45 subjects were assessed for cognitive anxiety on the Sport Competition Anxiety Test. Two months later they observed a person performing a new motor task which required high cognitive processing to be performed well. After this observation, 22 subjects were randomly assigned to a Mental Practice and 23 to a Control group. The former performed a cognitive rehearsal of the task, whereas the latter did not. None practiced the task physically before being tested. Analysis of variance showed that both errors and performance time interacted significantly with Mental Practice versus Control group scores and scores on the Sport Competition Anxiety Test. Among subjects who practiced mentally, those scoring low on cognitive anxiety performed significantly better than subjects who scored high. Further, the relationship between test scores of cognitive anxiety and performance for the total sample was analysed by different curvilinear regression models. The cubic model fitted the data better and accounted for a greater percent of variance on error performance explained by anxiety test scores (R = .39) than the linear correlation (r = .25). This cubic model formed a polynomial relationship between cognitive anxiety test scores and error in performance.