The most important finding of the present study is that action planning significantly predicted health protective behavior (i.e. fruit consumption) as well as the restriction of health risk behavior (i.e., high-caloric snack consumption). Our results showed a better model fit when action plans were added to the model with only attitudes, social influences, self-efficacy and intentions, indicating that the prediction of both types of behavior significantly benefited from the incorporation of action planning, thereby conforming our first hypothesis. When viewed in the light of the literature on other health behaviors, such as physical activity [20
], sun protection behavior [34
], and (vitamin) pill intake [60
], our findings with regard to fruit consumption support the notion that action planning may be an important strategy to promote health protective behaviors and suggest that current social-cognitive models on health protective behavior should be extended by incorporating volitional cognitions that facilitate the transition from motivation to behavior. Whereas most previous observational studies that found a behavioral influence of action planning failed to incorporate a measure of past behavior in the analyses, the present study accounted for the influence of past behavior in the extended analyses. Even after the inclusion of past behavior, which is generally the most powerful predictor of future behavior, action planning remained significant, which demonstrates that action planning significantly predicted behavior change. These findings corroborate results from several intervention studies, in which the formation of action plans has been shown to increase the performance of health behaviors [e.g., [20
]]. The interplay between action planning and past behavior was outside the scope of the present study. Thorough examination of this relationship would, however, be an interesting direction for future research, as this may yield important information on theoretical modeling and practical application of planning strategies in individuals with high and low levels of past behavior.
Our findings with regard to snack consumption verify these suggestions and broaden their scope to include both health protective as well as health risk behaviors. The present study is the first to explicitly compare the predictive value of planning in both types of behaviors and found that the predictive value of action planning was equally powerful in the promotion of fruit consumption and the restriction of snack consumption. These findings confirm our third hypothesis and indicate that one and the same type of planning can be applied in both types of health behaviors.
Other important findings pertain to the established mediating and moderating effects of action planning (hypothesis 2). The longitudinal correlational design of the present study allowed us to examine the nature of the influence that action planning exerts in the intention - behavior relationship. Our findings of full mediation in the fruit consumption study and partial mediation in the snack consumption study confirm our hypothesis and correspond to results of previous studies, in which both full [e.g., [32
]] and partial [e.g., [30
]] mediation have been found in various behaviors. The difference in mediating effects may pertain to the strength of the underlying intentions. Wiedemann and colleagues [64
] have demonstrated that the strength of the mediated effect of action planning increases along with levels of intentions. The relatively low intention with regard to restricted snack consumption, as compared to fruit consumption, may therefore have precluded full mediation of the intention - behavior relationship by action planning.
The results with regard to potential moderating effects of action planning partially confirm our second hypothesis. A positive moderating effect of action planning was demonstrated in the fruit consumption study, thereby replicating previous reports of moderation of the intention - behavior relationship [e.g., [30
]]. However, only a small trend with regard to the moderation effect was found in the snack consumption study. The insignificance of this effect may, again, be explained by relatively low motivation scores; the overall intention towards restricted snack consumption was substantially lower than the intention to eat sufficient amounts of fruit, which may have precluded the appearance of moderating effects of action planning in the snack consumption study. Besides the proposition to incorporate action planning in existing, traditional social-cognitive models, these findings provide suggestions on how and where to integrate the concept; action planning can tentatively be considered as a mediator as well as moderator in the intention - behavior relationship. It should, however, be mentioned that the present consideration of action planning as concurrent mediator and moderator, is at odds with the conceptualization of moderators as being unaffected by the status of a predictor variable [e.g., [65
]; but see [67
]]. In the present study, action planning was measured at T2 in order to adequately investigate its mediating influence. Although this measure may be tentatively considered as a proxy for a baseline measure of action planning, application of the latter would have resulted in a stricter conceptualization and testing of the moderation effect. This limitation should be taken into account when interpreting the current findings and future studies would do well to incorporate longitudinal measurements of mediating and moderating variables.
Furthermore, whereas the four previous studies used a similar type of action planning [i.e. implementation intentions; [24
]], the current study used a different approach. Instead of focusing on when, where, and how a goal-directed response will be implemented (i.e. eating fruit, not eating snacks), the formation of specific preparatory plans was emphasized. Although the former type of planning, i.e. implemental planning, has been subject of substantial research efforts to decrease to intention - behavior gap, the latter planning mode, i.e. preparatory planning, has also been shown to reliably predict a variety of health behaviors [e.g., [17
]]. Moreover, one of our previous studies compared the behavioral influence of both types of behaviors and found that preparatory planning outperformed implemental planning in the prediction of fruit consumption [69
]. Further, preferably experimental, research is, however, recommended to substantiate the present findings and optimize planning concepts and interventions for both health protective and health risk behaviors. In doing so, the application of coping planning as a protocol for restriction of health risk behavior may be reckoned with. Coping planning is a barrier-focused strategy that pertains to the identification of risk situations and the specification of suitable coping responses [70
]. As this strategy has been shown to reliably predict performance of health behavior in the face of barriers [20
] and has been successfully applied to the restriction of health risk behavior, such as smoking [73
] and binge-drinking [74
], comparison of the benefits of this and other types of planning may yield vital knowledge for the optimization of planning interventions.
Limitations of the present study need to be acknowledged. First, the lack of validity statistics with regard to the behavioral assessment of snack consumption should be mentioned and calls for caution in the interpretation of the results regarding this measure. Items used in the present study were part of a food frequency questionnaire to estimate total and saturated fat intake [54
] and although this questionnaire has been previously validated, there are currently no specific validity statistics available for the selection of items used to measure snack consumption.
Second, a relatively low explained variance of snacking behavior was found, indicating that other motivational, volitional, and/or environmental factors need to be taken into account for the prediction of snack consumption. The low explained variance is, however, not uncommon, as dietary behavior is generally not well-predicted with explained variances of 30% and higher being exceptions rather than the rule [76
]. Furthermore, although ultimately this study aims at optimizing the prediction of fruit and snack consumption, the primary purpose was to investigate the influence of action planning in the intention - behavior relationship. We therefore only took three other direct predictors of the behaviors into account (past behavior, intention and self-efficacy), whereas most previous studies included many more determinants, often resulting in higher explained variances [e.g., [49
]]. Third, data were collected from a random sample of adults that were all members of an existing internet research panel. As these respondents voluntarily participate in surveys and receive incentives for their participation, the degree to which the findings generalize to the Dutch population at large may be limited. However, the demographic characteristics of the participants in both study samples corresponded rather well to demographic distributions within the Dutch adult population [5
], rendering substantial reduction of the external validity of our results unlikely. Furthermore, attrition was found to be somewhat selective in the snack consumption sample as lower educated participants were more likely to drop out. This attrition bias may limit internal and external validity of the study. However, general attrition rates were equal in both study samples and the influence of educational level as a covariate was not significant. It is therefore unlikely that the main results of this study have been compromised as a result of attrition.