Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Inter-Subject Neural Reliability Increases with Conceptual Abstractness: Evidence Against Idiosyncratic Abstract Concept Representation
Poster Session D, Thursday, October 1, 4:30 - 6:30 pm, Wangari Maathai
Stephen Mazurchuk1,2, Jeffrey Binder1, Jiaqing Tong1, Joseph Heffernan1, Leonardo Fernandino1; 1Medical College Wisconsin, 2Mayo Clinic
Introduction: Understanding the underlying representational differences between concrete and abstract concepts remains a central question in the neurobiology of language. One influential view holds that knowledge representation is organized along an axis from sensory-motor (concrete) representations to language-derived (abstract) representations (Paivio, 1991)[MS1.1]. Related proposals further suggest that neural representations of abstract concepts are also more idiosyncratic—that is, more variable across individuals—because abstract meanings are less directly grounded in perceptual referents and thus less constrained by the neuroanatomy of sensory-motor cortical areas. This idea has been supported by functional neuroimaging studies that found larger inter-subject variability in the neural representation of abstract concepts (Botch & Finn, 2024; Wang & Bi, 2021).[LF2.1][MS2.2] However, other theories point to emotional and social features as key aspects of abstract concepts (Borghi et al., 2019; Kousta et al., 2011; Mazzuca et al., 2025), emphasizing that all concepts are ultimately grounded in experiential information. In this view, abstract concept representations are not necessarily more idiosyncratic. We investigated this issue by analyzing the across-subject reliability of whole-brain activation patterns for hundreds of nouns as a function of word concreteness in 3 large data sets. We also evaluated the extent to which neural activity related to word concreteness could be explained by the relative importance of sensorimotor and socio-emotional experiences to word meaning. Methods: FMRI data were collected during familiarity rating of a total of 806 unique nouns across three different experiments (n=44, 44, and 23 participants). For each word, we computed across-participant intraclass correlation coefficients (ICC) of spatial activation patterns and evaluated their correlation to concreteness ratings (Brysbaert et al., 2014). Separate analyses were conducted for the entire cortex, the default mode network [MS3.1][LF3.2][MS3.3](DMN), and regions showing high within-participant reliability[LF4.1][MS4.2]. We created a composite experiential score to predict concreteness by subtracting the average of ‘Social’ and ‘Valence’ ratings from the average of ‘Vision’, ‘Sound’, and ‘Manipulability’ ratings. Results: Surprisingly, concreteness was negatively correlated with across-participant ICC values in all 3 studies, for all ROIs (DMN: study 1, r = -0.65; study 2, r = -0.46; study 3, r = -0.38; all p < .0001), even when controlling for age of acquisition, semantic diversity, number of morphemes, number of letters, and word frequency. This indicates more consistent neural activation patterns across individuals for abstract relative to concrete nouns. We also found a very high correlation (r = 0.91) across the whole-brain activation map based on concreteness ratings and the one based on the composite experiential score. Conclusion: These findings challenge the idea that the neural representations of abstract concepts are inherently more idiosyncratic than those of concrete concepts. The results also accord with the observation that regions of the cortex associated with social and emotional processing (pSTS, temporal pole, and anterior cingulate gyrus) aligned with the abstract end of the concrete-abstract gradient. In contrast to the view that the concrete-abstract axis represents a continuum from experiential to language-derived knowledge, these results indicate that both ends of this spectrum are grounded in experiential features that differ in kind.
Topic Areas: Meaning: Lexical Semantics, Multisensory or Sensorimotor Integration