Four experimental groups and a control group, of five pigeons each, were trained to successively discriminate a vertical amber line from one oriented either 30 or 60 deg to the right. Depending on group, pecks were intermittently reinforced by either 2.5 or 5.0 sec access to the food hopper. During discrimination, the subjects were provided a second response key, a peck upon which provided a 5-sec time-out (TO) from either the positive or the negative stimulus. Frequency of response on both the TO key and the discrimination key was monitored throughout. After the discrimination training, all subjects were given a postdiscrimination generalization test. Although peak shift was not obtained, the results support the hypothesis that byproducts of successive discrimination learning are functions of magnitude of S+ reinforcement and degree of S+/S-separation. It was concluded that the results are consistent with a conflict theory of by-products of discrimination learning.In successive discrimination training, a subject is alternately presented two stimuli: S+, which is correlated with reinforcement, and S-, which is correlated with nonreinforcement or extinction. Three well-known "by-products" usually accompany the development of differential stimulus control in this procedure. These are peak shift, positive behavioral contrast, and S-aversiveness. 1 According to Terrace, each of these by-products disappears with extended discrimination training (Terrace, 1966(Terrace, , 1971, and none develops if the discrimination is acquired without "errors," that is, responses to S- (Terrace, 1966).According to Terrace (1971, 1972), by-products of discrimination learning originate in the occurrence of frustration-producing responses in the S-condition. However, this view has not held up well to experimental test (see, e.g., Rilling, 1977, pp. 469-470). Moreover, Terrace's theory cannot readily account for by-products occurring in the context of aversive conditioning procedures wherein S-is correlated with the nonaversive condition (Hearst, 1969, p. 36).In noting the foregoing problems, Hearst (1969) has questioned the parsimony of the S-aversiveness concept and, in the same stroke, made the case for a more general gradient-interaction model of discrimination learning such as the one developed by Spence (1937). Such a model is not without convincing empirical support (see, e.g., Hearst, 1968Hearst, , 1969 Terrace, 1971), it would seem to be incumbent upon proponents of gradient-interaction theory to show how the theory might be expanded to account for the foregoing byproducts.In this spirit, the purpose of this study was to investigate aspects of a general gradient-interaction theory, wherein it is proposed that S-aversiveness and related by-products of discrimination learning are functions of dispositional conflict at S-between inhibitory and generalized excitatory tendencies accruing to that stimulus as a result of discrimination training.The experiment below specifically examines two conflict-affecting variables: degree of S...