ABSTRACT
This study compared the relative efficacy of two strategies designed to mitigate psychological reactance in health campaigns by using reminders of behavioral autonomy: preemptive scripts, which appear before the appeal, and restoration postscripts, which appear after. Employing a mixed-model experiment with a 2 (threat to freedom: low vs. high) × 3 (reactance-mitigation strategy: control vs. preemptive script vs. restoration postscript) × 2 (health campaign topic: exercise vs. nutrition) between-subjects design and a within-subjects factor of time (immediate posttest measurement followed by a one-week delay), this study (N = 394) compared the effects of the two mitigation strategies on reactance, attitude, and behavioral intention at two points in time. Moderated mediation models indicated that the reactance-mitigation strategies equivalently reduced the degree to which reactance was experienced in response to increasingly threatening health appeals (relative to the control). This effect indirectly influenced behavioral intention via attitude change and remained after one week.
Disclosure of potential conflict of interest
We have no known conflict of interest to disclose.
Notes
1. The definition of a boomerang effect may be controversial given a lack of clear and consistent explication of it (Hamilton et al., Citation1993). In reactance literature, some studies explicitly examine a boomerang effect (e.g., Bessarabova et al., Citation2013) while others do not (see Rains, Citation2013, meta-analysis of reactance research), focusing instead on whether high-threat messages result in diminished message effectiveness (in terms of attitudes, beliefs, or behavioral intentions) relative to a low-threat message. The latter – a reduced-persuasion effect – is established by comparing high- and low-threat messages, and a boomerang effect is demonstrated by examining the effects of a high-threat message relative to a baseline or no-message controls (Kumkale & Albarracín, Citation2004).
2. Balance tests indicated that demographic variables were equally distributed across experimental conditions.
3. We found equivalent results when using indicator and sequential coding schemes.