Conditional discrimination of series containing either two or three rewarded trials

Abstract Series of trials differing only in the number of rewarded trials (2R or 3R series) were used to test rats' ability to learn conditional numerical discriminations. Although the series were conditionally related to initially non-numerical situations, the rats successfully learned the discrimination. It is concluded that these discriminations can plausibly be conceptualized as conditional counting. A pair of transfer tests given after the discriminations were well established suggested that, under these conditions, numerical information will overshadow other non-numerical sources of discrimination made relevant by the transfer shifts.