We next investigated the relation between FGM and the saccade landing position. We measured the deviation from the median saccade landing position for every stimulus position (Figure 2B) on every trial and selected the 25% of the trials where the saccadic endpoint deviated most to the left but still landed in the 2.5° target window (blue arrows in Figure 7B) and the 25% of the trials where the saccade deviated

most to the right (red arrows). BAY 73-4506 cost In the remaining 50% of trials the saccadic endpoint was relatively close to the center (green arrows). Figures 7C and 7D shows the spatiotemporal profile of V1 FGM in the trials with deviating saccades. If the saccade deviated to the left, FGM was higher on the left side of the figure and if the saccade deviated to the right FGM was strong on the right side of the figure (paired t tests, p < 0.05). In the trials where the saccade ended close to the center, FGM was more

HCS assay homogeneous (Figure 7E) and stronger (p < 0.05, see Supplemental Information). Accordingly, the strength of FGM in area V1 predicted saccadic accuracy (Figure S6). We observed similar effects in V4 where an increase of FGM on the left predicted that the saccade would deviate to the left, and an increase in FGM on the right predicted a deviation of the saccade to the right (Figure S6). These results suggest that the profile of FGM is read out for the accurate planning of saccades toward the center of the figure. The relative timing of the neuronal activity evoked by the line elements, the FGM and the attention effects provides insight into the chain of events underlying figure-ground segregation. To measure the timing of visually Thiamine-diphosphate kinase driven activity, we fitted a curve to the average visual responses and took the time point where it reached 33% of its maximum as an estimate of latency (Roelfsema et al., 2007) (colored traces in Figure 8A, see Supplemental Information). The latency of the visual

response in V1 was 40 ms and the latency in V4 was 52 ms, and a bootstrap analysis indicated that this latency difference was significant (p < 0.01). To measure the latency of FGM in the two areas, we fitted the same type of curve to the difference between the responses evoked by the figure and background (Figure 8A). The edge modulation in V1 had a latency of 60 ms and was followed by FGM in V4 at latency of 67 ms. These latencies were both later than the visual response in V4 (p < 0.05), and the difference between them was marginally significant (p = 0.06). Finally, the V1 center modulation occurred with a latency of 95 ms, significantly later than V1 edge-FGM and V4 FGM (both Ps < 0.05). An analysis of latency across individual recording sites confirmed these effects. Activity in area V1 started with the visual response, which was followed by edge-FGM (Figure 8B, p < 10−6, paired t test), which was, in turn, followed by center-FGM (Figure 8C, p < 10−4, paired t test).