Abstract
Whether the capacity of working memory (WM) is limited by a central resource and/or mechanisms that are shared across sensory modalities is controversial. To address this question, we conducted a dual-task experiment where tactile and visual WM load were varied separately (1, 2 or 3 items per modality), and memory was tested unpredictably for vision or touch on each trial. The visual and tactile contralateral delay activity (CDA and tCDA components) were measured over visual and somatosensory cortex during WM maintenance. Increasing memory load in vision or touch produced strictly modality-specific effects. The CDA increased for visual load increments from 1 to 2 and 3 items, and the tCDA increased with 1 versus 2 tactile items, but showed no further enhancement for 3 items. Critically, neither did tactile load influence the visual CDA, nor did visual load modulate the tCDA. There were also no crossmodal costs of increasing load in the other modality on visual and tactile WM performance. An additional behavioral study found crossmodal interference when WM load was increased beyond the capacity limits of tactile and visual stores, with impaired performance when 12 multisensory items had to be maintained relative to 6-item single-task baselines. This interference was uncorrelated with individual differences in visual or tactile WM capacity, but strongly predicted individual differences in tests of executive control functions involved in task/response selection (Stroop-, Simon-, and Psychological Refractory Period tasks). This suggests that crossmodal interference in WM tasks is not the result of shared capacity limitations across modalities, but instead reflects dual-task coordination costs. Overall, these experiments show that tactile and visual WM representations are maintained by modality-specific mechanisms with independent capacities, and that crossmodal interference arises primarily at the level of domain-general central executive processes that control these modality-specific stores.
Meeting abstract presented at VSS 2018