An experiment is reported which tests Fazey and Hardy's (1988) catastrophe model of anxiety and performance. Eight experienced crown green bowlers performed a bowling task under conditions of high and low cognitive anxiety. On each of these occasions, physiological arousal (measured by heart rate) was manipulated by means of physical work in such a way that the subjects were tested with physiological arousal increasing and decreasing. A repeated-measures three-factor ANOVA was used to test the hysteresis hypothesis that the performance x heart rate graph would follow a different path for heart rate increasing compared with heart rate decreasing in the high cognitive anxiety condition, but not in the low cognitive anxiety condition. The ANOVA revealed the predicted three-way interaction of cognitive anxiety, heart rate, and the direction of change in heart rate upon performance, with follow-up tests indicating that the interaction was due to hysteresis occurring in the high cognitive anxiety condition but not in the low cognitive anxiety condition. Other statistical procedures showed that, in the high cognitive anxiety condition, subjects' best performances were significantly better, and their worst performances significantly worse, than in the low cognitive anxiety condition. However, the results did not provide unequivocal support for the catastrophe model of anxiety and performance.