Figure 3.

The neuronal worth in V1

(A and B) Activity elicited by the stimulus with the discriminant segment in the RF (dark green, black) or not in the RF (light green, gray) in the presence (black, gray) and absence of a mask (dark and light green) in monkey D (A) and monkey M (B). The insets above illustrate the different stimulus conditions. The panels below represent Resp_diff, which is the difference in activity elicited by the conditions with the discriminant segment inside and outside the RF in the presence (gray area) and absence of masking (green area). The extra activity difference that occurs without masking is shown as a green area and corresponds to our measure of the neuronal worth.

(C) Integrated Resp_diff as function of the stimulus duration in the presence (black) and absence of a mask (green). The data points correspond to the shaded areas in the bottom row of (A) and (B). Lines represent linear fits and error bars the SEM across recording sites. The horizontal shift between the lines (red arrow) corresponds to the neuronal worth. Insets, distribution of the worth estimates determined for individual recording sites.

(D) Upper: SVM that maximally separates trials with the discriminant segment inside and outside the RF of the V1 recording sites in monkey D. Histograms, distribution of SVM scores across trials with the discriminant segment inside (dark green) and outside the RF (light green). We used a time window from 50 to 250 ms after stimulus offset (gray rectangle in B) on trials with shorter stimuli (≤153 ms). We performed a median split on the SVM score on trials with decaying activity elicited by the discriminant segment to compare the monkeys’ accuracy on trials with high and low V1 activity. We here selected trials in which the monkey was cued to report about the pair of curves in the V1 RFs (n = 1/3 of these trials). The accuracy was higher on trials with a higher level of decaying V1 activity (black bars) than if the V1 activity was low (white bars). n, number of trials in the analysis. The p-values indicate the result of a bootstrap test.

See also Figures S1 and S2.