the fact that it is a non-agglutinated form morphologically;
hence this element should be marked as [−Pred, −Aggl].
In the literature on Japanese, a number of opinions are available as to how nai
should be related to aru (see Hashimoto 1969; Yamaguchi 2004; Kato 1985; and
many others). But if the existential/possessive nai – the negative counterpart of
aru ‘be’ – is a functional element, as noted above, a reasonable analysis would be
that, as suggested by Kato (1985), the negative form nai is derived from ara-nai by
dropping the verb part ara-, owing to the fact that aru somehow lacks an inﬂectional
form *ara- in contemporary Japanese. It is worth noting here that archaic negative
expressions like ara-nu and ara-zu do inclde the verb form ara-, which must be
deleted when combined with nai. Note further that the verbal negator nai can only
have the te-form naku-te, unlike the verbal negation which can have both naku-te
and nai-de, because the verb is dropped.
The regular negator nai combined with verbs does not have the properties of a
lexical adjective, despite its adjectival inﬂection, and serves as a functional predicate.
It is plausible to say here that this type of nai acquires status as a functional
category via the process of decategorialization – a shift from [+Adj, +Pred] to [−Adj,
+Pred] (i.e. the loss of ‘adjective’ status), and further, to [−Pred] (the loss of ‘predicate’
status) In this case, the functional shift of nai has taken place while retaining
its inﬂection. (Note that major lexical categories are often grammaticalized into
minor categories while preserving their inﬂectional patterns; see e.g. Brinton and
Traugott 2005). Interestingly, there are cases where negative nai retains the status
as a lexical adjective, as exempliﬁed by warikire-nai ‘unsatisfactory’ and abunage
ga nai ‘without danger’.
(19) a. Ken wa [sono kettei o warikire-naku] omot-ta.
Ken TOP that decision ACC satisfy-NEG think-PST
‘Ken thought that decision to be unsatisfactory.’
b. *Watasi wa [Ken ni sono kettei ga warikire-nai-de] hosi-i.
1.sg TOP Ken DAT that decision NOM satisfy-NEG GER want PRS
‘I want Ken to want to drink sake.’
c. Watasi wa [kare no unten o abunage ga naku] omo-u.
1.sg TOP he GEN driving ACC danger NOM null think-PRS
‘I think his driving without danger.’
As shown in (19a, c), warikire-nai and abunage ga nai can be legitimately embedded
as a small clause predicate under omou (and warikire-nai cannot be embedded
under hosii, as in (19b)). The data suggest that negative nais associated with these
Lexical categories 65
expressions function as lexical adjectives, i.e. [+Adj, +Pred]. Nevertheless, their
morphological status diﬀers, as we can see from (20).3
(20) a. warikire mo si-nai b. (nan no) abunage mo nai
satisfy also do-NEG any GEN danger also null
‘not also be satisﬁed’ ‘also without (any) danger’
In both cases in (20), an adverbial particle can be added to the front of nai. In (20a),
the supportive verb suru occurs to the left of nai, which shows that nai is an agglutinated
element, i.e. [+Aggl]. In (20b), no supportive element is necessary, since
nai here is a non-agglutinated form, i.e. [−Aggl], just like the existential/possessive
nai. Both instances of nai in (20) are categorized as adjectives by the syntactic
criteria, as veriﬁed by (19a, c). This shows that nai can be a lexical adjective, regardless
of whether it is agglutinated or not, i.e. nai appearing in warikire-nai is speciﬁed
as [+Adj, +Pred, +Aggl], and nai appearing in abunage ga nai as [+Adj, +Pred, −Aggl].
It is worthy of note that Japanese has a deverbal predicate as well. The predicate
iru ‘need’ provides a case in point. This predicate shows verbal inﬂection. Even
though iru counts as a verbal, at least, in morphological terms (cf. Backhouse
2009), it cannot be embedded under hosii ‘want’.
(21) *Watasi wa [Ken {ga/ni} okane ga it-te] hosi-i.
1.sg TOP Ken {NOM/DAT} money NOM need-GER want PRS
‘I want Ken to need money.’
As discussed by Kishimoto (2005), any kind of lexical verb can appear in the embedded
clause introduced by hosii.4 Then, (21) shows that iru ‘need’ does not act
as a lexical verb, even though it behaves as a predicate with an argument structure
determining the thematic status of their arguments. This suggests that iru ‘need’
serves as a predicate devoid of its categorical property as a verb, i.e. [−V, +Pred,
3 These adjectival expressions do not have aﬃrmative forms, as in *warikireru and *abunage ga aru.
Abunage ga nai serves as an idiomatic adjective, so it can be easily embedded under omou even
though abunage bears nominative case marking. On the other hand, when nai serves as a gram
matical negator, it has an aﬃrmative counterpart, as (ia) shows, and the clause cannot be embedded
under omou ‘think’, as in (ib).
(i) a. Kare ga yuuki ga {na i/ar u}.
he NOM courage NOM {NEG-PRS/be PRS}
‘He {does not have/has} courage.’
b. *Watasi wa [kare o yuuki ga naku] omo u.
1.sg TOP he ACC courage NOM NEG think PRS
‘I think him without courage.’
4 In (21), any type of lexical verb can appear in the subordinate clause, because this clause involves
simple embedding. Thus, non self controlable verbs, as well as stative verbs, can be used as the
predicates of the embedded clause, as shown in (i).
66 Hideki Kishimoto and Satoshi Uehara
−Aggl], as opposed to an ordinary verb labeled as [+V, +Pred, −Aggl]. Furthermore,
the negative marker nai combined with iru ‘need’ behaves diﬀerently from the regular
negator nai (associated with ordinary verbs), which is identiﬁed as a functional
category, as shown in (22).
(22) a. *Watasi wa [Ken {ga/ni} okane ga ira-nai-de] hosi-i.
1.sg TOP Ken {NOM/DAT} money NOM need NEG GER want PRS
‘I want Ken to need money.’
b. Watasi wa [sono okane o ira-naku] omot-ta.
1.sg TOP that money ACC need-NEG think-PST
‘I thought that money unnecessary.’
Since ira-nai ‘need not’ can be embedded under omou, but not under hosii, the
negative nai associated with iru ‘need’ must count as a lexical adjective, i.e. [+A,
+Pred, +Aggl], illustrating that nai can be a lexical adjective even if it is a bound
form. The facts of the negative nai illustrate that there is no necessary connection
between the “lexical” versus “functional” distinction, on the one hand, and the
“agglutinated” versus “non-agglutinated” distinction, on the other.
Let us now turn to the discussion of the predicate sugiru, which carries the
meaning of ‘exceed’ or ‘pass’, for further illustration of the fact that categories
cannot be identiﬁed solely in morphological terms. Note ﬁrst that sugiru can be
used as a main verb, as in (23).
(23) a. Zikan ga sugi-ru. b. Hatugen no do ga sugi-ru.
time NOM pass-PRS statement GEN degree NOM excess-PRS
‘Time passes.’ ‘The statement is too excessive.’
Sugiru can also appear as part of compound verbs, as shown in (24).
(i) a. Watasi wa [gohan ga suguni deki te] hosi i.
1.sg TOP rice NOM immediately make GER want-PRS
‘I want the meal to be ready immediately.’
b. Watasi wa [soko ni hon ga at te] hosi i.
1.sg TOP there LOC book NOM be GER want-PRS
‘I want the book to be there.’
When hosii takes a control clause as its complement clause, however, it is not possible to emedd a
non self controllable verb, as in (ii).
(ii) *Watasi wa gohan ni [PRO suguni deki te] hosi i.
1.sg TOP rice DAT immediately make GER want-PRS
‘I want the meal to be ready immediately.’
In (ii), the dative argument serves as a controller that controls PRO. The unacceptability of (ii) comes
from the ‘self controllability’ condition imposed on control constructions.
Lexical categories 67
(24) a. Kuruma ga toori-sugi-ru. b. Ken ga gohan o tabe-sugi-ru.
car NOM go-pass-PRS Ken NOM rice ACC eat-exceed-PRS
‘Cars pass by.’ ‘Ken eats too much rice.’
The complex predicate toori-sugiru ‘pass by’ in (24a), where sugiru carries the sense
of ‘pass’, is a lexical compound which does not have a constituent structure transparent
to the syntax, but tabe-sugiru ‘eat too much’ is a syntactic compound with a
transparent syntactic structure. The soo suru ‘do so’ replacement test in (25), which is
often used to assess the syntactic transparency of verbal constituents (Kageyama
1993), indicates that the two kinds of compound verbs indeed have distinct constituent
structures.
(25) a. Ken ga gohan o tabe-sugi-ta. Mari mo soo si-sugi-ta.
Ken NOM rice ACC eat-exceed-PST Mari also so do-exceed-PST
‘Ken ate rice too much. Mari did so, too (=ate rice too much).’
b. Ken ga toori-sugi-ta. #Mari mo soo si-sugi-ta.
Ken NOM go-pass-PST Mari also so do-pass-PST
‘Ken passed by. Mari did so, too (≠passed by).’
The diﬀerence is further corroborated by the fact that sugiru expressing the sense of
‘exceed’ can be productively combined with any type of verb, while the verb sugiru,
which carries the meaning of ‘pass’, cannot.
When the excessive sugiru ‘exceed’ is combined with a verb, the entire complex
is often regarded as forming a syntactic V-V compound (Kageyama 1993; Yumoto
2005; Kishimoto 2009). Nevertheless, sugiru possesses properties diﬀerent from
those of other verbs appearing in syntactic compounds (e.g. kakeru ‘start’, naosu
‘repeat’). As shown in (26), sugiru ‘exceed’ can combine with a verb, a noun, an
adjective, an adjectival noun, or a negated verb to form a complex compound, but
the verb naosu ‘repeat’ can be directly combined with a verb only.
(26) a. Kare wa {tabe/kodomo/sizuka/isogasi/sira-na}-sugi-ru.
he TOP {eat/child/quiet/busy/know-NEG}-exceed-PRS
‘He {eats too much/is too childish/is too quiet/is too busy/knows too little}.’
b. Kare wa {tabe/*kodomo/*sizuka/*isogasi}-naosi-ta.
he TOP {eat/child/quiet/busy}-repeat-PST
‘He repeated {the act of eating/*child/*quiet/*busy}.’
Arguably, sugiru is the only verbal predicate (or to be more precise, the only verballyinﬂecting
predicate) that can be combined with elements other than verbs to give
rise to syntactic compound predicates. Importantly, the excessive sugiru forming a
syntactically analyzable compound with a verb, a noun, or an adjective is devoid of
its verbal property, i.e. it does not act like a verb. This is evidenced by the fact that it
cannot be nominalized by kata-suﬃxation.
68 Hideki Kishimoto and Satoshi Uehara
(27) a. ryoori no tabe-{kake/naosi}-kata
dish GEN walk-{start/repair}-way
‘the way of {starting to eat the dishes/eating the dishes again}’
b. gohan no tabe-sase-kata
rice GEN eat-CAUS-way
‘the way of making (someone) eat rice’
c. kare no nagu-rare-kata
he GEN hit-PASS-way
‘the way of his being hit’
d. *kare no {kodomo/karu/tabe}-sugi-kata
he GEN {child/light/eat}-exceed-way
‘the way of his {being too childish/being too light/eating too much}’
Since the nominalizing suﬃx -kata ‘way’ can be attached to a verbal element, compound
verbs can be nominalized with the addition of the suﬃx -kata, as in (27a). The
causative and passive suﬃxes sase and rare, which can be labeled as [+V, +Pred,
+Aggl], can also host this suﬃx, as in (27b) and (27c), showing that the morphological
status of bound versus free form does not aﬀect the possibility of katasuﬃxation.
Nevertheless, V-sugiru compounds cannot be nominalized with the suﬃx
-kata, as in (27d). In this connection, observe that kata-nominalization is permitted if
sugiru is not part of a syntactic compound, as in (28).
(28) a. (kuruma no) toori-sugi-kata b. zikan no sugi-kata
car GEN go-pass-way time GEN pass-way
‘the way of cars’ passing by’ ‘the way of time passing’
c. hatugen no do no sugi-kata
statement GEN degree GEN exceed-way
‘the way of making too extreme statements’
The fact that V-sugiru compounds in (27d) resist kata-nominalization suggests that
the morphological dependent sugiru appearing in syntactic compounds does not
function as a lexical verb, having the status as [−Pred, +Aggl]. In the light of the
data (27), it is reasonable to state that the predicate sugiru forming a syntactic
compound verb has undergone grammaticalization, and now serve as a functional
element.
In this section, we have discussed the issue of a correlation between the classiﬁcation
of lexical categories and their morphological forms. In Japanese, auxiliaries
following a main predicate are formed into a morphologically tight unit via agglutination.
On the basis of certain syntactic criteria, it has been seen that auxiliaries fall
into either lexical or functional categories, which suggests that the categorial status
of predicative elements does not strictly correlate with a morphological distinction
Lexical categories 69
between “agglutinated” versus “non-agglutinated” forms. This in turn shows that
morphological criteria are not necessarily reliable for the purpose of drawing a line
between lexical and functional categories in Japanese.
4 Lexical categories from the cognitive-typological
perspective
Cognitive linguistics places a focus on the nature and functional motivations of
linguistic categorization. Linguistic typology (or simply, typology) examines attested
patterns of variation across languages. This section will present an analysis of
lexical categories in Japanese representing the collaboration of the two approaches.
4.1 Basics in the cognitive-typological approach to categorization
Usage-based cognitive linguistic investigations have repeatedly found that gradience
is observed, when examined without contrary predilection, in almost every aspect of
linguistic phenomena. In analyzing lexical category structures and organizations,
cognitive approaches thus do not assume, a priori, “classical models” of categories,
in which categories are discrete and phenomena are either inside of or outside of a
given category (Lakoﬀ 1987; Langacker 1987). Gradience, or lack of discreteness,
however, does not mean chaos without any generalizations. It often implies “prototype”
motivations in category organization (Rosch 1978; Taylor 1989). A prototype is
a privileged subset of members of a category that represent the best exemplars of the
category. The theoretical concept of a prototype is often relevant to and shared by
typological approaches to linguistic categories.
Typological or cross-linguistic approaches are characterized by their acceptance
of the fact of linguistic diversity that linguistic categories of particular languages
can be irreducibly language-particular. Croft (1991, 2001) presents a theory based
on a prototype approach to cross-linguistic grammatical patterns and has developed
a universal deﬁnition (in the form of markedness patterns in typology) of the
major lexical categories that is not constrained by the peculiarities of a particular
language. To see how this theory can be applied in typologically characterizing
lexical categories in Japanese, Uehara (1998) is introduced here. To indicate those
points of departure of the cognitive-typological approach from the previous traditional
and generative approaches, Uehara (1998) surveyed formal and structural
criteria used to identify the so-called ﬁve “major” lexical categories of Japanese
in eight past analyses, all of which explicitly discuss their grammatical behaviors
(Hashimoto 1948; Kuno 1973; Martin 1975; Teramura 1982; Kageyama 1982; Miyagawa
1987; Shibatani 1990; Ohkado 1991). Having pointed out some problems arising
70 Hideki Kishimoto and Satoshi Uehara
from language-particular aspects of their grammatical behaviors, Uehara discusses
what contributions the cognitive-typological theory can make. Speciﬁcally, he adopts
Croft’s treatment of pragmatic (propositional) functions and reinterprets the formal
and structural criteria used in the previous methods in cognitive-typological terms to
provide a typological characterization of the lexical category structure of Japanese.
The ﬁve categories investigated as the “major” lexical categories in Japanese are
Noun, Adjectival Noun, Adjective,Verbal Noun, and Verb,5 and capitalization is used
throughout this section to indicate that these categories in question are languagespeciﬁc
categories assessed in terms of some language-internal properties. These
ﬁve categories are assumed to all be lexical categories with a relatively large
membership, more or less distinct from one another in terms of their linguistic structural
and/or semantic properties. They are termed the ﬁve “large” lexical categories
(compared with words in minor lexical categories like doodoo ‘digniﬁed’ or onazi
‘identical, the same’) in this section, to save the title “major” until they, or some
other lexical categories, are assessed to be appropriate for it in cross-linguistic
terms.
As one of the problems arising from language-speciﬁc aspects of lexical categories,
Uehara (1998) points out that the criteria for deﬁning these basic lexical categories
are based on concepts which themselves are not well deﬁned. Deﬁning some language-speciﬁc
properties of the Japanese language in terms of the concepts applicable
to, say, English, does not secure sound characterizations of them, but rather
brings about confusing situations for descriptive purposes. This is observed in
apparent contradictions among the criteria proposed for the same AN category
membership. One such pair is shown in (29). Each criterion is followed by those
works that propose it.
(29) a. ANs can inﬂect (-na prenominally, -da sentence-ﬁnally).
[Hashimoto 1948, Kageyama 1982]
b. ANs do not inﬂect and need a copula (e.g. da), and take na before N.
[Kuno 1973, Teramura 1982, Martin 1975, Miyagawa 1987, Shibatani 1990]
With the same criterion of inﬂection, ANs are inﬂectional according to some (notice
the use of hyphens) in (29a) while ANs are non-inﬂectional in others (notice the
same forms without hyphens) in (29b). This suggests that it is necessary to know
what kind of grammatical behavior in Japanese each analysis treats as “inﬂection”
before inﬂection can be used as a criterion for category membership of the language.
In fact, it is of prime importance in characterizing the overall structure of lexical
5 These category labels are used throughout this chapter to be consistent, but diﬀerent labels (not to
mention their names in Japanese) are found for the same categories in diﬀerent analyses. For
instance, the term “Nominal Adjective” is used in Kuno (1973) and its Japanese close equivalent
“meiyōshi” in Teramura (1982) for “Adjectival Noun” here.
Lexical categories 71
categories in Japanese to examine what counts as inﬂection in Japanese and deﬁne
it in terms of some intra-linguistic facts, which is done in the next section.
The other, more fundamental problem found in the previous approaches to
lexical categories in Japanese, according to Uehara (1998), is the lack of a principled
basis of choosing the criteria for lexical categorization. The criteria for a single category
of Adjectival Nouns, for example, are various and numerous; even when the
same criteria are collapsed together there still remain 19 diﬀerent criteria by eight
analyses. This lack of a principled basis virtually allows a scholar to privilege certain
preferred criteria, whose data are hard to interpret typologically. This part of the
section will brieﬂy illustrate this and explicate how Croft’s (1991, 2001) theory helps
one to reinterpret them in cross-linguistic terms.
After presenting his data as to the inﬂection criterion (see (29b) above) for his
argument for a certain lexical feature (cf. Section 3) for ANs in Japanese, Miyagawa
(1987) presents, as another piece of evidence, the data reproduced in (30), to show
that “the conditional (ke)reba attaches to V and A, but not to AN and N” (pp. 43–45):
(30) a. AN *sizuka-reba ‘if quiet’
b. N *sensei-reba ‘if a teacher’
c. A utukusi-ke-reba ‘if beautiful’
d. V tabe-reba ‘if (you) eat’
This part of his data can illustrate two accounts on which typological considerations
can be implemented. The ﬁrst concerns the theory of typological markedness
(Greenberg 1966). By simply mentioning the form of the conditional as “(ke)reba”,
Miyagawa dismisses a structural diﬀerence between As and Vs, which can be observed
in his data in (30), that ke-reba is for As while reba is used for Vs. In other
words, for the forms indicating the same function of conditional, As always require
an extra marking ke to the form for Vs. In the typological markedness terms, Vs are
structurally “unmarked” (or “the least marked”, in case there are more than two
types) for conditionals while As are structurally “marked” (or “more marked” in
comparison). This markedness criterion is crucial in linguistic typology for discovering
cross-linguistic patterns, and is used in cross-linguistic approaches to lexical
categories as well, as will be seen for those in Japanese later (see Croft 2001 for
discussion of the other, “behavioral potential” markedness criterion, which is also
used in typology but is omitted here due to space constraints).
The other point concerns the principle governing the selection of criteria for
lexical categorization in cross-linguistic studies. Although Miyagawa’s criterion
above says “the conditional (ke)reba does not attach to AN and N” as seen in (30),
it does not mean that they never take conditionals. They do, and the conditional
form of the copula nara(ba) is used with them instead. So Miyagawa’s paradigm in
72 Hideki Kishimoto and Satoshi Uehara
(30) above could be supplemented with the data in (31) below for a descriptively
fuller picture for cross-linguistic comparison.
(31) a. AN sizuka nara(ba) [quiet COP] ‘if quiet’ (cf. 30a)
b. N sensei nara(ba) [teacher COP] ‘if a teacher’ (cf. 30b)
What this criterion of conditionals by Miyagawa boils down to, then, is the same as
his ﬁrst one in (29b): whether some lexical item inﬂects itself or needs a copula in
predication. Interestingly, this criterion as to the forms of lexical items in predication
is among those criteria most commonly used (often in disguise) by the previous
analyses surveyed, and at the same time it is one of the criteria used in Croft’s
cross-linguistic approaches to major lexical categories as well. The diﬀerence
between the two, most naturally, is that some speciﬁc forms of the language (e.g.
(ke)reba in (30) above) are used in the former, while in the latter such criteria are
phrased in general, cross-linguistically applicable terms. Thus the criterion in question
is rendered in the cognitive-typological approach as the (structural) markedness
criterion of the lexical roots in the constructions indicating the pragmatic function of
“predication”.
In addition to predication, the cognitive-typological approach employs “reference”
and “modiﬁcation” as the pragmatic functions directly relevant to the cross-linguistic
identiﬁcation of major lexical categories. Again, constructions indicating reference
and modiﬁcation are found among the groups of structural criteria commonly used
and/or (often implicitly) repeated in the surveyed previous analyses of the ﬁve large
lexical categories in Japanese. They are of course expressed in more or less languagespeciﬁc
ways, such as “can function as subject (or object)” or “can take case particles”
for reference and “take na prenominally” (as in (29) above) for modiﬁcation.
The pragmatic function as the basis of major category deﬁnitions is a propositional
act, and it is thus referred to as “propositional” function hereafter in this
chapter. A propositional act is analogous to an illocutionary speech act, but it is a
speech act that structures information inside the proposition rather than modifying
the proposition as a whole (Searle 1969). In performing a speech act, the speaker
must perform a series of propositional acts. The most important of these are reference,
predication (since these are required for every proposition), and (to a lesser extent)
modiﬁcation. The reference function is deﬁned as “to get the hearer to identify an
entity as what the speaker is talking about” (Croft 1991: 52). The predication function
is to say “what the speaker intends to say about what he is talking about (the
referent)” (Croft 1991: 52) to ascribe something to it, and the modiﬁcation function is
“to enrich the nominal image by an additional feature” (Croft 1991: 123).
Croft’s (1991) cross-linguistic study on lexical categories found some patterns
repeated across languages in the interaction of propositional function and semantic
class. The semantic classes of objects, actions, and properties (taken originally from
Lexical categories 73
the traditional approach) are only a small subset of the semantic classes of words
found in human languages, and are deﬁned in terms of four semantic properties of
valency, stativity, persistence, and gradability (see Croft 1991). It is found that there
is a prototypical correlation between the semantic class of object and the function of
reference so that words denoting objects cross-linguistically take the prototypical
markedness pattern for nouns, in which their lexical roots are the least marked for
reference. A similar prototypical markedness pattern holds between actions and
predication for verbs and between properties and modiﬁcation for adjectives.
Such a prototypical correlation is most conspicuously revealed and illustrated in
the structural markedness patterns in English shown in Table 3 below, slightly modiﬁed
from Croft (1991:67) with English examples from Croft (1991:53):
Table 3: Typological correlations with English examples (from Croft 1991)
Reference Modiﬁcation Predication
Objects UNMARKED NOUNS
(vehicle)
genitive, PP’s on nouns
(vehicle’s, of/in/etc. the vehicle)
predicate nominals
(be a/the vehicle)
Properties deadjectival nouns
(tallness)
UNMARKED ADJECTIVES
(tall)
predicate adjectives
(be tall)
Actions action nominals, complements,
inﬁnitives, gerunds
(destruction, to destroy)
participles, relative clauses
(destroying, destroyed)
UNMARKED VERBS
(destroy)
4.2 Toward a formal deﬁnition of katsuyō-inﬂection in Japanese
This section addresses a basic, language-speciﬁc question postponed earlier: what
does it mean to say that a word inﬂects in Japanese? The word inﬂection is a translation
of “katsuyō” in Japanese, a traditional grammatical term. Past analyses by
and large agree that Nouns do not inﬂect while Adjectives and Verbs inﬂect in
Japanese. As for words in the AN category, however, as we saw above in the survey
of past analyses in (29), there are two competing positions: one posits that ANs
inﬂect and the other that ANs do not. This is illustrated below by the most basic
(i.e. the present, aﬃrmative, declarative) forms6 of representative examples of each
6 This form is attested by hearsay evidential markers, soo da ‘it is said that’ and to iu koto da ‘it is
the case that’.
(i) Kirei( ) da {soo da/to iu koto da}.
pretty COP {it.is.said.that/it.is.the.case.that}
‘{It is said that/It is the case that} (it) is pretty.’
Also, the unmarked form prototypically appears before prototypical conjunctive clause particles like
kedo ‘although’, and kara ‘because’.
(ii) kirei( ) da {kedo/kara} . . .
pretty COP {although/because}
‘{Although/Because} (it) is pretty, . . .’
74 Hideki Kishimoto and Satoshi Uehara
category, where the use of hyphen indicates the inﬂectional status (note the zero
pronominal nature of the Japanese language, where complement nouns can be
implicit and predicates alone can form sentences, indicated by the capitalization of
the ﬁrst letter, in the example sentences in (32) and below):
(32) a. N Hon da. [book COP] ‘(It) is a book.’
b. AN Kirei(-)da. [pretty-INFL/COP] ‘(It) is pretty.’
c. A Huru-i. [old-INFL] ‘(It) is old.’
d. V Tabe-ru. [eat-INFL] ‘(It) eats (it).’
This disagreement as to the inﬂectional status of ANs, together with the controversy
as to whether to establish a separate category of ANs, is observed not only in the
survey above, but throughout the history of syntactic category analyses of Japanese,
which dates back to FUJITANI Nariakira (1778). Fujitani ﬁrst recognized ANs as a
category called ari-sama, which is a subcategory of sama, which corresponds
roughly to the Adjective category. Since then, two prominent traditional grammarians,
Hashimoto and Tokieda, for example, took opposite sides (see Mizutani 1951 for
a summary), and the controversy has continued up to the generative approaches
(Kageyama 1982; Miyagawa 1987) as well, as seen above in (29). Therefore, what
counts as katsuyō-inﬂection needs to be made clear before it can be used as a
criterion in any lexical category analyses of Japanese.
The important question to ask is whether there is any linguistically salient contrast
among the four kinds of predicates in (32). Uehara (1998) answers in the aﬃrmative
to this question, and demonstrates that the contrast between N and AN on
the one hand and A and V on the other is a very salient one. The contrast shows up
when one starts looking at the forms of predications with more marked functions
than the basic, unmarked ones. Among notable instances of contrast is the reduplication
for emphasis of predicates, which are shown in Table 4.
Table 4: Lexical roots in unmarked and emphatic predicates
Unmarked predicates Reduplication for emphasis
N Hon da. ‘(It) is a book.’ Hora! Hon, hon! ‘See! (It) IS a book!’
AN Kirei da. ‘(It) is pretty.’ Hora! Kirei, kirei! ‘See! (It) IS pretty!’
A Huru i. ‘(It) is old.’ Hora! *Huru, huru! => Huru i, huru i! ‘See! (It) IS old!’
V Tabe ru. ‘(It) eats (it).’ Hora! *Tabe, tabe! => Tabe ru, tabe ru! ‘See! (It) DOES eat (it)!’
One contrast observed above is that the ﬁrst elements (i.e., lexical roots) of N and
AN predicates can stand alone in the emphatic predicative function while those of
A and V predicates cannot. That is to say, A and V roots are “bound” morphemes
unlike N and AN roots.
Lexical categories 75
The same can be observed also in question predicates, direct or indirect, most of
which involve the question particle ka. Table 5 below shows the most obvious case
of contrast, in which in casual colloquial speech the lexical roots are used in questions
without the question particle ka but with the question intonation.
Table 5: Lexical roots in unmarked and interrogative functions
Unmarked predicates Predicates in casual-style question
N Hon da. ‘(It) is a book.’ Hon ? ‘Is (it) a book?’
AN Kirei da. ‘(It) is pretty.’ Kirei ? ‘Is (it) pretty?’
A Huru i. ‘(It) is old.’ *Huru ? => Huru i ? ‘Is (it) old?’
V Tabe ru. ‘(It) eats (it).’ *Tabe ? => Tabe ru ? ‘Does (it) eat (it)?’
The same pattern of contrast can be observed in the use of lexical roots with the
question particle ka, with ka sira ‘I wonder if . . . .’ (mostly by female speakers) and
with ka doo ka ‘whether or not. . . .’ (indirect/embedded questions), only the ﬁrst of
which is shown below in (33) due to space limitations:
(33) a. N Hon ka? ‘Is (it) a book?’
b. AN Kirei ka? ‘Is (it) pretty?’
c. A *Huru ka? => Huru i ka? ‘Is (it) old?’
d. V *Tabe ka? => Tabe ru ka? ‘Does (it) eat (it)?’
The above examples demonstrate that the second elements of A and V predicates,
i and ru (u in case of consonant-root Vs) respectively, are more tightly joined with
their ﬁrst element (i.e. “root”), than are those of N and AN predicates (i.e. da).
The other group of predicates that exhibit the same pattern of contrast are
epistemic modality predicates, which are, again, predications with marked functions.
They include daroo ‘it is probable that’, desyoo ‘it is probable that (polite)’,
ka mo sirenai ‘may be’, ni tigainai ‘surely, must be’, and mitai (da) ‘seem like’. Only
the pattern with the lexical roots with daroo ‘probable’ is shown in (34) due to space
limitations, but the same pattern holds with the other modality markers as well:
(34) a. N Hon daroo. ‘(It) is a book, probably.’
b. AN Kirei daroo. ‘(It) is pretty, probably.’
c. A *Huru daroo. => Huru i daroo. ‘(It) is old, probably.’
d. V *Tabe daroo. => Tabe ru daroo. ‘(It) eats (it), probably.’
All the data presented here demonstrate that though all the basic predicate types in
(32) seemingly are composed of two elements, there is a salient diﬀerence between N
76 Hideki Kishimoto and Satoshi Uehara
and AN predicates on the one hand and A and V predicates on the other in the connectedness
between the two elements of their predicates. In other words, the roots of
A and V unmarked predicates are more “bound” to their following elements than
those of N and AN predicates. With this contrast in mind, one can now safely use
the “-” notation as in (35) below:
(35) a. N Hon da. [book COP] ‘(It) is a book.’
b. AN Kirei da. [pretty COP] ‘(It) is pretty.’
c. A Huru-i. [old-INFL] ‘(It) is old.’
d. V Tabe-ru. [eat-INFL] ‘(It) eats (it).’
With this distinction in boundness as the language-internal basis for determining
inﬂectional categories (A and V) and non-inﬂectional ones (N and AN) in
Japanese, one may now proceed to typological characterization of the lexical categories
in Japanese in the next section. However, there are two points worth noting
here about this boundness distinction of the language, which will have direct
relevance to our later discussions. These two points are discussed in the next two
subsections, respectively.
4.2.1 Gradience in boundness
In the previous section we have focused on the salient contrast in boundness of
lexical roots between N and AN on one hand and A and V on the other (see (35)
above). However, this by no means implies that the morphological bound/free distinction
itself is discrete. Instead, as will be seen below, the degree of boundness
varies even among the members of inﬂectional categories (of bound lexical roots)
and of the other, non-inﬂectional categories.
Both lexical roots of A (e.g. huru-i ‘old’) and V (e.g. tabe-ru ‘go’) are bound (i.e.
‘not free’) compared with those of Ns and ANs, in that their inﬂectional ending (-i
and -(r)u, respectively) cannot be dropped in the context where the lexical roots of
the latter can stand alone. A diﬀerence in the degree of boundness, however, shows
up between the two inﬂectional lexical categories and indicates that A roots are the
freer of the two. This is shown by the fact that the so-called “Adjectival conjugational
ending drop construction” (Konno 2012) is available, which expresses “the speaker’s
immediate reaction to a given situation in which he/she is involved at the time of
utterance” (Konno 2012). In that construction, as its name indicates, A roots are
used without their inﬂectional ending -i. No construction with the same function is
available for V roots, and even when occurring in similar functions they always
require their ending –(r)u. This contrast is shown in (36) below:
Lexical categories 77
(36) a. A Huru! ‘Old!’ (<= huru-i)
b. A Dasa! ‘Uncool!’ (<= dasa-i)
c. A Kimoti-waru! ‘Disgusting!’ (<= kimoti-waru-i)
d. V Wakar*(-u)! ‘(It) makes sense!’ (wakar-u ‘understand’)
e. V Simi*(-ru)! ‘(My teeth) aches!’ (simi-ru ‘soak through’)
It should be also noted that a phonological restriction renders many V roots
inherently bound, unlike A roots. The lexical roots of Verbs may end with consonants
(as in k of ik-u ‘go’ and r of wakar-u ‘understand’) or with vowels i or e
(as in oki-ru ‘arise’ and tabe-ru ‘eat’). The mora structure of Japanese does not allow
syllable-ﬁnal, non-moraic consonants. In other words, no lexical roots of consonant
root Verbs can stand alone without their inﬂectional ending.
The existence of the inﬂectional ending drop construction for A but not for V,
together with the language’s phonological restriction on the V (consonant) roots,
suggest that the degrees of boundness of the bound A and V lexical roots can be
shown as A > V, where the left side of > is freer.
Moving on to the other side of the boundness distinction, namely, N and AN,
one sees some similar diﬀerence in the degree of boundness between the two noninﬂectional
categories. This diﬀerence in boundness can be observed in their occurrences
with case-marking particles. N roots can be directly followed by case-marking
particles, while AN roots cannot and require a nominalizing suﬃx sa just like A
roots, as in (37), where the notation *(sa) means that sa is obligatory in this context.
(37) a. N hon {ga/o} [book {NOM/ACC}]
b. AN kirei-*(sa) {ga/o} [pretty-*(ness) {NOM/ACC}]
c. A huru-*(sa) {ga/o} [old-*(ness) {NOM/ACC}]
This indicates that between the two non-inﬂectional categories of free lexical roots,
N roots are the freer of the two (N > AN).
Table 6 below summarizes the above discussion:
Table 6: Degrees in boundness of lexical roots
inﬂectional ending
drop construction
marked forms
of predication
directly with
case-marking particles
N (ok) ok ok
AN (ok) ok *
A ok * *
V * * *
Table 6 clearly demonstrates that the bound/free distinction is a matter of degree
and the degrees in boundness of lexical roots of the four lexical categories can
78 Hideki Kishimoto and Satoshi Uehara
be shown in the form of a cline in (38) below (functional motivations for the cline
will be discussed later):
(38) N > AN > A > V
This boundness cline provides a possible explanation of why some researchers,
traditional and generative alike, postulate AN to be inﬂectional along with A and
V: they assign the boundness level between N and AN for the non-inﬂectional/
inﬂectional division, while others assign the one between AN and A.
As discussed in Section 4.2, however, the current analysis concurs with the latter
(i.e. AN, as well as N, is non-inﬂectional). This is because the katsuyō-inﬂection in
question is concerned with predication function, while the use with case-marking
particles indicating the boundness level between N and AN is concerned with reference
rather than predication (see the discussion above).
4.2.2 The boundness distinction as a cardinal formal property of Japanese
This contrast in boundness of lexical categories is such a characteristic property of
the predicate structure of the language that it captures a generalization, and thus
facilitates the description, of not only all the predicate patterns enumerated in
Section 4.2, but also other structural patterns of the language, of which only two
are noted here.
The ﬁrst represents the interaction of lexical categories with the pragmatics of
the language found in the diﬀerence between default/masculine and feminine styles
in colloquial speech. The following in Table 7 is an illustration of the diﬀerence, by
using a sequence of a predicate + a sentence ﬁnal particle of assertion, yo ‘I tell you’:
Table 7: The boundness distinction and a style diﬀerence
Default/Masculine style Feminine style meaning
N Hon da yo. Hon yo. ‘(It) is a book, I tell you.’
AN Kirei da yo. Kirei yo. ‘(It) is pretty, I tell you.’
A Huru i yo. Huru i wa yo. ‘(It) is old, I tell you.’
V Tabe ru yo. Tabe ru wa yo. ‘(It) eats (it), I tell you.’
Here also, we ﬁnd one pattern for Nouns and Adjectival Nouns, and another for
Adjectives and Verbs – a contrast which now we are able to capture by the epithets,
“non-inﬂectional” and “inﬂectional” groups.
The other point concerns the productivity of lexical categories of the language:
this boundness of the A and V roots is such that it has made these categories closed
Lexical categories 79
classes while making the other, “free” (non-inﬂectional) roots the only source for
new word coinage, even though cross-linguistically inﬂection does not always block
new members and such major categories in other languages are often open classes.
That is, when new terms are coined in Japanese, they become members of the A or V
classes with great diﬃculty, and instead become members of open classes such as
AN and N. A striking example of this pattern can be seen in the formation of loan
words in the language. There are a large number of English adjectives that have
been borrowed into the Japanese language, but they seldom, if ever, become Adjectives,
but ANs. Ritti ‘rich’ (39a) and ereganto ‘elegant’ (39b) below are among such
examples. In (39), the corresponding adnominal forms, which follow the AN but not
the A pattern, are shown in square brackets.
(39) a. Ano hito wa ritti {da/*-i}. [ritti {na/*-i} hito]
that person TOP rich rich person
‘That person is rich.’ ‘a rich person’
b. Hun’iki ga ereganto {da/*-i}. [ereganto {na/*-i} hun’iki]
atmosphere NOM elegant elegant atmosphere
‘The atmosphere is elegant.’ ‘an elegant atmosphere’
This is true also of English verbs borrowed into Japanese. They follow the very
productive Verbal Noun pattern, i.e. VN + suru. The following are only some of the
numerous instances of VN words from English7:
(40) kopii suru, katto suru, anaunsu suru, hitto suru
copy do cut do announce do hit do
‘to copy’ ‘to cut’ ‘to announce’ ‘to make a hit’
In fact, the boundness of A and V roots has been a characteristic of the Japanese
language throughout its history, so that one does not have to wait for the recent
borrowings from English for the evidence of their closed class status. In its long
history of contact with the Japanese language, Chinese vocabulary has ﬂowed into
the language, and loan words from Chinese constitute a major part of the vocabulary
of the Japanese language, most of which are Ns, ANs and VNs (see Uehara
(2003) for a scenario of the historical development of ANs into major category in
present-day Japanese).
7 There is another less productive, but still useful colloquial pattern of N+r u (e.g. kopir u ‘to copy’
from kopii ‘copy’ and memor u ‘to take notes’ from memo ‘notes’). See Uehara (1998: Chap. 4) for
more examples.
80 Hideki Kishimoto and Satoshi Uehara
(41) a. AN kantan na, kyodai na, kimyoo na, yuudai na
‘simple’ ‘gigantic’ ‘strange’ ‘magniﬁcent’
b. VN soodan suru, syoyuu suru, baisyuu suru, kenkyuu suru
‘to consult’ ‘to possess’ ‘to purchase’ ‘to research’
Thus, the existence of the two categories that are often noted as categories “unique”
to the Japanese language (Shibatani 1990), can be ascribed to this boundness of the
two inﬂectional categories, Adjectives and Verbs.
4.3 Typological characterization of lexical categories in Japanese
Now that what counts as inﬂection in Japanese is clearly laid out, the structural
criteria in Japanese for the major propositional functions of reference, modiﬁcation
and predication can be summarized as follows:
Reference: whether or not it takes nonzero morpheme(s) such as sa ‘-ness’ when
used with case-marking particles
Modiﬁcation: whether or not it takes nonzero morpheme(s) such as no or na before N
Predication: whether it inﬂects, or does not inﬂect and requires non-zero morpheme(s)
such as the copula or the dummy verb in the predicate
4.3.1 The ﬁve “large” lexical categories
The criteria listed above are used to typologically characterize the structural properties
of the ﬁve lexical categories in Japanese, which are summarized in Table 8 below
(the least marked forms (present, aﬃrmative indicative) are listed):
Table 8: Structural properties of ﬁve “large” lexical categories in Japanese
Reference
(take case particles)
Modiﬁcation
(before N)
Predication
(inﬂect)
N ROOT ROOT + no ROOT + da
VN ROOT (+suru + koto/no) ROOT + no/suru ROOT + da/suru
AN ROOT sa ‘-ness’ ROOT + na ROOT + da
A ROOT sa ‘-ness’ ROOT i ROOT i
V ROOT (r)u + koto/no8 ROOT (r)u ROOT (r)u
8 Some may argue that the stem form (root (i)) of the Verbs (e.g. tutum i ‘package’ from tutum u ‘to
pack’) be in the V reference cell in Table 8. Refer to Uehara (1998: Chap. 2), which points out that the
stem forms are often bound (e.g. *tabe from tabe ru ‘to eat’ and *k i from k uru ‘to come’) and that
when they are free, they typically refer to some objects involved in the actions (e.g. tutum i above) or
some aspects of them (e.g. kaer i ‘the time/occasion of one’s returning’ from kaer u ‘to return’),
rather than to the actions themselves, which are referred to by the form in the cell in question in
Table 8 (i.e., ROOT (r)u + koto/no).
Lexical categories 81
The above table of the structural properties of the ﬁve “large” lexical categories
in Japanese demonstrates two respects in which propositional functions interact
with Japanese lexical category organization. First, Table 8 shows no diﬀerence in
structural markedness between modiﬁcation and predication,9 which indicates that
modiﬁcation has not played a major role in the category organization of the language
in addition to what predication has already. This can be considered to be a
reﬂection of the fact that the modiﬁcation function is optional and secondary compared
with the two major propositional functions, both of which are required in
every proposition. Present-day Japanese thus can be characterized as one of the
languages in which this secondary status of modiﬁcation is structurally observed.
(This point and its relevance to the non-prototypical statuses of A and AN in the
language will be discussed later.) Second, following from the ﬁrst, reference and
predication have played major roles in the language’s lexical categorization: Noun,
at the top of the list of large categories in Table 8, is the least marked for reference
and others are progressively more marked as they go down the list. The markedness
order is reversed for predication, for which Verb at the bottom is the least marked. In
other words, the boundness degree order in (38), which is the determining factor of
the order of paradigms in Table 8, is motivated by the propositional functions of
predication and reference. The most “bound” V roots and their structural paradigm
are motivated by predication function, while the freest N roots and their paradigm
are by reference function. The other lexical roots and their paradigms of intermediate
levels of boundness are in between in that regard.
From the comparison of Table 8 with Table 3, where Croft’s cross-linguistically
attested, prototypical lexical category paradigms are exempliﬁed with English noun,
verb, and adjective, the following points regarding the ﬁve Japanese lexical categories
are observable as their functional-typological characteristics (for detailed discussions
of each point with concrete examples, refer to Uehara 1998):
(a) N exactly follows the prototypical nominal paradigm in the Objects row like
English nouns, where the root is unmarked for reference but is more marked
for modiﬁcation and predication. VN is identical with N in its forms and markedness
pattern, the only diﬀerence being the existence of an additional pattern
with suru ‘do’ for VN. In other words, VNs are Ns with the additional ability to
compound with suru into compound level Vs.
(b) Neither AN nor A in Japanese exactly follows the prototypical adjectival paradigm
in the Properties row, which English adjectives follow: ANs are like English
adjectives in reference and predication functions, but in modiﬁcation, ANs diﬀer
from them and need an extra morpheme (like Ns). In that sense, ANs are structurally
“nouny” adjectives (Wetzer 1992). On the other hand, Japanese As are like
9 This structure for the present day Japanese can be ascribed to the loss of the formal distinction
between adnominal and predicative forms that used to exist for many Verbs and Adjectives in Old
Japanese. See the relevant discussion in Uehara (1998).
82 Hideki Kishimoto and Satoshi Uehara
English adjectives only in modiﬁcation (the least marked form of the inﬂection).
Unlike English adjectives, Japanese As (like Vs) can stand by themselves (i.e.,
without taking the copula da) in predication. In that sense, As are “verby”
adjectives (Wetzer 1992). As are as unmarked for reference as for modiﬁcation.
(c) Except for modiﬁcation, V in Japanese follows the prototypical verbal paradigm
in the Actions row, which English verbs follow: Vs thus resemble English verbs
in predication (unmarked) and in reference (more marked), but, unlike them,
Japanese Vs are unmarked also in modiﬁcation without requiring any extra morpheme
(i.e., the same form as in predication).
Table 8 furthermore makes possible some observations regarding the way the ﬁve
categories are inter-categorically related and, accordingly, how they are organized
overall.
Verbal Noun constitutes a non-prototypical subcategory of Noun: VNs belong to
the Noun category because they possess the same structural paradigm with Ns, and
they are less-prototypical members of the Noun category because VNs in the additional
structural pattern (ROOT + suru) are most marked for reference (ROOT + suru
+ koto/no). Adjectival Noun sides with N (rather than A): AN and N share the same
markedness pattern (nonzero morphemes) in both modiﬁcation and predication
(and in the latter they even share the same form da). Thus, AN and N together can
form a super-ordinate category in contrast with A and V, which are unmarked for
both modiﬁcation and predication. In the super-ordinate category subsuming N and
AN, Ns are prototypical members while ANs are less prototypical ones, since only Ns
are unmarked for reference.
Adjective and Verb present no contrast in structural markedness (both inﬂecting
themselves without requiring any extra morpheme) with respect to predication and
modiﬁcation functions. In reference only, the two diﬀer and V is the more marked. A
and V in Japanese, unlike adjectives and verbs in English, are thus structurally very
similar to each other, which suggests that they together form a superordinate category
of which the two are subcategories, as opposed to the other three categories.
Thus, simply by looking at the structural characterizations of the ﬁve large
lexical categories, we ﬁnd the two super-ordinate categories of N (subsuming VN)
and AN on the one hand, and V and A on the other. One may notice this major
division corresponds to the non-inﬂectional/inﬂectional distinction based on the
boundness as discussed in the previous sections. In other words, the boundness
distinction constitutes the ﬁrst “cut” into the lexical category structure of Japanese,
thus characterizing it as the language with two, not three, “major” categories, of
which the ﬁve large lexical categories are sub-classiﬁcations. It should be pointed
out in passing, however, that this situation of Japanese is typologically not rare,
where adjectives do not structurally constitute an independent category of its own,
but become a subcategory of other major categories, and that it is not totally unmotivated
considering the secondary status of modiﬁcation compared to predication
Lexical categories 83
and reference. In Chinese and Thai, for example, adjectives behave like ‘stative
verbs’ (i.e. a subcategory of verbs), not requiring a copula in predication. In Korean,
adjectives and verbs both inﬂect and share their inﬂectional endings (Uehara and
Kumashiro 2007).
One of the two major lexical categories of Japanese subsuming the non-inﬂectional
subcategories are termed the “Nominal” here, since N represents its prototype, following
the cross-linguistic markedness pattern for nouns, while AN and VN represent
extensions from N. The other major category subsuming the inﬂectional subcategories
is termed the “Verbal” because Verb basically follows the prototypical
pattern for verbs and represents its prototype.
The dichotomous Nominal/Verbal distinction draws the line between ANs and
As in Table 8. Figure 1 below summarizes the discussions and schematically represents
the overall lexical category structure of Japanese:
Figure 1: Lexical category organization in Japanese
4.3.2 Other approaches and other minor/non-prototypical subcategories
In the light of Table 8 and Figure 1, which capture typology-based structural characterizations
of the 5 large lexical categories of Japanese and overall Japanese lexical
category organization, respectively, one can readily assess various other approaches
to them. They diﬀer in which of the structural criteria are given greater weight than
others, which semantic/functional criteria are additionally taken into consideration,
and which granularity levels of distinctions are to be accounted for. Furthermore,
with the 5 constructional paradigms in Table 8 as basic templates, one can also
characterize other, minor lexical categories of the language as non-prototypical subcategories
of them.
84 Hideki Kishimoto and Satoshi Uehara
As noted in Section 2, Tokieda (1950) argues explicitly against setting up Adjectival
Noun as an independent lexical category. But his argument, which leads to
categorizing all ANs into his taigen category, the Nominal category in Figure 1, indicates
that he weighs the commonalities between N and AN in the predication function
over their diﬀerences in the reference and modiﬁcation functions. In contrast,
Hashimoto (1947) and Kageyama (1982), which describe AN as inﬂectional in (29a),
attach no weight to the formal identity between AN and N in predication. Sakuma
(1951) goes further to collapse A and AN together by giving more weight to semantic
and functional, rather than structural, criteria, as indicated by the following quote
from Sakuma (1951: 54): “it is reasonable to posit one lexical category by grouping
Adjective and Adjectival Noun together by setting aside the morphological diﬀerences
between them, because they resemble each other so completely in their
meanings and functions (translation by SU).” He sets up a lexical category called
seijōgo, which subsumes the two as its subcategories, and calls A the ﬁrst seijōgo
and AN the second seijōgo.
Some propose to take ﬁner-grained distinctions into consideration to argue
for an analysis that stands apart from the popular ﬁve major categories approach.
One such analysis is found in Muraki (2012) and his previous works cited therein.
Muraki’s analysis can be characterized as an extension of Sakuma (1951) in that he
posits the third “adjective” category on top of his ﬁrst and second “adjective” categories,
which respectively correspond to ﬁrst and second seijōgo in Sakuma’s
approach and A and AN in Table 8 and Figure 1 above in the current approach.
His third “adjective” takes no (instead of na) in modiﬁcation and the copula da
in predication, but is not used in reference (Muraki 2012: 149). He lists the following
examples in their forms with no:10
(42) sinku no ‘crimson-colored’, batugun no ‘outstanding/unrivaled’,
gokaku no ‘evenly-matched’, marugosi no ‘unarmed’,
mayakasi no ‘fake’
His third “adjective” can be characterized as a non-prototypical subcategory of the
Noun category, lying on its border with the AN category, in the current approach: it
takes the forms of the N paradigm in Table 8 except it lacks the form in reference.
Muraki’s classiﬁcation of the group of words as the third “adjective” after A and AN
as the ﬁrst and second ones, respectively, reﬂects their structural properties: among
his three “adjectives”, they are the least diﬀerent from N.
10 See also “precopular nouns” in Martin (1975). Words of Muraki’s third “adjective” category listed
here as well as many others such as nama ‘raw’ and ippai ‘full’ are discussed in Uehara (1998:
Chap. 3) as those lexical items which are translated into adjectives in English but take no instead of
na in modiﬁcation.
Lexical categories 85
Some other non-prototypical subcategories can be characterized as such, not by
lacking a form or two of one paradigm, but by having forms of more than one paradigm,
listed in Table 8. Uehara (1998: Ch. 3) discusses what he calls double-function
cases of many words (typically denoting abstract concepts or human attributives)
among ANs, such as heiwa ‘peaceful/peace’ and kenkoo ‘healthy/health’. They are
ANs (taking na in modiﬁcation) but take no as well in modiﬁcation and/or can take
case-marking particles in reference as in (43), which indicates their ambivalent
status between AN and N. Such ambivalent AN/N words constitute a non-prototypical
subcategory of the Nominal category.
(43) a. heiwa na kuni b. kono kuni wa heiwa da.
country this country TOP COP
‘a peaceful country’ (AN) ‘This country is peaceful.’ (AN)
c. heiwa no sisya d. heiwa ga dai-iti da
messanger NOM ﬁrst COP
‘a messenger of peace’ (N) ‘Peace comes ﬁrst.’ (N)
Also among non-prototypical members of the Nominal category are those words
conventionalized with forms of the AN and VN paradigms, such as manzoku na/suru
‘satisfactory/get satisﬁed’ and nangi na/suru ‘diﬃcult/have diﬃculty’ as in (44)
below (See also the discussion of such Sino-words in Chapter 3 [Kobayashi, Yamashita
and Kageyama, this volume]). In a similar vein, some lexical roots (e.g. atataka
‘warm’, okasi ‘funny’) can be characterized as having been conventionalized with
forms of A and AN to various degrees, as in (45), where their forms in modiﬁcation
are followed by those in predication in square brackets (See also Uehara (1998:89)
for the list of such words).
(44) a. manzoku na kotae b. sono kotae ni manzoku suru
answer the answer with
‘a satisfactory answer’ (AN) ‘be satisﬁed with the answer’ (VN)
(45) a. atataka {-i/na} hizasi [hizasi ga atataka {-i/da}]
sunlight sunlight NOM
‘warm sunlight’ (A/AN) ‘The sunlight is warm.’ (A/AN)
b. okasi {-i/na} hanasi [sono hanasi wa okasi {-i/*da}]
story the story TOP
‘a funny story’ (A/AN) ‘The story is funny.’ (A)
The existence of such non-prototypical members/subcategories demonstrates
that they have as their prototypes the ﬁve “large” lexical categories and their structural
paradigms in Table 8. They are in turn prototypically organized as represented
in Figure 1, where Noun and Verb are their prototypical cores, respectively motivated
by propositional functions of reference and predication.
86 Hideki Kishimoto and Satoshi Uehara
5 Conclusion and future research perspectives
In this chapter, some issues surrounding the Japanese lexical categories (parts of
speech) have been discussed. In the literature on Japanese, words have been classiﬁed
in a number of diﬀerent ways, and the currently available proposals concerning
their categorization diﬀer signiﬁcantly depending on what kind of perspective has
been adopted. In traditional Japanese grammar, parts of speech are deﬁned with
an emphasis on their morphology. The point of divergence in traditional Japanese
grammar lies in the morphological status of words, in relation to their categorization,
and its major controversies concern the status of adjectival nouns and the treatment
of minor categories (such as auxiliaries). On the other hand, the generative and
cognitive-typological approaches aim to characterize categories from the perspective
of language universals, although the former often takes them as a given while the
latter motivates them by cross-linguistically attested data. The major diﬀerence
between the two views lies in the fact that the generative approach takes lexical
categories to be discrete, while the cognitive-typological approach does not a priori
assume them to be, admitting of them being non-discrete entities, which often form
a continuum.
In the generative framework, lexical categories are deﬁned via cross-categorical
features, which are claimed to be universal (i.e. they are regarded as valid across
languages). In Japanese generative grammar, there have been issues on how the
lexical categories should be deﬁned in terms of cross-categorial features, since the
language has more major lexical categories than can be deﬁned by the simple
feature system [±V, ±N], due to the presence of adjectival nouns and verbal nouns.
Further, Japanese is an agglutinative language, where bound predicative elements
(regardless of their categorial status) form a morphologically tight unit with the
main predicate. Thus, some but not all predicative elements traditionally classiﬁed
as dependent auxiliaries can be identiﬁed as major lexical categories. Besides, function
words (belonging to functional categories) can be free or bound morphemes.
These facts suggest that morphological criteria do not necessarily constitute reliable
heuristics to distinguish between the major lexical and the functional categories of
Japanese predicative elements.
In the cognitive-typological approach, words (and their categories) may be
gradient, and the structural organization of ﬁve lexical categories in Japanese can
be identiﬁed with reference to criteria utilized in a prototypical approach. Using
cross-linguistically applicable markedness criteria, it has been shown that Japanese
can be characterized as possessing two major lexical categories (inﬂectional and
non-inﬂectional categories), which have the ﬁve lexical categories as their subcategories.
The dividing line between the two major categories is the boundness
criterion, which is a linguistically salient contrast in morphological boundness and
is used to deﬁne the inﬂection of the language. Inﬂectional and non-inﬂectional
Lexical categories 87
categories are unmarked for the two major pragmatic functions of predication and
reference, respectively.
It is worthwhile to note that nouns and verbs are taken as constituting rudimentary
categories in all the approaches, which is naturally expected, given that both
are likely universal categories cross-linguistically (cf. Whaley 1997). By contrast,
some minor categories, which do not appear to be common or universal typologically,
give rise to a number of controversies. In this regard, Japanese morphology
is not necessarily eﬀective in evaluating the category memberships; traditionally,
categories are distinguished placing emphasis on morphological criteria, but as
noted earlier, there is no tight correlation between the morphological distinction of
“agglutinated/bound” and “free” and the distinction of “lexical” versus “functional”.
In addition, lexical items sometimes seem to have gradient properties or
could be ambiguous in their class membership, in which case category labels cannot
be assigned uniquely.
There are many theoretical issues that are worth pursuing for future research.
The most prominent and yet fundamental issue concerns the criteria for distinguishing
categories. The criteria primarily used for this purpose diﬀer among the three
approaches. Traditional Japanese grammar gives precedence to morphological criteria
over syntactic ones. Generative grammar uses both morphological and syntactic
criteria equally to distinguish categories, but cognitive grammar adopts semantic
and functional criteria as well as structural ones. Then, the question inevitably arises
as to what would be the optimal way of determining the status of lexical categories.
Furthermore, in the classiﬁcation of lexical categories, super-categories (such as
taigen (nominal) and yōgen (predicative), which groups verbs, adjectives and adjectival
nouns together) and sub-categories (as seen in the ﬁner classiﬁcations of
particles) are often posited. This raises the question of what level of classiﬁcation
is appropriate for capturing basic generalizations in Japanese grammar, and it
still remains to be seen how ﬁne-grained criteria should be used to identify lexical
categories. It would also be interesting to see whether the generalization on lexical
categories in standard Japanese applies to various dialects of Japanese, since
dialects sometimes show curious properties that help us clarify issues that are not
readily resolved only by considering standard Japanese (see e.g. Kudō 2004).
Acknowledgments
We are grateful to Taro Kageyama, Masayoshi Shibatani, and John Haig for valuable
comments and suggestions, which led to signiﬁcant improvements of the ﬁnal version
over earlier versions of the present chapter, in particular, in its coverage. While
the two authors are responsible for the whole chapter, Kishimoto mainly drafted
Sections 2 and 3, and Uehara drafted Section 4. Part of the work derives from the
88 Hideki Kishimoto and Satoshi Uehara
collaborative research project “Syntactic, Semantic, and Morphological Characteristics
of the Japanese Lexicon” (Project leader: Taro Kageyama, National Institute for
Japanese Language and Linguistics, 2010–2015).
References
Abney, Steven. 1987. The English noun phrase and its sentential aspect. Cambridge, MA: MIT disser-
tation.
Backhouse, Anthony. 2009. On the status of nai in modern Japanese. Journal of Japanese Linguistics
25. 59–74.
Baker, Mark. 2003. Lexical categories: Verbs, nouns, and adjectives. Cambridge: Cambridge University
Press.
Bloch, Bernard. 1946. Studies in colloquial Japanese I: Inﬂection. Journal of the American Oriental
Society 66. 97–109.
Brinton, Laurel and Elizabeth Closs Traugott. 2005. Lexicalization and language change. Cambridge:
Cambridge University Press.
Carnie, Andrew. 2007. Syntax: A generative introduction. 2nd edition. Malden, MA: Blackwell.
Chomsky, Noam. 1970. Remarks on nominalization. In Roderick A. Jacobs and Peter S. Rosenbaum
(eds.), Readings in English transformational grammar, 184–221. Waltham, MA: Ginn.
Croft, William. 1991. Syntactic categories and grammatical relations: The cognitive organization of
information. Chicago: University of Chicago Press.
Croft, William. 2001. Radical construction grammar: Syntactic theory in typological perspective.
Oxford: Oxford University Press.
Emonds, Joseph. 1985. A uniﬁed theory of syntactic categories. Dordrecht: Foris.
Fujitani, Nariakira. 1778. Ayui shō: Yosoinokata [Suﬃxal dependents: Inﬂectionals]. In Masao Takeoka
(ed.), 1961–1962. Fujitani Nariakira zenshū. 2 vols. Tokyō: Kazama Shobō.
Fukui, Naoki. 1986. A theory of category projection and its applications. Cambridge, MA: MIT disser-
tation.
Greenberg, Joseph H. 1966. Language universals: With special reference to feature hierarchies. The
Hague: Mouton.
Hashimoto, Shinkichi. 1934. Kokugohō yōsetsu [An outline of Japanese grammar]. Tokyo: Meiji Shoin.
Hashimoto, Shinkichi. 1948. Kokugohō kenkyū [Studies on Japanese grammar]. Tokyo: Iwanami Shoten.
Hashimoto, Shinkichi. 1969. Joshi Jodōshi no kenkyū [Studies on particles and auxiliaries]. Tokyo:
Iwanami Shoten.
Hopper, Paul J. and Elizabeth Closs Traugott. 1993. Grammaticalization. Cambridge: Cambridge University
Press.
Kageyama, Taro. 1982. Word formation in Japanese. Lingua 57. 215–258.
Kageyama, Taro. 1993. Bunpō to gokeisei [Grammar and word formation]. Tokyo: Hituzi Syobo.
Kageyama, Taro. this volume. Chapter 14: Lexical integrity and the morphology-syntax interface. In
Taro Kageyama and Hideki Kishimoto (eds.), Handbook of Japanese lexicon and word forma
tion. Berlin: De Gruyter Mouton.
Kato, Yasuhiko. 1985. Negative sentences in Japanese. Sophia Linguistica 19. Tokyo: Sophia University.
Kishimoto, Hideki. 2005. Tōgokōzō to bunpō kankei [Syntactic structures and grammatical relations].
Tokyo: Kurosio Publishers.
Kishimoto, Hideki. 2007. Negative scope and head raising in Japanese. Lingua 117. 247–288.
Kishimoto, Hideki. 2008. On the variability of negative scope in Japanese. Journal of Linguistics
44. 379–435.
Lexical categories 89
Kishimoto, Hideki. 2009. Hobun o toru dōshi to keiyōshi: Kontorōru to jōshō [Complement taking
verbs and adjectives: Control and raising]. In Taro Kageyama (ed.), Nichi ei taishō keiyōshi/
fukushi no imi to kōbun [The meanings of adjectives and adverbs and their constructions],
152–190. Tokyo: Taishukan Shoten.
Kishimoto, Hideki. 2013. Verbal complex formation and negation in Japanese. Lingua 135. 132–154.
Kitahara, Yasuo. 1981. Nihongo jodōshi no kenkyū [A study of Japanese auxiliaries]. Tokyo: Taishukan
Shoten.
Konno, Hiroaki. 2012. The Japanese adjectival conjugational ending drop construction: From a syntaxsemantics
interface perspective. Gengo Kenkyu 141. 5–32.
Konoshima, Masatoshi. 1973. Kokugo jodōshi no kenkyū [A study of Japanese auxiliaries]. Tokyo:
Ōhūsha.
Kudō, Mayumi. 2004. Nihongo no asupekuto tensu mūdo taikei: Hyōjungo kenkyū o koe te [The
tense-mood-aspect system in Japanese: Beyond standard Japanese studies]. Tokyo: Hituzi
Syobo.
Kuno, Susumu. 1973. The structure of the Japanese language. Cambridge, MA: MIT Press.
Lakoﬀ, George. 1987. Women, ﬁre, and dangerous things: What categories reveal about the mind.
Chicago: University of Chicago Press.
Langacker, Ronald W. 1987. Foundations of cognitive grammar, volume 1: Theoretical prerequisites.
Stanford: Stanford University Press.
Martin, Samuel E. 1975. A reference grammar of Japanese. New Haven: Yale University Press.
Matsushita, Daizaburō. 1924. Hyōjun nihon bunpō [Standard Japanese grammar]. Tokyo: Kigensha.
Matsushita, Daizaburō. 1930. Kaisen hyōjun nihon bunpō [Revised standard Japanese grammar].
Tokyo: Tyūbunkan.
Mikami, Akira. 1953. Gendai gohō josetsu [Prolegomena to modern usage]. Tokyo: Tōkō Shoin.
Mizutani, Shizuo. 1951. Keiyōdōshi-ben [An account of adjectival nouns]. Kokugo to Kokubungaku
[Japanese Language and Literature] 28(5). 31–47.
Miyagawa, Shigeru. 1987. Lexical categories in Japanese. Lingua 73. 29–51.
Monbushō. 1947. Chūtōbunpō kōgo [Intermediate grammar: colloquial language] Tokyo: Chūtōgakkō
Kyōkasho.
Muraki, Shinjirō. 2012. Nihongo no hinshi taikei to sono shūhen [Japanese parts-of-speech and their
surroundings]. Tokyo: Hituzi Syobo.
Ohkado, Masayuki. 1991. On the status of adjectival nouns in Japanese. Lingua 83. 67–82.
Okutsu, Kei-ichiro. 1978. Boku wa unagi da no bunpō [The grammar of “I’m an eel”]. Tokyo: Kurosio
Publishers.
Ōtsuki, Fumihiko. 1889. Gohō shinan [Instructions for usage]. In Genkai [Word Sea] (Dictionary of
Japanese).
Rodriguez, João. 1604–1608. Arte da lingoa de Japam. Translated as Nihon dai bunten [A grand
book of Japanese grammar] by Tadao Doi, 1955. Tokyo: Sanseido.
Rosch, Eleanor. 1978. Principles of categorization. In Eleanor Rosch and Barbara L. Lloyd (eds.),
Cognition and categorization, 27–48. Hillsdale, NJ: Erlbaum.
Sakuma, Kanae. 1951. Asahi shin kōza: Nihongogaku [Asahi new series: Japanese linguistics]. Tokyo:
Asahi Shinbunsha.
Searle, John. 1969. Speech acts: An essay in the philosophy of language. Cambridge: Cambridge
University Press.
Shibatani, Masayoshi. 1990. The languages of Japan. Cambridge: Cambridge University Press.
Takezawa, Koichi. this volume. Chapter 13: Inﬂection. In Taro Kageyama and Hideki Kishimoto (eds.),
Handbook of Japanese lexicon and word formation. Berlin: De Gruyter Mouton.
Taylor, John R. 1989. Linguistic categorization: Prototypes in linguistic theory. Oxford: Clarendon
Press.
90 Hideki Kishimoto and Satoshi Uehara
Teramura, Hideo. 1982. Nihongo no shintakusu to imi I [The syntax and the meaning of Japanese I].
Tokyo: Kurosio Publishers.
Tōjō, Gimon. 1841. Tama no o kuri wake [Classiﬁcation of words]. Osaka.
Tokieda, Motoki. 1950. Nihon bunpō kōgohen [Colloquial Japanese grammar]. Tokyo: Iwanami Shoten.
Uehara, Satoshi. 1998. Syntactic categories in Japanese: A cognitive and typological introduction.
Tokyo: Kurosio Publishers.
Uehara, Satoshi. 2003. A diachronic perspective on prototypicality: The case of nominal adjectives
in Japanese. In Hubert, Cuyckens, René Dirven, and John R. Taylor (eds.), Cognitive approaches
to lexical semantics, 363–391. Berlin: Mouton de Gruyter.
Uehara, Satoshi and Fumiko Kumashiro. 2007. On’in keitai no mekanizumu: Ninchi on’in keitairon no
apurōchi [The mechanism of morpho-phonology: An approach to cognitive morpho-phonology].
Tokyo: Kenkyusha.
Wetzer, Harrie. 1992. “Nouny” and “verby” adjectivals: A typology of predicative adjectival constructions.
In Michel Kefer and Johan van der Auwera (eds.), Meaning and grammar: Cross linguistic
perspectives, 223–262. Berlin: Mouton de Gruyter.
Whaley, Lindsay J. 1997. Introduction to typology: The unity and diversity of language. Thousand
Oaks, CA: Sage Publications.
Yamada, Yoshio. 1908. Nihon bunpōron [On Japanese grammar]. Tokyo: Hōbunkan.
Yamada, Yoshio. 1936. Nihon bunpōgaku gairon [Introduction to Japanese grammar]. Tokyo: Hōbunkan.
Yamaguchi, Akiho. 2004. Nihongo no ronri [The logic of Japanese]. Tokyo: Taishukan Shoten.
Yumoto, Yoko. 2005. Fukugō dōshi hasei dōshi no imi to tōgo [The meaning and the syntax of
compound and derivational verbs]. Tokyo: Hituzi Syobo.
Lexical categories 91
Handbook of
Japanese Lexicon
and Word Formation
Edited by
Taro Kageyama
Hideki Kishimoto
ISBN 978-1-61451-275-2
e-ISBN (PDF) 978-1-61451-209-7
e-ISBN (EPUB) 978-1-5015-0081-7
ISSN 2199-2851
Library of Congress Cataloging-in-Publication Data
A CIP catalog record for this book has been applied for at the Library of Congress.
Bibliographic information published by the Deutsche Nationalbibliothek
The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliograﬁe;
detailed bibliographic data are available on the Internet at http://dnb.dnb.de.
6 2016 Walter de Gruyter Inc., Boston/Berlin
Cover image: Hideki Kishimoto
Typesetting: RoyalStandard, Hong Kong
Printing and binding: CPI books GmbH, Leck
♾ Printed on acid-free paper
Printed in Germany
www.degruyter.com