Study 1 manipulated both self-relevance and mood in one design and assessed their additive and interactive effects on information processing. Based on dual-process models of persuasion, this study assessed message elaboration, i.e., the extent to which participants engage in systematic processing, in two ways: (1) we manipulated argument quality in a health message in order to assess message elaboration, and (2) assessed the number and profile of issue-relevant thoughts (cf. Petty and Wegener 1999). We predicted that a positive mood would increase sensitivity to the quality of the arguments in a health message only when this information is self-relevant, i.e., under conditions of high personal vulnerability to a health risk, and thus increase persuasion for strong arguments, and decrease persuasion for weak arguments. In contrast, under conditions of low vulnerability, a positive mood was expected to induce heuristic processing, and thus decrease a differentiation between strong and weak arguments.
Participants and design
A total of 121 students at a university of a large city participated in the experiment. The sample consisted of 51 male and 70 female participants with a mean age of 22 years (SD = 2.40). Participants volunteered and were randomly assigned to experimental conditions. The design was 2 (vulnerability: high vs. low) × 2 (mood: positive vs. negative) × 2 (argument quality of persuasive message: weak vs. strong) between-subjects.
Procedure and independent variables
As a cover story, participants were told that they would take part in a computerized survey about personal values and experiences, Repetitive Strain Injury (RSI) and health. Participants were first presented with a short introduction about RSI and the “Boston RSI Sensitivity Test”, and were led to believe that this test measured their risk of developing RSI related health problems in the future on the basis of a personality test. Participants completed the test and were told that the computer analyzed their responses. Next, they received false feedback regarding their vulnerability to RSI on their computer screens. Participants were informed that their risk of developing RSI related health complaints was either quite high (high vulnerability condition), or quite low (low vulnerability condition). Next, participants read a message that described the negative health consequences of RSI (e.g., chronic complaints in arms and hands). This information was held constant across conditions.
Mood was manipulated following the procedure described by Raghunathan and Trope (2002). Participants were asked to recall three positive or negative events they had recently experienced. They were asked to describe each event in a few sentences, indicating in particular the details of the event that made the experience positive or negative for them personally. To boost the credibility of the mood manipulation in the present health context, participants were told that research has shown that personal experiences from the past can have a profound impact on health, and that the present research was designed to further test this possibility. Next, participants were exposed to the persuasive message, which was presented as a letter submitted to a journal of health psychology. The message described RSI prevention training as a way to reduce the risk of experiencing the negative health consequences of RSI, and was supported by five weak or five strong arguments that were selected in a pilot study (cf. Petty and Cacioppo 1986). Examples are: “Research has demonstrated that knowledge of RSI-prevention strategies improved physical and psychological well-being” (strong argument), and “The institute that developed the RSI-prevention training would not put the training on the market unless it was effective” (weak argument). After reading the health message, participants completed the manipulation checks of vulnerability, mood, and argument quality, and the thought-listing and attitude measures (in this order). At the end of the experiment, participants were thanked for their participation and were extensively debriefed, in order to make sure that they understood real purpose of the experiment. None of the participants guessed the real purpose of the experiment, or expressed any suspicion regarding the experimental procedure.
Six items assessed the effectiveness of the vulnerability manipulation on a 7-point scale. Examples are: “The chance that I will develop RSI related health problems is high” and “Due to my sensitivity to RSI I am more prone to RSI related health problems” (Cronbach’s alpha = .90). The effectiveness of the mood manipulation was assessed by 9 items from the Profile of Mood State (POMS, Wald 1984), measuring the extent to which participants felt specific emotional states, such as feelings of depression, and relaxation. Negative items were reverse-scored; higher scores yield more positive moods (Cronbach’s alpha = .75, see Wald 1984 for a complete listing of the items). Three items assessed perceived argument quality of the health message, measuring on a 7-point scale how strong, supportive and sensible participants rated the arguments (Cronbach’s alpha = .88). Correlations between manipulation checks did not differ as a function of experimental manipulations.
Cognitive Processing. Participants completed a thought-listing task, in which they were asked to write down the thoughts that came to mind while reading the health message (cf. Petty and Cacioppo 1986). No time limit was set; participants could take all the time they needed. Two independent judges categorized issue-relevant thoughts in thoughts confirming (Range 0–4; Kappa = .72) message content, e.g., “I am definitely interested in a RSI prevention training” or “I never knew RSI was that serious”, and thoughts discounting message content, e.g., “I feel the risk of RSI is exaggerated in this message” or “I am not convinced about the benefits of the proposed RSI prevention training” (Range 0–4; Kappa = .71), and neutral thoughts, e.g., “The text reminded me of a book I read some time ago” (Range 0–3; Kappa = .70). Disagreements between judges were solved through discussion. Issue-irrelevant thoughts were not analyzed (9% of total; cf. Petty and Wegener 1999). Confirming thoughts correlated negatively with discounting thoughts, r = −.28, p < .01, and with neutral thoughts, r = −.18, p < .05. Discounting thoughts also correlated negatively with neutral thoughts, r = −.26, p < .01.
A thought-listing index was created (cf. Sherman et al. 2000) using the following formula: (confirming issue-relevant thoughts + 1)/(total issue-relevant thoughts + 1). Index-scores that approach 1 indicate higher levels of agreement with the health message; scores that approach 0 indicate lower levels of agreement.
Attitudes. Participants’ attitudes toward RSI-prevention training were assessed by a semantic differential scale with 4 items, using a 7-point scale, including how valuable, and interesting participants thought this training was (Cronbach’s alpha = .76). Attitudes were significantly correlated with the thought-listing index, r = .45, p < .001, and correlations did not differ as a function of experimental manipulations.
The manipulation checks of mood and vulnerability were subjected to a 2 (vulnerability: low vs. high) × 2 (mood: positive vs. negative) analysis of variance (ANOVA), to ascertain that the manipulations of vulnerability and mood were successful in producing the intended effects without producing unwanted side effects. The 2 × 2 ANOVA on the manipulation check for vulnerability revealed the expected main effect for vulnerability, F(1, 117) = 15.74, p < .001. Participants in the high vulnerability conditions felt more vulnerable to RSI (M = 3.65, SD = 1.48) than participants in the low vulnerability conditions (M = 2.65, SD = 1.24). Mood did not affect perceptions of vulnerability to RSI, either as a main effect, F(1, 117) = 2.69, p > .10, or in interaction with the vulnerability manipulation (F < 1). A similar ANOVA on the mood-measure revealed a main effect for mood, F(1, 117) = 4.38, p < .05. Participants in the positive mood condition reported a more positive mood (M = 3.37, SD = .49) than participants in the negative mood condition (M = 3.13, SD = .76). Vulnerability did not significantly affect mood, either as a main effect (F < 1) or in interaction with the mood manipulation (F < 2). A one-way ANOVA on the manipulation check of perceived argument quality revealed that, as expected, arguments were perceived as stronger in the strong arguments condition (M = 4.30, SD = 1.24) than in the weak arguments condition (M = 3.70, SD = 1.28; F(1, 119) = 6.93, p < .01). In sum, these results confirm that the manipulations of mood, vulnerability and argument quality were successful.
Thought-listing. A 2 (vulnerability) × 2 (mood) × 2 (argument quality) ANOVA on the thought-listing index revealed a significant main effect for vulnerability, F(1, 113) = 8.88, p = .004, and a significant main effect for mood, F(1, 113) = 3.88, p = .05. Overall, participants had a more positive thought valence under conditions of high (M = .65, SD = .29) compared with low vulnerability (M = .52, SD = .23), and under conditions of positive mood (M = .63, SD = .26), compared with negative mood (M = .54, SD = .27). These main effects were qualified, however, by the hypothesized three-way interaction between vulnerability, mood, and argument quality, F(1, 113) = 4.83, p < .05. In line with predictions, simple effects analyses revealed that positive mood affected the extent of systematic processing, as indicated by the increased differentiation between strong and weak arguments on thought valence, only when vulnerability was high F(1, 113) = 8.98, p = .003 (in all other conditions, F < 1). Relevant means are displayed in Table 1.
Attitudes. A 2 (vulnerability) × 2 (mood) × 2 (argument quality) ANOVA on the attitude toward RSI management training again revealed a three-way interaction between vulnerability, mood, and argument quality, F(1, 113) = 4.58, p < .05. Similar to the results for the thought listing task, simple main effects analyses again revealed a differentiation between strong and weak arguments for positive mood participants, but only under conditions of high vulnerability, F(1, 113) = 4.48, p < .05 (in all other conditions, F < 1). Relevant means are displayed in Table 1. In sum, only participants in high vulnerability, positive mood conditions reported more positive thoughts and more positive attitudes toward the recommendation when argument quality was strong rather than weak. This sensitivity to argument quality was not observed in low vulnerability, and negative mood conditions.
The present findings support the hypothesis that a positive mood increases systematic processing of a threatening health message, particularly when the information is relevant to the self. When participants felt vulnerable to a health risk, i.e., under conditions of high self-relevance, a positive mood increased sensitivity to the quality of the arguments in the health message. This increased differentiation between strong and weak arguments is a reliable indicator of systematic message processing (Petty and Wegener 1999), and was observed on high vulnerability participants’ thoughts and attitudes regarding the health message. In contrast, when a health message was not relevant to the self, mood had no effects on information processing or persuasion (also see Das et al. 2003; De Hoog et al. 2005). These findings are a first demonstration that the effects of mood on the processing of self-threatening information are moderated by self-relevance. When a threatening message is relevant to the self, a positive mood may help receivers to ‘face the facts’, and engage in systematic processing. When a threatening message has no direct self-relevance, a positive mood is likely to promote hedonic concerns, and prompt heuristic message processing.