The aim of the present study was to investigate how a person's inferential processes concerning the causes of his own physiological arousal would affect the affiliative tendency, anxiety level, and the frequency of GSR nonspecifics. 28 male high school students served as Ss. Each S was given a lactose-filled pill and led to anticipate an “second experiment” which ostensibly needed a painful injection. Ss in the first group were led to believe that the pill would arouse them. Ss in the second (control) group were told that it would cause symptoms irrelevant to arousal. Ss in the third group were informed that it would make them feel relaxed. It was predicted that Ss in the first group would show lessened affiliative tendency and anxiety compared to the control group, as they were allowed to attribute their arousal to an extrinsic source. It was also predicted that Ss in the third group would show greater affiliative tendency and anxiety than would the control group as a belief that arousal had been artificially, reduced would lead to an inference that they were affected in a particularly strong way by the stimulus. Major findings were as follows ; (1) Ss who could attribute arousal to the pill were less anxious than Ss in the control and the third group when they were led to anticipate a stressful situation. (2) Ss in the first group showed lessened autonomic arousal compared to Ss in the third group. (3) No relationship was found between the induced arousal processes and the affiliative tendency. A tentative interpretation for the obtained difference in the actual autonimic arousal was suggested on the basis of Beck's model. Some implications for application of “attribution therapy” to clinical practice were also suggested.