The early search results, before the first saccade, also served as a control for the possibility that the attention results in the late search fixations were influenced by differences in scan paths to the different attended stimuli in the RF. The responses of cells in both the FEF and V4 were modulated by feature attention, even when the animal selleck inhibitor was planning an eye movement to a stimulus outside the RF. Figures 2A and 2B show normalized firing rates averaged across the entire populations of FEF and V4 sites during target and no-share fixations in early search, i.e., prior to the first saccade after the array

onset. The response to the targets in the RF was significantly larger in comparison to the same stimuli on trials when they were the no-share stimuli in the RF, in both the FEF and V4, although the stimuli in the RF were matched across these two conditions. Thus, both areas show feature attention effects on their responses. Although both areas showed feature attention effects, they began earlier in the FEF than in V4. The effect of feature attention began with a latency of 100 ms after search array onset in the FEF (Wilcoxon signed rank test, p < 0.05), versus a 130 ms latency after search array onset in V4 (Wilcoxon signed rank test,

p < 0.05), and this latency difference was significant (two-sided permutation test, p < 0.05). Very similar results were obtained using a mutual information measure. We also measured the latencies of the effects of attention at each individual recording site. The cumulative distribution of latencies for the sites is mTOR inhibitor shown separately for the FEF and V4 in Figure 2C, and the distribution is clearly shifted to earlier times in the FEF (Wilcoxon rank-sum test, p < 0.05). There was one site in each area with an early attention latency of 40 ms, but the FEF site is obscured by the V4 distribution line in the figure. The cumulative distributions do not reach 100% in either out area

because many cells in each area did not have a significant effect of attention at any latency in this analysis (Wilcoxon signed rank test, p < 0.05). The median latency was 240 ms in the FEF, and it was not measurable in V4 because less than 50% of the V4 sites showed a significant effect of feature attention. To rule out the possibility that the shorter latency of attention effects in the FEF was due to the larger magnitude of attention effects in that area, we recomputed the latencies using a subset of sites with similar magnitudes of attention effects in each area. We only considered sites in each area with a 10%–30% increase in response to the target versus the no-share stimulus, in the period of 120–220 ms after the onset of the search array. We matched the sites in the FEF to the same number of sites in V4 with similar effect sizes (43 sites in both areas).