';' I

SEMANTIC
1~I-IEORY
•
»>

RU'TH M, KEMPSON
In this ierie»:
1.I\CT\lltEIt IN I.INGUI~TICS
1'. 11. ~I.\TTIIE\\'S '\(",/>'11I1111:.1' lJNIV~:ltg\TY 01' I.ONIlOI"
nF.It:-;"1t1l C()~IItIE .-II/IU/
~ TIlf.OUOIt.\ IIYSOS llistoricul Linguisíics
~

¡
t,.

"~
&
~:
}'

I~
~~
~(
r:&
~.,
k¿~ \

r"'
{:~

§1
~'
.\

~
!~
I~:
I~::
i~~
,~:
'<~
:.~ \
~~
\

~:
.~;.~
~.: \
::: I

.~.
~~
:~: \ CAMURIDGE UNIVERSITY PRESS
':
~: CAM D nr D GE
LONDON' NEW YOHK' fv\l:LDOUTlNE
.;.

:~;
~:'
.
:~
.,
;::
:~
':::/"'\.
"
 Introduction

Aspects o/ the Theory o/ Synt ax ch. 1, hut Sl'C also C'ltolllsky {,all.t:/wgr (/11I/

Mind (enlarged edition), and Ilach Syntactic TIrt'tJ,y ch. 11. 'I'he prohlelll
the nature of linguistic evidence has not hccn ¡:ivt"1l mucl, d('lailnl al/cnlion
01'
hy
2
trans(ormational Iinguists hut sce Sampson The ¡'-O'III '1 LfllI.t:/w.t:r ch .. 1,
Labov 'Methodology', Fillrnore 'On ¡.;en<:rativity' and I'!ouscholdl'f 'On Explanations of word meaning
argumcnts from asterisks'. The bctwcc.i compcrcncc and pcr-
distinction
formance is discussed in Chomsky's Aspects of the Tlreo,y 'if Syntax, ami in
more detail in Lyons and Wales [eds.) /'s)'dru/i1lKl/istir l'aper¡ (Sl'C in particular
Fodor and Garrctt 'Sorne rellections on cUlllpctelln' and IH:rfllflllanec').
.\ Chomsky's account of this division has hCCTl attuckcd hy 1I yrncs 'On cum-
r.
¡.
municative cornpetence', and more n'el'ntly hy (;. l.akofT amI nthcr s (scc t lu-
f rccommendcd reading for chapter .¡).
~
1n the first chaptcr 1 suggestcd SOl/le prcliminury conditionsof udcquucy
·t.$,f for scmantic theories by which particular theories could be tested, and
it rnight seem that we are now in a position to consider the detailed
t
•~. mcchanism of some proposed thcory. llut there was one large and
unwarrantcd assumption in the way that thcse cOllllitions wcre spccificd ;
it was assumcd that the rclationship hctwcen a word ami what it is uscd
to imply and that between a sentence nnd what it is uscd to imply
~: presented no problerns, ano were not a matter for debate. But, quite to
~
~: the contrary, an explanation of these rclationships - i.e. the problem of
': what we mean when we rcfer to the rncaning that a word or scntence
has - is the classical problem of scmantics, the problem indeed on
r~
,. which scrnantics has traditionally [oundercd. Since any formal represen-
~ tation of sernantics will implicitly present one particular solution to this
~
:, problem, we cnnnot usefully consider the dctuils of such a theory except
in the light of a eoherent account 01 meaning. •
There are three main ways in which linguists und pliilosophers have
attcrnpted to construct explanations of mcaning in naturallanguage: (o)
IS
.~ by defining the nature of word rneaning, (h) by defining the nature of
p.
sentence meaning, and (e) by explaining thc process of eommunication.
~
.~ In the first wa}:, word meaning is taken as the eonstruet in terms of
which sentence meaning and communication can be explained; in the
sccond, it is sentence meaning whieh is taken as basic, with words
characterised in terms of the systernatic contribution they make to
',:
sentenee mcaning; and in the third, both sentence and word meaning are
explained in terms of the ways in which sentencés and words are used in
the act of communication. It is no coincidence that there are these three
types of explanation. In the first place, thcre c1early is a relation between
words and objects. We use words lo refer to objccts, and to actions
J~~; (eonsider such words as ClIP, liorse, tooman, grodllate, cooking, noeeping,
•
~.
~~
. 10
.~ 11

1<:-
':~~ -,
 !
I
Explanations o/ toord mcaning Al 2.1 Meaning and rejerence
I
Ihinkinc), ami the explun.uion of this n-l.u ion is induhitabf y thc t ask of
there is a one-to-one correspondence hctwcen name and object: Ior
semantics. Similarly scntcnces are uscd lo dcscr ihc cvenrs, bc licfs,
opinions, ami it is unquestionably t he task of scmant ics lo cxplnin thc
I
I
exarnple the name The j>arllu/I(;ll rcfer s to thc objcct the Parthenon in
Athens, the name Ruth Kempsou refcrs lo th~ individual who wrote this
nature of the rclation betwccn scntcnces ami the statcs of alfairs those
book. This rclationship betwccn word and objcct is called the rclation-
sentences describe. Finally, since lunguagc is the vchicl« by mcuns of
which we effect cornmunication, it is argllahle thut the intcrprct.uion of
I ship of reference, and there is a long tradition of equating the problem of
meaning with the problem of rcference. According to this view, known
language should be explaincd in tcrrns 01' its rolc in communication.
as extensionalism beca use of its trcatmcnt of meuning in terms of the
Moreover these thrce nspects of Illeaning, word mcuning, scntence
objects, callcd extcnsions, to which thc items of the language refer,'
rneaning, and comrnunication, are reflcctcd in diflercnt uses 01' the word
the meaning of a word can be explained in tcrms of the relation bctween
mean, Corrcsponding to cxplanut ion (11) is :
that word ami ohject or objccts to which it refcrs. J ust as propcr narnes
(1) Supererogatory rncans 'supcrfluous'. rcfer to individuals, it has hccn suid , common nouns rcfer to scts of
(2) Spinster means 'unmurricd woru.ur'. individuals, verbs rcfer to aetions, adjcctives rcfer to properties -of
individuals, and udverbs refer lo propcr ties of actions. Thus, for exarnplc,
Corresponding to (11) is : it would be said that thc rclation bctwccn the express ion Ruth Kempson
\

and the individual Ruth Kcmpson is directly comparable with the

(3) The sentenee Ja/lles murdcrrd ,HII,\' mcans t hat somcone
relation betwccn the word mice and the scts of objects which can be
called [ames delibcrntely killcd snmeonc called Mnx. •
referred to by the use of that word; ami moreover both of these relations
In these two uses, the word mean has a lIleaning approximaling lo are said lo be siruilarly comparable lo the relation which holds between,
indica/f. Butthe word /IIf(lI/ is uscd in a ditlcrcnt scnsc in the [nllowing say, the word red and the scts of objects which have the property of
conversationbetwccn two spcakcrs, A and Il, a sense which corrcspnnds redness, and too to the rclation which holds betwccn the word quickly
to ~xplanatidn (e): and the sets of actions which liavc the appropriatc propcrty of speed.
IIowever, assuming for the momcnt that we can acccpt the clnirn of
(4) A: Are yOIl going lo bcd soou? hornogeneity in connection with thcse difTerent relations, there are a
D: Whut d'you mcnn? number of reasons to believe that any theory of meaning which attcmpts
A: 1 rnean that I'm t ircd , and thc sooner yOIl go lo bcd, the to explain all aspects of word meaning in terms of rcferenee is mistaken.
sooncr 1 can. 1n the first place, there are a nurnbcr of ernbarrassing counter-examplcs:
In this case, mean is attributable lo speukers und has the sume Illeanillg even if the relationship of refcrence can be said to hold between a word
as the expression intend /0 indicate. 'I'hus we have at least thrce possible such as imagination and sorne cluss of abstract objccts which constitute
starting points from whieh to construct un explanation of meaning - the acts of imagination, therc is no .sense in which words such as and, not,
signification of words, the interprctarion of scntcnces, or what a speaker tohether rcfer to anything. And '0111 prepositions present a similar prob-
is intending to convey in ucts of conununicatiou. Of thcse thrce uses, lern, Wliat does o/ rcfer to? What , in this very sentenee, docs the pre-
¡:
:. most traditional explanations of mcaning constitutc u n nttcmpt lo cxplnin position in rcfer to - or indced uery or toltat] NOl only is there a large
meaning in terms of the naming rclurion which holds bct wcen a word and non-homogeneous c1ass.i)f:exceptions, but there are a number of
and its object, and it is this that wv shull hr conccrncd with in this anornalies in explaining more straightforward cases. The relationship of
chapter. reference which holds betwccn expressions and non-cxistcnt objects will
be the sarne : it is therefore hard for a theory which explains rneaning in
2.1 Meaning and refer enc«
I T'he rnost well-knowu.cxposu ion is Russcll I ()OZ. hui chis view has rcccntly
Thc narning rclat ion hct wccu a word ami it~ ohjL'C! is most becn expounJcJ 1I~lIin hy Duvidson (ef. Duvidson 196711) within a trurh-
transparent with proper narnes, thc paradiglll case of nallling. llcre bused rheory of meunj ng (cf. J,I below).

12
13

<

ª
 R:~p'lallaliolls 01 uord I/ltallillg 2.1 Meaning aud reference

terms of rcference to uvuid prnlil'lillg SyllOIl)'lIly betwccn ;111 of the This is not lo dcny that thcre are problema in the nnalysis of refercncc.
following: the pterodactyl, tlte unicoru tlu: [trs! 10(11I1//1/ 'o ///11/1 (J/I the On thc contrary. the solution tI; -thc prohlcm of opacity in particular
moon, And, for the sarne reason, 0111 cxprvssion SIIl'h ;IS tlte jirst /IUIII lo rcmains an opcn OI1C, and sonH:thing of a n issue for philosophers.
land on thc moon will he prcclictcd 10 he quite diílercnt in k ind frorn the But it <loes cast doubt on t hc 'assumption thut any solution to the'
expression the first tooman lo 1(/11/1011 1!tf./II1J1J/1, hccuusc only in t lie former problerns of reference autnrnnt ically provides a solution to thc problern
case is there a rcferent to which t he cxprcssion cm st.md in a referring of meaning.
relationship. Prohlerns arise cvcn in un .malys is of common nouns
which refer to a set of ohjccts. ror in what srnsc can t hc rc be s.rid lo be a 2.2 Thc imagc theor y of mcaning
,consistent identifiable rclat ionsh ip of rcfcrcncl' bct wccn t hr wor d igunna Another solut ion to thc prohlcm of cxplaining the naturc of .
and a set of objects to whicl: it rcfcrs in C,) (7r wurd rncaning, which has an cqunlly long 'tradition, is to cxplain the
mcaning of a word in tcrms of rhc iougc in thc spcnkcr's (or hearcr's)
(5) Iguanas are not very COlll11Il111.
bruin. '¡'he problcm here is to know what form thc imagcs tuke, 'I'hc
(6) Are iguanas cxt inct?
most obvious point is that these'illlages cannot be visual. For suppose
(7) .Professor 11ra ncst a \\'1 1I is looking for igllallas,
rny image of a triangle is an cquilatcrul: '
In {S), the word might be suid to refcr to a class of objccts, viz. igu.mas ;
but in (6), the word either refers to a cluss of objccts or a null class,
apparently depending on the unswcr to the qucst iou.i And in (7), the
problem is no less acure: for on one intcrprct.uion there lIlay be, say,
two specific iguanas that I'rofessor Hruncstuwm is Iuoking Ior, but on
another interpretation he lIlay just be looking without thcrc ncccssarily
being any such objcct. On this intcrprct.uion, it makes no sensc to Fig. I
question whieh objects does thc word igual/a rcfcr to? This problern
arises in a large number of cases (culled 'opaque coutcxts'), following If this is said to constitute Cor me the meanmg oC the word triangle, then
vcrbs such as beliet:e, soan] nnd /1Of>e; ami t hcse present a notor ious eithcr triaugle has to mcan cquilatcrul triangles only, or triangle has tobe
problern to anyonc attelllpting tu providc un .m.rlysis of rcfercnce. said lo be arnbiguous according as the image is cquilateral, isosceles, or
;'
. Furthermore, iC we return to the paradiglll case of rcferring, propcr scalenc. For cuch of thcse is nuitually exclusive. In a similar vein, an
names, we fino an important diílcrcncc bctwccn thcse and an)' uther owncr of an alsatian ma)' havc a radically different image of dogs from an
syntacric category. Thollgh in proper n.uncs thcrc is a OIlC-tO-OIlC owncr of a miniaturc poodlc, but it is not obvious that they thereby'
correspondence betwecn word ami ohjcct , it is 1101 ohvious t h at propcr spcak a diffcrent language. Thcrc is no image corrcsponding to what i~.
narncs have any meaning ut all, ror it m.rkcs no se nse lo ask '\Vhat is the sharcd betwccn dogs, ami none either which has just those fcaturcs
meaning of thc expression NIJIIIII C/IIl/lIsII.l'?': nnc can nnly usk '\V11O docs shnrcd by all triangles. Ancl this is just one of the m:1I1y problerns Iacing
the cxpression NOIIIII C/lIJlIIshy rcfer to?' Th is slIggesls al t hc very lcast a SImple illlage theory of mcuning. Furthcr problema are prescntcd by
that a semantic account of prnpcr n.uncs should not be like that of other the fact that (a) one Illay llave more than ore image for a single expresa-
words, Ilut if this is so, thcn the original ussumpt iun of homogcneity in ion, and (b) two expressions may havc the sarne image. Thus the ex-
the semantie properties of proper nouns uud thc other categories, pression a tlred child may cvoke cithcr an irnage of a child (notice that
cornrnon nouns, vcrbs, udjectives, udvcrbs, and so 011, was a mistakc. there is no visual image neutral as between a boy or a girl) curlcd up and
nearly aslecp, or an image of a child stamping its Coot and screaming.
I Sentences such as Ihese pose ,',,,asiderahle prohlcrns fnr " rheory of
According to an account of rneaning which cquates the meaning a word
reference, as WilllCSS 11,,· larg,· a"'''"111 of pl.ilosophi<:al re 011 the
litcr.uu
subjrcl. has with an image, any word which relates to more than onc image is

If 15
I Ii

",
 Explallaf¡ons o/ word meal/il/g 2,3 Meaning ami concepts

predicted to be ambiguous. But despitl' t hc prcdict ion, t hc vxprcssinn nnd hencc vucuous. \t docs no more than substituto Ior the problem
a tired child is not ambiguous. On the othcr hand, sliould t \\'0 cxpressions tcrm meaning thc equally opaque tcrm conrept . It dOl'~ not providc un
bear the same irnagc, the image theory of ll1eanillg prcdicts that they will explanation of the rcquircd kind (cf. p, I ubovc), If mcaning is to be
be synonymous. Dut many cxpressions hnve the same ill1age: (/ tired explained in terrns of conccpts, it is cssewtial that the term (Ol/tf!>/
child, an unliappy child, (/1/ tlngry cluld , (/ [uture tyrant maya" cvoke the itself be given a rigorous dcfiuitiori. '
identical image of a child stamping its foot und screarning. Yct thcse Sapir's Swiss contcrnporary, de Suussurc, goes sorne \Vay towards
cxprcssions are by no mc.ins synOIl)'1I10US, /\s I huvc alrc.uly partly providing such a charaeterisat ion, FOI" though, like Sapir, de Saussurc
indicated, an image thcory of meaning faces the ndclirional problcm of talks freely of conccpts, he strcsscs that the concept (the word he uses is :
speaker variation. The ill1ages we havc of what might be rcferred to by signijié) an clement .stnnds for is solcly due to its valuc in the systcrn:
any word may not only v;lry fmlll occasioll to occ.isiou, hut since they 'Language is a system of intcrdcpcmlcnt terms in which the valuc of
arcdepcndent on our expericnce are certuin to dill'cr in lI1any dctuils, if cach terrn results solcly Iron: the simultancous prcscncc of orhcrs'. Oh
not radically in substance from those 01' othcr pcople. 'I'akc t hc wor d this basis he would have said thnt thc word bachelor, for exumple, has the
lecture for example. To those who give lccturcs, t hc wnrd Illight call IIp mcaning it docs solcly by virtue nf the ol,hl'r itcms in the systcm to
animage of an audience of bctwccn, S;¡y, twenty and one hundrcd which it is rclatcd - spinster, tooman, huslmnd, boy, Similar/y, right
people staring up at one wretchcd individual who, per haps sclf-con- across the vocabulary. 'J'hus eaeh of the mcrnbers of the following sets of
sciously, walks up and down in front of t hcm. Ilut ror t hosc who have words stands in a ccrtain rclut ion one to another (Iabclled valeur), which
never lectured, the image is more likcly to be that of one pcrson droning is itself a determinant of the intcrprctution of thc word:
on, often boring, sometimos incomprehensible, with thc accoll1panying
sensation ofhaving to fight feclings of dro\\'siness, Such ditlercnt images have criticise angry mother
should, if they correspond to the ll1eaning associatcd with the word glve pralse h;¡ppy uncle
, lecture guarantee that cornmunicarinn bct wccn two such grollps of lend accuse calm aunt
people using the word lecture would be impossible beca use cnch group borrow assess ' plcascd grandparent
. , has radically different illlages of Iecturing ami hcnce diílcrcnt concep- rcnt blame annoycd , nephcw
" tions of the meaning of the word lert ure, Worst of al/, there are lIlany hirc rcprirnand upset COUSIIl
words wit,h whieh it is irnpossible to nssociatc any image at all - and,
or, because, therefore, cte. Yet they are by no mcnns rncnninglcss. It is not of eourse obvious that this observation saves the problem of
defining concept from vacuity, for it is not clear how the inter-relation- . '.
2·3 : Meaning and conccpts ship of value (valeur) ami mcuning (signifié) can itsclf be tcsted, In any •
The standard retrcat Iróm thc extreme [orm of the image case, de Saussure's account is 01)CII to objections similar to those .raiscd
view of méaning is to say that thc ill1ages are not visual; but, if so, it is against both a refercnce theory of mcaning and an image theory of
not obvious what claim is being m.ulc. Considcr for cxample the sug- rneaning. In particular, words such as and, because, or, etc., are counter-
gestion that 'the speech clernent 'housc' is thc symbol, Iirst ami Iorcmost, exarnples lo this vicw, for it is not clear whether their intcrpretation can
r
I~
not of a single perception, rior cven of t hc notion of a particular objcct, be analyscd in tenns of conccpts, 1t will not do to suggest that the
but of a 'concept', in other words, of a convenicnt capsuh; of thought meaning of and is thc concept of co-ordination, for what is co-ordination
that embraces thousands of distinct cxpcr icnccs :11)(1 that is ready to other than by joining by and] Sirnilar/y with or: it is meaningless to
takc in thousands more' [Sapir 1921: IJ), \Vhat is involvcd in this clairn explain or as having the conccpt of d isjunction for its meaning, when in
that a word has as its Illeaning a 'eonvcnient eapsule of thought'? If this order to explain disjunction one nCl'ds to rcfer to oro And the general
~ .is a retraclion from nn im;¡ge theory nf ll1eaning, as it is. then it is ;¡ problern remains: to expbin meaning in terms only of concepts is
l~
J,. retraction fmm a specilic, fnlse clailll to one that is entirely untestable unempirical.
:~ o" I

~:' i6
~ 17
~
.;
"

"
,~
.'~.
 I Exfilallatiol/J

,
2.J.1
01 toord

Componential
/l/eal/il/g

al/(/!.IIj'is
The assuuipt inn 01' SYSll'III;ltit: rdatiollships of Illcalling
of more primitive semantic cornponents,
of sernantic cxplanation
2. J M fa 11¡IIC

one is transferring
from word mcaning onto the componcnts
a/lll

the burdcn
(oIIa!'t,

-betwecn words is however indcpcnch-nt orthe prohlcin of cxplaining the which together, in dilfercnt ccmbin.uions, constitute word 'mcnninga.
basis ofthese relationships ; ami a considcruhl« .unuunt nf dctailcd work Indccd what rcrnuins to be cxplaincd in such .ornponcntial analyses, as
on the structure of the vocabulary has hccu done in rcccnt ycars, Many we shall see bclow (6.J.2), is the relationship bctwccn the words of :1
linguists have turned lo what has bccn cullcd componcnt ial unalysis to given language and the apparcntly indepcndcnt components. Wc may
giye an explicit representation of thc systcmut ic rcL,t ions hct wecn words, glibly say that thc word spil/sler has a meaning which is a cornplcx of the
01\ this view the meanings of words are an.rlyscd /lot as un itn rv concepts sernantic componcnts [FEMALE), [HUMA'N], [ADULT], [NE,VER
but as complexes made \11' of componcnr« 01' nll':lning wh ich are t hem- MARIU ED], bUI the central problcrn is the relationship betwcen the
selves semantic primitives.t In this vvin, ,I/>/",I'/n 11Iight he a/lalysed as a word //11/1/(/1/ and the componcnt [HUMAN], and so on for the rcrnain- ...
1\,
'r:
~ sémantic complex rnnde \Ir of t hc Ic.uuns (vquivulent lv callcd com- der of the vocabular y. Unfortunately thc account of the sernantic
l'
.; ponents.! or markers) [FEMALE), [!\!E\'I':I{ i'vl,\ IUUED). [A DUL'I'], cornponcnts themsclvcs givcn by linguists using the methodology of
L;;
. [HUMAN). This form of analysis was uscd in par! icular hy anthropolo- componential analysis is often no more substantial than de Saussurc's-or ..
giSts seeking to give an accuunt of kiuship teflni/lology in various Sapir's charactcr isation of word mcaning. For cxample, it has been
i
" cultures. For exarnplc, the distinetion hetwccn mothrr ami aunt in clairncd that scmantic fcaturcs are not defined in terms of 'physical
~
I, ;' English might be made cxplicit if the tcrms wcre nnalvsed as contrasting properties and rclations outside the human organism' but are symbols
"
,.:': cornplexes of the componcnts [rE 1\r¡\ LE], [1' ¡\ rU~N'I' O F), [Crr [LD 'for the interna] mechanisms by rncuns of which such phcnornena are
$
" OF].) Such componential annlysis is not nI' COIII'$e rcstr ictcd to k inship conceived and conceptualised' [Bierwisch 1970: 181). Katz, one ofthe
~ terms: it can be applied in many arcas 01' thc vocahularv, Fnr cx.unple, central exponcnts of scmantics within transforrnational grammar, gives
~
,;; the distinction betwccn murder and !lil/ can be stalcd cxplicitly und a much more detailcd account, but with little more substnnce:
.;
:: economically if murder is anulyscd as havillg :1 Illealling which is a
A semnntic murkcr is a thcorcticul cnnstruct which is inrendcd tu rcpf'cscnt a
~ cornplex of components rcprcscnring intcntion , cnusution, ami dcath,
concept that is part of the scnse M morphernes and othcr constitucnts: of,
~ and kill as having a cornplcx of only t hc COlllpOllents representing natural languagcs, By a concept in this cnnncction we do not mcan irnages or
causation and death. In a similar \\'ay, g17'r and t alte can be shown mental ideas or particular thou~hts. Concepts ... are abstract entities. They
:::.
.', 'to be distinct by virtue of their cO/ltrasting coruplcxcs of components do not bclong to the C(J/lSciOIl~cxpcricnce of anyonc, though they' may he
::; representing causation and change of owncrship. By this mcuns, de thought about , as in our thinki/l~ ahout the concept of a cirelc. Thcy arenot
~..
"
individuatcd by pcrsons : yOIl and '1 Illay think about the same concepto .
','
'}
Saussure's concept of valeur char;¡cterising the rclationship which a
(Katz 1972: J8)
~'
'i
word holds to other words in the svstcm can be stuted cxplicitly in terms
~, '
of related but distinct componcnt cornplcxcs, ancl indccd thc main Ncither characterisation <loes more. than say that the meaning of a word
¡,
~~ value of cornponential analysis lies in t hc econoll1y of st.ucmcnt of thcsc is a (complex) concept, and this wc rejected as. vacuous. Ye,t Katz
::
~ relationships which it allows, roundly dismisses this rejcction, inaintaining that 'it is quite unreason- ...
, " However in so stating the intcr-rclationships hel\\'ecll words in terms able to insist at the outsct ... on a cl.rrification of the ontological under-
"
) pinnings of the notions of concept and proposition as a precondirion for
~
J
1 For a more dctuiled diseussioll 01' 111<' sl"IIIS ,,1' II11'S~ (()IllIHlI'~IlIS, cf, 6.2 accepting the cxplanations of semantic propertics and relations givcn by
below.
~ 2 I shall use rhc rcrm srrnnntic ((JIII!UW,.", wit h syst cmut ir arllhiguity, holh
a theory employing 'semantic markcr' .. .' (Katz 1972: 39). l lowever
~' to menn rhc sect ion of thc ovcr-u] l grilllul1ar in ",hid. t he scrnunt ic Katz' mere disrnissal of the attack is not warranted. In the first place, as
~ generalisurions are sturccl, :tnd to nll':ln tlH: (.'Il'flll·ntary sl'fl1:mlic units we have already secn in pnrt, difTering accounts of the nature of rneaning
~ isolalcd hy cOlllponcnl i,,1 '1I\"lysi"
~ make difTerent claims nbout what constitlltes the proper don,"aiQ of
~ J Cf. Lehrcr 1')74 ,,"d Ll'lTt. "175 1'01' ,kl"iJc.d """I,'St's "r En~lish ,,"d
~ olher kinship lennillol,,!:il'S. semantics. And the setting up of a formal scmantic theory 'as one
~
~
1'8
J 19
1
t
 ¡-, _ 1

;-

;!:
~ E.\pla1/atiollS o/ uord /l/eol/llIg Recommended rÚII/iIlC

component of a gener:Jllinguistic theory, pre:;upposes :J solution to this 'On denoting' (reprinted in Olslicwsky (ed.) l'roblems ;11 the I'hilosophy' 01
x
.::' question: thus any such theory, as indccd t hc one Katz providcs, must 1,(/IIg/lage); hut this view was until .reel'.IItl)' thought to he untcnnblc Ior
beirnplicitly prcsenting sorne clairn as to t lie nuture of IlIeaning - and ir rcasons brought [orward 1>)' Frcgt·: seco Frt'gc '01>cr Sin n und Iledcutung'
(trunslarcd as 'On sense and r efcrcnce"), and Dummetr's detailcd account of
this c1aim is unfalsifiablc, thcn the theory itsclf hccorncs unclllpirical.
Frcge in fo'rege: I'hiloso,"¡y 11/ 1.1I11.t:IIII,~I', Russcll's analysis has al so "bccn
.Secondly, in defining me:Jning in tcrrns of mental constructs, Katz' auacked for indepcndcnt rcasuns hy Strawson : in this connection see the
'theory has no apparent place for nn cxplunation of the relation betwecn recommendcd rcading for chaptcr 1). 1Iowever the rcduction of the problcm of
o·, .,
aword and some object that it may be uscd to rcfer lo, or of thc relatian meaning to the problem of reference has becn re-introduced, both by David-
J between a sentcncc and the sta te of uíluirs it describes. Each of these son (see Davidson "Truth and mcaning') and within possible-world semantics
~ (sce Hintikka 'Scmantics for propositional auitudes' (reprinted in .Linsky
~i .rclations has.been mcrely reduced to un untcstahle
cxpression and a mental construct.
rclnt ion bctwecn an
And in the case of English worrls
(ed.) Reference and Modalily», On rhe prohlems of analysing sentenccs' such
~ as Are iguanas extincti , studunt s should r efer to the lirerature on cxistcnce:
~ -such as human where there is a one-to-one correspondence between thc sce for exnmplc Moorc 'Is cxistcnce a prcdicatc!', I'car s' and Thomson's
~ word and a semantic component, wc are given no explanatíon al all other articles of thc sume name (rcprinred in Strawson (ed.) I'hilosopltical'Logic),
j . than the bald staterncnt that the rncnninp of human is thc conccpt repre- Quine 'On what thcr e is' (in From a.Logical Point o/ Vietu). On the problema ¡'

~X
~ sented by [HUMAN). 'I'his is not only no! cxplanalory as an account of prescnted tu theories of rcfcr cnce hy opaq.!e contexts sce Linsky (ed.]
~ Reference and Modality, For a linguislie npprouch to the problcm of opacity,
~
~ rneaning, but it is also quite unfnlsifiablt,. So, while t hc mcthod elllployed
sce Hall-I'artce 'Opacity, corcfer cnce .and pronouns' and 'Opacity and scopc', . ,

aQ in componential analysis may be uscful (and J s!t:J11 indeed be rnaking There is considerahle disagreement ovcr the scmanric propcrties 01' propcr
~~ use of it in later cha¡Jlers of this book), the theoretical unclcrpinnings names : [or a representativo set of views sce Searlc .'-)'"eechActs ch. 7, Kripke '.
:~
S . provided by an account which incorporates a definition of sernantic 'Nam.ing and ncccssity', Dummctr's rt'ply to Kr ipke in Frege : I'hi/oso/,/'y o/
components in conceptual terrns are not (hose of a falsifiablc thcory. Jf Language, and Mates '011 the scmant ics of proper namcs',
.;
sernantics is to be part of an cmpiricnl scicncc, such an account rnust be The classic statcment of an idcuiional thcory of rneaning is g·lvell. by
;, 2,2
"

l.
rnade more substantial. Lockc, whose exposition is r epr iutcd in Lehrer ano Lchrer (eds.) Theory of
~~
,¡ Meaning . The twentieth-ccntury linguistic represcntative of this view is
In attempting to unravel the conccpt of menning in naturnllanguagcs
~
,\ Sapir: see Sapir's La I/gI/agt. .
.~ by considering the nature of rneaning, we sccm lo be in a paradoxicnl
~ position: characterising Illeaning merely in tcrrns of conccpts is unex-
~.,
2,3 De Saussurc's account of semant ics, r,iven in Cours de Linguistique
~. planatory, and characterising meaning in terrns of rcference scerns to Gé1/érale, is now mainly remembered for the emphasis he gave to structural
:j .: relationships between lexical iterns. Two Iorrns of semantic description can
. enter into toa many problerns to be a convincing solution, This is not of
.'
~
,.i, course to suggest that problema in thc :Jnalysis of rcfcrcncc can be be associated with his work: dcscription in terms of so-called lexioal fields,
e
.::; dismissed, but merely that the rclationship of reference does not provide
and, much later, description in terrns of componential
account of work done on lexical fiel¡ls und of componential
analysis. An excellcnt .
analysis is 'given in
¡~ 'an adequate basis for an explanation uf word mcaning. We rnust now Lehrcr Semantic Fields and Le,\'ica/.~'/TIIrtl/re, and she summariscs the work of
I~ turn to the second of tire thrcc lIIain possibilitics that 1 oullined carlicr _ Trier who was the mast well-known carly e.xponent of analyses in lcrms of
anaccount of sentcnce rncaning. Icxical fields, For a uscful survey ,of work 0'1\ semantic ficlds, se.e Vassilyev,
1
¡ 'Thc theory of sernalllic ficlds: a s\lrvcy'.
~
RECOJ\1MENDED HEADING 2.3.1 A oetailed introduction to lhe analysis of kinship terms is given in
Burling's Mal/'s Mal/y Voires, which incluoes an analysis of the set of English
2.1 THe probJem of iuentifying propcrrics of IlIe~ning \\'itb properties of kinship tcrrns, This hook is illciuelltaJl)' an excellent account of componential
reference is almost inv~riahly includcd in int roductions to philosophy of analysis and the problcrns in its application. Other refcrences inclutle ~9lfn~-
language: see for c:-camplc Ihe introduction lo I'~rkinson (ed.) Tlte Tlteory 11 bury 'The stnlctural 'analysis of kinship semantics', Goodenough 'Compo-
.. : Mean;ng. and Alslon Philosophy of LOI1¡:lIogt! ch. I (which ~Iso gives a critical nential analy~is ami the study of mcaning', and Conklin 'Lexicographical
cxposition of both behavioristic thellries 01' lIle~ning :lnd idealional Ibeorics) . treatment of folk taxonomics'. For an account of items such as give and /ake
Thc most famous attempt to analyse IIlcanin/j in such terms is Hussell's and other relateu ilellls sec lIrndix Compol/en/ial Ana/ysis o/ Gel/eral Vocaól/-

20
21
 .(~ ..• -L'
~
Explanations o/ word meauing

lar)'. For general problems in thc meth"d"logy of componenrial unalysis see -,

Nida Componential ¡l"alysis (1Mra1/i1/K. 'Th« first application of the methods '.

of c.omponential analysis within the frarncwork of transforlllational grammar 3

was Katz and Fodor 'The structure of a semantic theory' (rcprinted in
Rosenberg and Travis [eds.] Readings in the Philosopliy 01 La1/guage). The Meaning and truth
most complete discussion of problems in semantics hy Katz is arguably
Semanlic Theory, but for a further detailcd defence of his position that a
sernantic marker (equivalenl to a pr imirivc component of mt'aning) necds no
characterisation other than a conceptual onc sce Knt z 'Logic and language:
an exarnination of recenl criticisrns of intensionalism' s. {¡o This pos it inn is
also defended, against a truth-hascd rhcorv of IIlcaning, by ll ar mnn in
'Meaning and semantics'. For Iurrhcr rcfcr cnccs 011 CO/llpollt'ntial analysis,
see the recornmended reading Ior chapt cr (•.
).1 Tarski's truth definition ano scntcnce rncaning
,
"
'.
The attempt to cxplain (he hasis of meaning in terrns of the
relationship between a word and the object :t may be used to rcfcr to lías
a long tradition in philosophy. Recently however a rather different '.'
starting point for a theory of mcaning has bcen advocated. In line with
work in logic by Tarski, a logician whosc theory of truth put forwardin
"
1933 is now widely held by philosophers, it has been suggested .that it is
the characterisation of wh»t we mea n by talking of the rneaning of Q
~
le~ sentence that is the basis for a semantic thcory, and not the characteri-
'F sation of word mcaning. Tarski proposed for formal languages con- .
.'~ ¡'
structed by logicians, that a definition of truth could be given .for 'a
language if for each sentence of the language! a rule schema can predict
correctly the formula "

S is true if and only ir p

where S is the name of the sentence (i.e. a mere spelling ?ut of the
sequence of symbols making IIp the sentence) and p the conditions
which guarantee the truth of that sentence. The significance of this
suggestion for logic is of course not our concern here. However it has •...
been claimed (Davidson 1967a) that Tarski's formula for a theory ol.· , .
0., ..:'

truth can also provide the basis of a theory of meaning for natural
languages. The suggcstion is that to know the meaning of a sentence-is
to know under what conditions that sentence would be true: 1'0 take
Tarski's classic example, lo know the meaning of the string' of words
making up the sentence SlIOW is tohite is to know what conditions-have
'.,
In fact he was referring not to sentences of natural Ianguage- but to "
staternents ¡¡fa formal logic. Cf. 3.4.1 below for a discussion of rhe dis-
tinction between sentence and statement,

22
23

..

"