On 10/14 at 10:00am at ILC N400, Tyler Knowlton (UPenn) will give a talk entitled “Non-conservative quantifiers are unlearnable”. Tyler will be joining us over zoom, and everyone is welcome! The abstract of the talk and the zoom link can be found below.

Join Zoom Meeting
https://umass-amherst.zoom.us/j/91355871706

Title: Non-conservative quantifiers are unlearnable
Abstract: Perhaps the most well-known typological universal in semantics is the “conservativity” of quantificational determiners: roughly, the observation that only the noun phrase a quantifier combines with matters for the truth of the sentence. For example, only fish (and their properties) matter to the truth of every fish swims. Accordingly, no language has a “non-conservative” quantifier like equi, where equi fish swims means “the fish and the swimmers are numerically equivalent.” The hypothetical equi fails to be conservative because the number of swimmers also matters. This robust cross-linguistic generalization has been argued to reflect a fundamental fact about quantifier semantics, as opposed to general cognitive constraints, communicative pressures, or historical accident. In particular, conservativity has been used to question the standard, relational conception of quantifier meanings. But if conservativity has a specifically linguistic source, then non-conservative quantifiers are predicted to be unlearnable. I’ll present a series of experiments that bear out this prediction. Adult participants fail to learn two novel non-conservative quantifier meanings, even when explicitly taught, but succeed at learning their conservative counterparts. Moreover, this learnability asymmetry disappears when the intended meanings are taught as novel verbs instead of novel quantifiers. These results suggest that the conservativity universal is tied to learnability, which supports semantic theories on which conservativity reflects a deep fact about the human language faculty.