Abstract
Although attention is known to modulate neural activity, there has been much debate on whether it modulates tuning width, or preferred features, or gains. While some studies show evidence for changes in tuning width and preferred features without change in gains, other studies show evidence for gain modulation only. Here, we adopt a computational approach and ask what is the best way in which attention can modulate neural activity so as to maximize visual search performance. Our simulation results predict that all forms of modulation occur, but their utility varies with task difficulty due to target-distractor discriminability. While modulation of tuning width contributes little in easy tasks where the target and distractor are very different, its contribution increases in difficult tasks where the target and distractor are similar. The opposite trend is shown for gain modulation, whose contribution decreases with increasing task difficulty. This suggests that the conflicting experimental observations in the field may be due to differences in tasks and in their difficulty. This calls for new experiments that systematically investigate neural modulation as a function of task difficulty.