Abstract An asynchronous model for the dynamics of neural networks admits learning behaviors characteristic of classical and operant conditioning provided that appropriate plasticity algorithms are chosen. Stimulus generalization and discrimination can also be observed. Studies of such psychological phenomena are carried out by computer simulation of networks with designated sensory, association, and motor neurons, and the results are compared to those for live subjects. Various prescriptions for plasticity are investigated, including those corresponding to reward, punishment, and unlearning routines. These are characterized by their effect on network stability as quantified by a newly proposed stability measure.