Two experiments were conoucted to examine the effects of congitive variables on GSR activities. In Experiment Ⅰ,four groups of subjects (8 each), named as HF, LF, HV, and LV, received different conditioning trials and extinction trials. They differed from each other according to the probabilities of shock (high or low) and the schedules of shock reinforcement (fixed interval or variable interval). The data during both conditioning trials and extinction trials indicated that there was non-significant difference either between High Probability Group and Low Probability Group, or between Fixed Interval Group and Variable Interval Group, so far as the GSR amplitude was concerned. Thus, the theories that the greater the probility of quantified threat, the greater the fear response, and the more unpredictable the source of threat, the greater the experimental anxiety were not supported. Experiment 2 was undertaken to test two assumptions derived from Ellis´ theory relating implicit verbalization to emotional arousal. The 11 subjects in Agitated Group were instructed to silently read sequence of sentences designed to elicit negative effect (to increase anxiety) and the other 11 subjects in Soothed Group, read sequence of sentences designed to elicit positive effect (to decrease anxiety). No signigicant difference was found between these two groups. The assumption that differernt kinds of self-verbalization serve to elicit different GSR deflections was not supported. The experiment was also designed to investigate the effect of sentence types (Observation, Inference and Evaluative Conclusion) on GSR activities. A 2×2 analysis of covariance of data yielded non-significant difference between mean GSR amplitude elicted by Inference sentences and Evaluative sentences as a whole and that elicited by Observation sentences. Ellis´ theory that "The anxiety response assoicated with sequence of sentences originates not with the observation, but with the inference, and the evaluative conclusion" needs further investigation.