Chierchia Not come Children x[-come(x)] Childrenk (after the kind type-shift that is obligatory to acquire argument status) Children didn’t come. childrenk x[-come(x)] <e,t> e -come(childrenk) =-(come(childrenk)) =-(Qx[R(x,childrenk)&Q(x)] x[come(x)] ) =-(Qx[R(x,childrenk)&Q (x)] x[come(x)]) x[come(x)] come(x)
Krifka (simplified) x[-come(x)] x[children(x)] Come Children Children didn’t come. x[children(x)] x[-come(x)] <e,t> <e,t> x[-come(x)] x[children(x)] = -come(x[children(x)]) not allowed in standard Montague grammar!!! = -(come(x[children(x)])) =-(Qx[children(x)&Q(x)] x[come(x)] ) =-(Qx[children(x)&Q (x)] x[come(x)]) x[come(x)] come(x)
Conclusion Carlson builds type-shifting into predicates. Chierchia applies local type-shifting to nouns with a small detour via kinds. Krifka applies local type-shifting to nouns. > Narrow scope is always accounted for by local type-shifting and doesn’t presuppose that bare nominals always refer to kinds. > General constraint on covert type-shifting: apply it as locally as possible.
The empirical validity of a locality constraint on type-shifting
Do bare nouns take wide scope? If they do, there is no reason to assume a locality constraint on type-shifting... The answer... YES! NO! Min Que The rest of the world (or close to it…) English (Carlson), Spanish (Espinal and McNally 2010 and references therein), Hungarian (Farkas and de Swart 2003), Russian (Geist 2010), Albanian (Kalluli 2001), Hebrew (Doron 2003), Hindi (Dayal 2003, 2004), Mandarin Chinese (Yang 2001, Rullmann & You 2006), Indonesian (Chung 2000, Sato 2008), Javanese (Sato 2008), Turkish (Bliss 2003), Brazilian Portuguese (Schmitt & Munn 1999)
How to go about testing scope? > A first attempt Every boy read a book. wide a. There is a book that every boy read. b. Every boy is such that he read a book. narrow Why is this not a good format for test items? Because every situation that makes a. true will also make b. true.
How to go about testing scope? > A better attempt John didn’t read a (single) book. wide a. There is a book that John didn’t read. b. John read no book. narrow Why is this a better format for test items? Because a. can be true in situations in which b. is not true.
A small classroom experiment Deze diagnose heeft ons doen inzien waarom hij sommige dwangideeën heeft, zoals altijd de eerste willen zijn (op de trap, in bad, aan tafel...) of woedebuien (omdat hij dingen niet begrijpt) of irrationele angsten (zoals steeds denken dat er bijen rond zoemen, terwijl het soms maar een grasmaaier is). Hoe ouder hij wordt, hij is nu bijna acht jaar, hoe duidelijker het autisme wordt. Does this necessarily mean that he doesn’t understand anything? omdat hij dingen niet begrijpt because he things not understand Ik vind het absoluut niet leuk dat hij moet huilen vanwege mij. En dat is wel een aantal keren op een dag, omdat hij dingen niet mag of dat hij juist iets moet (naar bed gaan bijvoorbeeld). Ik weet dat het er bij hoort, maar leuk is anders. Nu kan ik er weer even tegen. Does this necessarily mean that he’s not allowed to do anything? omdat hij dingen niet mag because he things not may
Setting-up the bare nominal test items A. B. A. B. This last sentence is truth-conditionally only compatible with a wide scope reading of colleagues. Task: judge the naturalness of the last utterance with respect to the rest of the dialogue on a scale from 0 to 5. Rationale: subjects should not accept a continuation in which Flynn contradicts himself.
Further design of the experiment An experiment that would only look at the acceptability of bare nominal items would be meaningless. Why? Because we wouldn’t know what the numbers meant. Our baseline Given that we were testing whether bare nominals could scope above negation, we needed an item that could not. > Negative Polarity Items
Further design of the experiment Experiments also need control items and fillers. Why? Control items are used to check whether people are actually sensitive to the phenomenon one is testing. Filler items are used to try to distract subjects in such a way that they don’t discover what the experiment is really about. Our control items > Singular indefinites Our fillers > See example
Further design of the experiment > Overview of the number of items: 2 NPI items 2 Singular indefinite items 3 Bare plural items 5 Fillers > Participants and procedure: Questionnaire was put online. Included a number of questions that would allow us to weed out non-native speakers. Total number of relevant questionnaires: 63.
Results: statistics Paired t-tests There’s a (significant) difference between the NPI items and the BP items. There’s a (significant) difference between the BP items and the SI items. There’s a (significant) difference between BP1 and BP2. /
Do bare nouns take wide scope? There is ground to assume that bare nouns can take wide scope. > This means that the general narrow scope behaviour cannot be derived solely by forcing covert type-shifting to apply locally. > Covert type-shifting turns out to be less constrained than might seem at first sight.