Rats treated with scopolamine (0.5 mg/kg SC daily) during the acquisition of a discrimination task with symmetrical negative reinforcement (light-go, noise/light-no go) showed a learning impairment, with both active and passive avoidance deficits. In the initial stage of such training, however, fewer passive avoidance errors and more active avoidance errors were made by treated animals if active avoidance pretraining had occurred in the no-drug state. A similar experiment using the same stimulus arrangement with asymmetrical reinforcement (no punishment of intertrial, and no go signal, responses) showed a scopolamine effect consisting mainly of increased responding to extinction signals and during intertrial intervals, with little or no active avoidance deficit. Furthermore, interactions due to changes in treatment conditions in successive stages of training were minimized in the latter task, suggesting that the effects of the shift-no shift factor on distribution of errors in the early stages of active-passive avoidance learning were unlikely to have been due to a genuine drug dissociation. Overall, these results and others obtained previously in the same and related tasks tend to rule out some unidimensional explanations of antimuscarinic effects, e.g., response disinhibition (an exclusively motor deficit) or impairment of stimulus sensitivity (an exclusively sensory deficit). The data rather confirm the notion of a sensorimotor drug bias leading to a shift in response prepotencies depending jointly on stimuli, responses, and response consequences. Prior learning history and behavioural compensation for adverse treatment consequences at the reinforcement level may interact with the sensorimotor bias so as to produce "set perseveration" (perseveration of response tendencies).