The present experiments with goldfish were attempted to examine the effects of US and CS intensities on classical conditioning of shuttling response. Every subject was given daily 10 classical conditioning trials, each of which consisted of a 10 sec CS immediately followed by a 0. 4 sec US. In Experiment I, 80 subjects were given either of eight intensities (0, 0. 2, 0. 4, 0. 6, 0. 8, 1. 0, 1. 2, 1. 4V/cm) of US. The results showed that, performance remained at low but substantially higher than zero in the low US intensites (0, 0. 2, 0. 4V/cm), then increased up to 0. 8V/cm, thereafter unchanged (Fig. 1, 2, 3). It was suggested that, with considering high level of general activity, some care should be taken when the CS control group is used in similar experiments. In Experiment II, 50 subjects were given either of five intensities (0, 25, 50, 75, 100V) of CS. The results showed that each asymptotic level of performance referred to a linear function of the CS intensity (Fig. 4, 5, & 6). The results seemed to contradict with the prediction from the Rescorla-Wagner model, which assumes that the CS intensity does determine the rate of conditioning, but not its asymptotic level.