The aim of this study was to investigate whether the eating desire would be lower in the presence of facial expression of an obese than of a normal-weight eater in participants who were or not themselves obese. Normal-weight and obese participants assessed their desire to eat liked and disliked foods. These foods were presented alone and with a normal-weight and obese eater expressing pleasure, disgust, or neutrality. Results showed that, compared with a normal-weight eater, perceiving an obese eater decreased the viewer's desire to eat, whatever his/her facial expression. Thus, pleasant faces of normal weight but not of obese eaters increased the eating desire. Furthermore, the influence of eater's facial expressions did not differ as a function of the participants' BMIs. These data were discussed in the framework of the embodiment theory of emotion and of their implications in terms of nutritional education, either by enabling people to learn to like certain unpalatable foods or by helping them moderate their food intake simply through the sight of an obese eater.
Social cues are important determinants of eating behavior. It has been shown that individuals eat more and spend more time eating in the presence of others than alone (1,2). However, recent studies have also revealed the important role of other people's emotional facial expressions on our own eating behavior. When participants perceived another eater expressing either disgust or neutrality, their desire to eat liked foods decreased (3,4,5). In contrast, when the eater expressed pleasure, the desire to eat disliked foods increased. In the theoretical framework of the embodiment theories (6), this is explained by the perception of other's emotional facial expression that induced the same emotion in the observer by an automatic imitation of this facial expression. This imitation activates the mirror neuron system that produces the sensorimotor experience in the observer of perceived emotion (7). For example, it has been shown that the same brain areas (the left anterior insula and the right anterior cingulate cortex) were activated both when the individuals observed disgust in others and when they experienced disgust themselves (8).
However, the influence of the other on eating behavior may also differ as a function of the eater's BMI. The purpose of this original study was to examine whether individuals have less desire to eat with obese people than with normal-weight people. No study has attempted to examine this topic. In our diet-conscious society, which preaches the ideal of thinness, obesity is considered as a form of social deviance (9), and obese individuals are repeatedly targets of social stigmatization (10). Obesity is indeed associated with a wide variety of negative characteristics such as laziness, lack of willpower, or ugliness (11). It is also associated with eating disorders (12) and represents a serious risk to health (diabetes, hypertension, heart disease, etc.). Thus, watching obese people should also activate the idea that too much eating is bad for health, even if the eater is expressing pleasure. Due to negative characteristics attributed to obese people and their eating-related behavior, we may assume that participants have less desire to eat when in the presence of obese compared to normal-weight eaters, and this even if he/she expressed a positive emotion.
However, people's attitude toward obese people may depend on the group to which they themselves belong, i.e., whether they are or are not themselves obese. Various studies in social psychology have shown that people identify more with and are more influenced by the members of their own group (e.g., race, gender) (in-group members) than those of other groups (out-group members) (13). Thus, in line with this in-group effect, viewing an obese eater might decrease the desire to eat, but to a lesser extent in obese participants (in-group advantage) than in normal-weight participants (out-group). However, as discussed further latter, an alternative assumption might be that there would be no in-group advantage for obesity. There is evidence that obese and normal-weight individuals share the same negative stereotypical beliefs about fat people (10). Consequently, the decrease in the desire to eat when faced by an obese eater could be similar whatever the participant's BMI.
In sum, our study was aimed to investigate whether the influence of emotions (disgust, neutrality, pleasure) expressed by another eater on the participant's own desire to eat varies as a function of the eater's BMI. We expected that eating desire would be lower in the presence of the facial expression of an obese rather than of a normal-weight eater. Furthermore, we crossed the group membership of the eaters and participants in order to highlight the any potential in-group advantage (obese participants/obese eater). The in-group advantage would be observed if the eating desire of obese participants was greater in the presence of an obese eater than of a normal-weight eater. However, obesity is also associated with other symptoms, such as depression (14), restriction (15), and hyperempathy (16). We therefore measured these possible moderator variables. We also measured the degree of hunger in order to predict distortions in the evaluation of eating desire.
Methods and Procedures
Seventy-five subjects (41 women and 34 men aged from 19 to 37 years (28.5 ± 4.9)) were recruited in Clermont-Ferrand through advertising in the local newspapers. Based on their BMI, they were assigned either to the obese group (BMI >30) or to normal-weight group (BMI between 19.5 and 24.9). There were 37 normal-weight participants (20 women, BMI = 21.1 ± 1.8, and 17 men, BMI = 22.6 ± 1.6), and 38 obese participants (21 women, BMI = 34.1 ± 3.1, and 17 men, BMI = 33.7 ± 3.4). None of the participants were receiving medical treatment. They participated for a payment of €10.
The participants were seated in front of a computer screen in a laboratory setting. The stimuli consisted of photographs of six different food products presented on the computer screen: three liked (chocolate, French bread, cream cake) and three disliked (cooked sausage with vegetables, kidneys, black pudding). These liked and disliked foods have been pretested by Gil et al. (17). In this study, young adults rated the intensity of the emotion “to like,” aroused by each photograph of 13 different food products, on a 7-point scale, from 0 “I do not feel this emotion” to 6 “I feel it very strongly.” The three liked foods used in the present study received the highest mean ratings, and the three disliked foods the lowest mean ratings (Table 1).
Table 1. Means and standard deviations of the evaluation of the “to like” emotion toward the food pictures on the 7-point liking scale, from 0 “I do not feel this emotion” to 6 “I feel it very strongly”
There were also photographs of eight eaters seated at a table: two men and two women for each BMI. Each eater expressed three different emotions: pleasure, neutrality, and disgust. Each of the six food products was presented with an eater expressing each facial expression (Figure 1). For the four normal-weight eaters, the emotional faces had been pretested and selected for the quality and the intensity of the expressed emotions in Rousset et al.'s (4) (Table 2). The emotional faces for the obese eaters were tested in the present study. Sixteen obese models, eight men (age = 30.3 ± 3.4, BMI = 33.6 ± 2.8) and eight women (age = 28.4 ± 4.2, BMI = 32.8 ± 3.7) were photographed expressing three different emotional expressions (i.e., pleasure, neutrality, and disgust) in the same conditions as those used by Rousset et al. (4). These photographs were evaluated by 32 additional participants (age = 23.47 ± 4.1, BMI = 21 ± 1.9). They rated each face on the basis of nine emotions (anger, disgust, fear, sadness, guilt, surprise, interest, satisfaction, pleasure, and neutrality) on a 7-point scale from 0 “the face does not express this emotion” to 6 “the face strongly expresses this emotion.” The photographs obtaining the highest emotional scores were selected. However, unlike in the case of the normal-weight models, it was difficult to find two obese females who portrayed all three emotional expressions well. We were therefore obliged to select three obese women instead of two in order to obtain the right scores for each emotional expression (Table 2). Furthermore, the selected photographs of obese eaters were assessed by a Facial Action Coding System coder (18).
Table 2. Means and standard deviations of the intensity of the perceived emotions of the four normal-weight eaters and the five obese eaters which best expressed disgust, neutrality and pleasure on a 7-point scale, from 0 “the face does not express this emotion” to 6 “the face strongly expresses this emotion.”
In addition, the participants were asked to complete four questionnaires. The Beck Depression Inventory 13 (19) which assesses depressive symptoms on 13 items. This questionnaire provides a depression score between 0 and 39, allowing a classification of individuals as no depression (0–3), mild depression (4–7), moderate depression (8–15), or severe depression (>16). The Dutch Eating Behavior Questionnaire (20) assesses restrained eating on 10 items and provides a score of restriction between 0 and 5. The empathy quotient (EQ) (21) assesses the ability to understand and share another person's feelings. The EQ ranges from 0 to 80. Between 0 and 32 the EQ is weak; between 33 and 52, it is average; between 53 and 63, it is above average; and between 64 and 80, the EQ is very high. The last questionnaire assesses the sensation of hunger with a hunger score of between 1 and 7 (22).
The participants completed the four questionnaires presented in random order, and performed the eating desire test. The questionnaire-test order was counterbalanced.
In the eating desire test, the participants assessed the intensity of their desire to eat the food products shown in each of the photographs, on a horizontal nonstructured scale (from the left, “I have no desire to eat,” to the right, “I have a great desire to eat”). The scores varied between 0 and 10. They were two series of photographs (food presented alone vs. food presented with an eater), with the series order being counterbalanced across subjects. The first series consisted of the six different photographs, one for each food product presented alone (three liked and three disliked). The second series consisted of 72 different photographs (4 × 3 × 6), i.e., four eaters (two normal-weight and two obese, one of each sex) for each of the three emotional expressions (disgust, neutrality, and pleasure), and for all six food products. Within each series, the photographs were presented in random order.
Depressive symptoms scores
In our study, the mean depressive symptoms score for the obese participants (mean = 7.04) indicated that they were suffering from mild depression whereas the normal-weight participants were not depressive (mean = 2.48). The ANOVA run on these scores with the participants' BMI and sex as between-subject factors found a main effect of BMI, F(1, 71) = 17.24, P < 0.001, but no main effect of sex, nor any interaction between BMI and sex, F(1, 71) = 0.06, F(1, 71) = 0.001, respectively, all P > 0.05. This confirmed that the depression score was higher for the obese than for the normal-weight participants. As indicated by the Pearson correlation coefficient, there was a positive correlation between the BMI and the depression scores, r = 0.43, P < 0.01: the higher the BMI score was, the greater the symptoms of depression.
Restrained eating scores
The ANOVA carried out on the restrained eating scores revealed a main effect of BMI, F(1, 71) = 20.32, and of sex, F(1, 71) = 13.76, all P < 0.001, but no interaction between these two variables, F(1, 71) = 0.24, P > 0.5. Thus, the restrained eating score was higher in women than in men (2.83 vs. 2.34). Similarly, the obese participants had a higher restrained eating score than the normal-weight participants (2.88 vs. 2.29). There was a positive correlation between the BMI and the restrained eating score, r = 0.46, P < 0.01, with BMI scores corresponding to more restrained eating behavior.
The ANOVA run on the EQs indicated a main effect of BMI, F(1, 71) = 6.56, and of sex, F(1, 71) = 6.85, all P < 0.05, but no interaction between these two variables, F(1, 71) = 0.88, P > 0.05. Thus, women had a higher EQ than men (33.09 vs. 29.29), and obese participants than normal-weight participants (33.05 vs. 29.33). There was a positive correlation between the BMI and the EQ, r = 0.27, P < 0.05, with higher BMIs corresponding to higher EQs.
The ANOVA carried out on the hunger scores revealed neither a main effect of sex, F(1, 71) = 0.23, or of BMI, F(1, 71) = 3.30, nor a BMI × sex interaction, F(1, 71) = 0.02, all P > 0.05. In line with these results, there was no correlation between the BMI and the hunger score, r = −0.15, P > 0.05.
To summarize, the BMI was significantly correlated with depressive symptoms, restrained eating and empathy. Consequently, these three different scores were introduced as covariates into the subsequent analyses of variances. However, only depression had a significant effect. The other two covariates were therefore removed from the analyses and the presentation of the results.
Desire to eat food products in a nonsocial context: Foods presented alone
An analysis of Covariance (ANCOVA) was run on the eating desire score for the food presented alone with the participant's BMI and sex as between-subjects factors and food category (liked vs. disliked) as within-subjects factor, and depression as covariate. This ANCOVA revealed a main effect of depressive symptoms, F(1, 68) = 4.66, P < 0.05, but no significant interaction involving this factor. However, the correlation between depressive symptoms and the desire to eat food presented alone did not reach significance, r = 0.16, P > 0.05. The results also indicated a significant main effect of food category, F(1, 68) = 3.90, P = 0.05, and a significant food category × participant's sex interaction, F(1, 68) = 4.93, P < 0.05, but no significant main effect of sex, F(1, 68) = 1.16, P > 0.05. Thus, this interaction indicated that the desire to eat the liked foods tended to be higher in the women than in the men (6.56 vs. 5.81), F(1, 73) = 3.10, P = 0.08. In contrast, the desire to eat the disliked foods was similar in both the women and the men (2.06 vs. 2.56), F(1, 73) = 1.08, P > 0.05. There was no main effect of BMI, F(1, 68) = 0.14, P > 0.05, or any interaction involving this factor. Thus, the desire to eat foods did not differ as a function of the individual's body weight.
Desire to eat food products in a social context: Effects of facial expressions
We calculated the difference between the scores obtained for the desire to eat food presented with an eater (social context) and food presented alone (nonsocial context). A general repeated ANCOVA was run on these difference scores with food category, facial expressions, and eater's BMI and sex as within-subjects factors, participant's BMI and sex as between-subject factors, and depression as covariate factors. This ANCOVA revealed no main effect of depressive symptoms but instead a significant depressive symptoms × food category × facial expression interaction, F(2, 140) = 3.18, P < 0.05. To study this interaction, we performed a linear regression analysis designed to investigate the proportion of variance accounted for by depressive symptoms. This analysis was conducted using depression as a predictor variable of the desire to eat the liked or disliked foods and was run for each emotional facial expression. The results showed that depressive symptoms only significantly predicted a decrease in the desire to eat the disliked foods when the eaters expressed disgust. Thus, the more depressive the participants were, the more sensitive they were to the emotion of disgust expressed by the eater toward disliked foods and the more their own desire to eat decreased, R2 = 0.06, P < 0.05. Depression scores were not significantly correlated with the other variables.
The ANCOVA also revealed a significant interaction between food category, facial expressions, eater's BMI, eater's sex and participant's BMI, F(2, 142) = 2.86, P < 0.05. In contrast, the effect of participant's sex did not reach significance and was not involved in any interaction. Consequently, this factor was removed from the subsequent analyses which were then performed separately for each facial expression.
In the case of the disgusted faces, the ANOVA performed on the difference scores between food presented with an eater expressing disgust and food presented alone revealed no main effect of participant's BMI, or any interaction involving this factor. In contrast, there was a significant main effect of eater's sex, F(1, 73) = 5.64, P < 0.05, but no interaction involving this factor. The one-sample t-tests revealed that the difference scores were significantly lower than zero both for the men eaters, t(74) = −4.89, and for the women eaters, t(74) = −5.46, all P < 0.001. Thus, the desire to eat was lower when the eater expressed disgust, and more so when the disgust was expressed by a woman (−0.81) than a man (−0.69). In addition, there was a significant interaction between food category and the eater's BMI, F(1, 73) = 6.92, P < 0.05 (Figure 2). This interaction indicated that the desire to eat the liked foods in the presence of a disgusted face again decreased more in the presence of an obese eater (−1.06) than of a normal-weight eater (−0.78). This phenomenon was observed for all the participants, both normal-weight and obese. The difference scores were significantly lower than zero for both the normal-weight eater, t(74) = −4.08, and the obese eater, t(74) = −4.46, all P < 0.001. In contrast, the desire to eat the disliked foods did not differ depending on the eater's BMI, F(1, 74) = 0.93, P > 0.05. The difference scores were significantly lower than zero in both the BMI groups, i.e., the normal-weight eaters, t(74) = −4.15, and the obese eaters, t(74) = −3.86, all P < 0.001. Thus, compared to food presented alone, a disgusted face decreased the desire to eat both liked and disliked foods, but, in the case of the liked foods, did so to a greater extent when the disgust was expressed by an obese eater.
In the case of the neutral faces, the ANOVA run on the difference of expressed eating desire between food presented with an eater expressing neutrality and food presented alone indicated no main effect of participant's BMI, nor any interaction involving this factor. The only significant interaction was between food category and eater's BMI, F(1, 73) = 4.61, P < 0.05 (Figure 3). For the disliked foods, the difference in desire to eat in the presence of neutral faces and food presented alone was similar for the normal-weight eater (−0.24) and the obese eater (−0.23), F(1, 74) = 0.07, P > 0.05. The one-sample t-tests revealed that the difference scores were significantly lower than zero for the normal-weight eaters, t(74) = −2.04, P < 0.05, and marginally lower than zero for the obese eaters, t(74) = −1.71, P = 0.09, thus indicating that, in the case of the disliked foods, both the normal-weight and obese participants' desire to eat decreased in the presence of neutral faces when compared to a presentation of the food without an eater. In the case of the liked foods, the desire to eat was lower in the presence of an obese eater with a neutral facial expression (−0.42) than in the presence of a normal-weight eater (−0.23), F(1, 74) = 4.78, P < 0.05. The one-sample t-tests revealed that the difference scores did not differ from zero for the normal-weight eaters, t(74) = −1.57, P > 0.05, but were significantly lower than zero for the obese eaters, t(74) = −2.40, P < 0.05. Thus, the desire to eat the liked foods decreased in the presence of an obese eater for all the participants, both normal-weight and obese.
Finally, in the case of the pleasant faces, the ANOVA run on the difference scores revealed a significant main effect of eater's BMI, F(1, 73) = 4.14, P < 0.05, but no other main effect and no interaction (Figure 4). Thus, whatever the food category and the participant's BMI, the desire to eat was greater in the presence of a pleasant facial expression worn by a normal-weight eater than by an obese one (0.33 vs. 0.20). The one-sample t-tests revealed that the difference scores were greater than zero for the normal-weight eaters, t(74) = 2.73, P < 0.01, but that they did not differ from zero for the obese eaters, t(74) = 1.72, P > 0.05. Therefore, compared with food presented alone, the desire to eat both the liked and disliked foods increased when only the normal-weight participants were in the presence of a pleasant face.
Our study showed that the participants' desire to eat was affected by the emotions expressed by the other eaters. Compared with foods presented alone, neutral and disgusted faces decreased the participants' desire to eat, whereas pleasant faces increased it. This result is consistent with those obtained by Barthomeuf et al. (3) which showed that viewing a person expressing pleasure in eating increases the desire to eat whereas seeing a person expressing disgust decreases it. According to embodiment theories (6), the effect of the perception of facial expressions on the perceiver's behavior is explained through his/her automatic imitation of the perceived emotions. As set out in our introduction, the afferent feedback from imitation induced the perceived emotion in the observer. However, the fact that the effect of the eater's facial expressions was modulated by the status of the eaters (obese or not) and the category of foods (liked vs disliked) partly calls into question the conceptualization of the embodiment process in terms of a simple automatic imitation mechanism. Indeed, our study showed that the photograph of a pleasant face still increased the desire to eat the liked foods but only when the face was that of a normal-weight eater. The disgusted faces also decreased the desire to eat both liked and disliked foods, but this decrease in the case of the liked foods was greater when the eater was obese than when he/she was normal-weight. Consequently, the eater's BMI played a critical role in the effect of emotional facial expressions on the desire to eat. This BMI-related influence may be the consequence of the stereotypes associated with obese people. Some studies have shown that normal-weight participants considered obese individuals to be less intelligent, less willing, more angry, ashamed, and greedy than normal-weight people (10). Obesity is also associated with eating disorders (12) and a risk to health. Thus, watching obese people should also activate the idea that too much eating is bad for health, even if the eater is expressing pleasure. The characteristics attributed to obese people should therefore have moderated the process of embodying other people's emotions by activating a negative assessment of the food or the food-related behavior. This explained why the participants were less eager to eat in the presence of obese eaters.
In addition, our results revealed no in-group advantage for the obese participants. Instead of feeling an increase in the desire to eat in the presence of an obese eater as was observed for the normal-weight participants in the presence of another normal-weight eater (3,4), the obese participants were negatively influenced by the presence of individuals belonging to the same category. Numerous studies have showed that stigmatized groups develop a positive, collective identity that fosters both awareness of discrimination and group pride and which allows them to sustain their self-esteem (23). However, unlike other stigmatized groups, obese individuals internalize stereotypes that are prejudicial to their own group and exhibit no in-group bias (24). In conclusion, our study demonstrates that the desire to eat varies not only as a function of the presence of others and the emotions that they express, but also as a function of their BMI.
Independently of this BMI-related effect, our study also highlighted the role of depressive symptoms in the obese participants, who were more depressed than the normal-weight participants. Furthermore, as the depression scores increased, the desire to eat the disliked foods in the presence of a disgusted face decreased, thus suggesting a greater sensitivity to the facial expression of disgust. This finding is consistent with studies showing that the judgment of the intensity of negative facial expressions is greater in depressed individuals (25). It is therefore particularly important to continue to pay special attention to the role of depression in obese participants, even though the obese participants in our study were suffering from mild depression and not severe depression. What is more, our results could have practical implications in terms of nutritional education. For example, viewing eaters expressing a pleasurable attitude toward foods which they do not like, such as vegetables, could lead people to like and eat them. This is consistent with Christakis and Fowler's study (26) that revealed a phenomenon of societal progression in obesity. For example, a person's chances of becoming obese increased by 57% whether he/she had a friend who became obese. This emphasizes the critical role of the social influence on food-related behavior. However, in order to fully understand and made reliable the results found in our study, we have to evaluate the effect of the eaters' emotional expression in a life situation as with real people sharing a meal with guests. In our study, we measured the desire to eat rather than the level of food intake, thus assuming that there is a strong link between these two variables. The desire to eat is effectively a good predictor of food intake (27). However, it is questionable whether the obtained effects would be similar in a food intake situation. We have also used validated pictures of facial expressions. However, in the conditions of a real meal, the effect of the emotion expressed by the other may changed in the dynamic of social interaction. Future research could attempt to generalize these results to the case of real food intake.
We thank Laurie Mondillon for assessing the obese facial expressions with the Facial Action Coding System, and Aurore Bernard for the taking of photographs of obese eaters. This research was supported by a grant from the French National Research Agency (ANR Blan06-2-145908 FaceExpress).