This study investigates the effects of feature-based attention on responses of direction-selective neurons in the middle temporal area (MT) of macaque visual cortex to attended stimuli inside the receptive field. Redirecting attention between the preferred and null direction of transparent random dot motion patterns caused a mean modulation of responses of approximately 32%, about half of what was observed when the two directions of motion in the receptive field were spatially separated allowing feature-based and spatial attention to work in concert. This is consistent with models of visual attention that interpret the attentional modulation of a neuron as the combination of all attentional influences, treating stimulus location simply as another feature.

Based on single cell recordings in monkey, it has been suggested that neural activity can be related directly to psychophysically measured threshold behaviour. Here, we investigated in humans whether evoked potentials correlate with behavioural measurements like discrimination thresholds and reaction time. Subjects were asked to report the perceived direction of object motion stimuli which contained variable amounts of coherent motion. Simultaneously, we recorded evoked potentials with a multielectrode array, or measured the reaction time. We show here that motion coherence had a strong influence on both amplitude and latency of the evoked potential. Stronger motion signals evoked stronger and faster cortical responses. The latency reduction of the motion onset response with increasing coherence correlated very well with the concurrent decrease in reaction time. Taken together, these results suggest that temporal integration is an important step in analysing motion signals to generate a reliable behavioural response. We stimulated a two-dimensional array of correlation-type motion detectors with the same motion sequences, and analysed the distribution of local motion signals according to signal detection theory. Performance resembled that of human subjects when the decision strategy was optimized so as to exclude small signals and, in particular, when the ideal observer had some knowledge about a region of interest in which the object was to be expected.

No abstract