| dc.description.abstract |
Understanding how humans construct meaning – and the nature of conceptual representations that support this process – remains a central question in cognitive science. Embodied approaches propose that semantic knowledge is grounded in perceptual and motor (modal) systems, whereas distributional (amodal) approaches emphasize linguistic co-occurrence as the primary basis of meaning. Increasing evidence suggests that these perspectives are not mutually exclusive: conceptual representations appear to have a hybrid structure, drawing on both sensorimotor experience and linguistic knowledge. However, key questions remain unresolved, including the circumstances under which sensorimotor information is activated, when such activation is functionally relevant for comprehension, and how sensorimotor and linguistic systems interact during real-time processing. Although theoretical frameworks suggest that these contributions depend on task demands, contextual salience, and processing depth, direct empirical tests remain limited. This thesis addresses these issues by examining how linguistic and sensorimotor information jointly influence behaviour across multiple tasks and contexts. Through four studies, I investigate the activation, functionality, and interaction of sensorimotor and linguistic factors in two conceptual domains – colour and space – chosen for their contrasting perceptual properties and theoretical relevance for embodied accounts.
Study 1 examines when colour-based sensorimotor traces are activated, manipulating stimulus type (words vs. sentences), contextual salience (fillers, sentence distribution, sentence position), and grammatical structure (tense-induced psychological distance). Using a compatibility paradigm, the study shows that colour effects are stronger for isolated words than for sentences and are sensitive to contextual salience introduced by fillers. In contrast, grammatical manipulations of psychological distance do not modulate the effect, suggesting that colour simulations are fragile and strongly dependent on surface-level task features rather than deeper semantic structure.
Studies 2 and 3 investigate the spatial domain within a visuospatial working-memory framework to test sensorimotor activation under cognitive load. Study 2 manipulated the type and modality of spatially associated words. The findings show that such words do not alter the content of working memory but do influence spatial recognition even under high cognitive load. However, Study 3 provides no evidence that the content of visuospatial working memory facilitates or interferes with linguistic comprehension. Thus, although spatial simulations are activated during demanding tasks, they do not appear functionally relevant for language processing in this paradigm.
Study 4 directly examines the contribution of embodied and linguistic experience using sensorimotor estimates and distributional semantic models to human behavior. Both factors contribute to behaviour during language processing. Sensorimotor experiences are most strongly engaged when linguistic scaffolding is minimal (Study 2) and remain detectable in shallow and deep linguistic tasks (Study 1, word-based experiments; Study 3). When linguistic scaffolding is available through sentence-level context and compositional structure, both linguistic and embodied information contribute, with linguistic factors showing a modest advantage (Study 1, sentence-based experiments).
Overall, the thesis supports a hybrid, dynamically weighted view of conceptual representation in which linguistic and sensorimotor systems operate flexibly and in complementary ways, adapting to specific task requirements. |
en |
