Fig. 1.

Results of the human visual statistical learning experiments under passive and active conditions. (A) Illustration of a typical display presented in a trial of the familiarization phase during the human statistical learning experiments. In the active condition, one of the two scenes (a grid with its shape arrangement) was the Target, the other was the Distractor, and the participant learned to choose the Target across multiple trials with feedback. During the passive condition, the participant just inspects the two scenes for 6 s without a task. (B) Human discrimination performance in the active condition during familiarization. Average performance was computed in 10-trial bins, except for the last bin, which contained 12 trials. (C) Comparing human learning of conditional-, joint-, and single-element statistics in the passive and the active conditions after the familiarization session. The y axis shows average fraction of correct responses across subjects. Learning effects remained highly significant in the active condition, albeit with a significant reduction in learning the joint statistics (joint test). Chance performance in the control condition (equal single) indicates that the observed learning effects are specific and not due to general improvement due to increased overall familiarity. In B and C, dashed horizontal line shows chance performance, error bars represent SEM. (mean ± SEM). *P < 0.05.