I A-PDF Merger DEMO : Purchase from www.A-PDF.com to remove the watermark
PRAGM AT ICS
Ie thit etks: p. n, M^TTHEWS ,l...
674 downloads
3910 Views
87MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
I A-PDF Merger DEMO : Purchase from www.A-PDF.com to remove the watermark
PRAGM AT ICS
Ie thit etks: p. n, M^TTHEWS ,llor$clqgy Sccond cdition a. coi|rlr, AtN.l r. M. xEMPsoN Sc',,;nfk Tlaory r. B\ Norl Eittoti..l Lirgrittkt in Ling'itlkt J. ALLrooD,L.G ANDErsoNrnd o. DAHL 'dttt o. B. rr\ Tb Pht*t o! Siout this loud Don't do it nou, ttttt NOWI
These two kinds of deictic usagecontrast with the non-deictic \rsq*a of the same words or morphemes. Some examples will help to rn61" the three-way distinction clear; in the following the a case5q,. ge.rlrral usages,the b casessymbolic usages,and the c casesnon-deicttc usages: (30)
(3 r)
\3 2) (33,,
(:+)
a. You, you, but not you, are dismissed b. What did yor say I c. You ctn never tell what sex they are nowadays a . This finger hurts b . Tiri ciry stinks I met trrr'Jweird guy the other day a . Push not zot{r, but nora) b . Let's go roar rather than tomorrow Notl;, that is not what I said Not ,ia, one, idiot, that one b . Tlar's a beautiful view Oh, I did this and rla, a . Move it from lhere to there b . Hello, is Harry rlere ? c. Thete we go
(Note that, in most cases,the three kinds of sentencesonly favour the three kinds of interpretation.) Here are some further contrasrs betweenjust two of the usages,each labelled a, 6 or r as before:
(3s) (3 6 ) (37 ) (3 8 )
a.
((ln responseto: "Who wants another?")) / do b. ((ln response to: "Wilt thou have this woman ro thv wedded wife ?")) 1 will b . I did it ten years ago c . Harry had done it ten years cgo b. John lives opposite c . John fives opposite Bill b. We can't see the chimp because it's behind rhe tree When Harry's front axle buckled. he was behind a truck
A few brief comments on eachofthese: in (35a)the pronoun I is used gesturally to self-nominate from a group, in (fSb) it just has the symbolic usage; in (36b) the word ago places the time at which the b6
relative to the time of speaking, in (36c) the time is --ridn occurred ln ]]i"ri". a ttt. time at which the events in the narrative occurred is understood )"'lss oooot;t"Gna ,qually aeafty ' aroufld thc corner, etc ) to the place of utterance, in (lzc) it is relative to Bill's )l'Jr,ir. chimp on the opposite side of the L"ii.n. It tggUl Oeiizd locates the (l8c) in it locates Harry at the rear end participants, the rle from of the truck' perhaps the most important distinctions in the use of i- Tl,"ra "." but they are not the only ones' As we shall see when deictic terms, we shall need *,e consider discourse deixis, within non-deictic usages usages.' All the c from non-anaphoric anaphoric to distinguish but non-deictic interpretations, most natural in their are, casesabove term where some is usage anaphoric usages. An alsonon-anaphoric (ot class of objects) that some Dicksout as referent the same entity picked Thus, in the following, ie can out. orior term in the discourse naturally be interpreted as referring to whoever it is that Jo'lrn refers
(3e)
John came in and he lit a fire
We will return to anaphora, but just note here that it is perfectly possible,as Lyons (r977a:676) points out, for a deictic term to be usedDoti anaphoricatlyand deictically- For example, in: \40)
I was born in Lotdon and have lived trt€le evea since
t/ralerefers back to whatevet placeLondoz refers to, but simultaneously contrasts with here on the deictic dimension of space, locating the utteranceoutside London. Note that it is also quite possible for the gesturalusage to combine with the non-deictic anaphoric usage too: (4r)
| .!t
a frnger: this ore
Here tir's oze refers to whatever a.76ngerrefers to, but simultaneously must be accompanied by a presentation of the relevant finger. Clearly the proliferation ofdifferent kinds of usage of deictic terms is a source of considerable potential confusion to the analyst. The following summary of distinctions may help to keep them clear I thc d.ictic ' On. w,y of thinkina.bout rh.sc non-d€ictic us.g€s is to think of t.rms .s beins r.lstiviz.d to th€ t.xt instcsd of to thc situ.tion of uttcrance. In th.t wey, anaphoric usagcsc.n b€ s.cn to b. r.lgt.d to vrrious non.naphoric non-d.ictic us.s.s,..8. to shifts in dcictic interprciat;on duc to indirect discou.s€. Sce Anderson & Kccnrn, in press; also Fillmor., t98r.
6?
2.2 Desc/iPtiae aqqroaches
Deixis \4 2 )
D;frerent usagesoJ deiclic telnrs a. gestural |. deictit-. b. symbolic 2. nofl-deictic. c. non-anaphoric d. anaphoric
These difliculties are compounded when the phenomenon of deictic projection, or shifts from the egocentric centre, ate taken i11o account; and they are further multiplied by the interaction of tle semanticsof non-deictic categorizationsof (especially)spaceand timr with deictic modifiers. We shall now take up each of the five major categoriesof deixis in turn: person, time, place,discourseand social deixis,in order to illustratethe complexitiesthat arise.An appreciation of these complexitieswill indicate how involved and unexploredthe phenomenonof deixis really is and how the philosophicalapproaches to indexicalscan handle only a small proportion of these problems, Person deixis As speakers switch, so the deictic centre, on which the rest of the deictic system hangs, is itself abruptly moved from participant to participant. The difficulties that a Martian or child might have with such a system are neatly illustrated in the following Yiddish story: 2.2, I
A melamed [Hebrew teacher] discovering that he had left his comfortable slippers back in the house, sent a student after them q'irh a note for his wife. The note read: " Send me your slippers with this boy". When the student asked why he had written "your" slippers, the melamed answered: "Yold! lf I wrote 'my' slippers, she would read 'my' slippers and vrould send her slippers.What could I do with her slippers ? So I wrote 'your' slippers, she'll read 'your' slippers and send me mine". (Rosten, r968: 443-4)
? joke above depends' of course' on r ,vinson, in prep ) The Yiddish bel-weensource.andspeaker' which becomesimmefi" ii.,in.,io" a^."1.,oertinent if one reads aloud ''?-ol.u.t, tn. U"sicgrammaticaldistinctionshere are the categories were producing a componential nf frrst, second and third person. I f we r 968: 47q-8I ) of pronominal systems' see Lyons, (for whicn inalysis to need for the known systems would ii" i""aut"" that we seem person' speaker inclusion ( * S); for second for first clucially include: ( and for third person, speaker and inclusion +A); pcrson,addressee (see (-S, Burling, r97o: t4-r7; lngram, -A) exclusion "ddr"""." person is quite unlike first third note that 1978).It is important to to any specific correspond does not or s""orrd person, in that it (Lyons, rg77at 638). event Da icipant-role in the speech of manifestations obvious Pronominal systems, which are the most (Ingram, r978)' person,generallyexhibit this three-way distinction But some pronominal systems exhibit as many as fifteen basic pronouns(ignoring honorific alternates)by superimposing distinctionsbasedon plurality (dual, trial and plural), genderand so on. Here it is important to see that the traditional category of plural is not symmetrically applied to first person in the way it is to third: roedoes not mean plural speakers in the same way that ,rlPy means more than one third person entity (Lyons, I968: 277)' In addition, in many languages,there are two first person 'plural ' pronouns, corresponding '. ' and ' we-exclusive-of-addressec to ' we-inclusive-of-addressee perhaps it is but directly, in English manifested distinction is not This indirectly : for the contraction from let s to let's only seems felicitous if the ar is understood inclusively, as illustrated below (Fillmore, r97rb): (43)
Uc\
Let's go to the cinema ?Let's go to see you tomorrow
person deixis is reflected directly in the grammatical categories of person, it may be argued that we need to develop an so independent pragmatic framework of possible participant-roles, that we can then see how, and to what extent, these roles are grammaticalized in different languages. Such a framework would
Other languageshave pronominal systems much richer than the English one: in Japanese,Pronouns are distinguished also with respect to sex of speaker, social status of referent and degree of intimacy with referent, so, for example, the sccond person pronoun
note that the speaker or spokesman can be distinct from the source of an utterance, the recipient distinct from the target, and hearers
? Thus it can be argucd that in en8lish the s€ntence Bi i. i' to com. ia aor') g.ammstically €n.odcs (smongst oth.! things) thal th. recipient is nol th€ tarset (Biffic is), in contns. to Bi !, con. in ,ou wher. rccipieDr ind tarset are coincid€nt. (The exampl€ comes from Gazdar, 19794.) Rut see also
Although
or bystanders
distinct from addresseesor targets, and that sometimes are grammaticalized in non-obvious ways (see
such distinctions 68
exsrnplc (so) b€low.
o9
2,2 Des.liPriae aPPloaches
Deixis Aizi can be glossed 'you, addressedby this intimate male speaker' (U y e n o , r9 7 r: r6 -1 7 ; H a rada, 1976: S I I); and vi l l ageTami l hasuo to six singularsecondperson pronouns accordingto degteeof relative rank betweenspeakerand addressee(Brown & Levinson, I978:206). We shall return to some of these facts below when we considei socialdeixis.Here we should simply notethat thesevariousdistinctions are often encoded in verbal inflections in an isomorphic manngl. Sometimes, though, morphological agreement can make furthsl distinctions not overtly made by the pronouns themselves.A simplq example of this occurs in languagesthat draw their polite second person singular pronoun from their plural one, where there will be no overt distinclion betweensecondpersonsingular polite and second person plural pronouns. Here, 6nite verbs will agree in both cases with the superficially plural pronoun. But with neminal predicatcg the distinction is morphologicallymarked: such predicatesagreewit\ the real-world number of the referent (Comrie, t97s). So, in French, (45) is ambiguousas to whether there is one or more addressecs,bur (46) can only be addressedto a single addressee: (45'
(+6)
Vous parlez frangaisi Vous €tes le professeur?
In a similar sort of way, as Fillmore ( r 97 r b) notes, the editorial are of, for example, the New Yorher takes plural verb agreement(thus .oe ate not &e an\, but in the reflexive the underlying singularity shows through in phrases like as for owsef. Finally, as we have noted, pronouns are often used non-deictically; but the actual variety of usescan be sholvn to be far greater than one would easily imagine (W a ts o n , r9 7 5 ; Sa c k s ,r 9 7 6). In addition to pronouns and agreeing predicates, person or participant-role is marked in various other ways. As is well known to anthropologists,kinship terms, and other kinds of title or proper name, often come in two quite distinct sets, one for use in address (as vocatives in second person usage) and the other for use in reference(i.e. referring to individuals in third person role). Even when the lexemesare the same, they may be zred very differently in addressand reference(see e.g. Beck, rgTz: zgofr for Tamil usage), or only a sub-set of the referenceterms may be used in address.The latter is the casewith English kin terms - one can say both Fleary rs my uncle and Henry rJ n1,,couJrn, but only Hello, Uncle ! not, in modern .70
Australian English, Hello, Cousin! Further' in some (as opposed are up to four distinct sets of primary **" (a) a set of vocative terms' (b) terms: kin ]], .".."iut "r.lppt"-entary) first person possessivefeature 1"."i ri Ltt .'*nl"h have an implicit (c) a set of terms which have etc')' 1,""'-"""n '^, rnother's brother', 'your mother's brother (i'e mean l""""tj o..tl. o"ssessivefeature " possessivefeatures person third terms which have 1,"-i fal *i "f ""1 " Australian Some etc)' or her mother's brother', li"'-"-"rt" have ' triangular ' kin terms' such that a term X even l;;"";;"" i"altiaual .xis only usableifr is (say)the speaket'sfather a"nTtiig "" Such suppletive sets of terms Ini-tt.- uaar.""..'s grandfather' in what are essentially terms ih"r"for" "rr"od" person-deictic features (see Heath et al'' r982)' for reference,not address interesting grammatrcal category' agaln are an general in Vocatives phrases that refer to the addressee' underexplored. Vocatives are noun incorporatedasthe arguments orsemantically butarenot syntactically prosodicallv from apart set tather pr"al"ui"; thev are ll" !:d" ii " can be divided into Vocatives them' ofa sentencethatmayaccompany as in (47), and addresses' as in (+8) (Zwicky' r"ll", o. "o--ooses, r974)1
ttjltr"t." -"ndard
Q7) (+8)
Hey you, yow jvst scratched my car with your frisbee rr." ir"itt i". Madam, r'othing is as good nowadays
symbolic The distinction is precisely that between gestural and usages,applied in this domain. Summonses are naturally utterancecan be initial, indeed conversation-initial(seeSchegloff, I97za), and their own in (see Chapter acts speech 5) of as independent thought of right. Addresses are parenthetical and can occur in the sorts summons all Not occupy' can parentheticals locations that other in forms can be used as addresses(e.g hey you in (47) cannot occur that all be it may (+8)), although in the slot occupied by Madam addressescan be used as summonses(Zwieky, rg74l.79r) Vocative forms in difierent languages appear to be highly idiosyncratic and complex. Note that greetings,partings and various 'ritual' formulae (e.g. 6/ers1ou said after a sneeze) can be thought of as vocative in natute. A further point to note in connection with person deixis, is that whcre face-to-face contact is lost, languages often enforce a distinct mode of, for instance, self-introduction Thus, whereas in a
t, z l)escliqtioe aqPloaches
Deixis face-to-face meeting I can say -I'rzJoe Bloggs,on the telephone I rn15, say This is Joe Bloggs or Joe Bloggs is speahingwith third person v616 agreement(but seeSchegloff, rg79a); in contrast in Tamil we would have to say on the telephone the equivalent of Joe Bloggs om speah;ng, with first person verb agretment. ln conclusion, it should be noted that the two basic participanl roles, speakerand addressee,are not the only ones that can becorn. involved in grammatical distinctions. Various languages(e.g. the Philippine languageSamal) have demonstratives(discussedbeloyl that specify location near other participants - in this caseattending but not speaking parties, and present but non-participating parties. The Australian languageDyirbal has an entirely separatealternativc vocabulary to be used in the presenceof'taboo' kinsmen, whethrr or not they are participants (Dixon, r97z: 3zf). Moreover it 15 common in many languages(e.g- German; Hymes, 1974:56) f61 mother to say to father, in the presenceof little Billie, somethinglike:
(+s)
Can Billie have an ice-cream, Daddy ?
taking the point of view, for the purpose of vocative selection, of ths audience. These distinctions make it important that we do nor confuse, as is often done in the linguistic and philosophical literaturc, the categories of addresseeand hearcr. (Incidentally, note that as so often in the analysis of deixis, these various examples involve the overlapping organizations of the Iive basic categories of deixis: thus greetings usually involve tempofal, person and discourse deixis; demonstratives both space and person; vocatives both person ano social deixis: and so on.) In addition to speaker, addressee and audience (third person, being, of course, definable in terms of the first two), there are a number of further distinctions in person deixis that probably need to be made. We know that, interactionally, important distinctions are often made between overhearers, unratified vs. raiified participants, those of the latter who are addressees and those who are non-addressed participants, and so on (see Goffman, r976l. z6o; Goodwin, rg79a, t98I). Also, as we have noted, we sometimes need to distinguish speakc! from source and addressee from target. Thus if the air-hoitess announces
(so)
You are to fasten your scat-belts no\r'
sheis the speakeror spokesman,but not the sourceofthe instructions,
to be encoded in rhe use of the infinitive form' In ond this seems (e g a chief) nor 7l]""r1, i^ formal ceremonies,neither the source were necessarily (e.g. the i"-r"rt., lt"*?t !Ht-,"".-t:11-spirits) will perhapsbe found reflected distinctions )'li l.ril-" "fthese t",' ( ^- ^-^.L-another /"--u"t language or categories ol in the g.ammatica,l lrvinson' In PreP r' Time deixis Both time and place deixis are greatly complicated by the with the non-deictic conceptualinteractionof deictic co-ordinates these aspects of deixis in zation of time and space. To understand good understanding of the depth it is first necessary to have a general, but these topics semanticorganization of space and time in Fillmore, lie beyondthe scopeofthis book (seethough, Leech, I969; rg75; Lyons, r977a: Chapter I5). Briefly, though, the bases for systemsof reckoning and measuring time in most languages seem to be the natural and prominent cycles of day and night' lunar months, seasonsand years. Such units can either be used as measures, relative to some fixed point of interest (including, crucially, the deicticcentre), or they can be used calendrically to locate events in 'absolute' time relative to some absolute otigo, ot at least to some part of each natural cycle designatedas the beginning of that cycle (Fillmore, r 975). tt is with theseunits, calendricaland non-calendrical, 2.2,2
that time deixis interacts. Like all aspectsof deixis, time deixis makes ultimate referenceto participant-role. Thus as a first approximation (but see below), zoar can be glossedas 'the time at which the speaker is producing the utterancecontaining zora'. lt is important to distinguish the moment of utterance (or ins cription\ or coding rirrte(or CT) from the moment of reception ot teceioing liare (or RT). As we noted, in the canonical situation of utterance, with the assumption of the unmarked deicttc centre,RT can be assumedto be identicalto CT (Lyons ( I977a: 685) Complexities arise in calls this assumption deictic simultaneity). thc usageof tense, time adverbs and other time-deictic morphemes wherever there is a departure from this assumption, e.g. in letter writing, or the pre-recording of media programmes. In that event, a decision has to be made about whether the deictic centre will remain on the speaker and CT, as in (5I), or will be projected on the addresseeand RT, as in (52) (Fillmore, t975):
2.2 Descriqtiae aqqlodches
Deixis (5r) (s2)
a. This programme is being recorded today, Wednesday {p"1 rst, to be relayed next Thursday b. I write this letter while chewing peyote a. This programme was recorded last Wednesday. April 1sg,1o be relayed today b. I wrot€ this letter while chewing peyote
Linguisticconventionsmay often specifythe proper usagein situations where RT is not coincident with CT. For example, the Latin 'epistolary tenses'used pasttensefor eventsincluding CT, pluperfect for events prior to CT - in other words the deictic centre was projected into the fu tu re, the recipients ' RT ( Lakoff, r 97o: 847). Bur we shall have to skirt these issueshere (see Fillmore, r975). There are a number of aspectsof 'pure ' time deixis, where the.e is no direct interaction with non-deictic methods of time reckoning. These include tense (to be discussed below), and the deictic time adverbs like English zou.,, lhen, soon, recettly and so on. We can improve on our previous glossfor zoze a generalized conversational implicature ftorn the expression an -F-to the assumption that the mcntioned F is not closely related to the speaker. In contrast (ro5) *'ill only implicate ( r o6) if ( r o5) oc c ur s in t h e p a r t i c u l a r s o r t o f s e t t i n g i l l u s t r a t e d i n \ r oJ ) : (ros) (ro6) (roz)
The dog is looking very haPPY Perhaps the dog has eaten the roast beef A: What on €arth has happened to the roast beefl B: The dog is looking very haPPl
The implicature in ( I o6) is thlus palticulalizcd (see Smith & Wilson, rg79: r7rff for an account of horv implicatures like this might be calculated). Now most of the floutings or exploitations of the maxims are particularized, in that, for example, ironies require particular bac[ground assumptions to rule out the litcral interpretations Butitcould !'
is so i ns i s r r hat the d's ti n.r nn lndeed, Sper be. & W i l s on ( for r h.om i ns ) fundament.l that t w o qui te di fl er ent k i nds of r eas oni ns ar e €m pl oy ed, both of *hi.h clnnot be s ubs um .d w i thi o s s i ns l e th€or y of i m Pl i c a( ur e W hr l e clsim ids that the s tander d i m pl i c atur es ar e deduc ti ons fr om a s i nSl c m ai ,m ot Rcl€vanc., b.ck S.ound as s um pti ons and uhat i \ s ar d, the) s uPA.s r r har ' fisur es of sp€.ch ' ty pi c al l y i nv ok e i m ages and as s oc i a( i ons of a qu'r .
r 2b
claimed that metaphorslike (I08) or tautologiesIike ( to9) .ocrhrpsbe they convey in a relatively context - independent way 5"1.t *ft", (r"8) ('og)
England is a sinking shiP
r- .nv case,it is clear that the two dimensionscross-cut: for example' ,ttut arisefrom observingthe maxim of Relevanceare iiiirpfi.",ut"t only with respectto the l,r,i"ututir"a, since utterancesare relevant Thus B's response in (ro7) I]'ri"ul", topi" or issue at hand juxraposition to A's question in ( r o7)' l,l"flcates( t o6) by virtue of its that are both iut the important Point here is that thoseimPlicatures the maxims and are generalizedhave a special derivedfrom observing theory For it is these in particular which ;porrance for linguistic conteol of linguistic *iit U. nrta to distinguish from the se,nantic will be routinely associated implicatures such because expressions, with the relevant expressionsin all ordinary contexts' We have been using the term implicature loosely to refer to what (and we shall Gricewas careful to designateconaetsationalimplicature will misunderstanding where no the shorthand use to continue general be a to implicature the tertn intended in fact result).But Grice coverterm, to stand in contrast to what is said or expressedby the truth conditions of expressions,and to include all the kinds of pragmatic(non-truth-conditional) inferencediscernible. In addition thento convetsationalimplicatures, i.e those calculatedon the basis of the maxims, Grice envisaged an entirely dilTerent kind of nontruth-conditionalinferences,namely conventional irnplicatures. Conventionalimplicatures are non-truth-conditional inferencesthat are zo, derived from superordinate pragmatic principles like the maxims,but are simply attached by convention to particular lexical itemsor expressions-Grice providesjust two examples:the word Dal hasthe same truth-conditional (ot truth-functional) content as the word and, with an additional conventional implicature to the effect thatthere is some contrast between the conjuncts (Grice, I96I); the other example is the word thereJorewhich Grice holds contributes nothing to the truth conditions of the expressionsit occurs within (Gtice, r975: 44). Other examplesthat have been suggestedare the meaningsof eoen(Kempson, r975; Karttunen & Peters, r979) and yat (Wilson, r 975). Conventional implicatures can be expected to contrast with '7
3.2 Reoisions,Ploblems a\d applicatio,ts
C omt erca tional imq li ca tul e
systematic inferences of intriguing complexity (?.:*." & L€vinsor, r978), Indeed, there may well be a generalprinciple here: for ever., kind of mutually assumed constraint on laoguage usage, there will I a corresponding set of potential inferences that come about eithe. from the speaker observing or flouting the constraint. If this is sothere are many kinds of implicature yet to be discovered. The proliferation of speciesof implicature reminds us of the poinr made in Chapter r, namely that recent developments in pragmatie have as a consequencea 'hybrid ' view of the nature of meaning.Tt c total signi6cation or communicative content of an utterance may be divided, accordingto Grice (r975), along the lines indicated in Figute 3. r. Here we see that the truth-conditional content of an utterance (what, in Grice's special sense,is said) may be only a small pan of its total meaning, and as we explore other topics in pragmatics we shall continue to add further kinds of pragmatic inference to t[i5 inventory. GenetalizedQuontit!implicatures 3,2.4 One ofthe greatattractionsofthe notion ofconversationalimplicatu6, as was indicated in 3.o, is that it promises to simplify semantics substantially.For instance,the proliferation ofsensesof lexicaliterns can be avoided by noting that implicatures often account for differenr interpretationsofthe sameitem in different contexts; thus, aswe havc seen, the 'and then' interpretation of and can be attributed to the maxim of Manner. But to demonstrate how far-reaching the contributions from Grice's theory can be in this regard, we shall need to expressmore rigorously our understanding of how the maxims work, so that wc can accurately predict some range of implicatures, show how these could be mistaken for aspectsof the sez.reof the linguistic expressions involved, and demonstrate the substantial benefits that accrue to semantics if these mistakes are not made and the systematic ellects of implicature are taken into account. Here we shall concentrateon some generalizedeuantity implicatures, as these seem at present to be the best understood (thanks especially to the work of Horn, ry72' rg73 and Gazdar, rg79a). Following Gazdar (rg7ga), we shall consider two specific and rmportant sub-cases: scalar Quantity implicatures, and clausal Quantity implicatures.
of a set of linguistic alternates' or linguistic scale consists the samegrammatical category' which can ."p."""i"nt "f "",iu" by degree of informalioenessor semantic l.l".a ir, titt"u. "rder " an ordered set Such a scale will have the general form of by angled brackets) of linguistic expressions or scalar e9 €2, e3... e11tas l n:
(e, e", c", .'. en)
etc', in a sentential frame A we obtain +,hereif we substitute €r, or ?r
A!::l'-li::':i.',::::T^lli*if '1fl':ilt: ;5i*"tr-ro'-"a ""1::i":: A(e"),etc'but not viceversa'For example'takethe f,,'i;l(;J.nt"il"
all and some'These form an implicarional scale iriittt ot"t,in"* i-toil, (rIg) (i e' ,o.r>, becauseany sentencelike (rr8) entails is true (t l9) is true also) but not vice versa:
*han.u",
(lI8)
(r t8) (r r9,
All of the boys went to the Party Some of the boys went to the PartY
general predictive rule for l.,low, given any such scale, there is a if a speaker asserts deriving a set of Quantity implicatures, namely in the ordered that a lower or weaker point (i.e. a rightwards item that a highet implicates he then obtains, a scale on set of alternates) Thus fl o' obtain does set) the ordered (leftwards in point or stronger the not all that implicates (r conversationally I9) one if one asserts quite compatible it is though so even party; is this boys went to the wiih the truth of (rr9) that (Ir8) is also true, as shown by the non-contradictorinessof (r zo): (rzo)
Some of the boys went to the party, in fact all
We may formulate this generally as a rule for deriving scalar implicatures from scalar predicates: Scolatimplicatutes:Given any scaleof the form (e ,. eg,-e3' ' ' ' en), if a sieaker assertsA(?J, then he implicates : -A{",)' i{ l' ni""), then he implicates A(e2)and - A('r), and in "".""t" general,if he assertsA(en), then h€ implicates - (A(e.-I))' (A(e'-z) aod so on, uP to - (A(€r)) For the scalar implicature to be actually inferred, the expression that gives rise to it must be entailed by any complex sentence of which (r zr )
it is a part. Thus the utterance of
j.z
C oneer sotional imPlic atur e a natural language will contain reference to pragmatic components of meaning. Secondly, syntactic rules seem to be sensitive 1q .r.1, elements of meaning. Consider again, for example, (t r2) and (rr3) above: ,! takes the singular form ofthe verb, uous the plural forry1 but uozr does not take a plural noun phrase after the verb irre if it is being used to refer to a singular addressee. Consequently, as r,e. noted in 2.2.5, there is a morphologically encoded distinction berween the ooal that is genuinely plural and the oolJ that actually refsls 1, a singular referent, which appears just with nominal predicates (ses Comrie, r975 for cross-linguistic data here). \ow in some languagq5 which have additional honorific devices, the morphology requils5 that all items referring to a particular person be at the same honorific level. Thus most Tamil speakers would lind ( t r 6) ill-formed or unacceptable because the honorific level in the subject does not agree with the honorific level in the predicate: ( r I 6)
I ?talaivarcolraanka headman-honorific says-supcr-honori6c
Such honorific levelsare not alwaysencodedby regular morphologicr,, elementsor otherwise in the form of the linguistic items. So rulesof morphological agreement have to refer to the conventional implrcatures that specify the degree of tespect offered by the speakerto the hearer or the referent. But in that casesyntax is not autonomous with respectto pragmatics, a claim that most linguists *'ould resist. The only way to escapesuch a conclusion is to generatesentences with, for example, unacceptablecollocationsof honori6cs, and then have an additional set of pragmatic filtering rtles that ape standard morphologicalprocesses.The machinery for such a solution hasbeen explored by Gazdar & Klein (r977). It is not an elegant solution tn that such pragmatic filters would do a tremendous amount of work in a language rich in honorifics, of a sort that would normally be thought ofas standardlymorphologicalor syntactic(for the magnitude of the problem see e.g, Harada, I976, on Japanese) The issue i. important becausethe inter-relation between conventional implcature and syntax is one of the clearest areas where pragmatics impinges deeply on grammatical processes.In fact little thought has yet been given to the implicalions that such inter-relations have for the overall organization of a theory of grammar. Recently a new and different interpretation of the nature of r3 0
Retisions, ptoblems and applications
fall . ,---fional implicature, and the linguistic phenomena that t'-i,-',," (I975' & Peters Karttunen by put forward been ,.oo., has they suggest that the cote examples of the lllr'-rrr.nii"rrv usuallv described as Pretuppotit;on are reallv best llll'o."non vl'l.-^ conventional implicatures, and they outline a formal ^. *t,nt^ the framework of Montague grammar' The theory ]I",J"", Chapter 4, and we note its existencehere lliJ"r"".a extensivelyin terminological confusion ln fact the phenomena they lllj"i-i"" l"-"";u" ftuu. quite different Properties from the conventional (see Levinson, ,']Li.rru."" of items like those in (rro) and (rIr) in this book the term.will be retained,as Grice intended' ,Oin'il ""0 i^, rhe inf"tenc"" associatedwith such items' We have described the kinds of implicatu'e central to the literature' thete do seem to be additional nonbut, as anticipated by Grice, produced by different maxims or of inference kinds conventional For example, we shall see below that usage. language orinciples of that produces implicatures principle of itformativeness ,tr"r" i" u to the maxim of Quantity (Atlas those due sometimesin conflict with principles of politeness that produce & Levinson, I 98 t ), and there are
I implicat.d
I ll
non'convcntionally
convcnt ionally
con"crsationally,
non-conv.rsationally
r--_]
B.ncrally
perticul.rly
Figure 3.r Kixds oJ communicational conlent r! I... impl;cetcd
by the maxims of conv€rcition.
!3 r
C omtet sa tiona I impl ic a tul e A(er)' and ,S knows that not A(er),. Now empirically the inference from, for example, (r tq) is to the stronger . S f,""*"if,", ,S),, and this is a generar fact about th. s."l-ar i-oil""ir."".-b,i*t "",?, ui.a. of Quantity implicature seen
r",-,rJ"i.i"I"#; ;:'.'r::I#l# ;i:,,*":::fi: "]"!:il:
some stronger statement obtains. Why this should U. ."-"1., ,* of the many mysteries in this €lrea (see Atlas & Levinson, r98I for discussion). We now turn to clausal irr
rationis (with (tz6)
"
"risht "i_prin":::';;::l;":::r*
( reTea)rormu-
Clautal implicaturcr: .|/ S assprts some
complex expressron p which (i) containsan embedded sente;;;.";; iuio i""nn.. entails.norpresupposes q..and(iii),h;;:;';",'.;||"il",,r.
expressionr of roughry equal bfevity which contains g such that r docs entail or presut by' asserting p rather than r, S implicates ,h", h.tf"-"--q';^ ',1"n'
i.' h. i-;ri;;;;;;'i$:o:*"'*
*hether s is true or rarse'
The underlying intuition is this: if I use some linguistic expression that fails to commit me to some emb.dd.d ;;;;;";;:;';; ;$":".".. to anothet availablestronqer e:
;1""T ffi;::Jlil:i,l# ;"11'i,T",;H#; ::1,1:;;;{
posltton to make the stronger - sltatement' Thus if I say (Iz7) instead of (rz8). \r.Z-) (r2U)
I betieveJohn is awav I know John is awav
it is possible, for a ].-r:lt:*:,1* away. Or if I sav-
I know, that John is in fact nor
(I29)
The Russiansor the Ar havejust landedon Mars it is entairedthat one or ,n. o,n-ttt""n" party has landed on Mars' but I implicate ,rr* i, i"-p"""iui.,srrer is the Russianq and it is possible that it isnot the RuJil:;rfi;T k ow' This is becauseI have chosen ro utter ( r z9) in pr"f.r"r,"", fo. ",
*o"ra i,l"l;;; ;6',f :il"^.;j:,*: :l,H;?' ^gorwhich "o,"-r (rlo) The Russians and the I
Thustheuttering 0,"j""",,,:;;;;;,ni::1'j;::: " becausethe, whrch disjunct is ", true,
choice ofa disjuncrion has consequence that neither the -""_ of the r:mbedded sentences is entailed (or 136
j.z
Retisions, problems and applications
presupposed) by the whole' Since one could have chosen a stronger that did entail one or both of the disjuncts, one can be "*p."""ion taken not to be in a position to utter the stronger expression' Hence a statement of the form p or q generatesthe set of implicatures: {Pr, P - P, PC, p - q) i.e. it is epistemically possible that p, also that not ,r, also that q, also that not g. A sentence of the form p or g has these implicatures by reference to the availability of other sentenceslike p attd 4 or simply p or 4 which are stronger or more informative becausethey do entail p or qor both' Similar pairs of'stronger' and 'weaker' constructionsare illustrated i n (r3t): (a\ t,rongerIorm (r3r) 'P andq' 'sincet , q' 'a knowsp' 'a realizedp' 'a rcvealedP' p' 'necessarily
(b) wcahaform ', ot q' 'if r t h€nq' 'a belicves P' 'a ihouShtp' 'd saidP' 'po$iblv p'
(c) inqlicatu t ol(b) lPP,P - P, Pq,\ - 9l {Pp'e- p'Pq, P- q1 {Pt' P - t} (Pp' P - t) lPP' P - ,\ {Pl' P - Pl
Note that items that occurred in the list of scales in (rzI) with the appropriate scalar implicatures could reappear here with additional and stightly difierent clausal implicatures. For example, th€ utterance of possiblyt carries the scalar implicature 'not necessarily p'; but since porsr6/yp in contrast to tecetratily p do€s not entail p, there will also be a clausal implicature from the utterance of possrblyp to the efiect that the speaker does not know whether p is or is not the case (i,e. the set of implicatures {Pp, P - p} will arise). Or again, utterances of the form 2 or q will have the scalar implicature K - (p & q) and the clausalimplicatures {Pp, P - p,Pq, p - g}. So under rules (rar) and (rz6) even the more simple complex sentences may give rise to multiple Quantity implicatures. We are now in a position to show how the recognition of such generalized Quantity implicatures can help us to simplify semantics. At the beginning of thc Chaptet we outlined a pervasive problem in semantics: a large number of words seem to behave as if they either had a single sense that is protean (i.e' may change from context to context), or alternatively had a very large number of distinct but cl (necessarilyp, p, possiblyp> (ceitain that p, probablethat p, possiblethat p) (must, should,may) (cold, cool) (love, like) (none, not all)1r
To show that these regular scalar inferences are indeed implicatures we need now to produce a Gricean argument deriving the inference, for example, that (r 18) is not the case,from the utterance of (l rg) in co-operative cilcumstances. A short version of such an arqumenr might go as follows: (, z+)
The speakerS hassaidA(e");if S wasin a positionto stat€that a stronger item on the scaleholds - i.e. to asserrA(er) - then ,nor non€,. i.e. .somc'. is thc soutce " Th€ implic.tur. here, trom not alt ro of one inr.rprct.tion of €xlmpte (88) di$uss.d .bov€.
r3 4
he would be in breach of the first maxim of Quantity if he asEerted A(er). Sincc t the addressee assume thrt S i8 cooperating, an-J therefore will not violate the maxim ofQuantity I take it that S wishes to convey thst he i3 *itho,rt t "rtti.tg, tot in a position to state that the stionger item tr on the scale holds. and indeed knows that it do€s not hold
John says that some of the boys went
More generally,
(rzs)
and somewhat more explicitly:
(D (iD
S has said p There is an expression g, more informative than p (and thus q entailsp), which might be desirable as a contribution to the current purposes of the exch.nge (end here there isperhapsan implicit reference to thc maxim of Relevrnce) (iii) q is of roughly equal brevity to p; so S did not say p rather ihan q simply in order to be brief (i.c to conform to thc maxim of Manner) S f.""* ,tttt g holds but nevertheless uttered t he (iv) 6i"..ii would be in breach ;f the injunction to make his contribution as informative as is required, S must mean me' the addressee, to infer that S knows that q is not the case (K q), or at least that he does not know that 4 is the case (- Kq)
The important feature of such arguments to note is that they derive an implicature by reference to what has ttot been said: the absence of a statement A(er), in the presence of a weaker one, legitimates the inference that it is not the casethat A(er), via the maxim of Quantity' Another feature to note is that the inference is implicitly or explicitly rnodiffed; that is to sEy that from the utterance of epistemically A(e1)one actually infers 'speakerknows that not A(e,)', (symbolically, K : A(er)), rather thao just - A(ar). Hence what is conveyed is the speaker's commitment to his knowing that - A(cr)' This makes clear the pragmatic nature of the implication (referring to participants' knowledge states), and has imPortant implications for formalization' Following Hintikka ( r 96z) we may represent ' S knows that p' as Kp, and 'S doesn't know whether p' (or 'it is epistemically possible that p' ) as Pt. The two concepts K and P are then related just like the modal notions zecessaryand possiblc(seeAllwood, Andersson & Dahl, knows thatt iffit is epistemically P 1977. Iro)t i.e. Kp - t(S -' impossible, given what S knows, that not p). A final and related point to note about this Gricean argument is that it equivocates (and we here use the epistemic notation to good effect) between the inference and K - (A(er)), i.e. between 'S does not know that - K(A(er)) r 35
C ofloelsat;onaI impl i catul e For example, faced with the English words [o, and. warn. the semanticist might be tempted to claim that each covers a distinct and different (if approximate) span on some range of heat. It follows that it ought to be contradictory to say: (t32)
This soup is warm, in fact hot
just as it would be to say (rff)
rThis book is short, in fact long
But of course it isn't. The semanticist might then either claim that the meanings of natural language terms like warm are simply too vague or loose to engender contradictions, or he might ,h"t warm is in fact ambiguous betwcen a ,neither cold ,roa ",rgg"ri hoi, a"n". ( cl e a rl y n o t p o s s i b l e i n (r3 z )), and a.not col d,or.at l aast w arm, senre (the reading relevant for (r gz)). But the theorist who utilizes implicature has another kind of response: the scale of heat is not divided into discrete, labelled spans, but rather organized so that whar is hot is a special sub-case of what rs warm; thus a sentence of the form X r'sAot entails ,X is warm'. Consequently the terms form a scale (hot, warm), as in ( r z r ), and this predicts that to say X is wor^ conversationallyimplicates ,X is not hot'. But the implicature, tike all implicatures, is defeasible and thus is cancelled Uy,f," u"""rtion in (r3z) that X is in fact hot. Such an account b""o-. g"""*f "rn claim about the meaning of items in linguistic scales, " in g".,"ril items (when embedded in statements) enail their "rr.h Ioiver bounds (warm in a sentence will entail .at least warm') but ;^jl;_t" their upper bounds (roarrz implicates .not hot,). ^.r.ty The recognition of such scalar implicatures not only aids the understanding of the semantics of the general ,rocabulary in a language, but it also plays a crucial role in understanding ,f," :fogl""f, expressions in natural language, specifically the corinecti,rei the qusntifiers and the modals. The correct analysis of such terms is of course crucial to any semantic theory, but especially to those based on togical principles. Nevertheless understanding in this area was seriously hampered until the development of a theory of implicature, Forexample, it has long been noted that in many natur"l ;";;;", _disjunction appears to be ambiguous between an exclusive r-eading as in (r34) where it seems to be asserted that only one disjunct is true, and an inclusive reading as in ( r 35) where both iisjunct" U" ,*, ""r, r18
j.z
Reoisions,ptoblems and applicatiotrs
Mirabelle's in the kitchen or the bedroom The book is red or crimson the Thus the two hypothetical senseswould be: one and only one of the of or both V); one (exclusive, symbolically true disjuncts is account this v)' On (inclusive, symbolically disjuncts are true the addition of or both in (136) would serve to 'disambiguate'
(r34) (r35)
the sentence: Ronaldis a movie star or a politician,or both However Gazdar (r979a: 8z) argues that the ambiguity theory cannot be correct. For by standard logical equivalences the following correspondences can be established:
(rf6)
a (- P')& (- q') - U' v q) e( - t ) ( t q) & ( - {) ) v ( , & q) V Now given (I38) we would predict that there ought to be a reading of (t39) as ( I 4o) : (!37) (r:8)
(rl 9) (r40)
Ronald isn't a movie star or a politicran Either Ronald's not a movie star and he's not a politician, or he's both
But there does not seem to be such a reading, where what is asserted is that either both conjuncts are false or both are true't6 So the ambiguity view seems to provide the wrong predictions. But there's an alternative to the ambiguity claim, namely an implicatural account. For the scalar mechanism in (lzl) straightforwardly predicts that p ot q wlll standardly get interpreted as 'p V g'(i.e. exclusively) as follows. There is a scale (azd' or) where the sense of azd may be equated with logical & and the sense of oz with logical v (i.e. inclusive disjunction). Hence to sayp or q will implicate that the stronger item on the scale does not hold, i.e. - (p & q). But if we then conjoin the sense of p or q with the scalar implicature we *ith heavy ltrc6s oD 4 it is potsibl..o int.rprct (r39).r conv'vina " Although j ust thc s.cond diljun€t ot ( t {o}, as 6hown bv th€ po3sibi litv of !.vins : Ro'ald ttar OR o por.ician' hc't BOTH.This misht s€'m to rrSuc thtt ;sd't a ^oti or crnnot bc Gqu.tcd {'ith loeicd v, bcceuic - (, v q) & (, & s) is . contt.diction However, thi! sP.ci.l inte'action bctwe'n ltrcas 'nd ncgrtion i! quit€ 8e't€t.l - thus on. c.n !.v lto.rt do.'n't LIKE Mar'ho' h' LOVES .lthouah not likinS somcon..nt.ils not lovinS thcm. Th€ principlc hcrc ^r', secmsto b€ that sivcn s ecrlc (c,,er), if onc rssctts - A(cr) with heavv !r'ds on c', one c.n me.n A(c,). lt co'rld b€ cl.im€d that cr it hcrc nol |.t.d but .nrion d. For further rcmrrks, scc Gricc, 1978 and Horn' t978. ..rhu
r 39
C onaer t a t iona I imp lica tur e
g#i:;"9_ft:l.LY"ffi;;; "r,,i"E;;i;;;;;;;;" obtain the exclusive readins:
be considered univocat, and inelusiue, .h.-.J;:;;;r:;pli,",;on to a generatizedtype of implicatu; G*;;;;';rr", };fi,|* The modals provide another. crucial logical domain where impli_ cature provides essential insights. L".,y ,"a*n thought, as Horn (1973) has"""i".i"i'"? -Olr.r nicely documented, there has been considerable confusion abour the prop". i",.";;;;;;;r;;;rn.."r,,"" between the sentential operators zecessatg and possibleandthe related modals rrlst and may, etc. Thr in this way' consider (r4t); this seemsto imply (r;i.tobl"-t "ti"e 'I he gori a may in fact be a membet of the genushomo 1:1ll rt42) Thegori We might
a maynot in r"., u" ,n.,ni".'oi.i., *ii"Jo"_, "
thus led, as Ar roro wiig"";;;;be ;;,il:l,: fi lT.A":fi I,T,:l,,J.l ;,;i: ( r4 3 ) O p -9 _ 6 i.e. if p is possible,then rt is possible that not , But we will also want to allow is.necessarvmust also be possible,and thus we *,,1 I-lit:Ier rso adopr rhe axi om i n (r44): ",'-1": (r++) tre .: g p t.e. tt, ls necessary, then it is possible And as a matter of definition:
( r4s)
Dr,-O-p r.e. if p is necessary,then rt rs not posstbte that not ,
But putting these three axioms t,
atthe "u*,ai,,,i",,ii pl'J;:::::"* ff:li 1"";:l';:j::::iaterv (I+6)
(bv(r44)) - oe Jil, 9e t||, op+O-p t b v 'c"',iiiup".;,io., t , + ji) (iii) O p - - - Dp iUy from {r45) with
of double negations) (iv) therefore, trp -""3ortt"tt"n Clearly we cannot hold onto botl axtoms' and losicians have mostlv had the good 3lr.But r'vhat leads us to think that (I43) mrght be".;;;;":".;." a ralid inference?
trana e il";;;fi:T;'ff:J:[l;:"Jffi,,:ffi
r40
j.z
Reaisions,problems and applications
(the speaker knows that) the stronger 4p, will be to implicate that joe" not hotd, i.e. (or, strictlv, K - Cp)' But, by logical - !P not the case'i e' equivalence,if p is not necessary,then it is possibly in natural inference (r43) is a legitimate So - Op - I - 1t. a logical than rather implicature an viewed as is if it t".rfr.g., as: it therefore us rephrase Let inference. an utteranceofthe form Op conversationallyimplicates - !p' and thus bv logical equivalence,O - p A considerable amount of confusion in early attemPts to formalize modal logic might have been avoided if the distinction between logical consequence and conversational inference had been available
(r+z\
(see Horn, r 973). Turning now to clausal implicatures, note that p ot g has the following implicatures : (r+8)
ImPlicoturesof 'P or q' scalar : K- ( t &4) clausal:{Pr, P * P, Pq, P - q\
The clausal implicature explains the intuition that it would be extremely misleading to utter (r49) if one knew that Claude was in the dining room: (r+s)
Claude'seither in the dining room or in the study
for by (r48) the qtteranceof (I49) has the clausalimplicature that for all the speaker knows he may be in either room. Thus if one knows that t, one does not co-operatively convey that by stating t ot qi the use of the disjunction rather conveys that one has grounds for believing one or the other disjunct but does not know which. By accounting fot the fact that the utterance of a disjunction thus effectively conveys much more than its logical sense, the thcory of implicature once again makes it possible to retain the simple logical analysis of or as inclusive disjunction while accounting for the divergence from that analysis in actual use. Similar remarks can be made about conditionals. \try'hateverthe correct semanlic analysis of conditionals is (and there is now good reason to think that natural language if ... ther. c nnotbe equated with logical -r, the material conditional - see Gazdar, tgTga:. 8g-7\, a number of particularly troublesome features can be accounted for by means of implicature. By our rule (rz6) we can predict (r5o): r4t
C onaercationaI imp I ic a turc For example, faced with thc English words ,o, and warm, the semanticist might be tempt€d to claim that each covers a distinct and different (if approximate) span on some range of heat. It follows that it ought to be contradictory to say: (r32)
This soup is warm, in fact hot
just as it would be to say (!33)
iThis book is short, in fact long
But of course it isn't. The semanticist might then either claim that the meanings of natural language terms like warm are simply too vague or loose to engender contradictions, or he might th", parm is in fact ambiguous between ,neither a colJ oor"ugg""i ho*,, (clearly not possible in (r3z)), and a.not "..r"" cold,or,at least warm, sense(the reading relevant for (r3z)). But the theorist who utilizes implicature has another kind of response: the scale of h";;;"-;r, divided into discrete, labelled spans, but rather organizea i". is hot is a special sub-case of what is warm; "o,i"i.,n th; a sentence of the form X ri lro, entails .X is warm'. Consequently tt t..-"-fJ..r, u scale(hot, warm), as in (rzr), and this predicts that to" suy X i, *i_ conversationallyimplicates ,X is not hot,. But the implicature, like all implicatures, is defeasible and thus is cancelled Uy,f," in (r32) that X is in fact hot. Such an account ""..r,i"n b"io_" g"rr"."l claim about the meaning of items in linguistic ""r, " scales, i" g.r,".;i ;""1, items (when embedded in statements) cntail their loiver bounds (warm in a sentence will entail .at Ieast warm,) but merely i;;;;rr" their upper bounds (raarz implicates ,not hot,). The recognition of such scalar implicatures not only aids the understanding of the semantics of the general .,rocabulary in a language, but it also plays a crucial role in uiderstanding ii";l'"gi*r, expressions in natural language, specifically the corinecti.,es] the quantifiers and the modals. The correct analysis of such rerms is of course crucial to any semantic theory, but especially to those based on logical principles. Nevertheless understanding in this area was seriously hampered until the development of a theJry of i-pli;;;;". Forexample, it has long been noted that in many natur"l;;;;;*, -disjunction appears to be ambiguous between an exclusive r'ead-ing as in (r34) where it seems to be asserted that only one disjunct is true, and an inclusive reading as in (r 35) where both iisjunct" U" ,.r,., ""r, r38
3-2 Reoisiont, Ploblems and ap,licatiorrs (rf+) (r35)
Mirabelle'sin the kitchen or the bedroom The book is red or crimson
Thus the two hypothetical senseswould be: one and only one of the disjuncts is true (exclusive, symbolically V); one or both of th€ disjuncts are true (inclusive, symbolically v). On this account the addition of or both in (136) would serve to 'disambiguate' the sentence: (rf6)
Ronald is a movie star or a politician,oi both
However Gazdar (r979a: 8z) argues that the ambiguity theory cannot be correct. For by standard logical equivalences the following correspondences can be established : q) - Q 'v qle( - f ) &( *( ( p) ( q) ) v ( t & e) ( p v & c) Now given ( r 38) we would predict that there ought to be a reading of (r39) as ( r 4o) :
(r37) (r38)
(r:g) it;;i
Ronald isn't a movie star or a politician Either Ronald'snot a movie star and he's not a politician'or he's both
But there does not seem to be such a reading, where what is asserted is that either both conjuncts are false or both are true.ts So the ambiguity view seems to provide the wrong predictions. But there's an alternative to the ambiguity claim, namely an implicatural account. For the scalar mechanism in (t2t) straightforwardly predicts that p ot q will standardly get interpreted as 'p V q' (i.e. exclusively) as foltows. There is a scale (azd, or) where the senseof azd may be equated with logical & and the sense of or with logical v (i.e. inclusive disjunction) Hence to say p or q will implicate that the stronger item on the scal€ does not hold, i.e. - (, & q). But if we then conjoin the sense of p or g with the scalar implicature we $ Althoush $'ith h€.vy strcs! on ot it is Po.sibl€ to intcrprct (r39).s conveving just th! scconddis.iuncrof(rro), asshown by the po$ibilityofs.vins: Ronard k 't a nooi. t,a. OR a Politi.ia^, h.'t BOTH This miaht sccm to .rauc rh.i o' cmnot bc cqu.tcd sith logicrl v, b€crure - (t v 9) & O & ?) i' . contr.dicrion. How.vet, ahisspcci.l interaction b€tw.cn strcts.nd ncgation ir quit€ gen€ral - thus on. c$.ty Ha/rr do.tn'', LIKE Ma ha,h. LOVES i.,, .lthough not likins 3omconc.nt.ils not loving them. Thc principl. h.t se.ms to be that 8ivcn. rc:lc (c1,c1), ifonc:rscrts ' A(cJ *ith h.avv.tres.' on .,, one crn mc.n A(cr). h could b€ claim€d thtt cr is hcrc not 6rd but rzrh.r n ntion d. For futthcr i.mirki, 6cc Gric., t97t rnd Hom, 1978.
r 39
C onaelsa t ionaI impI i ca tu/ e
j.z
obtain the exclusive readins:
Reoisions,problefls atd applications
:..;t;;;;;;;ffi#;ri"lff"r,:"?";i"Y"1;J1"""fi :
be considered uniaocal, and exclusive interpretation being due to a generalized,u'-"-"'::':"' ,the pe ol rmplicature (see Gazdar, r979a: z8-8:). The modals provide another-crucial logical domain where impli_ cature provides essential insights. Orr"r m"ny fo",."l thought, as Horn (1973) has nicely "".,,uri.i-oi documented, therc has been considerable confusion about the prop"" ;",".or"i"ii"n between the sentential operators "'i,n...,"o"rr zecessaty and possibleand the related modals rzzs! and may, etc- The probtem-s *ir" i.' iii, *"r.'i".ria* ( r4 r); th i s s e e msto i m p l y (r4 2): !tnr) (t42)
The gorilla may in fact be a member of the genushomo
Thegorillamaynot in f"., b", ,.,..,';;;';;i"
*"Jiiin"_"
might thus be led, as Aristotle was on occasion, to take I: the following as a basic axiom (where D = necessarily,O : possibly): (i43) o t* o -, i.e. if p is possible,then it is possible that not , But we will also want to allow
bepossibre, *" *il,"",3xiT;::lTff:"ill;*, "* ".Jir,",' ( r44) Do
op
- is necessary, i.e. ifp then it is possible And as a matter of definition: (I4s )
op lil, (r, !e or,Jorp
an utteranceofthe form Op conversationallyimplicates - Op' and thus by logical equivalence,O - p A considerable amount of confusion in early attempts to formalize modal logic might have been avoided if the distinction between logical consequence and conversational inference had been available
(rtl')
(seeHorn, rg73). Turning now to clausal implicatures, note that p o/ { has the following implicatures: Im,licatutet of '9 or q' scalar : K- ( ?&q) clausal:( Pt , P - P, Pq, P - ql The clausal implicature explains the intuition that it would be extremely misleading to utter (r49) if one knew that Claude was in the dining room:
(r+8)
(r+q)
Dt r - g! p r.e. ifp is_necessary,then it is not posstDlethat not ,
But puning these thra" u*io-, tog.ther we arrive immediately absurdity that, if t is necessary, then it is ";; "";";;"-,"',, (rq6) *
(the speaker knows that) the stronger 6p, will be to implicate that j*,,rot tota, i.e. - !p (or, strictly, K - !p). But, by logical if p is not necessary,then it is possibly not the case' i'e' -"quiual"nce, p. So (r43) is a legitimate inference in natural Ap -.9 rather than a logical t".,g|r.g", if it is viewed as an implicature inference. Let us rephrase it therefore as:
at the ".
(bv(,44)) t b y t , + t ii
(iii) O p - - np 1by "oii.upo";tion from (r+s) with of double negations) (iv) therefore, Do -t[oott""t"n Clearly we cannot hold onto botl axtoms' and logicianshave mostlv had the good ;;;;;;." *"0" us to thinlt that (r43) might be a,atid inferenc. ?*r,:::^11:,
Dand e f; ";;'i ;: T;,'ff il:::J,I,1"""i:: iI".,"#;
Claude'seither in the dining room or io the study
for by ( r 48) the utterance of ( I49) has the clausal implicature that for all the speaker knows he may be in either room. Thus if one knows that p, one does not co-operatively convey that by stating, o/ q; the use of the disjunction rather conveys that one has grounds for believing one or the other disjunct but does not know which' By accounting for the fact that th€ utterance of a disjunction thus effectively conveys much more than its logical sense' the theory of implicature once again makes it possible to retain the simple logical analysis of or as inclusive disjunction while accounting for the divergence from that analysis in actual useSimilar remarks can be made about conditionals. \y'y'hateverthe correct vmantic analysis of conditionals is (and there is now good reason to think that natural language it .,. thet eannotbe equated with logical --+,the material conditional - see Gazdar, r91ga: 8g-7\, a number of particularly troublesome features can be accounted for by means of implicature. By our rule (rz6) we can predict (r5o):
140
r4l
C onztersa tional impl i ca tw e (t5o)
Clautal imrlicatules ol,if p then q' { P b ,P-p ,p q ,p -q l
Therefore to say (r5r) is to implicate that one does not have any reason to think that Chuck has actually already got a scholarship or to think that he will defnitely give up medicine: (r5I)
lf Chuck has got a scholarship,he'll give up medicine
Some have thought that the hypothetical implications associatedwith the use of if .-. then should be built into the meaning of the conditional. But the problem is that such implications _ like all the others we have discussed in this section - are defeasible. Tbus if we embed (r5r) in the discourse context indicated in (r5z), rhe clausal implicatures evaporate:
( rsz)
A: l'vejust heard that Chuck has got a scholarship B: Oh If Chuck has got a scholarship, he;ll give up -dear. medicine
Hence consistently associatedbut neverthelessdefeasibleaspects of the meaning of the conditional can be explained by implicature. If such hypothetical implications were built into the semanticsof the conditional, the usage in (r5z) would force us into yet another ambiguity claim. Theexistenceof a number of differentkindsof euantity implicature, including scalar and clausal, gives rise to a projection problem for implicatures,i.e. the implicatures ofcomplex expressio.rs not be equivalent to the simple sum of the implicatures of all -"y the parts. Consider, for example, the fact, discussedin 3. r, that implicatures can be suspendedby explicit mention in f-clauses, as in: (t53) Some,if not all, of the workerswent on strike Here there should be a scalar implicature (r54) due to rorr.e(by.ule ( r z r) a b o v e ): (tS+)
(all the workerswent on strike) K i.e. S knowsthat not all the workerswent on strike
But there should also be the clausal implicature (rSS) due to the phrase if not a// (see the prediction in ( r5o)):
( rss)
r42
P(all the workers went on strike) i.e. it is possible, for all S knows, that all of the workers went on strtxe
3-2 Reoisions,prcblems aqd applications (I55) are inconsistent, and it Now the two implicatures (I54) and implicature (r55) effectively clausal seemsintuitively clear that the observations' ( cazc€lsthe scalar implicature r 54). On the basis of such mechanism (or projection cancellation) Gazdar (rg79a) sets up a Let the follows'r6 as dcsigned to model implicature cancellation, adding by assessed com;unicative content of an utterance U be to sequentially the distinct semantic and pragmatic inferences of U the that the context C, where C is understood to be the set of beliefs the speaker is committed to at the point when U is uttered' On ,ri,"rurr." of U, first the entailrvents (or semantic content) of U are added to the context (here we might add: only if they are themselves consistent with all the propositions in C; otherwise Participants will an.lyse U as a Quality flout and expect an appropriate implicature)' Next, all the c/azsal implicatures are added that are cot'JtSten'with the content of C (now augmented with the entailments of fI), inconsistent clausal implicatures simply being rejected and not added to the set of propositions in context C. Only now c n scalar implicatures be added, just in casethey in turn are consistentwith the context as already incremented by the entailments and clausal implicatures of U. This mechanism will correctly predict that the scalar implicature (I54), being assessedafter the clausal implicature (r 55) has been added to the context C, will be rejected as inconsistent with what has already been accepted. Thus on Gazdar's account deleasibility is captured by making implicatures acceptable only if they are consistent with entailments and other implicatures that have priority. Note that this mechanism also explains why implicatures may be overtly denied as in (r56): Some of my best friends are drug-addicts, in fact probably all
(t 56)
for the entailments of the second clause, being added to the context first, will cancel the implicature due to Jotte. Gazdar's mechanism appears to be perfectly general and to operate on sentences of arbitrary complexity. For example consider (r 57): '
G.zd..'.hod€l is. mod.l of whrt individual sP..k€r. ar€committed to' ind rhe wry in *hich this cencels implicrturcs. It do€! not csPtur. int.r.€tiv. .3p€ctsof what rnly bc mutually t.kcn fot 8r.ntcd.s convcrlrtion prGc€d3 It .ko makcs th. wronS pr.dictioN whh resPcct to tropcs or .rploitrtions of the mrximt, whcrc implic.tures oftcn crnccl cntrilm€nts. But it sccms to op€r.t€ w.ll within th. limit€d, but imPortant, domain of gcncralized conv€ruationcl irnplic.turc..
r 43
C onverca tional impIi ea tur e (rSZ) (rS8)
Someofthe Elgin Marbles are fakes,and either the rest ofthem ar€ too, or they're inferior originals (i) K - (all of the Elgin Marblesare fakes) (ii) P(rheresrof the Elgin Marbles are fakestoo) (iii) P (the rest of rhe Elgin Marbles are fakes too) (iv) P(the rest of the Elgin Marbles are inferior originals) (v) P (the rest of the Elgin Marbles are inferior originals) -
Here (i) is a scalar implicature due to some, and the rest of the implicatures are clausal ones due to the disjunction in the second conjunct of(r57). Note that implicatures (i) and (ii) are inconsistenr, so the scalar implicature (i) will be cancelled, and the sentence as a whole will have just the implicatures (iit(v). We also now have some account of the problem raised by Sadock (1978: zgr) and discussedabove in connection with example (75;. Implicatures are saidtobe non-dctachaDleand therefore not defeasiUte simply by substituting a synonymous cxpression for the expression that gives rise to them. But consider that the meaning of rightmost scalar items like sorzeis consistent with leftmost items iike a/iin such scales. It follows that sorzehas the semantic content paraphrasable as 'at leastsome'or'some if not all '. Therefore (r59) and should fr6o) bc synonymous, and ergo by the principle of non-detachability thcy should sharethe sameimplicatures.Dut they don't. However, we n