Slide Slam A3
The roles of language and semantics in gesture imitation following left-hemisphere stroke
Haley C. Dresang1,2, Rachel C. Metzgar1, Aaron L. Wong1, Laurel J. Buxbauum1,3; 1Moss Rehabilitation Research Institute, 2University of Pennsylvania School of Medicine, 3Thomas Jefferson University
Introduction: Dual-route models of reading and word repetition distinguish between a semantic route and a direct non-lexical route that bypasses semantics. Semantic information often confers benefits, but individuals with damage to the semantic route can still accomplish many tasks by relying on the direct route. Similarly, dual-route models of action distinguish between a semantic and a non-semantic visuo-motor route to execute gestures. This has been tested by comparing meaningful versus meaningless gesture imitation. However, it remains debated how this system responds to damaged semantic integrity or access (e.g., via lexical input). This is significant due to the common co-morbidity of apraxia, aphasia, and semantic impairments after left-hemisphere cerebrovascular accidents (LCVA) and the utility of gestures in language rehabilitation. In this study, we investigated the role of semantics and language in gesture imitation for patients with LCVA and the brain regions associated with benefit of semantic and linguistic information in gesture imitation. Methods: Participants were 18 neurotypical controls and 42 adults with chronic LCVA. Patients were excluded for severe language comprehension impairments (<4 on WAB comprehension), but were not otherwise selected for presence of aphasia, apraxia, or semantic impairment. Lesion analyses included a subset of 39 patients. Participants completed three gesture imitation tasks: (1) named meaningful, (2) unnamed meaningful, and (3) meaningless gestures. Named versus unnamed examined linguistic benefits, while unnamed versus meaningless examined semantic benefits. The outcome variable was hand posture (HP) accuracy, the most sensitive component of gesture imitation, which captures when the shape or movement trajectory of the hand and/or wrist is incorrect. Mixed-effects models examined group (controls vs patients) by task (named, unnamed, meaningless) interaction effects on HP accuracy, with random effects of subject and item. Support vector regression lesion-symptom mapping identified lesions associated with reduced benefit of linguistic information (named residualized against unnamed) or semantic information (unnamed residualized against meaningless) on HP accuracy. Voxelwise statistical significance was determined using Monte Carlo permutation analyses (10,000 iterations, p<0.05, clusters with >500 contiguous voxels). Results: There was a group-by-task interaction for HP accuracy, where patients were less impaired for named (β=1.25, p<0.001) and unnamed (β=1.05, p<0.001) compared to meaningless gestures. Controls showed better accuracy for named compared to unnamed (β=1.02, p=0.04) or meaningless gestures (β=1.39, p<0.001). Lesions associated with reduced benefit of the linguistic label (named vs unnamed) included middle and posterior temporal regions, basal ganglia, and uncinate fasciculus. Lesions associated with reduced benefit of semantics (unnamed vs meaningless) included middle and inferior frontal gyrus and precentral regions. Conclusion: For controls, linguistic cues conferred additional benefit on gesture imitation beyond meaning, whereas for patients, meaningful gestures conferred a benefit but there was no additional benefit of linguistic cueing. These patterns parallel findings of the benefits of semantic versus linguistic cues for action naming in people with and without aphasia. Reduced benefits of language and semantics were associated with distinct lesion sites, consistent with the roles of prefrontal regions in semantic retrieval/control and ventral stream regions in comprehension. Future work will examine cross-domain benefits of semantic information for LCVA patients.