| Research suggests that concepts are distributed across brain regions specialized for processing information from different sensorimotor modalities. Multimodal semantic models fall into one of two broad classes differentiated by the assumed hierarchy of convergence zones over which information is integrated. In shallow models, communication within- and between-modality is accomplished using either direct connectivity, or a central semantic hub. In deep models, modalities are connected by cascading integration sites with successively wider receptive fields. Four studies provide the first direct test of these models using speeded behavioural tasks involving feature inference and pattern completion. Shallow models predict no within- versus cross-modal difference in either task, whereas deep models predict a within-modal advantage for feature inference, but a cross-modal advantage for pattern completion. Study 1 investigated the prevalence of within- and cross-modal feature correlations in a large database of feature production norms. Studies 2 and 3 used relatedness judgments to tap participants' knowledge of relations for within- and cross-modal feature pairs. Study 4 was a dual feature verification task. The pattern of decision latencies across Studies 2 to 4 is consistent with a deep integration hierarchy.Keywords. semantic memory, multimodal representations, binding problem, embodied cognition. |
