Slide Slam B1
Event associated experiential features are involved in event representation at brain regions preferentially representing event concepts
Jiaqing Tong1, Jeffrey Binder1, Stephen Mazurchuk1, Coin Humphries1, Lisa Conant1, Leo Fernandino; 1Medical College of Wisconsin
Introduction: Investigation of concept representation in the brain has concentrated on object concepts, with far fewer studies of event concepts. A previous fMRI study  showed that event concepts engage left temporal-parietal areas more than object concepts, yet the mechanism underlying this preferential representation is unclear. Embodiment theories of concept representation claim that cortical regions involved in processing perception and action experiences also represent concepts. Here we investigate the hypothesis that differences between event and object concept representation in the brain reflect underlying differences in their experiential content. We used a previously validated 65-dimensional experiential model of conceptual content [2-3] together with representational similarity analysis (RSA) on a large fMRI dataset. Methods: 39 healthy adults each participated in 3 fMRI sessions, during which event and object nouns were presented visually in a jittered event-related design, and participants rated the familiarity of each word on a 1-3 scale. The stimuli included 40 items in each of 4 event categories (social, nonverbal sound, verbal sound, and negative events) and 4 object categories (animals, foods, tools, and vehicles). Each word appeared 6 times across the 3 sessions. Multivariable regression was used to generate a t map for each word. To define regions of interest (ROIs), whole-brain searchlight classifiers were applied separately to the 4 event and 4 object categories, and the event and object cross-validation classifier accuracy maps were contrasted. A 10-mm radius patch was placed around local peaks in each significant cluster to form ROIs. RSA multiple regression was then applied on all 320 words using the neural representational dissimilarity matrices (RDMs) from these ROIs as dependent variables and model RDMs representing experiential features in the conceptual model as independent variables. In addition, a categorical model coding the 8 categories was included to assess the alternative hypothesis that neural representation depends primarily on category membership. Single sample t-tests were applied to the regression coefficient of each experiential feature, which were thresholded using FDR at α < 0.05. Results: Comparison of classifier accuracy maps yielded 13 ROIs where accuracy was higher for events compared to objects, located in left inferior frontal cortex, left superior temporal sulcus, left supramarginal gyrus, left medial prefrontal cortex, left posterior cingulate cortex, and right inferior frontal gyrus. Within these ROIs, the similarity structure of neural activations was correlated with model similarity structure mainly for features related to social (Social, Communication), spatial (Scene), causal (Caused, Consequential), temporal (Short), sound (Low), and motivation (Needs) experiential content. The categorical model was not significantly related to neural activation at any ROI. Discussion: Regions preferentially engaged by event concepts show sensitivity to experiential content that distinguishes events from objects. These results suggest that differences between event and object concept representations in the brain reflect underlying differences in their experiential content. References 1. Bedny, M., S. Dravida, and R. Saxe, Cogn Affect Behav Neurosci, 2014. 14(3): p. 891-901. 2. Binder, J.R., et al., Cogn Neuropsychol, 2016. 33(3-4): p. 130-74. 3. Fernandino, L., et al., bioRxiv, 2021: p. 2021.03.16.435524.