Terrapins at MACSIM ft. Valentine

On modality, conditionals, quantification and attitudes.

Saturday October 3, the 12th MACSIM – Mid-Atlantic Colloquium for Stuidies in Meaning – takes place at Rutgers University in New Brunswick, NJ. The Terrapin presence is strong. In addition to four posters – from Yunhui Bai (Philosophy), Matthew Shi, Mal Shah and Elizabeth Swanson (Linguistics) – plus one student talk from Sarah Boukendour (Linguistics), the invited faculty talk this year is by Valentine Hacquard. Their titles and abstracts are below.

• Valentine Hacquard / Modal force in acquisition [invited talk]
• Sarah Boukendour / Perfect conditions: When the sufficient becomes necessary [talk] 
• Yunhui Bai / Epistemic modals: Accommodation and reflection [poster]
• Mal Shah, Tyler Knowlton, Robert Hopkins, Justin Halberda, Paul Pietroski and Jeffrey Lidz / Individual differences reveal similarities in meaning [poster]
• Matthew Shi / Restricting the semantics of interrogative disjunctors [poster]
• Elizabeth Swanson / No Fun: Complement desirability is a limited cue to attitude verb meaning [poster]

Valentine Hacquard - In this talk I will present various studies that converge in showing that children master the force of possibility modals as early as age 2, but struggle with necessity modals until they’re at least 4 or 5 years old, seeming to understand them as if they expressed possibility. I will discuss potential sources for their difficulty (linguistic, pragmatic, and conceptual), and speculate on what might be particularly challenging about universal quantification in the domain of modality.

Sarah Boukendour - An utterance of "If you complete the course, you may use the laboratory" can suggest that completing the course is not only sufficient but necessary for permission. This conditional-perfection inference varies with context. I propose that its support depends on two conditions: the exchange calls for a complete contribution, and the speaker is understood to be in a position to provide it. Cooperative interpretation can then support taking a contribution with no indication of partiality as complete. The conditions apply both when a speaker reports terms and when the utterance helps establish them. Knowledge can support the first contribution; practical authority can support the second without being reduced to knowledge. The proposed explanation of variation concerns the contribution's role: supplying the relevant terms in full or addressing only one circumstance. Questions are one source of evidence for that role, not a required part of the account. The proposal also predicts parallel contextual effects on the restrictive interpretation of universal sentences such as "Everyone who applied is on the list." Existing experiments motivate the contextual distinctions but do not yet test their proposed combination across uses and constructions.

Yunhui Bai - Satisfaction theory says that a sentence S presupposing p means that S's local context needs to entail p. Essentially, this is saying that the local context c needs to already accept p (Veltman, 1996): c[p]=c. But given the update semantics by Veltman, must p's sole impact on the context is to check if p is accepted by c; this makes it a presuppositional item without any assertive effect. I suggest that this can used to explain the indirectness constraint observed for must: when its prejacent is directly verifiable via observations, it will automatically become common ground and thus must has a trivial effect; when the evidence is indirect, the common ground is not automatically updated, but using must can force the prejacent be accepted via presupposition accommodation.

Malhaar Shah, Tyler Knowlton, Robert Hopkins, Justin Halberda, Paul Pietroski, and Jeffrey Lidz - Philosophers and linguists have objected to theories which take meaning to be psychological objects, like concepts, because they have assumed that meanings are invariant, while concepts vary from person to person (e.g., Frege 1918). This has been corroborated by psychological evidence showing variation in words like furniture (e.g., McCloskey and Glucksberg 1978). We provide evidence that there is sometimes remarkable stability in how certain words are interpreted. We show this using quantifiers. In sentences like Most of the dots are blue; most is consistently interpreted as a procedure comparing the cardinality of the subset (blue dots) to the cardinality of the superset (all dots), rather than to the complement set (non-blue dots). Previous work (e.g., Knowlton et al. 2021) performed only between-group comparisons of the interpretation of such words. In these within-individual experiments, we show that the differences between individuals’ performance is best explained by differences in their extra-linguistic factors, such as their facility with estimating large numbers, rather than the procedure they assign to most. We also show that the estimate of their facility with estimating large numbers can come from visual tasks or from auditory tasks, supporting our contention that this is a measure of modality-neutral meaning.

Matthew Shi - In some languages, forming alternative questions (AltQs) requires a special disjunction marker (Haspelmath, 2007; Mauri, 2008). In Mandarin, AltQs are formed with the morpheme háishì. In this paper, I motivate a new approach to its semantics. Contra previous proposals, I argue that háishì is not a symmetric connective (cf. Huang, 1991; Erlewine, 2025); instead, it expresses an asymmetric mapping from a proposition p and a question Q to form a new question, roughly ‘Q or p?’, which is like Q, except that it also has p as a possible answer. When Q is a polar question, its positive answer q forms an AltQ together with p: “q or p?” To support this proposal, I observe empirical asymmetries in háishì constructions involving the distribution of polar question particles and the acceptability of wh-questions. Given the restrictive semantics observed in Mandarin, we also call for revisiting cross-linguistic interrogative disjunctors (e.g., Basque, Finnish, etc.) to see if they instantiate similar constraints on disjoining questions.

Elizabeth Swanson - How do children learn the meanings of propositional attitude verbs, which describe mental states? By age 3–4, children differentiate belief verbs (think, know) from desire verbs (want, like, need) (Harrigan et al., 2019; Perner et al., 2003). One hypothesis is that children are sensitive to semantic properties of the complements these verbs select (a learning process similar to Resnik, 1996). Another hypothesis is that children make use of differences in syntax between belief and desire verb complements (Hacquard & Lidz, 2019, 2022). Prior work has suggested that semantic cues, specifically noun phrase co-occurrences, are a poor source of information about which attitude verb a speaker used. However, we reexamine the role of semantics in attitude verb learning, focusing on a different cue that may be more predictive: the desirability of the complement. In two experiments, building on a paradigm used by Dieuleveut et al. (2022), we provided adult participants with the complements of frequent English attitude verbs from the Gleason corpus (stripped of tense/aspect information; e.g., “I want to call Michael” became ‘Daddy calling Michael’) and asked participants to rate whether the event sounded fun for a preschool child. Desire verb complements were rated as fun more often (M=0.67) than belief verb complements (M=0.49), but there was significant variation both across attitude verbs in a subclass and between items for a particular verb. Furthermore, ‘need’ complements were rated as undesirable, failing to align with other desire verbs, but agreeing with prior work in which necessity modal complements were judged less desirable than possibility modal complements (Dieuleveut et al., 2022). These results indicate that although desire verb complements are rated as somewhat more desirable than belief verb complements on average, desirability alone is unlikely to be a sufficient cue to attitude subclass. Learners would have to average across many different complements of the same verb—which span the full range of desirability ratings, as judged by adults—and then make use of a mean difference of ~17%. We will also present pilot results from a third experiment examining adults’ ability to use syntactic frames vs. semantic information to guess which attitude verb a speaker used. Overall, we conclude that if semantic cues play a role in attitude verb learning, they likely need to be combined with syntactic cues.