Because of the many conditions (23 figure positions and 2 attention conditions) we averaged responses

across recording sessions on different days. We only included recording sites with at least three figure-detection and three curve-tracing sessions in the analysis (the average C59 wnt in vitro number of sessions in every task was 5). This resulted in a sample of 59 recording sites in V1 (18 in monkey B, 15 in monkey C, and 26 in monkey J) and 46 in V4 (18 in monkey B, 9 in monkey C, and 19 in monkey J). The average number of trials per condition was 484. We did not observe significant differences in visual responsiveness between the figure-detection and the curve-tracing tasks (paired t test; V1 and V4, p > 0. 05). We separately analyzed the V1 responses evoked by the texture background, the figure edge and the figure Enzalutamide center. To create the space-time plots in Figures 4C, 4D,

6C, and 6D we first computed for every RF the distance of the figure center to the RF center in all stimulus conditions and rounded these distances to the nearest multiple of 0.5°. We then computed for every figure position (0.5° steps) an average response across RFs. In area V1, background responses were obtained by averaging across conditions where the RF-center was separated from the nearest figure edge by at least 2°. Edge responses were averaged across conditions where one of the two edges fell in the RF, and center responses across conditions where the RF-center was within 0.5 degrees of the figure center. In V4, we did not distinguish between edge and figure responses, because the large RFs typically did not fit entirely within the figure. Instead, we compared figure positions where the figure completely or partially PD184352 (CI-1040) covered the RF to background positions where the figure was outside the RF. We took as background responses those conditions where the RF border was separated from the nearest figure edge by at least 2° (except for 5 recording sites with distance <1.8°, and 3 with distance <1°) and as figure responses those configurations

where the RF hot spot (point in V4 RF with maximum response) was within 2° from the figure center. To compute the population responses, we first normalized the responses before averaging across recording sites by subtracting Sp and dividing the result by (Pe-Sp) (Sp and Pe were defined above). As a measure of the reliability of the FGM across trials, we computed the FGM d  -prime: dFGM′=(y¯Fig−y¯Bck)/sAct, where y¯Fig and y¯Bck is the average response evoked by the figure and background, respectively, and sActsAct is the pooled standard deviation across the two conditions. For each recording site, we obtained an estimate of the FGM d-prime during figure detection and curve tracing by averaging d-prime values across the corresponding sessions. The model consists of areas V1m, V2m, and V4m (the subscript “m” stands for model). The input signal arrives in two maps.