Abstract Investigated the effects of fear arousal and sidedness variables on compliance with a dietary regimen. Experiments involved 202 women volunteers who were 10 per cent or more overweight, and aged 20‐60 years. Experiment 1 involved a 3 (low, medium, high fear) × 2 (single, multiple exposure to fear message) × 2 (one‐, two‐sided communication) design. The fear levels involved discussing the health hazards of obesity. Experiment 2 manipulated the fear‐message position relative to the recommendations (fear‐recommendations, recommendations‐fear, fear‐delay‐recommendations). Persuasive impact was measured via follow‐up weight checks at 2, 4, 8, and 16 weeks. Results indicate nonsignificant effects from sidedness and pre‐standardised fear levels. Using subjects' fear arousal ratings medium fear is significantly better (p <.025), supporting Janis's curvilinear hypothesis. A single exposure to the fear message is superior (p <.025) to multiple exposures, the interaction with time being highly significant (p <.001). Experiment 2 results indicate the optimum position for the fear message as immediately prior to recommendations (p <.025). Results support both cognitive and fear‐reduction hypotheses, but the latter is favoured.