Content uploaded by Jan Rijkhoff
Author content
All content in this area was uploaded by Jan Rijkhoff on May 17, 2019
Content may be subject to copyright.
© 2007 The Author
Journal Compilation © 2007 Blackwell Publishing Ltd
Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Word Classes
Jan Rijkhoff*
University of Aarhus
Abstract
This article1 provides an overview of recent literature and research on word classes,
focusing in particular on typological approaches to word classification. The cross-
linguistic classification of word class systems (or parts-of-speech systems) presented
in this article is based on statements found in grammatical descriptions of some
50 languages, which together constitute a representative sample of the world’s
languages. It appears that there are both quantitative and qualitative differences
between word class systems of individual languages. Whereas some languages
employ a parts-of-speech system that includes the categories verb, noun, adjective
and adverb, other languages may use only a subset of these four lexical categories.
Furthermore, quite a few languages have a major word class whose members
cannot be classified in terms of the categories verb–noun–adjective–adverb, because
they have properties that are strongly associated with at least two of these four
traditional word classes (e.g. adjective and adverb). Finally, this article discusses
some of the ways in which word class distinctions interact with other grammatical
domains, such as syntax and morphology.
1. Introduction
Differentiation between word classes can be regarded as an instance of one
of the most fundamental traits of human cognition: putting people or
things, but also more abstract entities such as words, into groups on the
basis of certain shared characteristics (categorization). Traditionally, the
following ten word classes are distinguished: verb (sit, go, read, etc.), noun
(dog, tree, table, etc.), adjective (blue, cheap, nice, etc.), adverb (here, today,
well, often; more on adverbs below), preposition (in, on, below, before, after,
etc.),2 numeral (one, two, etc.), article (the, a/an), pronoun (you, they;
someone, anyone; who, whose etc.), conjunction (and, or; if, because, etc.)
and interjection (shh, oh no, phew, hey, hmm, etc.). This categorization
is, however, rather biased towards word classes in the familiar European
languages and several typological studies have suggested that the traditional
set of categories mentioned above needs to be revised so as to be able to
account for certain word classes attested in other, often more ‘exotic’
languages (see, for example, Kuipers 1968; Broschart 1997).
710 Jan Rijkhoff
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
The term ‘word class’ is used in two ways. In the wide sense it covers
both grammatical and lexical word classes, in the narrow sense it only includes
lexical word categories. This article is mostly concerned with lexical
word classes in the languages of the world (parts -of-speec h system s).
Words that belong to a grammatical word class (also called ‘function
words’ or ‘empty words’) have little or no identifiable meaning and belong
to a closed paradigm with very few members (i.e. there are few other
words that belong to the same word class). articles, various kinds of
pronouns and conjunctions are all examples of grammatical words.
(1) Definition of the grammatical word ‘article’ according to the Long-
man Dictionary of Contemporary English (LDCE):
a word used before a noun to show whether the noun refers to a particular
example of something or to a general example of something. In English, ‘the’
is called the definite article and ‘a’ and ‘an’ are called the indefinite article.
lexical words, on the other hand, have a definable semantic content
(hence, they are also referred to as ‘content words’) and typically belong
to an open word class, that is, the number of words belonging to a lexical
word class is not fixed. English has four major lexical words classes: verbs,
nouns, adjectives and adverbs. In (2) all major lexical word classes are
represented.
(2) The BrazilianAdj studentN workedV hardAdv.
Often a lexical word has several meanings (senses), as is shown in the
definition of the noun ‘tree’:
(3) Definition of the lexical word ‘tree’ (LDCE):
1. a very tall plant that has branches and leaves, and lives for many years;
2. a drawing that connects things with lines to show how they are
related to each other.
It can be shown for many grammatical words, however, that they originated
as lexical words or phrases (a phenomenon known as grammaticalization;
Heine et al. 1991; Hopper and Traugott 2003). For example, verbs are often
the source of prepositions (e.g. Yoruba
∫
iV ‘use’ →
∫
iPrep ‘with’), comple-
mentizers (e.g. Ewe béV ‘say’ → béCompl ‘that’, as in ‘He saw that ...’),
subordinators or quotation markers (Lord 1989: 307–8; Heine and Kuteva
2002: 263), but as the grammaticalization of a lexical word or phrase does
not happen overnight, there is a long period in which it is not always
possible to draw a hard and fast line between lexical and grammatical words
(on gradience, see, for example, Sasse 2001; Aarts 2007). In English,
auxiliary verbs are good examples of words that fall somewhere between
the lexical and the grammatical end of the word spectrum:
(4) I am going to go to Amsterdam in an hour (or: I’m GOnna GO to Amsterdam
in an hour)
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
Word Classes 711
Whereas the second instance of the verb go in (4) is clearly lexical (LDCE:
‘to move or travel to a place that is away from where you are or where
you live’), the first occurrence (in the phrase be going to or be gonna) is
clearly less than completely lexical and mainly serves to indicate future
time reference.
As was already mentioned at the outset, this article is concerned with
word classes in the narrow sense, that is, we will only discuss the four
major lexical word classes verb, noun, adjective and adverb.
2. Some Recent Approaches to Lexical Word Classes (Parts-of-Speech Systems)
More than 2000 years ago, words were already categorized into distinct
groups by ancient Greek philosophers, but it has proven to be a rather
difficult task to come up with definitions for the various lexical word
classes that have cross-linguistic validity (Anward et al. 1997). This is
mainly due to the fact that word classes in individual languages are often
distinguished on the basis of certain language-specific criteria, which do
not necessarily apply to other languages. For example, to say that a noun
is a word that is inflected for number is quite irrelevant for the many
languages across the globe in which number marking is absent (Rijkhoff
2004: 146–52). Notice also that even languages with nominal number
marking have nouns that do not inflect for number (cf. treeN – treesN, but
goldN – *golds). It is not possible to use purely semantic criteria either.
Although verbs are typically associated with actions or situations (i.e.
temporal entities) and nouns with things (spatial entities), there are many
nouns that are used for non-spatial entities such as events (meetingN, wed-
dingN, funeralN and gameN), feelings (loveN, hateN) and other abstract entities
(linguisticsN, politicsN). Furthermore, a concept that is lexicalized as a noun
in one language may be lexicalized as a verb in another language (Evans
2000a). Recently, however, some new ideas have been proposed to deal
with word classes across languages by Croft (1991, 2000), Anward et al.
(1997), Baker (2003) and Hengeveld (1992) (see also Hengeveld et al.
2004, Hengeveld and Rijkhoff 2005).
Croft argues that verb, noun and adjective are not categories of par-
ticular languages but rather language universals in the sense that they
constitute what he calls ‘typological prototypes’. He focuses on the con-
structions that are used for the three universally attested communicative
functions, predication, reference and modification, rather than the word
classes associated with these functions (verb, noun and adjective, respec-
tively).3 In his view (Croft 2000: 84–5, 87):
Categories in a particular language are defined by the constructions of the
language. Moreover, the constructions are the primitive elements of syntactic
representation; categories are derived from constituents. [ . . . ] Constructions
define grammatical categories.
712 Jan Rijkhoff
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
The range of constructions in the universal-typological theory of parts of
speech covers constructions for predication, reference and modification. Most
important, it explicitly recognises that predication, reference and modification are
pragmatic (communicative) functions or, as Searle described them, propositional
acts [ . . . ]
Croft then argues that one should focus on the unmarked combination
of pragmatic function and lexical class and that in the case of parts-of-speech
the unmarked combinations are (Croft 2000: 88):
Any other combination of pragmatic function and lexical class is marked.
Thus, the English object word vehicle is unmarked for reference but
marked for predication (be a vehicle) and modification (vehicular, vehicle’s).
Croft’s approach has much in common with the ideas proposed earlier by
Hopper and Thompson (1984), who investigated properties of word
classes from a discourse perspective. They stated that the basic categories
‘noun’ and ‘verb’ are best viewed as ‘universal lexicalizations of the pro-
totypical discourse functions of “discourse-manipulable participant” and
“reported event”, respectively’, and concluded that ‘categoriality itself is
another fundamental property of grammars which may be directly derived
from discourse function’ (Hopper and Thompson 1984: 703).
Anward et al. (1997) propose a multi-dimensional method for the
cross-linguistic study of word classes. In their view, a combination of
phonological (lexical), morphological and syntactic criteria should be used
(feature-clustering) to distinguish between word classes. In the case of
maximal part-of-speech differentiation, we find that the members of dif-
ferent word classes can be distinguished on phonological, morphological
and syntactic grounds. At the opposite extreme, in the case of no part-
of-speech differentiation, there are no phonological, morphological or
syntactic differences between words. In between these two extremes, there
are six other logical possibilities (Anward et al. 1997: 173 –8):4
(5) lexical class pragmatic function
Verb predication of an action
Noun reference to an object
Adjective modification by a property
(6) PHON MORPH SYNT
+ + + Maximal part-of-speech differentiation
+−−Exclusively phonological part-of-
speech differentiation
−+−Exclusively morphological part-of-
speech differentiation
−− + Exclusively syntactic part-of-speech
differentiation
++ −Non-syntactic part-of-speech
differentiation
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
Word Classes 713
This template would then provide the basis of more complex parts-of-
speech systems, also allowing for subclasses, superclasses and overlap. Addition-
ally, it should be investigated how the various parts of speech cluster in a
language (e.g. Are there languages with nouns but without verbs?). The
authors point out that if we were to take into account all dimensions of
variation (notably the possibility of allowing for overlap and partial or total
inclusion), we should find an ‘enormous spectrum of logically possible
part-of-speech-related variation among languages.’ Because it is unlikely
that all logical possibilities actually occur, they state that it is the task of
typology to establish the limits on the cross-linguistic variation of parts-
of-speech systems (Anward et al. 1997: 179).
Baker’s (2003) proposals are formulated within Chomsky’s generative
grammar. In this syntactocentric approach, verbs are characterized as
words that always project a subject (or rather specifier in generative ter-
minology). Nouns are characterized by the fact that they bear a referential
index; they are the only words about which it makes sense to ask whether
or not its referent [or rather the referent of the phrase it is the head of,
as only noun phrases (NPs) can be used to refer] is identical with another
entity. Adjectives, finally, are regarded as the unmarked lexical category:
they lack both a specifier and a referential index. In Baker’s view, the
verb–noun–adjective distinction is universal, leaving no room for languages
whose parts-of-speech system deviates from the traditional lexical categories
that have dominated European linguistics since antiquity.
In Hengeveld’s approach crucial reference is made to the function(s) that
a lexical item can fulfill in certain linguistic structures without the speaker
having to resort to special grammatical measures such as relative clause
formation, as in (9) or (14), or the use of medial verb constructions, as in
(15) below. Here are Hengeveld’s (slightly reformulated) definitions of the
four categories verb, noun, adjective and (manner) adverb:
a verb is a word that can ONLY be used as the head of a clause (or rather ‘predicate
phrase’ in Hengeveld’s terminology; see Figure 2);
a noun is a word that can be used as the head of a noun phrase or ‘NP’ (called
‘referential phrase’ in the original publication; more on the label ‘noun phrase’
below);
an adjective is a word that can be used as a modifier of the head of a noun phrase;
a manner adverb is a word that can be used as a modifier of the head of a clause.5
(Hengeveld 1992: 68; see also Hengeveld and Rijkhoff 2005: 406–7)
The reason Hengeveld restricts himself to manner adverbs (such as hard
in She works hard) is that other kinds of adverbs, such as yesterday or hopefully,
−+ + Non-phonological part-of-speech
differentiation
+−+ Non-morphological part-of-speech
differentiation
−− −No part-of-speech differentiation
714 Jan Rijkhoff
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
do not modify the head of the clause, but larger units within the sentence.
Notice that Hengeveld (contrary to Croft) claims that word classes con-
stitute true categories of particular languages, giving special status to verbs,
which are defined as words that can only occur as predicates (head of the
clause). This is because in many languages members of other word classes
can also be used as the head of a clause, but for verbs this is the only
unmarked option.
For example, in Dutch an adjective such as langAdj ‘long, tall’ requires
the presence of a copula (i.e. an extra measure is necessary; in the example
below a form of zijn ‘be’) when it appears as the head of the clause:
Dutch
Because many other languages do not require an extra measure in the
case of a non-verbal predicate, Hengeveld’s definition of adjectives and other
non-verbal word classes (nouns, adverbs) leaves open the possibility that
they can also be used as the head of the clause without extra measures, as
in Mojave (a native American language of the Yuman family, spoken in
Arizona and California):
Mojave (Schachter 1985: 19)
This does not mean that Mojave does not distinguish between verbs and
adjectives. The following examples show that verbs but not adjectives must be
relativized (REL) when they modify the head of the NP (Hengeveld 1992: 75):
Mojave (Schachter 1985: 19)
The next section gives an overview of lexical parts-of-speech systems
in the languages of the world.
3. Parts-of-Speech Systems in the Languages of the World
The overview provided in the current section focuses on two important
aspects of parts-of-speech systems: (i) the number of lexical word classes in a
(7) [Die manN]NP isCop langAdj
that manNbe:3.SG.PRES tallAdj
‘That man is tall’
(8) [
ʔ
i:paN-c]NP homi:Adj-k
manN-SUBJ tallAdj-PRES
‘The man is tall’
(9) [
ʔ
i:paNkw-su:pawV-ny-c]NP . . .
manNREL-knowV-DEM-SUBJ . . .
‘The man who knows . . .’
(10) [
ʔ
i:paNhomi:Adj-ny-c]NP . . .
manNtallAdj-DEM-SUBJ . . .
‘The tall man . . .’
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
Word Classes 715
language (a quantitative consideration), and (ii) the nature of a lexical word
class, that is, whether its members are ‘rigid’ or ‘flexible’ (a qualitative
consideration).
3.1. LANGUAGE S WITH A SINGLE LEXICAL WORD CLASS
Whether or not there are languages with just one lexical word class is
a controversial issue, which is at least partly due to the fact that such
languages are so unlike the familiar, well-studied languages of Western
Europe and contradict the widely accepted claim that all languages have
distinct classes of nouns and verbs.6 Never theless, several linguists working
with lesser known languages have argued that there are indeed languages
with only one lexical word class, which comes in two varieties. On the
one hand, it has been claimed that certain languages only have verbs and
that reference to participants in an event is achieved by clause-like con-
structions, as in this example from Cayuga (an American Indian language
from the Iroquoian family):
Cayuga (Iroquoian; Sasse 1993b: 657)
The literal meaning, however, would be something like (Sasse 1993b:
657): ‘It became lost to him, it is his wallet, he is this man’ or ‘it losted
him, it walle t s him, the one who mans.’
On the other hand, there are languages that are deemed to have a
single lexical category whose members are extremely flexible in that
the same word can fulfill all the major lexical functions (head of clause,
head of NP, modifier of head of clause or modifier of head of NP)
without requiring any special measures. Thus, in Samoan (a Polynesian
language) ‘there are no lexical or grammatical constraints on why a
particular word cannot be used in the one or the other function’ (Mosel
and Hovdhaugen 1992: 73–4, 77; see also, for example, Broschart 1997
on Tongan):
. . . the categorization of full words is not given a priori in the lexicon. [ . . . ]
It is only their actual occurrence in a particular environment which gives them
the status of a verb or a noun. [ . . . ] What is given in the lexicon is not a
particular word class assignment, but the potential to be used in certain syn-
tactic environments as a noun or a verb.
[ . . . ] all full words which function as noun and verb phrase nucleus can also
be used as attributive modifiers.
The following examples show that la can occur both as the head of an
NP (12) and the head of a clause (13):
(11) a-hó-hto:’ ho-tkwe’t-a’ n e:kyÈh-okweh
PAST-it:to_him-become_lost it:him-wallet-be this he:it-man
‘This man lost his wallet’
716 Jan Rijkhoff
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
Samoan (Austronesian; Mosel and Hovdhaugen 1992: 80)
3.2. LANGUAGE S WITH TWO LE XICAL WORD CLASSES
Languages that have two lexical word classes also come in the two varieties
rigid and flexible. Many languages only have distinct classes of verbs and
nouns; adjectival notions and manners are expressed in various ways. For
example, Galela (a Papuan language spoken on the island of Halmaheira)
has verbs and nouns but no separate class of adjectives. Instead it uses a
construction headed by a participialized verb to express a concept such as
English ‘big’ (notice that the participle is formed by reduplicating the first
syllable of the verb lamo ‘be big’).
Galela (van Baarda 1908: 35)
The Australian language Ngiyambaa also seems to have just two major
lexical word classes. Like Galela, it has a class of true verbs, but in Ngiymabaa
the other word class does not consist of nouns, but comprises a group of
words that can serve both as the head of an NP and as a modifier. Even
though members of this word class do not all behave in exactly the same
way morphologically (cf. English, where not all nouns can occur in the
plural), the author sees no reason to distinguish more classes, because such
a distinction ‘would serve no descriptive purpose elsewhere in the grammar’
(Donaldson 1980: 70).
3.3. LANGUAGE S WITH THRE E LEXI CAL WORD CLASSES
Wambon (a Papuan language spoken in Southern Irian Jaya, Indonesia)
has separate classes, verbs, nouns and adjectives, but apart from one or two
exceptions it has no (manner) adverbs. Instead it employs medial verb
constructions (de Vries 1989: 49):
The category of manner adverbs can be so marginal because Wambon prefers
to use medial verbs as modifiers of other verbs in serial verb constructions in
which the modifying verb immediately precedes the modified verb. [...] Very
often the medial verbs specifying manner, are verbs which are derived from
adjectives by -mo [ . . . ].
(12) ‘Ua malosi le la
PERF strong ART sun
‘The sun is strong.’ (lit. ‘The sun strongs’)
(13) ‘Ua lale aso.
PERF sun ART day
‘The sun is shining today.’ (lit. ‘The day suns’)
(14) awi dohu i lalamo
his foot it be_big:PRT
‘his big foot’
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
Word Classes 717
For example, in (15) the verb matetmo ‘be good’ is derived from the
adjective matet ‘good’:
Wambon (de Vries 1989: 49)
Ngiti, a Nilo-Saharan language (Central Sudanic branch) spoken in Zaire,
also has three major word classes, but in addition to verbs and nouns it
has a word class that combines the modifier function of adjectives and
adverbs in other languages (Kutsch Lojenga 1994: 336):
There is no morphological nor a clear syntactic distinction between a class of
adjectives and a class of adverbs in Ngiti. The functional term modifiers is
therefore used [...] to cover a fairly large . . . class of words, containing about
150 items, which are neither nouns nor verbs and which all have a modifying
function in relation to different constituents.
In (16) the word àn
7
is used adjectivally to modify a noun meaning
‘light (of weight)’, whereas in (17) it serves as a manner adverb (modifying
the verb carry) meaning ‘easily, without effort’.
Ngiti (Kutsch Lojenga 1994: 338)
3.4. LANGUAGES WITH FOUR LEXICAL WORD CLASSES
Languages with four distinct lexical classes (verb, noun, adjective, adverb)
are attested across the globe, English being one of them. Adverbs differ
from the other word classes in that it is generally not possible to establish
which members of this word class are more or less prototypical. Instead
we find a variety of subtypes, each of which is equally ‘representative’ for
the word class adverb.
In English many adverbs are actually derived from adjectives: beautifulAdj
– beautifullyAdv, nice – nicely, polite – politely.
(15) Jakhov-e matet-mo ka-lembo?
they-CN good-SUPP.SS go-3PL.PAST
‘Did they travel well?’
(16) ngbángba n6-ìtdù 6s
4
àn
7
child RSM-carry:PERF.PRES light load
‘The child carried a light load’
(17) 6s
4
ngbángba n6-ìtdù àn
7
light child RSM-carry:PERF.PRES load
‘The child carried the load easily’
(18) a. I will give you the keys tomorrow.[tomorrow = adverb of time]
b. She often goes swimming
in the morning.
[often = adverb of frequency]
c. Does she speak German well? [well = manner adverb]
d. He was not there. [there = adverb of place]
718 Jan Rijkhoff
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
3.5. CLASSIFICATION OF PARTS -OF-SPEE CH SYSTEM S
The cross-linguistic overview of parts-of-speech systems provided above
is based on statements in language descriptions rather than ideas put
forward in theoretical discussions. Because Hengeveld (1992) collected
information about parts-of-speech systems (henceforth PoS systems) in a
representative sample of the world’s languages, it is hardly surprising that
his classification of PoS systems covers all the PoS systems exemplified in
Sections 3.1–3.4 (recall that we are only taking into account manner
adverbs here). It must be emphasized, however, that the various types of
PoS systems in Figure 1 should be regarded as reference points on a scale
rather than distinct categories. Because languages are dynamic entities,
they can only approximate the ideal types in this classification.
Hengeveld makes a distinction between language with a flexible PoS
system and languages with a rigid PoS system. In languages with a flexible
PoS system, some or all of the functions that are typically associated with
the four traditional (rigid) lexical categories are performed by members
of the same word class (Types 1–3). In languages with a rigid PoS system
(Types 4–7), these functions are distributed over distinct, non-overlapping
groups of words (for details see Hengeveld et al. 2004). The three flexible
word classes are called contentive, non-verb and modifier. A contentive
is the most flexible kind of lexical item. It can be used in all major lexical
functions: head of a clause, head of an NP, modifier of the head of the
NP and modifier of the head of the clause. Notice, however, that strictly
speaking there is no NP in languages with contentives (Type 1) or non-
verbs (Type 2), as the head of the phrase is not a true noun but a member
of a flexible word class. Members of the flexible class modifier (Type 3)
can be used to modify the head of a clause or the head of an NP, but for
the other lexical functions the language employs members of specialized
or rigid word classes (verbs and nouns).
A flexible word class is not some kind of union of two or more rigid
word classes, but a distinct category in itself, just like a rigid word class.
Fig. 1. Basic classification of parts-of-speech (PoS) systems (based on Hengeveld 1992; adverb,
manner adverb).
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
Word Classes 719
Perhaps the main semantic difference between rigid and flexible word
classes resides in the fact that members of a flexible word class seem to
lack properties that are highly characteristic for members of a rigid word
class. In other words, the fact that a contentive can be used in verbal,
nominal and adjectival function does not imply that it combines the
typical properties of verb, nouns and adjectives. Rather, a contentive is
neither a verb nor a noun or an adjective, precisely because it lacks the
characteristic properties of these rigid word classes (Rijkhoff 2003, forth-
coming). For example, Samoan (Figure 2: Type 1) has no words that are
coded as transitive, a prototypical verbal feature (see also below).
Hengeveld’s classification indicates that rigid word classes adhere to the
following hierarchy (Hengeveld 1992: 68):
(19) Parts-of-speech hierarchy: rigid word classes
Verb > Noun > Adjective > (manner) Adverb
According to this hierarchy, a language that employs members of a
certain rigid word class ‘down’ the hierarchy (e.g. adjective) must also
employ members of the rigid word classes ‘up’ the hierarchy (verb, noun).
How can we explain this hierarchy? Notice, for example, that some lan-
guages have distinct classes of nouns and adjectives (Type 3 in Figure 2),
whereas other languages have nouns but no adjectives (Type 4). Interest-
ingly, it seems that adjectives are only attested in languages whose nouns
denote a property that is specified as having a spatial outline (shape),
which means that these nouns do not require some extra measure (such
as the employment of a numeral classifier) to make them countable.7
Dutch
In contrast, Thai has no adjectives and its transnumeral nouns do not
seem to denote properties that are specified for a spatial contour (shape):
‘[Thai nouns] purely denote concepts and, for this reason, are incompatible
with direct quantification’ (Hundius and Kölver 1983: 166).8 In other
words, the concepts denoted by Thai nouns first need to be ‘individuated’
(by a numeral classifier) before they can be counted.
Thai (Hundius and Kölver 1983: 172; CLF = numeral classifier)
There is also evidence to suggest that a language can only have a true
(rigid) class of nouns if it also has a group of words that are coded as being
transitive (i.e. verbs).9 For example, all languages with nouns have transitive
(20) drieNum oudeAdj paraplu-s
three old umbrella-PL
‘three old umbrellas’
(21) rôm saam khan
umbrella(s) three CLF:long, handled object
‘three umbrellas’
720 Jan Rijkhoff
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
words, but they are claimed to be absent in Samoan, which does not
distinguish between nouns and verbs (Mosel 1991: 188; Mosel and
Hovdhaugen 1992: 724).
Apparently, a language can only have distinct classes of verbs, nouns and
adjectives if the words in that language somehow encode the prototypical
properties of temporal and spatial entities, that is, events and things
(Rijkhoff 2003). The prototypical event is an activity that involves an
agent and a patient; the prototypical thing is a concrete object. In other
words, it seems that a language can only have major, distinct classes of
verbs, nouns and adjectives if the lexicon contains (i) words that designate
a dynamic relationship between an agent and a patient (+transitive), and
(ii) words that designate a property that is specified as having a boundary
in the spatial dimension (+shape).
(22) Necessary conditions in the hierarchy of rigid word classes:
So far no attempts have been made to explain why some languages have
a distinct class of (manner) adverbs, whereas others do not (hence the question
mark in the hierarchy above), but it has been suggested that the adjectival
feature gradable might play a role here (Rijkhoff forthcoming).10
It was already mentioned at the outset that claims about languages with
a single lexical word class, whether flexible (Type 1) or rigid (Type 7), are
controversial (Section 3.1). For example, Mithun (2000: 397) has argued
that Cayuga, Tuscarora and other Iroquioan languages do distinguish
between verbs and nouns. In her view, words that can be identified as
verbs on morphological grounds, function as nouns semantically or syn-
tactically. The situation is complicated by the fact that verbs have been
lexicalized in different degrees:
Some morphological verbs have been so fully lexicalized as nominals that
speakers no longer use them as predicates and may even be unaware of their
literal verbal meanings. Others are never used as nominals. Still others have
two uses, one as a referential nominal, one as a predicate. (Mithun 2000: 419)
If it is true that Cayuga has a major class of verbs and a minor class of
nouns, this would mean its PoS system falls somewhere between Types 6
and 7 at the rigid end of the scale.
As to languages at the flexible end of the scale (Type 1 or 2), the most
detailed discussion of such a language concerns Mundari, an Austroasiatic
language of India (Munda family). Whereas Evans and Osada (2005) have
tried to show that Mundari has distinct classes of verbs and nouns (and a
closed adjective class), Hengeveld and Rijkhoff (2005) have argued that
Mundari is one of the languages with a single class of contentives, at least
VERB →NOUN →ADJECT IVE →MANN ER
ADVER BS
[+TRANSITIVE][+SHAPE][+GRADABLE?]
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
Word Classes 721
as far as its basic, non-derived words are concerned. They add that the
Mundari PoS system would occupy a position between Types 1 and 2 if
one also takes into account derived words, which can be used for all major
lexical functions, except head of the clause (i.e. derived words are non-verbs).
Details such as these illustrate the usefulness of a more refined classifi-
cation of PoS systems, as shown in Figure 2. The classification in Figure 1
has been expanded with two kinds of intermediate systems. A language is
considered to have an intermediate PoS system of the flexible type when
it has a word class that is compatible with two contiguous systems of the
hierarchy. This situation obtains in Mundari (Type 1/2), for example,
that has a class of derived words (non-verbs) that have fewer functional
possibilities than the class of basic words (contentives). A rigid language is
Fig. 2. Parts-of-speech systems, including intermediate types (Smit 2001; Hengeveld and
Rijkhoff 2005).
722 Jan Rijkhoff
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
classified as having an intermediate PoS system when the last word class in
the PoS hierarchy consists of a minor (smallish, closed) class of items. This
is true, for example, for Babungo, Bambara, Gude, Kisi and many other
sub-Saharan African languages with a minor class of adjectives, which all
have PoS systems of Type 5/6 (Rijkhoff 2004: 142).
4. Parts-of-Speech Systems and Grammar
Recent studies indicate that there are certain grammatical features that
correlate with the presence of a flexible word class in the PoS system of
a language (Hengeveld et al. 2004; Rijkhoff 2004; Hengeveld and
Rijkhoff 2005). It was already mentioned that words in a language with
an extremely flexible PoS system are not specified for the typical verbal
feature transitive. One might say that this is only to be expected: if
languages such as Samoan had words that were specified as being transitive
(denoting a dynamic action between an agent and a patient), this would
immediately characterize these words as verbs (i.e. a rigid word class8).
Neither do we expect to find words codified for grammatical number or
gender in languages without a distinct class of nouns, as this would make
the words ‘unflexible’. In other words, one may hypothesize that flexible
words in languages with a PoS system of Type 1 or 2 are always transnu-
meral and genderless. This appears to be the case for all languages with
the relevant PoS systems that we are aware of (Rijkhoff 2004: 42). Other
grammatical properties that are deemed to correlate with the flexibility in
the PoS system of a language are the absence of copulas and suppletion
(Hengeveld and Rijkhoff 2005: 421).
Finally, it has been shown on the basis of data from a representative
sample of the world’s languages that the word order possibilities of a
language are partly determined by the PoS system of that language. As
flexible words are not coded for a particular lexical function (head of
the clause, head of the NP and modifier of a head), it is often not clear
for the hearer how these words should be interpreted. It turns out
that in languages with a flexible word class, word order constraints
are used to signal what kind of function a flexible word has in an
actual utterance. For example, in quite a few languages with a distinct
class of adjectives, some adjectives occur before the noun, whereas others
follow the noun (doubling). This kind of word order variation appears
to be absent in languages with a flexible word class (Types 1–3),
apparently because it would lead to processing difficulties (Hengeveld
et al. 2004: 563–4).
5. Recent Publications
In addition to the more or less recent publications on word classes that
were already mentioned (Croft 1991; Hengeveld 1992; Anward et al.
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
Word Classes 723
1997; Baker 2003), there are various overview articles in anthologies,
handbooks or encyclopedias (see, for example, Schachter 1985; Sasse
1993b; Evans 2000b; Vogel and Comrie 2000, Haspelmath 2001). Indi-
vidual word classes are discussed in, for example, Bybee (2000), Lehmann
and Moravcsik (2000), Bhat and Pustet (2000). Several monographs have
been devoted to adjectives (e.g. Pustet 1989; Bhat 1994; Wetzer 1996;
Beck 2002; Dixon and Aikenvald 2004). Plank (1997) provides a useful
bibliography on word classes in linguistic typology.
Short Biography
Jan Rijkhoff’s main areas of research are linguistic typology, parts-of-
speech, lexical semantics (especially nominal aspect and Seinsart) and
grammatical theory, in particular semantic and morpho–syntactic parallels
between the NP and the sentence within the theoretical framework of
Dik’s Functional Grammar and its successor Functional Discourse Grammar
(Hengeveld and Mackenzie 2005). He has authored or co-authored papers
in these areas for Journal of Linguistics, Journal of Semantics, Linguistics, Stud-
ies in Language, Linguistic Typology, Functions of Language, Acta Linguistica
Hafniensia, Italian Journal of Linguistics (Rivista di Linguistica), Belgian Journal
of Linguistics, as well as various anthologies such as Approaches to the Typol-
ogy of Word Classes (Vogel and Comrie eds. 2000) and International Hand-
book of Typology (Haspelmath et al. 2001). His book The Noun Phrase
(Oxford University Press 2002/2004) investigates NPs in a representative
sample of the world’s languages and proposes a four-layered, semantic
model to describe their underlying structure in any language. It examines
the semantic and morpho-syntactic properties of the constituents of NPs,
and in doing so it shows that the NP word order patterns of any language
can be derived from three universal ordering principles. His current
research is concerned with the parts-of-speech hierarchy, the semantics of
flexible word classes, the relation between form and function of noun
modifiers, and various aspects of NPs in Functional Discourse Grammar. From
1990 to 1994 Rijkhoff was a core member of the EuroTyp project (funded
by the European Science Foundation) and in 1995 he held a fellowship from
the Alexander von Humboldt Stiftung at the University of Konstanz (Ger-
many). Before coming to the University of Aarhus (Denmark), where he
presently teaches, Rijkhoff was a visiting scholar at the University of Texas
at Austin (1997–1999). He holds a BA in Dutch language and literature
from the Free University and an MA and a PhD in Linguistics from the
University of Amsterdam (both in the Netherlands).
Notes
*Correspondence address: Dr. Jan Rijkhoff, Department of Linguistics, University of Aarhus,
Nordre Ringgade 1, Building 1410, DK-8000 Aarhus C, Denmark. Email: linjr@hum.au.dk.
724 Jan Rijkhoff
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
1 Abbreviations: 3, 3rd person; Adj, adjective; Adv, adverb; ART, article; CLF, classifier; Compl,
complementizer; CN, connector; Cop, copula; DEM, demonstrative; N, noun; NP, noun
phrase; Num, numeral; PAST, past tense; PERF, perfective aspect; PL, plural; PoS system, parts-
of-speech system; Prep, preposition; PRES, present tense; PRT, participle; REL, relativizer;
RSM, resumptive marker; SG, singular; SS, same subject; SUBJ, subject; SUPP, support verb;
V, v e r b .
2 The more general term is ‘adposition’, which also includes postpositions (Dutch de tuin IN
‘into the garden’) and circumpositions (Dutch OM de tuin HEEN ‘around the garden’).
3 As adverbs constitute a rather mixed bag of different subcategories, they are often ignored in
discussions of lexical word classes (see also below).
4 For a recent application of the multi-dimensional approach, see Francis and Matthews (2005).
5 The original definitions have ‘lexeme’ instead of ‘word’.
6 Cf. Evans (2000b) or Mithun (2000); see also the collection of articles a special issue on word
classes in the journal Linguistic Typology (volume 9, number 3; 2005).
7 Only entities with a definite outline can be counted. Notice that this does not mean that
adjectives cannot occur in a classifier language. In many languages numeral classifiers have
developed into markers of other grammatical categories such as definiteness, specificity or
topicality (Rijkhoff 2000, 2004: 51). In such cases the erstwhile classifiers no longer serve as
‘individualizers’ in the sense of Lyons (1977: 462).
8 A transnumeral noun is neutral with respect to number, hence the same form can be used to
talk about one or more entities.
9 Note that the presence of a set of transitive words in the basic lexicon is a necessary and
sufficient condition for a language to have a major, distinct class of verbs, but only a necessary
condition for a language before it can have a major, distinct class of nouns.
10 When we go from left to right in the hierarchy of rigid word classes, we see increased
specialization. In contrast, the hierarchy of flexible word classes shows a decrease in specializa-
tion (but an increase in flexibility): (i) Parts-of-speech hierarchy (flexible word classes): modifier
> non-verb > contentive.
Works C i t e d
Aarts, Bas. 2007. Syntactic gradience: the nature of grammatical indeterminacy. Oxford, UK:
Oxford University Press.
Anward, Jan, Edith Moravcsik, and Leon Stassen. 1997. Parts of speech: a challenge for typology.
Linguistic Typology 1–2.167–83.
Baker, Mark C. 2003. Lexical categories: verbs, nouns, and adjectives. Cambridge, UK: Cam-
bridge University Press.
van Baarda, M. J. 1908. Leiddraad bij het bestuderen van’t Galela’sch dialekt, op het eiland
Halmaheira [Manual for the study of the Galela dialect, on the island of Halmahera]. The
Hague, The Netherlands: Nijhoff.
Beck, David. 2002. The typology of parts of speech systems: the markedness of adjectives. New
York, NY: Routledge.
Bhat, Darbhe N. S. 1994. The adjectival category: criteria for differentiation and identification
(Studies in Language Companion Series 24). Amsterdam, The Netherlands: Benjamins.
Bhat, Darbhe N. S., and Regina Pustet. 2000. Adjective. Morphology: a handbook on inflec-
tion and word formation (Part 1), ed. by Geert Booij, Christian Lehmann and Joachim
Mugdan, 757– 69. Berlin, Germany: Walter de Gruyter.
Broschart, Jürgen. 1997. Why Tongan does it differently: categorial distinctions in a language
without nouns and verbs. Linguistic Typology 1–2.123–65.
Bybee, Joan. 2000. Verb. Morphology: a handbook on inflection and word formation (Part 1),
ed. by Geert Booij, Christian Lehmann and Joachim Mugdan, 794–808. Berlin, Germany:
Wa l t e r de Gr u y t e r .
Croft, William. 1991. Syntactic categories and grammatical relations: the cognitive organization
of information. Chicago, IL: University of Chicago Press.
Croft, William. 2000. Parts of speech as language universals and as language-particular categories.
Approaches to the typology of word classes (Empirical Approaches to Language Typology 23),
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
Word Classes 725
ed. by Petra M. Vogel and Bernard Comrie, 65–102. Berlin, Germany/New York, NY:
Mouton de Gruyter.
Dixon, Robert M. W., and Alexandra Y. Aikhenvald (eds.). 2004. Adjective classes: a cross-
linguistic typology. Oxford, UK: Oxford University Press.
Donaldson, Tamsin. 1980. Ngiyambaa: the language of the Wangaaybuwan of New South
Wales. Cambr idge, UK: Cambridge University Press.
Evans, Nicholas. 2000a. Kinship verbs. Approaches to the typology of word classes (Empirical
Approaches to Language Typology 23), ed. by Petra M. Vogel and Bernard Comrie, 103–
72. Berlin, Germany/New York, NY: Mouton de Gruyter.
Evans, Nicholas. 2000b. Word classes in the world’s languages. Morphology: a handbook on
inflection and word formation (Part 1), ed. by Geert Booij, Christian Lehmann and Joachim
Mugdan, 708–32. Berlin, Germany: Walter de Gruyter.
Evans, Nicholas, and Toshiki Osada. 2005. Mundari: the myth of a language without word
classes. Linguistic Typology 9–3.352–90.
Francis, Elaine J., and Stephen Matthews. 2005. A multi-dimensional approach to the category
‘verb’ in Cantonese. Journal of Linguistics 41.269–305.
Haspelmath, Martin. 2001. Word classes and parts of speech. International Encyclopedia of the
Social and Behavioral Sciences, ed. by Paul B. Baltes and Neil J. Smelser, 16538–45. Amster-
dam, The Netherlands: Pergamon.
Haspelmath, Martin, Ekkehard König, Wulf Oesterreicher, and Wolfgang Raible (eds.). 2001.
Language typology and linguistic universals: an international handbook, vol. 1. Berlin,
Germany/New York, NY: Walter de Gruyter.
Heine, Bernd, and Tania Kuteva. 2002. World lexicon of grammaticalization. Cambridge, UK:
Cambridge University Press.
Heine, Bernd, Ulrike Claudi, and Friederike Hünnemeyer. 1991. Grammaticalization: a con-
ceptual framework. Chicago, IL: The University of Chicago Press.
Hengeveld, Kees, 1992. Non-verbal predication: theory, typology, diachrony. Berlin, Germany/
New York, NY: Mouton de Gruyter.
Hengeveld, Kees, and J. Lachlan Mackenzie. 2006. Functional discourse grammar. Encyclopedia
of language and linguistics, 2nd edn, vol. 4, ed. by Keith Brown, 668–76. Oxford, UK:
Elsevier.
Hengeveld, Kees, and Jan Rijkhoff. 2005. Mundari as a flexible language. Linguistic Typology
9–3.406–31.
Hengeveld, Kees, Jan Rijkhoff, and Anna Siewierska. 2004. Parts-of-speech systems and word
order. Journal of Linguistics 40–3.527–70.
Hopper, Paul J., and Elizabeth Closs Traugott. 2003. Grammaticalization, 2nd edn. Cambridge,
UK: Cambridge University Press.
Hopper, Paul J., and Sandra A. Thompson. 1984. The discourse basis for lexical categories in
universal grammar. Language 60–3.703–52.
Hundius, Harald, and Ulrike Kölver. 1983. Syntax and semantics of numeral classifiers in Thai.
Studies in Language 7–2.164–214.
Kuipers, Aert. 1968. The categories verb-noun and transitive-intransitive in English and Squa-
mish. Lingua 21.610–26.
Kutsch Lojenga, Constance. 1994. Ngiti: a Central-Sudanic language of Zaire. Köln, Germany:
Rüdiger Köppe.
Lehmann, Christian, and Edith A. Moravcsik. 2000. Noun. Morphology: a handbook on
inflection and word formation (Part 1), ed. by Geert Booij, Christian Lehmann and Joachim
Mugdan, 732–57. Berlin, Germany: Walter de Gruyter.
Lord, Carol Diane. 1989. Syntactic reanalysis in the historical development of serial verb
constructions in languages of West Africa. PhD dissertation, University of California, Los
Angeles. Ann Arbor, MI: University Microfilms International.
Lyons, John. 1977. Semantics (2 volumes). Cambridge, UK: Cambridge University Press.
Mithun, Marianne. 2000. Noun and verb in Iroquioan languages: multicategorisation from
multiple criteria. Approaches to the typology of word classes (Empirical Approaches to
Language Typology 23), ed. by Petra M. Vogel and Bernard Comrie, 397–420. Berlin,
Germany/New York, NY: Mouton de Gruyter.
726 Jan Rijkhoff
© 2007 The Author Language and Linguistics Compass 1/6 (2007): 709–726, 10.1111/j.1749-818x.2007.00030.x
Journal Compilation © 2007 Blackwell Publishing Ltd
Mosel, Ulrike, 1991. Transitivity and reflexivity in Samoan. Australian Journal of Linguistics
11.175–94.
Mosel, Ulrike, and Even Hovdhaugen. 1992. Samoan reference grammar. Oslo, Norway:
Scandinavian University Press.
Plank, Frans. 1997. Word classes in typology: recommended readings (a bibliography). Linguistic
Typology 1–2.185–92.
Pustet, Regina. 1989. Die Morphosyntax des ‘Adjektivs’ im Sprachvergleich. Frankfurt am
Main, Germany: Lang.
Rijkhoff, Jan. 2000. When can a language have adjectives? Approaches to the typology of word
classes (Empirical Approaches to Language Typology 23), ed. by Petra M. Vogel and Bernard
Comrie, 217–57. Berlin, Germany/New York, NY: Mouton de Gruyter.
Rijkhoff, Jan. 2003. When can a language have nouns and verbs? Acta Linguistica Hafniensia
35.7–38.
Rijkhoff, Jan. 2004. The noun phrase (expanded paperback edition of 2002 hardback publica-
tion). Oxford, UK: Oxford University Press.
Rijkhoff, Jan. Forthcoming. On flexible and rigid nouns. Studies in Language.
Sasse, Hans-Jürgen. 1993a. Syntactic categories and subcategories. Syntax: an international
handbook of contemporary research (2 vols.), ed. by Joachim Jacobs, Arnim von Stechow,
Wolfgang Sternefeld and Theo Vennemann, 646–86. Berlin, Ger many: Walter de Gruyter.
Sasse, Hans-Jürgen. 1993b. Das Nomen – eine universale Kategorie? Sprachtypologie und
Universalienforschung (STUF) 46–3.187–221.
Sasse, Hans-Jürgen. 2001. Scales of nouniness and verbiness. Language typology and linguistic
universals: an international handbook, vol. ii, ed. by Martin Haspelmath, Ekkehard König,
Wulf Oesterreicher and Wolfgang Raible, 495–509. Berlin, Germany: Walter de Gruyter.
Schachter, Paul. 1985. Parts-of-speech systems. Language typology and syntactic description (3 vols.).
Volume I: Clause structure, ed. by Timothy Shopen, 3–61. Cambridge, UK: Cambridge
University Press.
Smit, Niels. 2001. De rol van derivatie bij lexicale specialisatie. MA thesis, Department of
Linguistics, University of Amsterdam.
Vogel, Petra M., and Bernard Comrie (eds.). 2000. Approaches to the typology of word classes
(Empirical Approaches to Language Typology 23). Berlin, Germany/New York, NY: Mouton
de Gruyter.
de Vries, Lourens. 1989. Studies in Wambon and Kombai: aspects of two Papuan languages of
Irian Jaya. PhD dissertation, University of Amsterdam.
Wetzer, Har rie. 1996. The typology of adjectival predication (Empirical Approaches to Language
Typology 17). Berlin, Germany: Mouton de Gruyter.