Month: November 2023

Causal dependence in actuality inferences

Prerna Nadathur

A range of complement-taking predicates give rise to surprising actuality inferences, in which a modally-embedded complement event is understood non-modally, as taking place in the evaluation world.  I argue that actuality inferences can be explained—and unified across predicate classes—on an approach in which the modality of participating predicates is analyzed in causal terms.  This talk focuses on an illuminating case study: enough and too predicates.

Hacquard (2005) observes that, like the ability modals at the heart of the puzzle (Bhatt 1999), enough/too predicates have aspect-sensitive actuality inferences.  Under imperfective marking, French enough predicates like (1a) are compatible with the non-actualization of their complements; their perfective counterparts (as in 1b) show the complement entailment pattern of implicative verbs like French réussir (‘succeed’, ‘manage’; 2).

(1) a. Juno était assez rapide pour gagner la course, mais elle n’a pas gagné.
Juno was-IMPF fast enough to win the race, but she did not win.
b. Juno a été assez rapide pour gagner la course, #mais elle n’a pas gagné.
Juno was-PFV fast enough to win the race, #but she did not win.
(2) Juno { réussissait / a réussi } à gagner la course, #mais elle n’a pas gagné.
Juno { managed-IMPF / managed-PFV } to win the race, #but she did not win.

Despite the contrast in (1)-(2), I argue that enough/too inferences are—semantically speaking—instances of implicativity.  I build on a causal account of implicative lexical semantics (Nadathur 2016, 2019) to show that enough/too actuality inferences arise just in case the compositional interaction between grammatical aspect, modal flavour, and the enough/too matrix adjective reproduces the semantic structure of an implicative: that is, where the matrix adjective denotes an actionable property which is causally involved in realizing the enough/too complement, and the perfective aspect induces an eventive interpretation of the matrix assertion.   Insofar as the implicative analysis explains the aspect-sensitivity of enough/too inferences, I suggest that it naturally extends to ability modals’ actuality inferences, when coupled with a causal approach to ability.

Towards a Structuralist Metasemantics for Number Words

Eric Snyder

According to non-eliminative structuralism, the referents of numerical singular terms, such as the numeral ‘two’ or ‘the number two’, are numbers, construed as positions within the natural number structure. However, a potential problem comes in the form of sentences like ‘{∅, {∅}} is the number two among the von Neumann ordinals’. If this is an identity statement, then its truth would seemingly require identifying the second position of the natural number structure with a particular set, thus giving rise to a version of Benacerraf’s famous Identification Problem. In response, Stewart Shapiro (1997) draws an analogy to expressions like ‘the Vice President’, which are ambiguous between denoting an office-holder (e.g. Kamala Harris) or an office (the office of the Vice Presidency). Similarly, Shapiro suggests that in ordinary arithmetic contexts, such as ‘Two is less than three’, we view positions as analogous to office-holders, while in other contexts, we view them instead as analogous to offices occupied by entities playing the role of numbers, e.g. {∅, {∅}}. However, this suggestion faces two serious challenges. First, what exactly is the nature of this purported ambiguity, and what empirical evidence, if any, is there for it? Second, even if we grant the ambiguity, we appear to get a revenge version of the Identification Problem anyway: just permute the positions within the natural number structure. The purpose of this talk is to defend Shapiro’s ambiguity thesis, by supplying the empirical support required, and explaining how, when appropriately understood, the semantics assumed does not give rise to a revenge form of the Identification Problem.