Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Causal reasoning in language depends on domain-specific semantic networks and not the frontotemporal language system: Evidence from causal reasoning about illness
There is a Poster PDF for this presentation, but you must be a current member or registered to attend SNL 2023 to view it. Please go to your Account Home page to register.
Poster B15 in Poster Session B, Tuesday, October 24, 3:30 - 5:15 pm CEST, Espace Vieux-Port
Miriam Hauptman1, Marina Bedny1; 1Johns Hopkins University
Language offers a powerful communicative channel by which we convey information about the causal structure of the world (Pinker, 2003). Linguistic description is particularly useful when causes are not directly perceptible. A key example is illness causality, where linguistic communication transmits culturally accumulated knowledge about imperceptible causes – whether biological (e.g., pathogen transmission) or psychosocial (e.g., divine retribution) (Legare & Gelman, 2008; Legare et al., 2012). But what is the role of the frontotemporal language system in causal reasoning? Does the language system compute causality, or does it instead relay causal information to non-linguistic domain-specific semantic systems that ultimately enable causal reasoning? Because illness affects living things, we hypothesized that causal reasoning about illness would recruit the non-linguistic semantic ‘animacy network,’ particularly dorsal precuneus (Fairhall et al., 2013a, 2013b; Deen et al., 2022; see also Warrington & Shallice, 1984). The present functional magnetic resonance imaging (fMRI) study uses a sentence reading paradigm with tightly controlled stimuli to distinguish between these possibilities. Participants (n=10) undergoing fMRI read two-sentence vignettes that either elicited a causal inference about illness, elicited a causal inference about the mechanical failure of inanimate objects, or did not elicit a causal inference, as confirmed by separate group of online participants (n=26). Illness and mechanical vignettes were matched across linguistic variables known to modulate activity in language regions (e.g., Shain, Blank et al., 2020), including length, frequency, surprisal, and syntactic dependency length. Noncausal vignettes contained the same sentences as in illness and mechanical vignettes, but shuffled across trials, such that causal and noncausal vignettes were perfectly linguistically matched. We included catch trials describing ‘magical’ events and used a ‘magic detection task’ that encouraged participants to engage with the meaning of the stimuli. The same participants additionally performed a language localizer task (Monti et al., 2009, Fedorenko et al., 2010; Kanjlia et al., 2016) and an animacy/social localizer task (Saxe & Kanwisher, 2003), enabling us to perform individual-subject univariate ROI analysis and multivariate analysis (MVPA). Reasoning about the causes of illness activated the dorsal precuneus of the ‘animacy network’ relative to reasoning about the causes of mechanical failure. By contrast, mechanical reasoning activated areas previously implicated in spatial and physical reasoning (e.g., Weiner et al., 2017). Illness reasoning also activated the precuneus relative to noncausal vignettes, which contained sentences from the illness reasoning condition but did not elicit causal inferences. A linear support vector machine classifier trained on multivariate patterns of activity across the whole cortex (10 mm radius searchlight) likewise distinguished between both illness vs. mechanical reasoning and illness reasoning vs. noncausal vignettes in bilateral dorsal precuneus. We also observed preferential responses to causality (illness + mechanical reasoning > noncausal) in bilateral supramarginal/angular gyrus. Critically, we failed to observe responses to causality in the language network. Reasoning about the causes of illness ‘borrows’ the animacy network and does not recruit language regions. Using illness reasoning as a case study, we offer evidence that causal reasoning elicited by language depends on domain-specific neural systems rather than the language system itself.
Topic Areas: Meaning: Discourse and Pragmatics, Meaning: Lexical Semantics