Two separate experiments are reported. They show the methodological difficulties and subsequent conceptual complications in Berkowitz's modified frustration-aggression theory and Zillman's theory of aggression. There was a marginally significant difference in the verbal measure of aggression favouring subjects aroused by frustration over subjects aroused by exercise, but not on the behavioral measure of aggression. Exp. II was conducted to test an alternative hypothesis, deduced from Fraisse's theory of emotion, which states that subjects respond more aggressively when confronted with unexpected than with expected annoyance. All six measures of aggression confirmed the hypothesis. The results of these two experiments indicate conceptual and methodological difficulties inherent in Berkowitz's modified frustration-aggression theory and Zillman's theory of aggression; they also suggest that Fraisse's theory of emotion presents a better model of aggression.