The relative effectiveness of five procedures for the training of individuals to perform tasks under stress was tested in a criterion situation, where subjects were requested to perform a visual search task under the threat of electric shocks. During training on the task, different groups of subjects received shocks of (a) criterion-level intensity; (b) milder than criterion-level intensity; (c) gradually increasing intensity; (d) randomly varying intensity. The last group (e) received no shocks at all. The results pointed to three conditions for the enhancement of training effectiveness: (a) minimal interference of exposure to stressors with task acquisition, (b) familiarity with stressors characteristic of the criterion situation, and (c) absence of unrealistic expectations about future stressors. However, none of the five training procedures meets all three conditions. Implications for the design of procedures whereby persons can be trained to perform proficiently under stress are discussed.