Reply to Hupé et al.: The predictive correlation of pupil dilation and relative dominance durations in rivalry is not a statistical artifact

We are encouraged by Hupé et al.'s (1) independent replication of our findings (2) and the fact that our data per se stands uncontested. Hupé et al. argue against our interpretation of these results, based on removal of short dominance durations from the analysis. Although it is no surprise that selective removal of data points can reduce the significance of any correlation, we appreciate the theoretic basis of their concern: Because pupil dilation persists for 1–2 s, short preswitch dominance durations may be disproportionately affected by the pupil dilation accompanying the previous switch event (figure 1F in ref. 2).