Abstract
Greater fear arousal is associated with greater engagement with persuasive messages, and negative information and events are more potent than their positive counterparts. Hence loss-framed persuasive appeals, which emphasize the undesirable outcomes of noncompliance with the communicator's recommendations, should elicit greater message processing than do gain-framed appeals, which emphasize the desirable outcomes of compliance. But a meta-analytic review (based on 42 effect sizes, N = 6,378) finds that gain-framed messages engender slightly but significantly greater message engagement than do loss-framed messages. This effect is apparently not a result of whether the appeals refer to obtaining or averting negative (e.g., “skin cancer”) rather than positive (e.g., “attractive skin”) outcomes.
Thanks to Lisa Benz Scott, I-Huei Cheng, Rama Jayanti, Brett Martin, Robin Nabi, Lijiang Shen, and Monique Turner for supplying primary-research information.
Notes
a The coding judgments, in order, are: topic category (1 = disease detection, 2 = disease prevention, 3 = other); gain kernel-state language (1 = desirable states, 2 = undesirable states, 3 = both desirable and undesirable states, 4 = indeterminate); loss kernel-state language (1 = undesirable states, 2 = desirable states, 3 = both desirable and undesirable states, 4 = indeterminate); and message-engagement assessment (1 = number of thoughts; 2 = memory for message; 3 = other; 4 = multiple assessments).
∗p < .05, ∗∗p < .01, ∗∗∗p < .001.
a Power for detecting a population effect size of r = .10, assuming large heterogeneity, with a random-effects analysis, .05 alpha, and a two-tailed test (Hedges & Pigott, Citation2001).
Each of these 42 effect sizes is based on a unique human sample (and distinct message pair) and thus is statistically independent of the others, with the exception of the six effect sizes from Shen's (Citation2005) two within-subjects experiments. Replacing those six cases with mean effects for Shen's Study 1 (r = −.018, N = 286) and Study 2 (r = .166, N = 252) yields 38 cases with results virtually identical to those from the analysis of 42 cases: mean r = .056 (N = 5,302), 95% CI limits of .014 and .098, p = .008; Q(37) = 78.4, p < .001.
By comparison, the overall k-weighted average effect of need-for-cognition on message-engagement outcomes, expressed as a correlation, is .15. The specific mean effects are .17 for information recall (k = 23), .16 for responsiveness to argument-quality variations (k = 11), and .10 for number of thoughts (k = 10; Cacioppo et al., Citation1996, pp. 229–231).
A version of this paper was presented at the annual meeting of the National Communication Association in November 2005.