Effects of Social Norms Information and Self-Affirmation on Sugar-Sweetened Beverage Consumption Intentions and Behaviors

ABSTRACT

The separate and combined efficacy of a social norms and a self-affirmation intervention to motivate decreased sugar-sweetened beverage (SSB) consumption was examined in 2 experiments. College students were randomly assigned to receive information about SSB consumption risks, norms, both, or neither. In addition, participants performed either a self-affirmation or control task. Self-affirmation only weakly affected SSB consumption intentions and behaviors. However, participants in Experiment 2 who received risks information, norms information, or both reported greater SSB reduction intentions than did those who received no information. Two weeks later, those who received both types of information reported more frequent behavior change preparations, and it appears this effect may have been partially mediated by the changes in intentions to reduce SSB consumption.

Acknowledgments

We thank Lisa Liang and Spencer Michalek for their help in carrying out this project.

Notes

The number of days between the intervention and the follow-up did not differ as a function of condition ( ranged .00–.03), and controlling for number of days since the intervention did not alter the pattern of findings reported for either the preparatory behaviors or SSB consumption measures.

Of course it must be acknowledged that the assumption that mediation analysis actually provides evidence of mechanisms of change has been questioned for a number of reasons, including (but not limited to) the failure of most such analyses to meet the assumption that the predictor variable has been assessed without error (Kline, 2015), the possibility that the mediator and outcome measure are both caused by an unmeasured variable (Kline, 2015), and the tendency to use suboptimal cross-sectional designs that do not clearly establish the temporal order of the variables (Kline, 2015; Tate, 2015). In this particular experiment, we utilized a “measurement of mediation model,” which is the minimal design needed to justify conducting a mediation analysis (Kline, 2015). That is, given that our predictor is an experimental variable (i.e., we manipulated the information that participants received), the proposed mediator was measured after participants were exposed to the experimental manipulation, and the outcome (preparatory behaviors) was assessed 2 weeks later, temporal precedence among the predictor, the proposed mediator, and the outcome is established, thus rendering implausible the alternative interpretation that changes in preparatory behaviors actually produced changes in SSB reduction intentions. However, it is nevertheless necessary to interpret the results of our mediation analyses with caution given, among other things, that we did not manipulate the proposed mediator (e.g., SSB reduction intentions), nor can we be certain that SSB reduction intentions and preparatory behaviors do not share an unmeasured cause (i.e., are related because both are caused by a third variable).

To examine this issue we divided our sample in each experiment at the median into low and high SSB users based on their baseline beverage checklist scores. We then conducted analyses that compared high and low SSB users as a function of both information and self-affirmation on all outcome measures in both experiments. We also conducted analyses examining the information and self-affirmation condition effects on the outcome measures for only the high SSB users. In all of these analyses we found only weak effects of self-affirmation alone ( ranged .00–.02) or in interaction with information condition ( ranged .00–.06). Further, the weak interaction patterns that did exist were inconsistent across outcome measures (i.e., did not consistently show greater receptiveness to risk information in the self-affirmation condition). Of course, these results must be interpreted cautiously given that this experiment was not designed to address this question (i.e., we did not specifically recruit large numbers of high SSB users and then randomly assign them to condition) and the means in some conditions are based on as few as three participants.

We believe the result that several participants actually took more than the one beverage they were invited to take attests to the fact that participants generally were not aware that we would be able to determine what type of beverage they selected.