Social norms comprise a common construct of several theoretical models currently widely used to predict health-related behaviors and inform the development of behavior change interventions. Despite this, social norms remain relatively inconsistently conceptualized across studies, perhaps explaining the inconsistent findings relating social norms to the key behavioral outcomes they are hypothesized to influence. The present study assessed social norms that have been termed 'descriptive', and asked about behavioral outcomes that were neither overly general (such as 'being active') nor highly specific (such as 'exercising at a criterion level of intensity for at least 30 minutes at a time at least five days a week for the last 6 months'). Results showed that all of the social norms examined showed at least a trend level of correlation with the particular behavioral outcome they were hypothesized to influence. In all cases but two, these associations were statistically significant. Further, with the exception of social norms predicting fruits and vegetable consumption, these associations remained significant after adjusting for social support for either healthy eating or for physical activity, two constructs established as consistent predictors of their respective behavioral outcomes [2, 3]. Acknowledging the cross-sectional study design, this suggests that social norms may be potentially important determinants of physical activity and eating behaviors, and that this influence may be independent of the effects of the more well-established predictor, social support.
These results are not entirely consistent with those of previous studies, in which social norms have been inconsistently associated with physical activity and healthy eating [9, 19]. This may be due to our efforts to conceptualize and measure social norms using items that had good face validity, were not too broad or complex, and were likely to be clear and easily interpreted by respondents. In the Povey study, for example, the researchers used a complex healthy eating index , and Chatzisarantis used a broader measure of physical activity (the Godin Physical Activity Questionnaire) . Our results are consistent with one of the few studies to have examined the contribution of both social norms and social support to predicting physical activity , which found that both constructs contributed independently to predicting leisure-time physical activity.
There are several potential explanations for the associations observed in this study between social norms and physical activity and eating behaviors. For example, women who observe many others engaging in particular physical activity or eating behaviors may come to view these behaviors are 'normative' or socially desirable, and may adopt the same behaviors due either to a positive attitude about the behaviors, a shared belief in their value, and/or a strong social urge to confirm and 'fit in' to society. Alternatively, women who engage in these behaviors themselves may consequently be more likely to come into contact with women who engage in similar behaviors. Due to the cross-sectional design of the present study, this reverse direction of effects cannot be ruled out, and the influence of social norms on behaviors should be confirmed in prospective and experimental studies.
It should be noted that in this study, descriptive norms were operationalized in relation to one referent group - neighbors/people known to the participant - while social support was operationalized in relation to another group - family/work colleagues/friends. This conceptualization was considered most theoretically appropriate with regards to the contexts in which the specific behaviors might occur (for instance, eating fruit/vegetables is more likely to occur with family/work colleagues/friends than with neighbors or others). However, we cannot rule out whether the independent role of descriptive norms and social support was observed because the two constructs tapped different referent groups, rather than because the mechanism for normative influence is independent of social support. Similarly, social norms may motivate family/colleagues' social support, a possibility that was not examined within this study. Future research investigating hypothesized mediating effects between social constructs such as these is warranted.
Strengths of this study include the large sample size and the examination of social norms-behavior associations after adjustment for the more commonly-assessed construct of social support. In addition to the cross-sectional design, limitations of the study include the self-report nature of all constructs, although established measures were used where possible (e.g., the IPAQ-L to measure physical activity). Social norms were assessed using single, non-validated items only. It is also important to note that the conceptualization of social norms in this study, that is, the extent of agreement with statements about whether other people (in general, or those known to the respondent) engage in particular physical activity or eating behaviors, represents only one of several means of operationalizing this construct. Whether social norms conceptualized in other ways are independently predictive of health-related behaviors (assessed with varying levels of specificity) remains to be investigated. Further, the majority of the social norms questions in this study asked about 'other women', rather than other people generally. The alignment between the gender of participants and of the referent group in the social norms question could have resulted in stronger associations than may be observed in mixed-gender studies. In addition, there was not always precise correspondence between the behaviours assessed with the social norms, social support, and behavioural outcome indicators. Finally, this study did not assess other types of social support (e.g., instrumental or informational support), or social constructs such as social ties or social capital; nor did it attempt to investigate a comprehensive theoretical model predicting the outcome behaviors. Inclusion of more extensive measures of social context was not possible within the constraints of the present broader study, in which a large number of intrapersonal, social and physical environmental variables were examined.