Social modeling of food choices in real life conditions concerns specific food categories

Eating is a complex social event, and the social context during a meal can have multiple influences on food intake. For instance, it has been demonstrated that the quantity of food consumed increases when eating with familiar others compared to eating alone, which is known as the social facilitation of eating (de Castro & Brewer, 1992). However, not only the mere presence of others, but also their consumption can have an impact on intake. Indeed, it has been shown that individuals adjust the amount of food eaten to the quantities consumed by their commensals (Vartanian et al., 2015). This phenomenon is called social modeling and involves using others’ eating behavior as a norm, for instance as an indicator of the appropriate amount of food to consume in a given situation. Social modeling appears to be very robust because it has been observed in both men and women (Cruwys et al., 2015) (with some evidence of a stronger effect for women (Herman & Polivy, 2010)), when eating with both familiar and unfamiliar partners (Cruwys et al., 2015; Kaisari & Higgs, 2015; Salvy et al., 2007; Vartanian et al., 2015), and independently of weight status (Rosenthal & Marx, 1979) and state of hunger (Goldman et al., 1991). Additional studies have demonstrated that social modeling can occur even in the absence of others, when participants are provided information regarding the quantity of food consumed by previous eaters (“remote confederate” studies) (Robinson, Benwell, & Higgs, 2013; Vartanian et al., 2013). In such studies, a norm of consumption is established via descriptive norms, which can be indirectly conveyed, e.g. via the presence of empty packaging, or conveyed via messages that report the consumption patterns of a majority of individuals (social-norm based messages).

While social modeling of food intake is well established, less is known about social modeling of food choices. Two reviews (Cruwys et al., 2015; Robinson, Thomas, et al., 2013) examined the literature on the impact of social modeling on food intake and choices, and both concluded that the available data is insufficient to draw conclusions about the robustness of the effect on food choices. Indeed, among 69 studies reviewed by Cruwys et al. (2015) on social modeling, only 11 examined modeling of food choices, among which 8 succeeded in observing the phenomenon. However, 3 studies did not find significant modeling effect (Hendy & Raudenbush, 2000; Pliner & Mann, 2004). For instance, Pliner and Mann (2004) reported social modeling of food intake but not of food choices. These authors suggested that food choices may be less influenced by others’ behavior than intake because individuals feel surer about their food likes and dislikes than the appropriate amount of food to consume in a given situation.

Pliner and Mann (2004) were also interested in the impact of food healthfulness on modeling, and they observed social modeling of intake for “unhealthy” (high energy density) cookies but not for “healthy” (low energy density) ones. To date, little is known about the strength of social modeling effects on “healthy” food items because the majority of studies have been done using high energy density food items and only a small number of studies have focused on modeling of low energy density food items. In a study by Hermans et al. (2009), social modeling of quantities of low energy density food (vegetables) was observed, but investigations of the social modeling of food choices of low versus high energy density food has been limited. Robinson and Higgs (2013) found that participants were less likely to choose low energy density food items when eating with a participant making “unhealthy” choices, than when eating alone or in the presence of a participant making “healthy” choices (Robinson & Higgs, 2013). Thus, social modeling of food choices was observed, but this influence was only present in the “unhealthy” condition. In another study conducted by Burger and colleagues, participants were led to believe that previous participants took either a “healthy” or an “unhealthy” snack through the provision of a descriptive norm (empty packaging) before having to make their own choice. Participants were more likely to choose the snack they believed others had chosen, both in the “unhealthy” and “healthy” norm conditions (Burger et al., 2010).

One feature of these studies is that they were conducted in a laboratory setting, which leaves open the question of whether the modeling of food choices occurs in real-life situations. To date, one study investigated modeling of choices in real life conditions, but this study was focused on modeling of vegetarian versus non-vegetarian dishes (Christie & Chen, 2018). Further research is needed to better characterize the effect of social modeling of food choices in real-life conditions, especially for meals composed of a broad range of food items including low and high energy density food items. In addition, there are other factors which could impact modeling effects in real life settings that are often not present in laboratory studies. For instance the majority of studies conducted in the laboratory included subjects paired with strangers, whereas in real life situations people are likely to eat with familiar others. To date, only a few studies have examined modelling among familiar participants, but the results suggest that modeling may occur both with familiar and unfamiliar individuals (Cruwys et al., 2015; Kaisari & Higgs, 2015; Salvy et al., 2007; Vartanian et al., 2015). Further investigations are needed to confirm these preliminary results and examine whether familiarity moderates modeling observed in real life settings.

The aim of the present study was to determine whether modelling of food choices can be observed in a real-life setting (a university self-service restaurant), examine whether modelling exists across a range of food categories (defined using consumer perception of the nutritional quality of the items) and whether familiarity with the person that serves as a model moderates any effects.