Abstract
As millions of people turn to social media for health information, better understanding the factors that guide health-related judgments and perceptions in this context is imperative. We report on two Web experiments (n > 400 total) examining the power of society’s widespread weight bias and related stereotypes to influence nutrition judgments in social media spaces. In Experiment 1, meals were judged as lower in nutritional quality when the person who recommended them (the source) was depicted as obese rather than of normal weight, an effect mediated by stereotypic beliefs about the source as a generally unhealthy person. Experiment 2 replicated this effect, which—notably—remained significant when controlling for objective nutritional information (calories and fat content). Results highlight spillover effects of weight bias that extend beyond person perception to color impressions of objects (here, food) that are associated with stigmatized attributes. Implications for everyday nutrition judgments and public health are considered.
Notes
1 Ten meal images were selected from various online sources and pretested to ensure they were perceived as ambiguously healthy, namely, black bean and cheese quesadilla, chicken bagel sandwich, chopped salad with croutons and dressing, corned beef with vegetables, fruit and cheese plate, grilled vegetable quesadilla, pancakes with blueberries and syrup, sliced beef with vegetables, spinach salad with bacon and hardboiled eggs, and a vegetarian sandwich with sweet potato fries. For added validity, some images contained a tally of “likes” common in social media. Likes had no effect on our dependent variables are not discussed further.
2 Thumbnail images were selected from a larger set of before-and-after weight-loss images after pretesting revealed them to be closely matched on facial appearance.
3 We employed this single-item measure of overall healthfulness rather than more specific questions about the nutritional content of foods (e.g., estimated calorie content) in light of evidence that responses to the latter vary widely and are often vastly inaccurate (Carels et al., Citation2007; Lansky & Brownell, Citation1982).
4 A mixed-measures ANOVA featuring source weight status as the between factor and food healthfulness judgments as the repeated factor was also significant, as was a separate mixed-measures ANOVA featuring perceived healthfulness of the source as the repeated factor (Fs > 9.30, ps < .001).
5 Continuous variables (weight-watching concern, exercise frequency, and BMI) were mean-centered prior to constructing the interaction terms. Gender was coded 0 = female, 1 = male. Weight-watching concern did not vary by experimental condition, neither here nor in Experiment 2 (Fs < 1, ns).
6 We solicited multiple ratings of perceived source healthfulness to allow participants to update their impressions based on additional information (i.e., recommended meals). Suggesting that the weight status manipulation was effective, the source was perceived as significantly less healthy when she appeared obese (M = 2.78, SD = 1.09) as compared to of normal weight (M = 5.08, SD = 0.95), t(224) = 16.91, p < .001.
7 We also tested an alternative mediation model featuring perceived food healthfulness as the mediating variable, which revealed a smaller but still significant indirect effect, equal to –0.13 with a 95% confidence interval of –0.29 to –0.01.
8 As in Experiment 1, a mixed-measures ANOVA featuring source weight status as the between factor and food healthfulness judgments as the repeated factor was also significant, as was a separate mixed-measures ANOVA featuring perceived healthfulness of the source as the repeated factor (Fs > 13.33, ps < .001).
9 As in Experiment 1, an alternative mediation model featuring perceived food healthfulness as the mediating variable revealed a smaller but still significant indirect effect (–0.14 with a 95% confidence interval of –0.26 to –0.05).
10 As an anonymous reviewer suggested, our observation that the present effect was mediated by explicit endorsements of the source as a generally healthy/unhealthy person is at least suggestive of more systematic processing of the weight status cue.