54 S T R U C T U R EASN D B E Y O N T )

marginal.If one switches to subjunctive, as in (51b), the presentationalsenterice

An important set of constructionsthat are known to license NPIs are questions.
acquiresan optative meaning (I hope/wish that there will be . . . ) and the sentence
And it seemsclear that scalar items in questionstypically get an inclusive interpre-
becomesgrammatical. Thus the presentationalfutureplus the subjunctivecan license
tation. For examPle:
NPIs.Now, if we embeda scalaritem in the constructionin (51b),the interpretation
onegetsis the inclusive one:
(48) Questions
Did Johnor Paularrive?
(52) Inealis constructions
a. #No; theybothdid.
a. Li ci sari qualcunoche saingleseo francese.
b. Yes,theYbothdid'
In-that-placetherewill be somebodywho knowsEnglishor French.
b. Ci sara'qualcunochesappiaingleseo francese!
If the question was interpreted.as "Did John or Paul but not both arrive?" the answer
(I hope)therewill be somebodywho knowsEnglishor French.
in (48a) would sound more natural than it does. The fact that, if they both arrived,
c. #I don't hopethat therewill be somebodywho knowsboth EnglishandFrench.
the natural answer is yes suggeststhat or is being interpreted inclusively.
Considernext the caseof modalities of permission:
Sentence(52a),with the indicative embeddedunder the future, is a (plain) predic-
tion and it has a prominent exclusive construal (if in the relevant place we find only
(49) Modalityof Permission.
to smokeor drink. peoplewho speakboth languages,it is unclear whether the prediction we made was
a. It is permitted/legal
(Expectation:it is permitted/legalto do both) accurate).With the subjunctive embeddedunder the future, the favorite construal of
or becomesinclusive, as witnessedby the oddity of the continuation in (52c).
b. You maYsmokeor drink.
In conclusion, a rather solid generalization seems to emerge from these
considerations:
If we seeposted somethinglike (49a), we don't expect to be fined if we do both'
Something similar happenswith (49b). However, this also readily admits of a read-
("You (53) Generalizationon SIs
ing in which or is construedas having wide scopewith respectto the modal
case we get a (strong) exclusiveness im- (Ordinary)scalarimplicaturesare suspended
in the contextsthat licenseany
may smoke or you may drink"), in which
the "inclusive" reading if "smoke or drink" (asa Neg Pol or Free-ChoiceItem).
plicature. Intonation matters. It favors
form an intonational phraseof their own. Speakersseemto be a little less clear about
So,we have, on the one hand, a class of specializeditems (identified by a specific
imperatives, which nonethelessought to be mentioned in this context becausethey,
morphologicalmakeup) whose distribution appearsto be sensitiveto contextswith
like the modals, are known to be licensors of (free choice) any.
a common semanticcoloring (an extensively studiedfact, which happenswith re-
(?) markablecross-linguisticstability). We now also seethat a certainwell-established
(50) Imperatives
pragmaticphenomenon,the calculation of scalarimplicatures, is also systematically
Getme Paulor Bill.
sensitiveto those very contexts (or to some closely related ones).Perhapsthe extent
is a to which generalization(53) holds hasn't been fully appreciatedin the literature on
The question is, If I get you both do I fulfill the order? It seemsto me that there
thesetopics. One would like to understandexactly why this is case.In addressing
sensein which the answeris Yes.
this question, I am going to stick to Downward Entailingness as the property that
The final caseI'd like to mention is the inealis mood, which has also been iden-
in comesclosest to providing a characterizationof the relevant contexts (while being
tified as relevantto NPI licensing (see,e.g., the discussionin Giannakidou 1997)
To check the relevant facts, we have awareof its problems).
interaction with suitable embeddingpredicates.
do'
to go to a languagethat is a bit richer in its morphology than English. Italian will
Consider the examPlesin (51): 2.3. Implicaturesof negativecontexts

Taking stock, we have gatheredpreliminary evidence in favor of the following two

(51) a. ??Cisara'qualcunocheha mai sentitonominarePavarotti.l0
ideas:(a) implicatures are (or, at least, can be) addedlocally and (b) once an impli-
Therewill be somebodywho everheardof Pavarotti
caturehas been added, upon being embeddedwithin a DE element,it typically gets
b. Ci sara'qualcunocheabbiamai sentitonominarePavarotti!
(I hope)therewill be somebodywho has-suslheardof Pavarotti. removed.Now a more careful consideration of the facts shows, I think, that what
actually goes on in DE contexts is not simple cancellation of implicatures.What
actually happensis that implicatures are in a way "recalibrated." Consider, for ex-
In (51a) we have a presentationalconsffuction, "there will be," with an embedded
ample,sentenceslike the following:
indicative future. The embeddedclause contains the NPI mai'evet,' and the result is