g., Fridlund & Cacioppo, 1986; Niaura et al., despite 1998; Waters et al., 2004). Data analyses The hypotheses and research questions were first tested using repeated measures analyses of variance (ANOVAs) with smoking cue as the within-subject factor and argument strength (manipulated to be low vs. high, not evaluated during the study) as the between-subject factor. Gender was then added as a potential moderator in the second set of repeated measures ANOVAs. Results We found no significant differences between participants in the two argument conditions in any demographic, individual differences, or smoking behaviors assessed in the study; none of these variables were controlled in subsequent analyses. Simple hypothesis tests Smoking urges. A significant quadratic main effect of smoking cue, F(1, 94) = 4.

2, p < .05, partial ��2 = .04, and a marginally significant linear interaction between smoking cue and argument strength, F(1, 94) = 3.8, p = .06, partial ��2 = .04, were found on smoking urges. The results are graphed in Figure 1. Smoking urges started at a relatively high level for smokers in both argument conditions (M = 3.4, SD = 1.2). Urges decreased after participants watched the no-cue advertisements (M = 3.3, SD = 1.3) and increased slightly after they watched smoking cue advertisements (M = 3.4, SD = 1.4), which produced a significant quadratic effect. The linear interaction resulted from the smoking urge after watching smoking cue advertisements. Whereas smokers in the weak argument condition significantly increased their smoking urge after watching smoking cue advertisements, Mpre cue = 3.

3, SDpre cue = 1.3 vs. Mpost cue = 3.5, SDpost cue = 1.4, t(47) = 2.6, p = .01, smokers in the strong argument condition did not, Mpre cue = 3.3, SDpre cue = 1.3 vs. Mpost cue = 3.2, SDpost cue = 1.5. Thus, smoking cues increased smoking urges only for advertisements with weaker arguments. Figure 1. Interaction between argument strength and smoking cue on smoking urge (p < .06). Heart rate. Smoking cue had a significant main effect on heart rate change (measured in beats per minute). Advertisements with smoking cues were associated with a larger heart rate reduction (M = �C1.2, SD = 1.9) than advertisements without smoking cues (M = �C0.4, SD = 1.6), F(1, 88) = 14.4, p < .001, partial ��2 = .14. Argument strength interacted with smoking cues on heart rate change, F(1, 88) = 6.

1, p < .02, partial ��2 = .07, such that the difference in heart rate change between no-cue and smoking cue advertisements was significant only for advertisements with weaker arguments, Mno cue = �C0.3, SDno cue = 1.5 vs. Mcue = �C1.5, SDcue = 2.1, t(44) = 4.5, p < .001. The magnitude of the heart rate change is consistent with previous Anacetrapib literature (Carter & Tiffany 1999; Kelly, Barrett, Pihl, & Dagher, 2004). Figure 2 presents the results.