Abstract
Attending to a feature enhances visual processing of that feature, but it is less clear what occurs to other unattended features. Single-unit recording studies in MT have shown a monotonic relationship between neuronal activity and the offset between the attended and neuron's preferred direction. Such a relationship should predict a monotonic suppressive effect in psychophysical performance. However, past research on suppressive effects of feature-based attention has remained inconclusive. We investigated the suppressive effect for motion direction and orientation in two experiments. In the motion experiment, participants discriminated a low coherence random dot motion (RDM) stimulus and a 0% coherent RDM stimulus. In cued trials, a cue appeared prior to the stimuli to indicate the likely direction of the signal, while in neutral trials, no cue appeared. Participants reported which interval contained the coherent motion signal. Critically, the cue was only partially valid, and on invalid trials the offset between the cued and signal direction was systematically manipulated. The orientation experiment used a similar design, with participants discriminating between a low-contrast grating embedded in noise and noise alone. A partially valid cue indicated the likely orientation of the signal grating. We measured discrimination accuracy as a function of the offset between the cued feature and signal feature. Consistent with previous research, performance was higher on valid than neutral trials. Importantly, we found lower performance for invalid trials compared to neutral trials, indicating a suppressive effect for unattended features. Interestingly, the profile of suppression for direction showed a 'rebound effect' such that maximal suppression occurred for offset ~90° but declined for larger offsets near 180°. However, the profile of suppression for orientation exhibited a monotonic function without the rebound. These results demonstrate that unattended features are suppressed during feature-based attention, but the exact tuning function depends on the specific feature.
Meeting abstract presented at VSS 2014