Abstract
Theories of object-based selection postulate that attending to a particular feature of a visual object should obligatorily result in the selection of the whole object, including its task-irrelevant features. Some studies provide evidence to support these theories, but others have found conflicting evidence of reduced responses to unattended features, suggesting that, in some cases, attention can filter out irrelevant object properties. Here, we investigated whether variations in perceptual load might account for these disparate findings, given that conditions of high load would require considerable attentional resources, leaving little in the way of residual resources to spread to task-irrelevant features of the attended stimulus. Observers performed a contrast discrimination task on one of two laterally presented gratings while fMRI BOLD responses in early visual areas V1-V3 were measured. Spatial attention to one of two gratings was directed by a central cue, while the difficulty of the contrast judgment was manipulated to be low or high perceptual load. To test whether perceptual load influences neural responses to task-irrelevant visual features within an attended object, we measured the strength of orientation-selective responses using a signal detection-based pattern analysis. Under low perceptual load, we found greater discriminability of orientation-selective responses for spatially attended than unattended stimuli. In contrast, we observed no attentional enhancement of task-irrelevant orientation responses under high perceptual load. These results demonstrated that whole-object selection is not mandatory, but rather, the selection of irrelevant features is influenced by perceptual load. More specifically, the results are consistent with perceptual load accounts of attention that posit that attention is likely to "spill over" to irrelevant information, including irrelevant feature dimensions of an object, under low perceptual demand, whereas irrelevant information is filtered under high demand.
Meeting abstract presented at VSS 2013