"Rats were run on a series of position-discrimination reversals in a one-unit Y-maze, either with a reversal after every ten trials (ten trials given per day) or after achieving a criterion of nine out of ten correct choices on a given day. While both groups achieved one-trial position-discrimination reversal, the course of the learning differed between the groups. The trials-criterion group improved as a rectilinear function of ordinal position of the problem, while the performance-criterion group improved as a curvilinear function of ordinal position of the problem.