Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Mapping Action Verb Meanings Across Cortical Networks with Representational Similarity Analysis
Poster Session D, Thursday, October 1, 4:30 - 6:30 pm, Wangari Maathai
Fangfei Li1, Matthew King-Hang Ma1, William Shiyuan Wang2, Manson Cheuk-Man Fong1; 1The Hong Kong Polytechnic University, 2The Chinese University of Hong Kong
Previous studies have demonstrated that processing action-related meanings can modulate the activity of cortical motor regions. However, it remains controversial whether this engagement reflects specific representations of action features within these regions. To understand the nature of motor network involvement, the present study uses Chinese action verbs to map and compare semantic representations in visual, motor, and language networks across the neocortex. Thirty-six young participants (17 females) with no history of neurological disorders were recruited. All participants were native Cantonese speakers and performed a lexical decision task in an event-related design. The stimuli were Chinese phono-semantic compound words with semantic radicals explicitly cueing an effector (hand, foot, mouth, nonaction). For each real word, a pseudoword was created using the same semantic radical. The experiment followed a 2 (lexicality: real word vs. pseudoword) x 4 (semantic radical category: hand, foot, mouth, and nonaction) design. Twenty-two stimuli were used in each condition, amounting to 176 trials in total. Functional images were acquired using a whole-brain multiband EPI sequence (TR = 800ms, TE = 37ms, flip angle = 52 degrees, voxel size = 2 x 2 x 2 mm3, 72 axial slices). Mass-univariate functional activation analysis and representational similarity analysis (RSA) were conducted using SPM, The Decoding Toolbox (TDT), and custom MATLAB scripts. Representational dissimilarity matrices (RDMs) were constructed for individual regions of interest across visual, language, and motor networks in the neocortex. Data from four participants were excluded due to low accuracy (< 75%), leaving a final sample of 32 subjects. Consistently, all regional RDMs across visual, motor, and language networks demonstrated a highly significant fit to the binary theoretical lexicality model (all mean point-biserial r_pb > 0.30, group-level t-tests on Fisher z-transformed values: t(31) > 5.69, pFDR < .001). This suggests that there are distinct representational clusters based on lexical validity across these networks. When testing for effector specificity, individual regional RDMs were compared against theoretical somatotopic models. While no brain region exhibited strict somatotopic selectivity, partial selectivity was observed across bilateral visual cortex, left language networks (BA44, BA45, and inferior parietal cortex), and the right primary somatosensory cortex (Hand > Mouth: all t(31) >= 2.35, pFDR < .05; Foot > Mouth: all t(31) >= 2.78, pFDR < .03). Interestingly, no regions exhibited a significant preference for the mouth effector compared to hand or foot effectors. These patterns indicate a coarse division among different effectors rather than fine-grained, fully differentiated representations of action features within these cortical networks. Taken together, our results suggest that the representations of action meanings are organized more strongly by lexicality than by action features. The partial effector-specific effects observed across visual, language, and sensorimotor regions argue against a strict somatotopic mapping of action verb meanings in the motor system. Instead, action meanings appear to rely on broader and partially abstract semantic representations across multiple neocortical systems.
Topic Areas: Meaning: Lexical Semantics, Multisensory or Sensorimotor Integration