Absolute versus relative coding affecting variability.

<p>(a) Predicted variability for a model in which variability is determined by the absolute value within the bounds [0, 1]. Blue curve indicates participants for whom the highest-valued tile was omitted, red the lowest-valued. The model predicts that the two curves will exactly overlap. (b) Predicted variability for a model in which variability is determined by the relative value within the range of all rewards experienced during learning. The model predicts that the two curves will be translated relative to each other. (c) Empirical standard deviations obtained in Exp 2.