Experimental rats were trained on multiple 2-odor discrimination tasks, whereas controls were given repeated sessions on Task 1 and then were tested on a novel 2-odor task. Experimental rats showed strong positive transfer across problems and approached errorless or near-errorless learning. Control rats maintained near-perfect performance on Task 1 but performed at chance on initial trials when tested with novel odors. Thus, the near-errorless terminal performance of experimental rats was a function their having been trained on multiple problems and was not simply the result of eliminating "disruptive response tendencies" (I. C. Reid & R. G. M. Morris, 1992). Results support the view that when rats are trained on a series of 2-odor discrimination tasks, they acquire a strategy or rule that allows them to solve new problems with few or no errors.