Abstract
Recent research strongly suggests that semantic information influences attentional allocation (de Groot, Huettig, & Olivers, 2015; Moores, Laiti, & Chelazzi, 2003). However, most experiments showing semantic contribution have rendered semantic information central (relevant) to the task. Therefore, whether semantic information automatically (as a default) influences attentional allocation remains an open question. Recent behavioral evidence from our lab showed that when participants are presented with three objects, two of which are semantically related, attention is biased towards semantically related objects (Malcolm & Shomstein, 2014). Here, we employed fMRI to elucidate the underlying neural mechanism responsible for this automatic semantic biasing of attention. We hypothesized that if semantic information biases spatial attentional selection, then in addition to semantic-based modulation of early visual cortex (V1-V3) we should observe semantic-based modulation of spatial representations in inferior-parietal sulcus (IPS0-2). Additionally, if semantic bias is distributed broadly across the network then we should observe semantic based modulation in object-selective cortex (LOC). Participants viewed three objects, with one just above the central fixation and one in each periphery, below the midline. One of the peripheral objects was always semantically related to the central object. After the objects were presented, a target and two distractors were superimposed on top of each object. Critically, the target appeared on all three objects equally, rendering semantic relatedness task-irrelevant. Responses to semantically related and semantically non-related objects were compared in each ROI. We observed higher responses to semantically related objects throughout the early visual cortex as well as LOC. Interestingly, responses in IPS were only weakly modulated by semantic relatedness. Non-univariate analyses corroborated the observed pattern of results. Together, the results strongly suggest that task-irrelevant semantic information automatically constrains attentional allocation by directly acting on object representations in LOC and V1-V3.
Meeting abstract presented at VSS 2016