Using conditioned suppression of barpressing to investigate the stability of a conditioned stimulus-unconditioned stimulus (CS-US) association, we gave water-deprived rats either a few pairings of the CS with a strong footshock US or many pairings with a weak footshock US so that barpress suppression in response to the CS was equated. Experiment 1 established training parameters that yielded this equivalence. Specifically, rapid acquisition to a preasymptotic level of responding with strong shock produced suppression comparable to the asymptotic level reached more slowly with weak shock. Experiment 2 showed that although equivalent performance was obtained from extensive conditioning with a weak shock or limited conditioning with strong shock, only extensive conditioning with weak shock resulted in retarded acquisition of an association between that same CS and a footshock level perceived as midway between the two initial training shock intensities as implied by asymptotic performance in Experiment 1. Experiment 3 demonstrated that the observed retardation in animals given many conditioning trials with weak shock was CS specific. Collectively, these findings indicate that the malleability of learned behavior is not simply a function of initial associative strength but is dependent on path during initial acquisition.