Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Cortical representation of quantification: The role of the left anterior temporal lobe
Poster Session B, Friday, September 12, 4:30 - 6:00 pm, Field House
Nigel Flower1, Liina Pylkkänen1; 1New York University
Quantifiers like ‘all’ and ‘some’ are fundamental to how we construct meaning in language. They appear in all known languages and even emerge in the first generation of new ones (Kocab et al., 2022), suggesting they reflect a core aspect of human cognition. Yet how quantifier meanings are computed in the brain remains unclear. A major challenge in studying the neural basis of quantifiers is that, while language processing is rapid, quantifier interpretation spans the full sentence. Neural activity tied to the quantifier’s composition must be separated from serial prediction effects. Using MEG, we addressed this by presenting full sentences in parallel, giving the system access to both quantifier arguments simultaneously. Prior work shows this method evokes sentence-level effects as early as 130ms (Fallon & Pylkkänen, 2024; Flower & Pylkkänen, 2024). 24 native English-speaking participants read 4-word English sentences flashed for 300ms using parallel presentation. The sentences contained quantified phrases as in “all cats are nice,” using either the Aristotelian quantifiers all, some, or no or the definite determiner the. Participants reported whether an immediately following stimulus was the same or different as the first stimulus, with mismatches created by swapping out one word (e.g., all dogs are nice). MEG data were analyzed from the first stimulus only. We focused on the left anterior temporal lobe (LATL), broadly implicated in semantic processing (Pylkkänen, 2019). Generalized Quantifier theory (GQ) posits that quantifiers encode set-theoretic relations between two sets (Barwise & Cooper, 1981). The LATL has been linked to Boolean set intersection (Poortman et al., 2016), and could therefore participate in quantifier interpretation under GQ. If so, LATL should compute quantifier meaning via set intersection of its arguments, predicting: all > some > no. A second hypothesis is that LATL activity reflects quantifiers’ referential properties. Restricted Quantifier theory (RQ; Knowlton et al., 2022, 2023) posits that all and no involve plural definite semantics, unlike existential some. Given LATL sensitivity to unique entities (Grabowski et al., 2001) and to domain-restricting determiners (Leffel et al., 2014), RQ predicts stronger responses for all and no than for some. A spatiotemporal clustering test over the LATL showed a significant main effect of determiner type at 210–410ms after sentence onset. Pairwise comparisons revealed two patterns: the elicited more activation than all Aristotelian quantifiers; and all and no both elicited more activation than some, consistent with RQ. A generalized linear model confirmed these effects weren’t driven solely by bigram frequency. In sum, our findings suggest that the LATL contributes to computing quantified meaning, with activity patterns supporting the Restricted Quantifier theory. Whole-brain analysis revealed no other significant quantification effects after correction. These results align with prior evidence of LATL involvement in referential processing (e.g., unique entities like celebrity names; Damasio et al., 1996) and inform representational theories of quantification.
Topic Areas: Syntax and Combinatorial Semantics, Reading