Pigeons were exposed to discrimination training procedures using one-key multiple schedules of reinforcement and to generalization testing along dimensions of the two training stimuli. In the first two experiments using intradimensional discrimination training, both extinction and signalled reinforcement suppressed key-peck rate in one component and produced positive behavioural contrast, but only the extinction-trained group showed peak shift. Interdimensional training was used in the next four experiments and again the effects of stimulus control were compared. Both procedures resulted in excitatory dimensional stimulus control around the stimulus associated with the unchanged component but only the extinction procedure resulted in inhibitory dimensional stimulus control around the conditioned inhibitory stimulus, during generalization testing in extinction. Dimensional stimulus control was also investigated using two further types of generalization test, viz. combined-cue and resistance-to-reinforcement. However, these did not in general add anything to the analysis based on generalization testing in extinction. The results were discussed in the light of Spence's theory of gradient summation, which they supported, and Terrace's account of the by-products of discrimination learning, which they did not.