Interface transparency and the psychosemantics of most

Experimental investigations of ambiguity: the case of most

“…3 we present novel data from two experiments that support the view that most is uniformly a superlative construction. Section 4 discusses the implications of our results for the theory of most, differences between the determiners most and more than half that arise from the experiments, and also previous work on most-notably the work of Lidz et al (2011), who conducted a very similar experiment but reported very different findings about the nature of most than we do here. We address several concerns regarding the experiment in Lidz et al (2011) and the conclusions that were drawn based on this work.…”

“…Moreover, Kotek et al (2011a) have argued that bare most in subject position is in fact ambiguous between a (preferred) proportional reading and a (latent) superlative reading. This contrasts markedly with the theoretical framework of Lidz et al (2011), who conduct a sentence verification study of bare most to adjudicate between different ways of describing the truth-conditional import of this form. Lidz et al assume that bare most is unambiguously proportional and, indeed, their results provide no indication that bare most might share any properties with the most-which, we should note, they do not explicitly discuss.…”

“…Hackl 2009;Pietroski et al 2009Pietroski et al , 2011Lidz et al 2011;Solt 2011;Kotek et al 2011aKotek et al , 2011b. More specifically, while it is uncontroversial that bare most in subject position can give rise to proportional truth conditions similar to the ones expressed by more than half, it has been argued in Hackl (2009) that the verification strategy associated with it is quite different from the one associated with 1 Bare most can also occur outside of the partitive frame, as in (i) below, and in that case the sentence tends to have a generic interpretation.…”

“…That is, there are two unrelated mosts: bare most is a lexical determiner while the most is a superlative construction built from a gradable predicate such as many or much and the superlative morpheme -est. A recent version of this view on bare most, proposed in Lidz et al (2011), is expressed in (3). On this analysis, bare most is analyzed as a quantificational determiner that takes two set-denoting expressions A and B and yields True only if the number of As that are Bs is greater than the number of As minus the number of As that are Bs.…”

Logic and Cognition: Special Issue of Best Papers of the ESSLLI 2012 Workshop

“…Does people's model checking strategy depend mostly on sentence form or do people rely on properties of the picture, such as whether it contains dots of only two colours or more? These questions follow up on Lidz's Interface Transparency Thesis (Lidz et al 2011), which states that "the verification procedures employed in understanding a declarative sentence are biased towards algorithms that directly compute the relations and operations expressed by the semantic representation of that sentence"-and for the English "Most dots are yellow", Lidz had argued on the basis of experiments with English-speaking subjects that people would most likely approximate the number of yellow dots and check whether it is larger than the number of all dots minus that of yellow dots. Tomaszewicz' experiments with Bulgarianand Polish-speaking subjects replicated Lidz's findings as far as most1 is concerned, whereas most2 is verified by stepwise selection: "there are more yellow dots than red ones and there are more yellow dots than blue ones and …"; and that participants' behaviour is triggered by the semantics of the two quantifiers, much more than by the properties of the picture, even if the latter could have delivered a more efficient strategy.…”