Fig. 1

Diagrammatic sketch of the asynchronous saccadic choice paradigm. Participants viewed a horizontally arranged array of three targets where the central target served as the fixation point. On each trial, the participant made a single speeded saccade in response to the sudden illumination of one or both of the peripheral targets. On single target trials (50%) only one randomised peripheral target was illuminated and the subject made a saccade to it. On double target trials, both peripheral targets were illuminated but with a small asynchrony (δ) between them, and the participant was instructed to foveate as rapidly as possible whichever target caught his attention. Critically, participants were not instructed to choose the target that they consciously perceived to have occurred first, but to allow their gaze to shift automatically to whichever target caught their attention, unconsciously guided by the bias afforded by the asynchrony. The value of δ on each double saccade trial was selected by an automatic adaptive algorithm that optimised the information gain about the underlying function¹⁸. Participants performed 400 trials in total after a training run of 50 trials which was not used in the analysis. Details of the performance of each participant are given in Fig. 3. See Methods for further details