This study (N = 396) utilized a 3 (restoration: prescript vs. restoration postscript vs. control) × 2 (language: forceful vs. non-forceful) × 2 (topic: exercise vs. nutrition) mixed-model design to examine the effects of the reactance process over time with attitudes and behavioral intentions used as covariates. Results indicate that compared to non-forceful language, forceful language was perceived as a threat and elicited reactance (measured by anger and negative cognitions). Interestingly, forceful language elicited more positive attitudes than did non-forceful language. Although not a true boomerang effect, behavioral intentions were greater at Time 1 and Time 3 than at Time 2. There were no other significant effects for attitudes or behavioral intentions over time. Restoration prescripts and postscripts were equally successful at reducing negative cognitions but not anger. However, when topics were split and examined individually, restoration scripts were successful at reducing negative cognitions for the exercise topic but not for the nutrition topic. This raises important theoretical concerns for psychological reactance theory (TPR; J. W. Brehm, 1966). Mainly, boundary conditions of TPR may need to be revised if the reactance process varies for particular topics.