Abstract
Attention to behaviorally relevant features has been shown to improve visual detection in psychophysical tasks and to alter feature-selective responses in the visual cortex. According to models of feature-based attention, attending to a particular feature leads to a multiplicative gain enhancement, centered about the attended feature value (Boynton, 2009; Maunsell & Treue, 2006; Treue & Martinez Trujillo, 1999). Here, we evaluated whether feature-based attention might modulate activity in early visual areas in the absence of visual stimulation with cue-induced expectation. Observers were asked to detect a low-contrast grating, which was briefly presented on 50% of all trials. Prior to each trial, a color change occurred at fixation, indicating the likely orientation of the upcoming target (45 or 135 deg) with 80% validity for target present trials. (The color-orientation assignment was reversed halfway through the scanning session to avoid low-level confounds associated with the color cue.) Multivoxel pattern analysis was used to predict the attended and presented orientation on valid target-present trials, and demonstrated above-chance classification accuracy in early visual areas V1 through V3 (~70% accuracy). Moreover, analyses revealed successful decoding of the anticipated orientation on target-absent trials (~65% accuracy), indicating reliable bias effects in the absence of a stimulus. Critically, training on valid target-present trials led to successful generalization and classification on target-absent trials (~60%), implying that the attentional expectation of a particular orientation led to orientation-specific biases in early visual areas. Our results contribute to current feature-gain models of attention, by showing that top-down biasing of orientation-selective responses can result from a feature-based attentional template when there is no relevant stimulus to attend.
Meeting abstract presented at VSS 2014