Tolias et al. reply:

We agree with Ferrera and Maunsell that V4 could potentially serve as a source of conventional motion signals. This is also supported by our own study1, in which we found that 15% of V4 neurons were directionally selective before adaptation (Rayleigh test for circular statistics). The difference between our estimate of the percentage of direction-selective neurons and the estimate of Ferrera et al.2 (they report around 33% direction-selective neurons) probably has a methodological origin. First, the random dot stimulus we used to test direction of motion selectivity had 60% coherence, as opposed to the 95–99.8% coherent stimulus used by Ferrera et al.2. Second, as we simultaneously recorded from many neurons, we did not optimize the stimulus parameters (aperture size, dot size and density and speed) individually for each cell. Third, we used different tests and thresholds for statistical significance. In addition, Desimone and Schein3, who used moving gratings and a stricter tuning index threshold (<0.3, response to null/preferred direction) than did Ferrera et al., reported 13% of V4 neurons to be direction-selective, in close agreement with our result. These reasons are likely to account for the difference in estimates of direction selective neurons. It is clear, however, that there is a significant number of V4 neurons (13–33%, depending on estimates) that are tuned to direction of motion even before adaptation. We therefore believe that Ferrera and Maunsell are correct in suggesting that area V4 can be a significant source of conventional motion signals before adaptation. Our paper shows that this role seems to be significantly enhanced after adaptation1.