Janek Guerrini

This paper makes two contributions to the study of the interpretation of nominals across Germanic and Romance languages. First, it shows that plural kind terms, such as English bare plurals (e.g., lions) and Italian definite plurals (e.g., i leoni), have definite, non-generic uses in sentences expressing generalizations that were traditionally thought to uniformly involve generic quantification. These non-generic uses explain why the distribution of kind-denoting plurals in sentences expressing generalizations is wider than that of singular indefinites, which can only appear in generalizations with a genuinely generic Logical Form (Sects. 3–5). Second, the paper draws on a contrast between English and Italian (and, to a minor extent, French) plural forms, both bare and definite (Sect. 5). This yields a new approach to the mapping between the form and the interpretation of nominals that combines elements from Chierchia’s (Natural Language Semantics 6:339–405, 1998) and Longobardi’s (Natural Language Semantics 9:335–369, 2001) frameworks. On the approach I present, English bare plurals can be mapped either to kinds or to properties; Italian definite plurals, when not referential, can only be mapped to kinds; and Italian bare plurals only to properties. This explains the behavior of these expressions in a wide number of contexts, most importantly in episodic sentences, which raise puzzles for extant accounts.

In this paper, I argue against two common claims about so-called privative adjectives like ‘fake’: first, I argue against the idea that their semantic complexity requires a richer notion of lexical meaning than the standard one (see, e.g., Del Pinal, 2018); second, I argue against the idea that ‘fake’ is a subsective adjective ‘in disguise’ and does not semantically negate its input (see, e.g., Partee, 2010). I propose that a fake P is (i) intended to resemble a P and (ii) is not a P. This makes correct predictions for multiple applications of ‘fake’, a task at which other theories fail. In cases of double application of ‘fake’, the interaction between its conjunctive meaning and the negation hard-coded into clause (ii) yields a complex meaning, compatible with a variety of objects, which aligns with intuitions about what should count as a fake N. While the core meaning of ‘fake’ is quite simple, its mode of composition bears some complexity. In line with Martin (2022), I propose that ‘fake’ can alternatively (a) combine directly with the noun via Functional Application or (b) saturate its property argument via an implicit, contextually provided variable via Functional Application and then combine with the noun via Predicate Modification. Mode of composition (a) is clearly visible in syntactic parses that only allow for Functional Application: for instance, in Italian, if pre-nominal, ‘fake’ can only directly take the noun as an input (cf. Cinque, 2010). Positing (b) correctly predicts readings where ‘fake’ is not apparently privative: ‘fake watch’ can designate a watch that is made to resemble a Rolex but isn’t one, i.e. a fake(-as-a-Rolex) watch. When the intersection between the $[\![ \mathit{fake} ]\!] (*\mathit{implicit\ argument}*)$ complex and the noun is empty, rescuing principles originally proposed by Partee kick in to rescue from vacuous modification: this explains why we can refer to a fake gun as a gun, as in the sentence ‘this gun is fake’. As a result, besides correctly predicting iterated ‘fake’, this theory provides clear predictions on when and how Partee’s pragmatic principles of noun modulation apply. I conclude the paper arguing that this view of privatives calls for a classification of adjectives in terms of their mode of composition, rather than in terms of their emergent entailment pattern.

Language has a rich typology of inferential types. It was recently shown that subjects are able to divide the informational content of new visual stimuli among the various slots of the inferential typology: when gestures or visual animations are used in lieu of specific words in a sentence, they can trigger the very same inferential types as language alone (Tieu et al., 2019 ). How general are the relevant triggering algorithms? We show that they extend to the auditory modality and to music cognition. We tested whether pro-speech musical gestures, i.e. musical excerpts that replace words in sentences, can give rise to the same inferences. We show that it is possible to replicate the same typology of inferences using pro-speech music. Minimal and complex musical excerpts can behave just like language, gestures, and visual animations with respect to the logical behavior of their content when embedded in sentences. Specifically, we found that pro-speech music can generate scalar implicatures, presuppositions, supplements, and homogeneity inferences.