Explaining adolescents’ cigarette smoking: A comparison of four modes of action control and test of the role of self-regulatory mode

Abstract

The present study compared how well four modes of action control (intentional, habitual, reactive and stereotype activation) explain adolescents’ cigarette smoking, and examined whether individual differences in self-regulation (locomotion and assessment tendencies; Higgins, Kruglanski, & Pierro, 2003) moderate the behavioural impact of the respective modes. Findings from a prospective questionnaire survey showed that (a) willingness, prototype perceptions and past behaviour–but not intention–predicted smoking behaviour, and explained 63% of the variance, and (b) the assessment mode of self-regulation moderated the past behaviour–future behaviour relation such that past behaviour had less impact on future smoking behaviour at high levels of assessment. These findings suggest that adolescents’ smoking is controlled by stereotype activation, habitual and reactive processes. Implications of the results for designing effective adolescent smoking cessation programmes are considered.

Keywords:

• action control
• adolescents cigarette smoking
• intention
• willingness
• prototypes
• automatic
• locomotion
• assessment

Acknowledgements

This research was supported by ESRC (RES-000-22-0260). We are indebted to Nanne de Vries for his formative input as Action Editor of this article.

Notes

Notes

1. Our rationale for selecting measures of past and future behaviour with different response formats derives from the shared method variance problem: the use of identical measures of past and future behaviour (e.g. number of cigarettes smoked during a certain time period) means that past behaviour and future share method variance–variance that is not shared by the other measures (intentions, prototypes, etc). This leads to overestimation of the strength of the past behaviour and future behaviour relation compared to the relation between other cognitions and behaviour (Ajzen, 1991). For instance, Conner, Warren, Close, and Sparks (1999, Study 3) showed that the variance explained by past behaviour differed by 77% when past behaviour and future behaviour were measured using the same scale as compared to different scales. Thus, our use of different scales to measure past behaviour and future behaviour permits a more valid correlational test for the relative importance of the different modes of action control than would have been the case if we had used identical measures of past and future behaviours.

2. We acknowledge that it would have been desirable to obtain test–retest correlations for the entire sample. Unfortunately, however, lack of time and resources precluded this option (data had to be collected during class time and without incentives for either the school or students).

3. We also analysed the data in three other ways to double-check our conclusions. First, when we dichotomised past behaviour into smoked versus not smoked and future behaviour into smoked versus not smoked, logistic regression indicated that the prototype interaction term (evaluation × similarity) was the only significant predictor of behaviour (β = 0.02, SE = 0.01, p = 0.05); intention, willingness and past behaviour were not significant predictors (β = 0.43, 0.60 and 1.17, SE = 0.29, 0.37 and 1.25, respectively, p > 0.10). Second, when we dichotomised past behaviour into regular smoker versus irregular/non-smoker and dichotomised future behaviour into smoked versus not smoked, then past behaviour predicted behaviour (β = 2.68, SE = 1.22, p = 0.03) and intention and willingness did not (βs = 0.48 and 0.42, SEs = 0.35 and 0.39, respectively, ps > 0.17); however, the prototype interaction term still had a marginally significant β (β = 0.02, SE = 0.01, p = 0.077). Finally, when we excluded past behaviour from the original linear regression analysis, willingness (β = 0.42, p < 0.001), intention (β = 0.34, p < 0.01) and the prototype interaction term (β = 0.30, p < 0.01) predicted behaviour. Our interpretation of these findings is that prototype perception (the stereotype activation mode) is the most robust predictor of future smoking behaviour across different types of analyses in the present study.