Figure 5.

A, General experimental method. Observers reproduced orientations from arrays consisting of four oriented white lines. Across all blocks, targets were sampled from a uniform distribution. B, The critical manipulation was the distribution of nontarget items during the adaptation blocks. In Experiment 1A, nontargets (green) were sampled from a bimodal distribution centered on oblique orientations (i.e., incongruent with the distribution in natural scenes), whereas in Experiment 1B, nontargets (orange) were sampled from a bimodal distribution centered on cardinal orientations (congruent with natural scenes). C, Biases estimated from Experiment 1A (left) initially (block 1) indicated repulsion from the cardinals (positive bias). Bias steadily decreased across the adaptation blocks (blocks 2–7). During the postadaptation block (block 8), response biases were significantly less repulsed from cardinal orientations than preadaptation. Conversely, when the stimulus distribution was consistent with the distribution of orientation in nature (Experiment 1B, right), bias estimates remained unchanged by adaptation. D, Comparisons between preadaptation and postadaptation biases highlight the significant reduction in repulsion away from the cardinal axes after exposure to the incongruent distribution only.