Experiment 1 has examined the effect of overtraining on shift learning in a simultaneous matching-(or nonmatching-) to-sample discrimination . Experiment 2 has examined how much overtraining is required for rats to form the concept of matching in a matchingto-sample discrimination . In Experiment 1, rats were overtrained on a simultaneous matching-to-sample discrimination or nonmatching-to-sample discrimination in a discrete threestimulus-presentation T-type jumping stand and were transferred to either a nonshifted (nonshift) or a shifted task (shift). Group Nonshift learned their subsequent shift task more rapidly than Group Shift, regardless of whether rats were tested on the matching-to-sample task or the nonmatching-to-sample one (symmetry of transfer effect) . Experiment 2 shows that the apparent asymmetry of transfer effect is observed at reaching criterion (OT-O) , whereas the symmetry of transfer effect is observed after 5 days, 10 days, 15 days, and 20 days of overtraining. These results indicate that overtraining is an obvious operational precondition for symmetry of transfer effect in matching-(or nonmatching-) to-sample discrimination in rats.Recently, untrained or derived relations among stimuli during a conditional discrimination have been studied in various approach methods with nonhuman subjects such as pigeons or rats. The various approach methods have been classified roughly into three categories. The first method is a shift-nonshift paradigm of Zentall and Hogan's design (1974) (Nakagawa, 1992a, Experiment 2 in 1993a, 1993b, 2000b, 2000c b, 2001 Urcuioli, 1977; Urcuioli & Nevin, 1975;Zentall & Hogan, 1974, 1976. In this paradigm, for example, Zentall and Hogan (1974) had trained some pigeons on matching and others on oddity with one pair of stimuli, and transferred them to a new pair of stimuli , either with same rule holding as in the first problem, or with the opposite. The pigeons that had not shifted to the opposite task performed Requests for reprints should be sent to Esho Nakagawa,