See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/290830649

A discourse approach to the cross-linguistic category ‘Adjective’

Chapter · January 1989

DOI: 10.1075/cilt.61.16tho

CITATIONS READS

128 1,489

1 author:

Sandra A. Thompson
University of California, Santa Barbara
143 PUBLICATIONS   20,122 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Multimodal analysis of Mandarin conversational interaction View project

All content following this page was uploaded by Sandra A. Thompson on 28 July 2017.

The user has requested enhancement of the downloaded file.

 ~l ~et..W~ t ~ 1 ~ 4,,#aN~'OIlS

~ ~~.' ~~>l ~~88r

:ahler, H. (1965) Gmmmntik der l1alra.m ",dolle.~ia. Wieshnden: Otto Ilnrrowitz.
:arttunen, L. (11)74) ·Until". I'I"(/ceedillgs oj tl'e Clriwgo Lill8l/i.Hir Society 10,
UniversilY of Chicago, In, 284-97.
and S. Peters (1979) 'Convcntional implic;llure'. In ('. Oh. and D. 1\. Dinncen
(eds), !»'IIta.\ a/l([ Semalllies.vol.ll.I
Press.
••.esll ••••osiliml. New York: Acadcmic
:Iare, J. (1958) EIIIJ/elllmg a/l([ f:'lIIl1'ieklrmg da Km'U'.ui,·ell f(olljlmk/icm im
Frrlllzosisclrell. Berlin: I\kadclllic Vcrlag.
:Iein, J. (1980) 'Die Kom:essivrclation als argulllelllations·thcorctischcs
Zeitllscllri[t /iir Gemrallistiselle Lillgl/is/ik, 8, 154-69.
:iinig, E. (1985a) 'On the history of concessive conncctivcs in English. Dil.chronic
and synchronic evidencc'. Ullnl/a, 66, 1-19.
(1985b) 'Konzessive Konjunktionen'. To nppear in A. v. Stcchow and D.
Wunderlich (eds), HI/I,dhl/dl der Semnll/ik. Kiinigstein: Atheniium.
(1985c) 'Where do concessives come fromT In J. Fisiak (ed.), /{isto,.ical
Semalltics, Historical Word Fomratioll. Berlin: Mouton.
(1986) 'Conditionals, concessive conditionals and coneessives: areas of conlrast,
overlap and neutralization'. In Traugoll et al. (1986).
and P. Eisenbcrg (1984) 'Zur Plagmalik von Konzessivsiitzcn'. In G. Stickel
(ed.), Pragmatik ill der (jmmma/ik. Dusseldorf: Schwann.
and 1. Van der Auwera (forlhcoming) 'C1;luse inlegllliion in Gcrnwn nnt! Dutch
condilionals, concessivc condilionals, and conccssivcs'. To appcar in Ilaiman
and Thompson.
.lImasser, II. (1952) /{llIIrllmd, da Selllasiologie. lleidclhcrg: Carl Winter.
ehmann, C. (1982) 71,ol/gllls Oil (immmaticalizatioll. Cologne: Arbeilen des
Kiilner Universalicn Projckls, 48.
erch, E. (1929) l1istorisc!1C' hWlzii.\;selre ,\)."I//X II. Leipzig: Reisland.
. evinson, S. (1983) I'ragmatic.~. Cambridge: Cambridge University Press.
yons, 1. (1967) '/\ note on possessive, exislential and localive senlences'.
FOlllldatiollS oj Lallgl/age, 3, 390-6.
tensing, O. (1981) 'Untersuchungen uber die Synlax der Concessivsiilze im AIt-
und Millelhochdeutsehen'. Doctoral diss .. Kicl.
tittwoch, A. (1977) 'Negalive senlences with /llIIi/'. CL.\', 13,41 (j-17.
lank, F. (1979) 'Exklusivierung, Idenlifizierung, rel<ltionale Auszeichnung.
Variationen zu einem semantiseh-pragmatischen Thema'. In J. Rosengren
(cd.), S"melre /I//{I I'mgma/ik. I.undcr gennanische Forschungcn 48. I.und:
( i1certlp.
'uirk, It (11)54) 'II,/, (·Oll,./,.I.li,,/, Ucla/im, ill Of:' I'oct/)'. NclV I laven, (·onn.: Yale
University Press.
eichenbach, H. (1947) UemelltI oJ.\).",holic Logic. New York: The Free Press.
aebe K. (1986) 'Causal nnd purposive clauses'. In A. v. Stechow and D.
Wunderlich (cds). Halldlwefl der Selllalllik. Kiinigslcin: Athcniiull1.
lump. G. T. (1985) 77,1' Semalllic Variability oj Absolute COlls/metiollS.
Dordrecht: Reidel.
oulmin, S. (1958) 77,1' Uses oj A,.g"melll. Cambridge: Cambridge University
Press.
raugulI, E. C. ter Meulen. A., Reilly, J. S. and Ferguson C. A. (cds) (1986) 011
COllditiOllals Camhridge: Camhridge University Press.
ipr, G. K. (1949) H""'/l1I Heht/I'io" /l1I/1/he !',.illl'iple of /'I'/IS/ I:JJiJrl. Cambridgc.
Mass.: Aduison-Weslcy.
B~ ~t.!).Ll~~,
CHAPTER 7
A Discourse Approach to the Cross-
Linguistic Category 'Adjective'
Problcm.
Sandra A. Thompson
! Sandra A. Thompson
~_ tigGisticS Dept.
Santa Barbara. CA 93106
i§
Hopper and Thompson (1984) addressed the problem of a possible
discourse explanation for the way in which the linguistic categories of
Noun and Verb manifest themselves in actual language use. In this paper I
will be concerned with the question of categorization of 'property
concepts', that is, those concepts referring to properties, qualities or
characteristics of referents .
In English, there is a fairly close correspondence between this set of
concepts and the linguistic category of 'adjective', but it is wel~k~lO~t~that
this is by no means the case for many other languages. In Ius JustIfiably
celebrated 1970 paper, published in 1977, Dixon explores the question
'how does it Ithat is, a language with either no Adjective class at all or only
a small non-productive minor class of Adjectivel express concepts that are
expressed through adjectives in languages, like English, which do have this
major class' (p. 20)?
What I would like to do in this paper is to take a close look at Dixon's
answer to this question, and supplement it with some findings of my own. I
will then propose a discourse explanation for these findings. This research
can he thus he seen as a contribution to the study of 'natural grammar', the
study of what communicative factors arc responsible for the design of
grammars.
Dixon introduces the notion of 'semantic type', suggesting that 'the lexical
Hems of a language fall into a number of "semantic types" (each item
belonging to just one type) ... Each semantic type has, in a I?articular
language, certain "norm" syntactic and mOrJ~hological.~ropertles. ~{/clr
member of the type exhibits the norm propertIes. In addltton, there Willhe
a number of "extensional" properties, each applying only to certai"
~.'.',
.:l
 _______
m~mbers ~f the typ~' (p. 25, emphasis original). (Today such a description
nllg~t be discussed III terms of the notion of 'prototype'.)
J)lx~m relates the notion of 'semantic type' to linguistic categories in the
followlll.g wa~: he a~sumes that each 'semantic type' has a basic or 'norm'
connection with a smgle part of speech (p. 27), and that since
alliangu<~ges.appear to have NoUl~and Verb but some lack a major
class Adjective, ... some semantic types must be associated with
diffe.rent parts of speech in different languages ... M<)'J'ION,
AffECT, GIVING, CORPOREAL le.g. lallgh, sneezel, and other
types seem almost always to be classed together -this is the class that
is in all languages called Verh. OIl.JE<TS, KIN, and olher types arc
almosl always classed together - this is the class Ihat is in all
languages called Noun. (p. 2H).'
Later he suggests 'what it means to say that a language has "a class
of ?djectives". Thi~ is a set of lexical ite~ls, distinguished on morpho-
logical ami syntactic grounds from the uJ1lversal classes Noun and Verb'
(pp.62-3).
Becaus~ Dixon assumes that each semantic type has, in a given
langua¥e, lis own "no~mal' grammatical properties, he proposes to isolate
the ulllversal Adjective types by looking at the 'basic members' of the
Adjective c1.a.ss)ust ~n En¥lisl~, and .t1~ento investigate their linguistic
category affiliations III Adjective-deficient languages. The types which
emerge, based on semanlic, syntactic and morphological criteria, are
these:
I DIMENSION - big,lillle, long, wide, ...
2 PHYSICAL PROPERTY - hard, heavy, smooth, ...
3 COLOUR
4 HUMAN PROPENSITY - jealolls, happy, clever, generolls, prolld,
5 AGE -I/ew, yo III/g, old, ...
6 VALUE - good, bud, pllre, delidolls, ...
7 SPEED - jil.\l, slow, quick, ...
Fo~ the remainder of this paper, I will take these seven types of concepts,
which express ~rorerties of entities, as definitional for the term 'Property
Concept. That IS, If a word expresses one of these types of properties in a
language, I will call it a Property Concept Word.
Dixon's findings, based on a non-random sample of seventeen lan-
guages,2 are summarized in (I ):
(I) Languages may have a category of Adjectives which can be identified
on language-internal morpho-syntactic grounds. No matter how small
or restricted this category is, it is likely to include at least these four
s _" --.b-H
types of Property Concepts in Dixon's list above: DIMENSION,
COLOR, AGE, and VALUE.
n
(2) Whether or not there is a category of Adjectives, the words expressing
Property Concepts tend to fall into categories which either share
many properties with the class of Nouns, or many properties with the
class of Verbs.
My own research, based on a further sampling (again not random) of some
forty languages, has confirmed these findings.
The problem, then, is this: It seems to be a fact that the distinction
belween Noul/.\' and V('fb,~ ilself is a language universal, where largely the
sallie sel of concepls are found in each of Ihese categories from language
to language (see Givlln 1979, 19.H4), Hopper and Thompson 19H4,
Jacobsen 1979, Schachler 1985 and references cited in these works for
discussion). Independently of whether or not there is an identifiable class
of Adjective, Ihe lJuestion which then arises is:
Why should a given set of concepts, namely Property Concepts, he
distributed across these two quite distinct lexical categories, namely
Noun and Verb, in the world's languages (as opposed, say, to being
exclusively treated by grammars of languages as a subclass of either
Noun or Verb, or as a separate class of Adjective)?
Before considering Ihe discollrsc proposal, let us hrieny consider some of
the sorts of evidence lhallead 10 these conclusions.
First, Ihere arc languages in whit:h Properly Concepls form a semantic
subclass of the category Verh. For example, in Acehnese (northern
Sumatra) (Durie 1985), there are a number of pieces of evidence showing
that Property Concepts do not form a separate lexical category from
Verbs. First, they may take optional Undergoer cross-referencing just as
other non-controlled Verbs do:
(2) gopnyan saket =geuh
3 sick =UNDGOEH
's/he's sick'
(3) gopnyan rhet =geuh
3 fall=UNDGOER
's/he is falling'
(4) gopnyan ka saket =geuh
3 INCHO sick =UNDGOER
's/he's gollen sick'
 (5) gopnyan ka geu~woe they take the same possessive suffixes and enclitic pragmatic particles
3 INCIIO AG-return (Karlsson 1983: 22). :rhus, in the sentence:
's/he has already returned'
~ (10) Auto on sininen
Third, any Verb can be used attributively with no morpho-syntactic ado: car:NOMCOP blue:NOM
'the car is blue'
(6) aneuk muda nyan
child young that
'that young child' the form si"i"ell 'blue' inflects for case and takes the copula,just as a Noun
would:
(7) ureueng pula pade nyan
person planl rice that
(11) iso-ssa auto-ssa
'that person planting rice' big-INESS car -INESS
'in the big car'
Finally, many non-controlled Verbs may take that 'very':
(12) Pekka on mies
(8) ureueng nyan carl;ng that P.:NOM COP man:NOM
person that clever very 'Pekka is a man'
'that person is very clever'
(9) Ibn galak that kcu jih Similarly, in a language with gender or noun class, such as Diyari
I like vcry DAT 3 (Austin 1981), spoken in South Australia, as well as a number of Indo-
'I like him/her very much' European and African languages, there is a class 'nominal', which can be
said to subsume 'Nouns' and 'Adjectives' as subcategories, with common
Other languages in my sample for which similar arguments can be made morpho-syntactic categories, such as case, number and gender or noun
include: class, though of course Noun and Adjective will always be kept distinct by
the fact that only Nouns, but never Adjectives, are subcategorized for
Aghem (Bantoid) (Hyman 1979) inherent gender.
Chinese (Chao 1968, Li and Thompson 1981) Other languages in my sample for which analogous arguments can be
Lakhota (Buechel 1939) made include:
Noni (Cameroon, Beboid Bantu) (Hyman 1981)
Thai (Noss 1964) Arabic
Turkana (E. Nilotic) (Dimmendaal 1983) Asmat (Papuan) (V oorhoeve 1965)
W. Makian (Papuan) (Voorhoeve 1982) Bantu languages
Wappo (California) the Chadic languages Hausa (Cowlln and Schuh 1976) and Margi
the Austronesian languages: Samoan (Marsack 19(2), Indonesian (Mac- (Hoffman 19(3)
Donald 1976), MokiJese (Harrison 1976), Palauan (Josephs 1975), Dyirbal (Dixon 1972)
UJithian (Sohn and Bender 1973), Woleaian (Sohn 1975), Kusaiean the Dravidian languages Kannada (M. Nadkarni, personal communica-
(Lee 1975) and Wolio (Indonesia) (Anceaux 1952).' tion) ami Malayalam (George 1971)
Fore (E. New Guinea) (Scott 1978)
That is, in each of these languages, whether or not a class of Adjectives can Israeli Hebrew (Rosen 1977)
be identified on language-internal distributional grounds, Property Con- the Indo-European languages Persian (Elwell-Sutton 1979), Polish
cept Words share many features with verbs. (Laskowski 1979), Icelandic (Glendening 1961), Latin, DUlch
Second, it is also well-known that thcre are a number of languages in (Donaldson 1981) and German (Schulz and Griesbach 1960)
which Property Concept Words pattcrn very similarly to Nouns. For Kalkatungu (Pama-Nyungan, Australia) (Blake 1979)
example, in Finnish, Nouns and Property Concept Words have much in Quechua (Adelaar 1977)
common: they are inflccled in the same way, for number and case, and Tagalog (Schachter and Otanes 1972).
 However, while both of these pallerns exist, that is, Property Concepts
sharing many features with Verbs and with Nouns, they do not necessarily
characterize languages, since it also happens, as in some West African
languages (for examplc, Kusaal (Gur) (Ladusaw 1985) and Yoruba (Kwa)
(f. Serzisco, personal communication» or Japanese (Y. Koide, personal
communication), for example, that within a given language some Property
Concepts will be categorized with features shared with Nouns, while
others will havc features shared with Verbs.
Finally, of course, there are languages, such as English, Fula (Arnott
1970), Jacaltec (Craig 1977), Cherokee (Lindsey and Scancarelli 1985)
and Manam (Austronesian) (Lichtenberk 1983), in which a class of
Adjective can bc identified which shares fcw or no obvious propcrties
wilh eilher Nouns or Verbs.
The fact remains, howevcr, that cross-linguistically, Property Concepts
may show up as either Nouns or Verbs (see Dixon 1977 and Schachter
1985 for further discussion).
BUI why should this bc so'! This is the question which I will try to answer
in the rest of this paper.
Giv6n (1979, 1984) has suggested a semantic factor underlying the
categorization of cognitive 'percepts', namely, time stability. 'Experiences
... which stay relatively stable over time ... tend to be lexicalized in
human language as nouns ... At the other extreme of the lexical-
phenomenological scale, one finds experiential clusters denoting rapid
changes in the state of the universe ... languages tend to lexicalize them
as verbs' 11984: 51-2, emphasis original). According to Givon, the class
Adjective, for languages which have it, occupies 'the middle of the
time-stability scalc' (1984: 52). But there are two problems with this
explanation.
First, while time slability may be involved in distinguishing prototypical
cxcmplars of the eategorics Noun and Verb in the world's languagcs, I
question its relevancc for the calegory Adjectivc, since, as Givlln admits,
the 'prototypical adjectival qualilies' are 'Ihose of stclble physical qualilies
such as size, shape, texture, color, taste, or smell' (1984: 53, emphasis
mine). In fact, all of Dixon's seven 'semantic types' which characterize the
class of Adjectives in English, except possibly SPEED, denote relative
stable characteristics. Thus, it does not appear to be true that 'adjectives
occupy the middle of the time-stability scale'.
The second problem with the time-stability explanation for the
categorization of Property Concepts is that it makes the wrong prediction
about a particular set of morphosyntactic facts. That is, if 'Adjectives'
indeed occupied the 'middle of the time-stability scale', then we would
predict that in languages where 'Adjectives' are a subclass of Verbs, they
should exhibit features related to the fact that they are more time-stable
than Verbs; similarly, we would also predict that in languages where
'Adjectives' are a subclass of Nouns, they should exhibit features relaled to
the fact that they are less time-slable than Nouns. Now, while support for
the former prediction is abundanl, support for Ihe laller prediction is nol.
Thus, as a subclass of Verbs, Property Concepts onen exhibit morpho-
syntactic evidence of their stat;!';t)', their greater lime-stability when
compared to prototypical Verbs; they are often constrained not to occur
with certain tense-aspect morphemes, for example.
But the primary way in which Property Concepts can be identified as a
subclass of Nouns different from prototypical Nouns is, as mentioned
ahove, Ihal Iheir gender calegories arc derived, while those of Nouns arc
Iypically inhercnl. I do not see a way to relale Ihis fact 10 the lower degre.c
of time-stability of Property Concepts as compared to Nouns. Rather, 1\
has to do with the use to which Property Concepts are put by speakers.
Givon's semantic explanalion in terms of time-stabilily, then, does not
help me to explain the facts given in ( I). In fact, while I believe that such
semantic features as the stativity of verb-like Property Concepts have
significant morphosyntactic consequences, I Ihink that the explanation for
the specific cross-linguistic facts in (I) may not be found in semantics at
all. I suggest instead that a satisfying account of the lexical categorization
of Property Concepts in the world's languages can only be stated in terms
of the use of Property Concepts in actual d;scourse.4
To support this claim, I have taken a first step by investigating the use of
Property Concept Words in two languages: English and Mandarin
Chinese. In English, of course, it is a straightforward maller to argue for a
class of Adjectives to which Property Concepts belong. In Mandarin, on
Ihe other hand, there is an extremely restricted category of Adjective, but
most Property Concept WOlds are indislinguishablc from Vcrbs (for
discussion, see Li and Thompson 191{I and Schachler 1(85).
So far, these are the only two languages whose discourse use of Property
Concept Words I have investigated in any depth. However, the results are
similar enough in these two genetically and areally unrelated languages to
suggest thai investigations of similar data in other languages will yield
similar results.5
For English, since Property Concepts correlate largely with the
category of Adjectives, I will describe my findings for this language in
terms of the shorter term 'adjective' rather than 'Property Concept'.
I have looked at more Ihan 100 pages of transcribed natural spontane-
ous conversational discourse with 301{ Property Concept words, or
adjectives. I did not counl Nouns used as modifiers, such as steak place or
 broccoli pie ,nor did 1count adjective-Noun combinations which 1took to
be single lexical items, such as blind llate.
~ 1found that English adjectives are used in essentially Iwo ways: see tahle
7.1. Let us look at some examples." Examples of adjectives playing a
predicale role fell into two synlactic categories, having to do with
peculiarities of English grammar. First, there were those in which the
adjective appeared as a predicale adjective with a copular verb, as in:
(13) and her parents apprently weren't even that wealthy (C I)
(14) I was getting kind of good at playing in Ihe rain, really (Frie 19)
(15) that golme so mad (11035)
(16) aren't some chapters of Understanding Media appropriate?
(Frie35)
Most of my examples of adjectives functioning as predicates (209, or 86
percent of the 242 instances of predicate function) were of this predicate
adjective type.
Predicating a property or an e~lahli~hed discourse rderenl: 79%(N - 242)
Inlroducing a new discourse rderent 21%(N=66)
Total: 100% (N - 308)
The other, minority (14 percent), syntactic context in which I found
adjectives functioning predicatively was that in which the adjective was an
attribute to a predicate nominal head noun which is relatively 'non-new-
information-bearing', Ihat is, a relatively empty or an anaphoric head noun
in predicale nominal posilion. Let us look carefully at one such example:
(17) (talking ahout a potenlial dale)
The last time she saw him which was three years ago he was pretty
good looking, and um, you know she says he's a very nice guy, he's a
real good persoll (HOZI)
In (17), the predicale nominals guy and pel:wn are anaphoric (that is, they
are interpreted as co-referential with a nominal earlier in the discourse)
and, hence, non-new-information-bearing, since their referent has already
been introduced much earlier in the conversation. Thus, while it is clearly
the case thaillice and WJOdare attrilJllti"e ad.ieclives grammatically, they
are counted here as predicates fill/ctiollally. Here are two further examples
of this situation:
(I R) But 1did have lots of fun up at Lehigh. That was a good school (C 14)
(19) (talking about apartments)
1/: Well, theirs is a lIice apartmellt really. (rrie25)
(17-19) show instances of an attributive adjective functioning as it
predicate because its predicate nominal head is anaphoric. But there are
also cases where an attributive adjective functions predicatively because
its predicate nominal head is (relatively) empty. Here is an example of
such a case:
(20) She and 1 are gonna go out and get drunk at four o'clock in the
afternoon. It's a religiolls thillg we're gonna have (Fra9)
In (ZO), a grammatically atlributive adjective (religious) is functioning as
the predicate of its clause, because its predicate nominal head is non-new-
information-bearing. The difference between (20) and (17)-( 19), how-
ever, is that this time the non-new-information-bearing head is an empty
noun (thillg).
Here is another example of this type of situation:
(21) (talking about wedding girts)
I said why don't you pack it away, and save it to give away to other
people for wedding gifts, and she said, 'Hey, that's a great idca' (C I9)
In each of these examples, the atlributive adjective is functioning
predicatively, since it modifies a non-informative predicate nominal head
noun. Paraphrases in which the adjective figures as the sole element after
the copula would often be appropriate ('their apartment is lIice really'), for
example.'
If English weren't so inclined to express predicative notions in the form
of noun phrases with empty or anaphoric heads, these clauses could
appear with predicate adjectives, and translations inlo other languages
could easily find them coded as predicates rather than as nOlln phrases.
What I have shown, then, is that the primary function of adjectives in
English, accounting for 79 percent of the adjective usage in my data, is as
the predicate of their clause, whether they are the sole predicate word or
an attribute to a non-informative predicate noun.
The second discourse function of adjectives, according 10 table 7.1, is to
introduce new participants into the discourse. As expected, the adjectives
serving this function are all attributive.
Let us consider an example.
(22) (Ialking about how to find something in the bookstore)
S: Hey, you got a fill/llY baggie.
If: (It all wraps up.)
 , ••
s: Your haggie is hetter than mine
II: Sure is (Frie2)
Here S brings up the participant fllllllY baggie, which has not been
mentioned before, and which figures in the successive discourse, at least
for one turn.
In (22), it is clear that both the adjective and the noun in the phrase
[I/IIIIYbaggie are functioning to introduce the new referent. That is, both
the item baggie and its property of being funny are new. Here is another
example of the same type:
(23) (lOll king about apartments)
Arcn't those fabulous; Tom Smith used to live there. And he had
black'(//ul-white-striped sheets in his bedroom (Frie21)
However, the majority of the instances in which the adjective is
functioning to introduce a new discourse referent in my data are cases
where the head noun is, again, lIoll-information-hearing, that is, either
empty or anaphoric. For example, consider (24):
(24) (talking ahout a movic)
and there's something really sad that happens (HG 13)
In (24), the new referent is something unspecified except for its property
of being sad. In the successive discourse, the speaker goes on to describe
the sad thing, so it is clear that it is being introduced as a new referent in the
discourse - the point is that it is the adjective which is carrying the burden
of tagging this new referent for subsequent use in the later discourse,
namely by identifying it as sad. This, then, is an example of the new
referent being introduced by an adjective plus an empty head noun.
Here is a similar example:
fl.'.
In (26) you can see an example of the other situation in which an
'j'
~~
adjective plays the major role in the introducing of the new referent: that in
which the new referent is introduced by an adjective plus an allapllOric
head noun:
(26) H: We've got to get umbrellas.
S: Why don't we get two, since there are two of us, good, that means
I can get a pretty fallcy felllillille olle (FrieI8-19)
In (26), we see that the new referent, a pretty fancy feminine umbrella, is
distinguished from other referents in this discourse only by the properties
••••••••••'••••••• •• n1lllllllr••••• _ ••• __ •••• ••••• 1~:---
conveyed by the adjectives. This new referent, then, is contrasted with
other similar referents. Once again, the new referent is being introduced by
the adjective(s), not by the head noun, which is anaphoric (or what Halliday
and Hasan 1976 call 'replacive'), and hence, non-new-information-
bearing.
Here is another examplc 10 illustrate the case of an adjectivc introducing
a new referent with an anaphoric head noun:
(27) (talking ahout apartments)
I don't know, I think we're made more for ethllic apartlllellts than
{)allish modemallartlllt.lI(S (Frie25)
Example (2M) is a particularly interesting illustration; the head noun
food is not, strictly speaking, an alH~phoric noun, since there had been no
earlier mention of food in the conversation. But it is non-new-information-
bearing in juslthe same way, given lhe verh eat just a few words earlier.
(28) (talking about acne)
Yeah, but what you eal ... if you eat grea,lYfood ... (HG5)X
In all of these examples, thc adjectives are playing lhe role of introducing a
new participant into the discourse.
Now, if adjectives are heavily involved in the introduction of new
participants into the discourse, then we would expect lwo corollaries: ( I)
we would expect the noun l'lhrases in which these refcrcnt-introducing
adjectives arc found to bc grammalically indefinite or non-referential; (2)
we would expect that the nominal positions in which these referent-
introducing adjectives occur will he just those recorded independently for
new participants.
Both of these predictions are borne out by the data. First, these referent-
introducing adjective phrases are all grammatically indefinite or non-
referential, with one exception. What I found hardly any examples of,
interestingly enough, were t!t1illite noun phrases with adjectives in thcm.
In fact, only Ollce among allthc 30Madjectives in my corpus did I find an
"
instance of an adjective re-idelltijyillg or distingllishing an already
introduced referent. This one example is shown in (29):
(29) (descrihing a movie)
and this boy is Jewish
(several turns later)
There's this one part between lhis lewish glly and the girl (HG 13)
Clearly the adjective in the phrase this lewish glly in (29) is functioning to
re-identifY, and not to introdllce, a discourse referent. The fact that these
cases, while perfectly easy to produce as examples, are so rare in these
 dala, though, serves to underscore my claim that the two major functions
of adjectives are their referent-introducing function and their predicating
function. The referent-disti1lRl/ishi1lR function turns out to be an
extremely rare function in actual conversntionnllanguage. a fact which is Predicating a property or an e~tllhlished discourse rderent: 7t%(N=24.1)
contr:""7 10 eX'p~cta~i~lI1 and which could not have been arrived at by Introducing a new di~cour~e rderenl 29% (N = 97)
examll1l1lgone s II1tulltons. TOlal: 100% (N = .140) I
The second prediction of my hypothesis is also borne out by the data:
the nominal posi.tions in which these attributive adjectives occur are in just
Here is a Mandarin example of the predicating function for Property
thos~ recorded mdependently for the introduction of new participants,
Concept Words:
and III roughl.y the same proportions (Du Bois, 19H 7): that is, patient (45
percenl). obhque (I () per cent) and the S of an intransilive clause (30
(30) xue yuwen fangmiande ... suanshi bijiao chixiallgyidian
percent), but never the t\ position of a transitive clause.
sludy lang. aspect can:be:considered relatively popular a:little
The discovery of this referent-introducing function of adjectives, then,
'studying languages could be considered relatively popular' (kh IR)
is intriguing in two respects. First, it is intriguing hecause it is unexpected.
Simply knowing that adjectives can function 'nllributively' does not allow
In .(3 I) and (32) .you can see examples of Ihe referent-introducing
us to predict that tirey, Hnd not their 'head nouns'. do the majority of the
function for Mandann Property Concept Words:
work of introducing new referents in the noun phrases in which they
occur. Thus Hn examinHtion of how people talk ha.s uncovered a hitherto
(31) (talking about Chinese children studying English in school)
underestimated use of adjectives, which is not the one usually discussed in
trealmcn1.s of contcxl-free artificial examples. Second. thi.s discovery is zai hen yukuaide qifell -zhong xue, na shi zui haode
inl riguing because of the finding that the noun in such constructions is so at very happy atmosphere-in leal'll, that he most good
often superfluous. II is reasonable, then, to describe adjectives as referenl- 'to learn in a very happy atmosphere, that's Ihe best' (kh22)
inlroducing:
Example (31) is parallel to the English (22) above: both the Property
So far, then, we have seen that adjectives in English function in
spontaneous. natural conversation to predicnte a property of a referent Concept word yukuaide 'happy' and the noun qifell 'atmosphere' are
int.roducing the new referent, an atmosphere which has the property of
and to introduce a new referent into the discourse.
bemg very happy.
Implicit in my discussion so far is a claim which I can now make explicit:
Mandarin also exhibits the other type of situation, in which the Property
these two functions may also be viewed as differing in the discourse status
Concept Word is taking most of the burden of introducing the new
of the referent to which a given Property Concepl is being ascribed. That
referent: (32) shows a Property Concept Word with an al/apllOric head
i.s, a predicaling Property Concept ascribes H property to a referent
noun:
already established in the discourse, while a referent-introducing Prop-
erty Concept ascribe.s a property to a new referent, one which has not
(32) (talking ahout entrance examinations)
heen estahlished in the discourse. Semantically, of course, one could say
lhal in bolh funclion.s Ihe Property Concepl serve.s as a predicatc (as suoyi women zhcnde shi xiwang ncnggou you yige hcn gcng
implied in analyses of attributive adjectives as being semantically related so we rcally be hope able have a very IJIOIT
to relative clauses); however, the data reported here clearly suggest that 'so we really hope that we can have a much 1110reideal exam'
grammars renect the discourse status of the referent to whom the property /iximlgde kaoshi
is being ascribed rather than Ihe semantic function of predication. ideal exam
For Mandarin, I have looked at similar data again of natural spon- (kh2S)
taneous language. The results, for 340 Property Concept Words, are
strikingly similar to those found in English, as shown in table 7.2. From the In (32), the noun kaos!li 'exam' has already been introduced into the
table it is clear, then. that the same two functions as I found for English discourse; the whole discussion at this point is about the entrance
adjectives can be seen with Mandarin Property Concepts, and with examinations. As we saw with the analogous English examples, the new
roughly the .same frequency in the data. Thi.s time, there were 110 in.stances referent is contrasted with similar referents; what is lIew here is an
of a Property Concept serving to rc-idclI/i!y an already established examination which has the property of being ideal.
referent. That is, there were no analogues to the English case of (29). Analogous to the English cases described above in which Ihe Properly
 b--
Concept Word was attributive to an empty head noun are instances in
Mandarin of a Property Concept word standing alone, that is, with a zero
head noun. About 3 percent of the referent-introducing Property Concept
Words in my data are of this type, of which (33) is an example: In this paper I have tried to suggest a discourse explanation for the way in
which languages tend to categorize Property Concepts. I have taken
(31) (talking abollt the grading of exams) Dixon's semantic study of Property Concepts as a starting-point, and
... geng zlumgyllo -de shi women yao tigao xuesheng shown that Property Concept Words tend to share features with both
even:more important-NOM COP we want raise student Nouns and Verbs cross-linguistically. I rejected a semantic explanation for
'... Ithel even more important Ithingl is that we want to raise the this fact, since semantically, Property Concept Words tend to denote
stable precepts like Nouns do, so that the widespread tendency to
xlIexi Yingyll-de nengli
categorize Property Concepts as Verbs remains unaccounted for. Instead
learn English-GEN ability
I have suggested Ihal a consideration of the jilllction of Properly Concepl
ahilily of our students to learn English'
Words in discourse might shed somc light on the pattcrns we find in the
way they are categorized in many languages. That is, given that Property
So far, what I have shown, then, is that in ordinary conversational
Concept Words share the predicating function with Verbs, and the
language, Property Concept Words, or adjectives if we're talking about
referent-illlrodllcing function with NOII/lS,this sharing of both verbal and
English, have exactly two functions, as listed in tables 7.1 and 7.2: one, to
nominal functions in discourse provides an explanation for the fact that
predicate a property of an established discourse referent, and two, to
Property Concepts will sometimes be categorized with morpho-syntactic
introduce a new discourse referent.
properties similar to those of Verbs, and sometimes with morpho-
As I mentioned earlier, I think that these same discourse functions for
syntactic properties similar to those of Nouns, while sometimes, since they
Property Concept Words will be found in any language; if this is so, they
are neither prototypical Nouns nor prototypical Verbs, they are categor-
provide the basis for an explanation of the facts about how languages
ized as a separate lexical category of Adjective.
categorize Property Concepts given in (I):

Property Concept Words (as exemplified in English and Mandarin)

function in spontaneous, natural conversational discourse (i) to
predicate a property of an established discourse referent or (ii) to Criteria for determining noun-like or verb-like tendencies for property concept
introduce a new discourse referent. I note that these two functions were words
not decided upon in advance of my investigation; rather they emerged
from examination of the data.~ I Do PCs havc nominal inncclion as predicates or as altributes'!
2 In the predicate role, how are PC's different from Nouns?
2 Property Concept Words share the predicating function with Verbs, 3 In the predicate role, how are PC's different from Verbs?
and the referellt-introdllcing function with Nouns (see Hopper and 4 In the attributive role, how is the PC attached to Ihe Noun?
5 Can the PC be used anaphorically?
Thompson 1984).11I
I'C~/III/H'likl' NI/I/m 1'0' /III/H'lik/' J'(',b,~
J This sharing of both Verhal and Nominal jilllctions ill discollrse I yes I no
provides an explanation for the fact that Property Concepts will 2 fewor no differences 2 nol similar
sometimes be categorized with morpho-syntactic properties similar to 3 not similar 3 Few or no differences
those of Verbs, and sometimes with morpho-syntactic properties 4 as Nouns are 4 as Vcrbs are
similar to those of Nouns, while sometimes, since they are neither 5 yes 5 nol without a Nonn, Pronoun
prototypical Nouns nor prototypical Verbs, they are categorized as a or RelativeMarker
separate lexical category of Adjective.11

I wish to thank Ilenning Anderscn, Johan Vllnder Auweru, JOlinBybce,Jack Du

Bois, Mark Durie, Matthew Dryer, Aryeh Faltz, Mike Hannay, Teun Hoekstra,
 Paul Hopper, Ekkehard Konig, Christian Lehmann, Lachlan MacKenzie, Hans-
Jiirgen Sasse, IIansjakob Seiler, Leon Stassen and Ilarrie Wetzer for helpful
comments on the ideas in this paper and Manny Schegloff for sharing some of his
conversational datll with me. None of these people is responsible for the use I have
made of their advice. I am also gratdulto Ihe Nethcrlands Institute for Advanced
Study for the fellowship year during which this paper was written.
I Interestingly, in some languages, such as those in the Yuman family, 'KIN'
concepts may in fact be best analyzed as Verhs (Ilaipern 1942).
2 See Perkins (19HO) and Bybee (19H5) for a discussion of the importance of
random sampling in cross-linguistic research. I acknowledge the points made by
these authors, and regret that both Dixon's smnple and the sample I collected for
thc present study cOllltlnot be rllndom.
3 The criteria which I used to determine whether Property Concept Words
shared features with Nouns or Verbs are listed in the Appendix.
4 This is not to deny Givlln's insightful remarks (I'. 55fl) on the likelihood that
more and less time-stable adjectives, in a language with this category. may he
realized as more noun-like and more verb-like respectively. Nor do I wish to deny
the likelihood of a semantic distinction underlying NOllns and Verbs along the lines
Givlln suggests (though, as I\. Faltz poiuts 0111 (personal communication) it is quite
possible that the notion of 'time stability' willlurn Ollt 10 he derived from a more
primitive distinction between events and cntities).
5 Clearly, much mOIl' work needs to be done on this question. II is conceivahle,
for exmnple, that the similarities that I report here between English and Mandarin
renect a similar 'induslriali7.ed urban' eultllral outlook, and thai data from a non-
industrialized non-urban speech conllllllnity wOllld yield different resllits.
6 Codes after each eXlllnple rder to locations in my transcripts.
7 Thus the semantic distinction which is the focus of Siegel (1980), between
'intersective' and 'non-inlersective' adjective readings, turns out, according to my
data, not to be exploited by ordinary uscrs of conversational English; in actual
conversation the distinction hetween examples such as her 'That lutist is good' I
'That is a good hllist' (p. 55) is essentially neutralized. For an attested example of
such a paraplmise, see Du Bois (19HO: 213, ex. (23».
8 In the transcription used here, three dots signal a pause and does not indicate
materialomitted.
9 I emphasize the necessit y of basing claims about the discourse basis for lexical
and grammatical categories on conversalional discourse data; it is well known that
the use of adjectives in written English, for example, is radically different fromlhat
in conversational (see, for example. Chafe 19H2). hilt I hypothesize that it is
conversational, and not written, language which has determined grammatical and
lexical regularities.
10 Nouns (or nominal roots) can function as predicales too, but, as discussed in
Hopper and Thompson (1984), thcre is ahundant grammatical evidence that this is
a marginal function for Nouns.
II I acknowledge that the stalistical predominance of the predicate function in my
data suggests that more languages should have Properly Conccpts categorized like
Verhs than like Nouns. Verification of this hypothesis musl await a further study.
AdeJaar, W. F. II. (1977) Tm.",a Queclllla: Gralllllll/r, Texts, /)ielionary. Ussr:
The Peter de Ridder Press.
Anceaux, J. C. (1952) 11u! Wolio Language. The Hague: N. V. de Nederlandse
Boek- en Steendrukkerij.
Arnott, D. W. (1970) 11,e Nominal and Verbal SyslenL~ of Filla. Oxford: Oxford
University Press.
Austin P. (1981) J1 Gramlllar of Diyari, SOIl/Ir/\ ,wm/ia. Camhridge: Camhridge
University Press.
Blake, B. J. (1979) J1 Kalka/lmgu Gramlllar. Pacific Linguislics B-57. Australian
National University.
lluecheJ, E., S. J. (1939) J1 Gramlllar of LakllVla. St Francis, S. Dak. Rosehud
Educational Society.
Bybee, J. (1985) MOIp!rology: J1 Siudy of lire RelatiollJ!rip between Form and
MeaninR. Amsterdam: John Bcnjamins.
Chafe, W. (cd.) (1980) 11lt ['ear Swries. Norwood, NJ: Ablex.
(1982) 'Integration and involvement in speaking, writing, and oralliteraturc·. In
Tannen (J 982).
Chao, Y. R. (1968) J1 Grammar of Spoken Clrinese. Berkeley and Los Angeles:
University of California I'rcss.
Cowan, R. and It Schuh (1976) .\{lOkell lIallJl/.llhaca, NY: Spoken Languages
Services.
Craig, C. G. (1977) n,e SII"IIClureof Jacallee. Austin, Tex.: University of Texas
Press.
Dimmendaal, G. J. (1983) nU! "/il/'kalla Lallguage. Dordrechl: Foris.
Dixon, R. M. W. (1972) 11,e Dyirbal Lallguage of Nort!r Queellsland. Cambridge:
Cambridge University Press.
(1977) 'Where have all the adjectives gone?' Sllidies ill Lallguage, 1.1, 1-80.
i Donaldson, B. C. (1981) DlI/clr Ueferewe Gmmlllar. The Hague: Nijhorr.
Du Bois, J. (1980) 'Beyond definiteness: the trace of identity in discourse'. In Chafe
( (980).
(1985) Competing motivations. In J. Haiman (1985).
(1987) 'The discourse basis of ergativity'. Lmiguage 63.4,805-55.
Durie, M. (1985) J1 Gmmmar of J1celrnese. Dordreehl: Foris.
Elwell-Sutton, L. P. (J 979) Colloqlliall'ersiall. London: Routledge & Kegan.Palll.
George, K. M. (1971) MI//I/YI//I/m (;1'111I11I'(1/' and Urader. Kottaymn: Nalional
Book Stall.
Giv(ln, T. (J 979) Ullderslallding Gmllllllar. New York: Academic Press.
(J 984) Syntax: A Functional alld Typological llltrodllelioll, vol. I. Amsterdam:
John Benjamins.
Glendening, 1'. 1. T. Icelalldie. Teach Yourself llooks. London: St Paul's House.
Itaiman, J. (ed.) (1985) leollicily ill sylltax. Amsterdam: John Benjal11ins.
Halliday, M. A. K. and R. Hasan (1976) Colresioll ill Ellglislr. London: Longman.
Halpern, A. A. (1942) 'Yuma kinship terms'. J1merimll J1mlrrol'ologisl, 44.
425-41.
Harrison S. (1976) Mokilese Refewl£'C Grammar. 1lonolulu: University of Hawaii
Press.
Hoffman, C. (1963) A Grolllmar of lire Marg; Lallguage. Oxford: Oxford
University Press.
 b-
I
Hopper, P.1. and S. A. Thompson (1984) 'The discourse hasis for lexical (1982a) 'The West Makian language, N. Molluccas. Indonesia: a fieldwork
categories in universal grammar'. Lallgl/age, 611,7113-52. report'. In Voorhoeve (1 982b).
Hyman, L. (ed.) (1979) Aglrem Grammatical Structl/re. Southern California (cd.) (1 982b) 71,e Makian Languages and 711eir Neighbors. Materials in
Occasional Papers in Linguistics 7. Los Angeles: University of Southern Languages of Indonesia 12, Pacific Linguistics D-46. Canberra: Australian
California. National University.
( 1981) NOlli Grammatical SI",clllre. Southern California Occasional Papers in
/,.inguistics 9. Los Angeles: University of Southern California.
Jacobsen, W. II., Jr (1979) NO/III al/(l Verb ill Nootkall. IIeritage Record 4.
Victoria: British Columhia Provincial Museum.
Josephs, L. (1975) I'alal/all Referellce Grammar. lIonolulu: University of Hawaii
Press.
Karlsson, F. (19!Ll) Fillllislr Grammar. Helsinki: Werner Soderstrom Osakeyhtio.
Ladusaw, W. A. (llJH5) 'The category structure of Kusaal'. I'roceedillgs of the
1:"Iel'I'IIIhA /III/wi Meelillg of Ihe lIerkeley Lillgui.l'/ics Socil'lY. Berkeley:
Berkeley Linguistics Society.
Laskowski, R. (1979) I'oillische Grammatik. Leipzig: Veh Verlag Enzyklopiidie.
Lee, K. (1975) KI/.mieall Reference Grammar. Hololulu: University of Hawaii
Press.
Li, C. N. and S. A. Thompson (1981) Malldarill Chillese: A FUIIClionalReference
Grammar. Berkeley and Los Angeles: University of California Press.
Lichtenherk, F. (19113) A Grammar of Mallam. Honolulu: University of Hawaii
Press.
Lindsey, G. and J. Scancarelli (1985) 'Where have all the adjectives come from?
the case of Cherokee'. I'roceedillgs of lire Eleventh Anllual Meelillg of Ihe
lJerkeley Lillguistics Sociely. Berkeley: Berkeley Linguistics Society.
MacDonald, R. R. (1976) Indollesiall Referellce Grammar. Georgetown,
Washington, DC: Georgetown University Press.
Marsack, C. C. (1962) Samoall. Teach Yourself Books. London: St Paul's House.
Noss, R. B. (1964) '/1/{/i Referellce Grammar. Washington, DC: Foreign Service
Institute.
Perkins, R. (19RII) 'The evolution of culture and grammar'. Ph.D. diss., SUNY at
Buffalo.
Rosen, II. (1977) COlllemporary Hebrew. The Hague: Mouton.
Schachter, I'. (1985) 'Parts-of-speech systems'. In Shopen (1985).
and F. Otanes (1972) Tagalog Referellce Grammar. Berkeley and Los Angeles:
University of California Press.
Schulz, D. and II. Grieshach (19611) Grammatik der /Jeutschell Spmche .Ismaning:
Max Iluchcr Vcrlag.
SCOIl,G. (1978) '[he Fore Lallguage of /'apua New Guillea. Pacific Linguistics
13-47. Canberra: Australian National University.
Shopen,1'. (cd.) (1985) Lallguage Typology and Sylllaclic Descriptioll. Cambridge:
Cambridge University Press.
Siegel, M. A. (1980) Caplllring lire Adjecli,'e. New York: Garland.
Sohn, H. (1975) Woleaiall Referellce Grammar. Honolulu: University of Hawaii
Press.
and U. W. Bender (1973) A l./.lithiall Grammar. Pacific linguistics C-27.
Canberra: Australian National University.
Tannen, D. (ed.) (1982) Spoken and Wrillell Language: Exploring Orality alld
Lilemcy. Norwood, NJ: Ablex.
Voorhoeve, C L. (1965) 'I1te Flamingo Bay Dialect of Ihe ASIIll/t Lang/lage. The
Hague; H. L. Smits.

View publication stats