2004 - Scalar Implicaturen Pi - 20

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 1

JL^L^|\ rrvrrLrL^rurEJ/ r\rLrrKl lr rnEl\(JMtNA/

/b S T R U C T U R EA
SN D B E Y O N T ) ANU lHt SYNTAX/PRACMATICS INTERFACE 77

Second,this domain-extendingfunction must be universally closedat somepoint in alternatives(as he is analyzing the Hindi "even one" construction). My condition is
just equivalentto his since being stronger entails being less likely. Second,Lahiri
the derivation.Crucially, such an operationof quantificationalclosureis subjectto a ,
condition (or, if you wish, carries a presupposition).Its result must lead to some- identifiesNPI scopeassignmentwith associationwith focus. This, ior Hindi, is plau-
thing stronger than the corresponding meaning with a plain indefinite. I will make sible sinceNPIs in Hindi are literally formed with the use of the focus particle even.
this idea more preciseshortly. Meanwhile, let me enunciatethe principle: But if we wanted to generalizeLahii' s proposal to other languages(which Lahiri
is
careful not to do), we face a problem. For although NPIs *" rorrrputiblewith focus-
(113) Strengthening/blocking ing, they do not require any special focus assignment(in the form of, say, a special
Domainexpansions mustbe universallyclosed.Suchclosuremustleadto stresscontour).The relevant alternativesare not evocatedvia focus. It's just a texi-
with respectto the meaningof the plain indefinite.
strengthening cal property of any that it contrastswith ordinary indefinites.2a
I mentioned also that with my proposed modification, domain expansion is
no
What quantificational closure over domain-expandingfunctions achievesis twofold. longer an absolute requirement. More specifically, it may just happenthat the base
First, we are freed from having to assumethat NPIs require a specific way of extend- domain(i.e., what I have been indicating as D) is already, ro to-rp.ak, the largest
ing the domain. Second, we can compositionally implement pragmatic strengthen- reasonable domain. In this casethe only "extension" of D to be consideredwill UL O
ing as a condition on quantificationalclosure.Let me illustrate.Supposewe have an itself (which is allowed since g is not required to be properly increasing).This, as
we
NPI in a positive sentencesuchas any manwalked in.lts basemeaning,accordingto saw, is in keeping with our intuitions. To put it differently, an NpI, in the present
our hypothesis,would be as in (114a).If we apply closureto it, we get (114b): modified view, does not necessarilysignal an actual, specific domain expansion;
it
signals,rather, willingness to expand the domain (an effective way of couching
its
in')
(114) a. Some*ipy'(man')(walked role,which I owe to A. von Stechow).Notice that even in the limiting casein which
in')]
b. VgeAIsomeulny'(man')(walked thereis no suitable expansionaround, the proposedsemanticsstill ,.ru", a useful
in')
c. somep'(man')(walked pu{pose'namely, that of signaling that the indefinite must wind up within the scope
of a DE operator (and hence it reducespotential ambiguities). Bu1 such an effect is
I am assuming that no form of quantification is ever possible without a domain re- not stipulated as the primary role of NPIs. It is rather a byproduct of how their se-
striction, and universalclosure is no exception.So A in (114b) rangesover the do- manticsworks out (and, more specifically, of the requirement that the use of NpIs
main expansionsthe speakermay be willing to entertain.Formula (115c) represents, leadsto formal strengthening).
instead, the sentencemeaning corresponding to any man walked in, with the plain One further remark: I think that the proper way of thinking of (113) is in terms
indefinite replacing the NPI (i.e., the competitor).Now, (114c) asymmetricallyen- of something like "blocking", as is familiar from much work in morphology. The
tails (1 l4b).23Hence,the condition on strengthening,which closuremust satisfy,is presuppositionto which NPIs are subjectputs them in an "elsewhere"relation with
not met. Consequently,the sentenceis ungrammatical' indefinites,on the assumptionthat the former are parasitic upon the latter. Intuitively,
Supposenow we embedour examplesentenceunder someDE operatoras in ir NPIs are a more marked version of indefinites (not dissimilar from, say, the relation
is false that any man walked in. The basemeaningis (115a).The presenceof nega- betweenpronouns and their clitic counterparts).This brings about a sort of blocking
tion gives us anotherpossible site for quantificationalclosure,that is, above nega- effect, whereby the use of NPIs is blocked by the indefinites whenever the formei
tion. The result is given in (115b). don't give rise to a communicational advantageover the latter. Distributed morphol-
ogy may provide a good framework for fleshing this out. For the time being, I am
(115) a. r Solnegp;
(man')(walked
in') not in condition of being less vague.
in') ]
b. VgeA [r Somegpv'(man')(walked Let me now try to be more explicit on how strengtheningis formally implemented.
c. 1 somep'(man')(walked
in') I will first describe its semanticsand then its syntax. (Readersnot interestedin for-
mal details may safely skip this paragraph).Imagine introducing in our logical form
Formula (115b) says,"For any (reasonable)way of expandingthe domain of quan- a variable binding operator oo with the following semantics:
tification it is not the casethat someman in the extendeddomain walked in." Clearly,
(115b) asymmetricallyentails(115c). Hence,the competitor of the NPl-sentenceis (116) Strengthening (/blocking)
weaker and the condition on closure is met. Moreover, the sentenceis grammatical. = VgeAll0ll,if VgeAllgllentailsllQ'll;ottrerwise
llOo,g0ll is undefined,
llOa,g0ll
Obviously the basic insight is just K&L's. However, I borrow from Lahiri (1998) where$' is identicalto q, with all occurrences of g removed.
the idea that NPIs must be assignedscope(in our version, the site of quantificational
closure). And, like him, I make this operation contingent upon strengthening.There Suchan operator is what we would use to close off a domain-expandingfunction.
are two differences with respectto Lahiri's proposal, one of form, the other of sub- Thesemanticsgiven in (116) is still a first approximation,which might be improved
stance.First, Lahiri requires that the sentencewith the NPI be "less likely" than its on in many ways,2sbut it is sufficiently precise to give the reader something to

You might also like