Search for a command to run...
Abstract Seven prospective studies of health behaviours containing eight prospective datasets testing the moderating role of intention stability on intention–behaviour and past behaviour–behaviour relationships were examined within the context of the Theory of Planned Behaviour. The measure of intention stability was based on lack of change in intentions between the two measurement time points. Across different behaviours, samples and time frames more stable intentions were associated with intentions at time 1 that were stronger predictors of behaviour at time 2. Intention stability did not consistently moderate the past behaviour–behaviour relationship. Across studies the frequency-weighted mean correlation between intentions and behaviour was substantially greater for stable (r + = 0.60) compared to unstable (r + = 0.27) intentions (Z difference = 6.65, p < 0.001). The past behaviour–behaviour correlation was also stronger for stable (r + = 0.50, p < 0.001) compared to unstable (r+ = 0.34, p < 0.001) intentions (Z difference = 3.12, p < 0.01). Perceived behavioural control was the variable most strongly related to stable intentions. Implications for understanding the role of health cognitions in the performance of health behaviour are discussed. Keywords: Intentionsintention stabilitypast behaviourhealth behaviour Notes Notes [1] The other measures of intention stability used by Conner et al. (Citation2000) were the sum of the absolute differences between scores on each item at the two time points, the number of items changing between time points, and the within-person correlation of items across time points. Where multiple intention items are available at both time points all these measures can be computed and an average calculated. This was only clearly applicable in the case of studies 1a and 1b where the application of the combined index based on four measures of stability produced substantively identical results. The index used here was strongly correlated with this combined index for studies 1a (r = 0.93, p < 0.001) and 1b (r = 0.95, p < 0.001). [2] Note that excluding attitude, subjective norms and PBC from these analyses did not substantively alter the findings for any study. [3] Following a suggestion from a reviewer, we further explored the lack of moderation effect for past behaviour. Two possibilities were explored. First, we tested if the impact of past behaviour was significantly weakened when intentions were strong (i.e. extreme) and stable. This was achieved by testing the significance of the three-way interaction between intentions, intention stability and past behaviour (after controlling for main effects and all possible two-way interactions among these variables; Aiken & West, Citation1991). In only one of the eight tests was a significant effect found (study 5: β = 0.20, p = 0.04). Second, following Conner et al. (Citation2000) we also tested the power of PBC stability to moderate the impact of past behaviour on future behaviour. It was only possible to perform these analyses in five out of eight cases. Only for one study was a significant effect found (study 6: β = 0.28, p = 0.01). [4] As a reviewer pointed out, these analyses can be problematic where stability is confounded with intention scores. We therefore compared the mean intention scores and the mean intention extremity scores (modulus of the intention score) for the high and low intention stability groups. For study 2 (intention: low stability M = 1.56, high stability M = 2.39, p diff = 0.01; extremity: low stability M = 2.12, high stability M = 2.63, p diff = 0.003), study 4 (intention: low stability M = 1.39, high stability M = 1.52, p diff = 0.05; extremity: low stability M = 1.39, high stability M = 1.52, p diff = 0.05), study 5 (intention: low stability M = 0.43, high stability M = 1.78, p diff = 0.001; extremity: low stability M = 1.54, high stability M = 2.49, p diff = 0.001), and study 7 (intention: low stability M = 0.85, high stability M = 1.53, p diff = 0.002; extremity: low stability M = 1.18, high stability M = 1.65, p diff = 0.001) these differences were significant. These comparisons for studies 2, 4, 5 and 7 should therefore be treated with caution. [5] It is worth noting that this anomalous finding may be attributable to the sample including respondents who were not sexually active for the duration of the study. The past behaviour–behaviour correlation across the sexually active portion of the sample was r = 0.39, p < 0.01, N = 49. [6] This significant difference remained even after excluding study 7 where the anomalous result for past behaviour was observed (footnote 5). Past behaviour–behaviour relationship: stable intentions (r = 0.51, p < 0.01), unstable intentions (r = 0.36, p < 0.01); difference (Z = 2.87, p < 0.01).