Fig 4. Effect of other options’ outcomes.

A) Rank bias as a function of training diversity. Y-axis shows the percent of trials involving similarly rewarded images in which subjects chose the image that ranked higher during learning. A subject’s image rankings were based on how many times the subject chose each image over the other images it was pitted against (best, second-best, or worst, see Methods for further details). The plot depicts individual subjects’ choices (circles), group distributions (violin), group means (thick lines) and standard errors (gray shading). B) Effect of reward histories on choice. The plot shows the log odds effect on choice of three types of reward. Own: differences in reward history of currently available images. Current alternative: differences in reward history when rejecting one available image in favor of the other available image. Other: differences in reward history when rejecting one of the available images in favor of any other image.