Causal Information on Children's Attitudes and Behavioural Intentions Toward a Peer With Obesity

There is much evidence that the prevalence of overweight and obesity among school-aged children has increased in recent decades across all countries, including Ireland [1]. Overweight and obesity in childhood is associated with a variety of physical [2] and psychosocial problems [3]. From a psychosocial standpoint, obesity is considered to be ‘one of the most stigmatizing and least socially acceptable conditions in childhood' [4], p. 1818[]. There is evidence that young children rate peers with obesity as less liked and less preferred as friends or playmates than they do average-weight peers or peers with other stigmatizing attributes (e.g., facial disfiguration, being wheelchair-bound) [5]. Research also shows that negative characteristics, such as being lazy and having few friends, are more likely to be attributed to children who are overweight [6].

Given common negative stereotypes of children who are overweight or obese, it is not surprising that they often experience peer rejection [7]. Children who are overweight are more likely to experience social marginalization [8], peer victimization [9], and weight has been cited as the primary reason for bullying in schools [10]. Studies have also shown that weight-based stigmatization is associated with low self-esteem [10], body dissatisfaction [11], low physical activity levels [12] and depressive symptoms [9]. Despite evidence documenting the negative consequences of weight bias, few published studies have specifically attempted to reduce stigma and negative attitudes toward children who are overweight or obese [13]. Interventions aimed at reducing levels of weight bias among children have met with mixed success [7,14,15]. This may be partly the result of an inadequate understanding of factors that contribute to weight bias in children and of demographic differences in the levels of its endorsement [16]. Factors that contribute to children's weight-based stigmatization need to be better understood to develop effective programmes to reduce weight stigma.

Attribution of causality has been suggested as an important factor in the formation of children's attitudes toward obesity. According to Weiner's attribution theory [17], affective responses to other people should be more positive when the cause of their medical condition or physical disability is uncontrollable than when it is controllable. When negative attributes of a person are seen as controllable, the individual is viewed as being personally responsible for them and thus will be more stigmatized than when the attribute is viewed as uncontrollable and the individual is not seen to be personally responsible. For instance, DeJong [18] investigated how adolescent girls' perceptions of an obese peer were influenced by their beliefs about the cause of her obesity. Findings showed that unless the obese target could offer an ‘excuse' for her weight (e.g., glandular disorder) she was less liked and evaluated less positively than a normal-weight peer. Similarly, another study found that the provision of low-responsibility information reduced children's tendency to hold an obese girl responsible for her condition [19]. However, it did not alter their liking for the obese girl, and liking was found to be related to the nature of the condition, and not to its perceived cause. In a similar study, children, aged 9-11 years, attributed less blame to an obese peer whose weight was attributed to medical causes. The provision of this medical information had little effect on overall attitudes, especially among older children [7]. In contrast, a study among 8- to 12-year-old children found that greater beliefs about the controllability of obesity were positively associated with the strength of belief in negative stereotypes about overweight individuals [20]. Taken together, these studies suggest that attributions about the causes of obesity may play a role in expressions of weight bias towards children. To date, little is known about the effectiveness of providing different types of information about the causes of overweight.

The examination of environmental factors in weight-based attitudes and intentions is important given that these factors are often the target of behavioural weight management interventions [21]. Recently, many public health campaigns such as the ‘Let's Move' [22] initiative in the USA or the ‘Little Steps' [23] healthy living campaign in Ireland emphasize the role of the environment in the childhood obesity epidemic. These messages may have an impact on youth and their behavioural intentions toward an overweight peer. However, no such study to date has examined environmental factors in weight-based attitudes and intentions.

Over the past decade, Ireland has experienced a rapid change in the social and physical environment, and research has shown that 19% of Irish children aged 9 years are overweight and 7% are obese [24]. This has important implications for the health of children. The current investigation extends our understanding of weight stigma by focusing on a sample of children outside of the USA.

The aim of the study is to examine the effect of different types of causal information on children's attitudes and behavioural intentions toward a peer presented as overweight. This study builds on previous research by investigating the efficacy of medical causal information versus environmental causal information in reducing weight-based stigmatization. The current study was informed by Bell and Morgan's [7] methodology. Children were randomly assigned to read a vignette of a peer in one of three conditions: overweight with no causal information, overweight with biological information explaining the overweight, overweight with environmental information explaining the overweight. In addition, all children read a vignette of a peer depicted as average weight. The current study presented children with female targets in the vignettes, as obese females experience a greater degree of social stigmatization of obesity than obese males [25]. In addition, weight-related stigmatization is more likely to affect girls than boys [26,27]. Therefore, given that females are particularly vulnerable to weight biases and experience greater societal pressure to adhere to the thin body ideal than males, it was decided to employ female targets in the vignettes.

A second goal of this study is to explore a possible social desirability effect in the reporting of behavioural intentions. The recent attempts to promote positive attitudes and reduce stigma toward overweight peers [28] may cause children to respond in a ‘socially desirable' manner. To explore a possible social desirability effect, children are asked to rate in a Shared Activity Questionnaire (SAQ) not only their own behavioural intentions toward the overweight peer (SAQ-self) but also their perception of their classmates' behavioural intentions (SAQ-other). Finally, given mixed findings regarding whether boys and girls express different levels of weight bias [14,29], the present study aimed to examine possible sex differences in attitudes and behavioural intentions toward an overweight female peer following provision of specific information about the causes of overweight.

The following hypotheses were tested:

• Children will rate the average-weight child more positively than the overweight child on the Adjective Checklist (AC) and the SAQ-self.

• There will be a significant effect of type of causal information presented about overweight (biological, environmental, no causal information) on children's ratings on the AC and the SAQ-self.

• Due to a potential social desirability bias, children's ratings of the peer who is depicted as overweight (biological, environmental, no causal information) will be lower (less favourable) on the SAQ-other than the SAQ-self.

Of the 176 participants, 92 were boys and 84 girls from the three most senior classes in the Irish primary school system (4th, 5th and 6th class) of an urban school and ranged in age from 9 to 12 years (mean = 11.23, SD = 0.89). Nearly all participants were Irish (97%). Parental information leaflets and consent forms were distributed by class teachers. Only children whose parent/guardian, class teacher and the school principal had consented to their participation and who had also provided written consent themselves were involved. Table 1 shows the demographic breakdown of the sample by school year, sex and experimental condition.

Children were randomly assigned to read a vignette about a girl presented as overweight in one of three conditions, with each indicating a different cause for the obesity: i) biological: overweight presented as the result of a biological condition (a glandular problem), ii) environmental: overweight was the result of environmental factors (poor diet and no exercise), and iii) no causal information: no information was presented about the cause of the girl's overweight. Furthermore, all children received a vignette of a girl presented as average weight (the control target). A between-within subjects design was employed in this study where each participant was randomly assigned to one of three obesity conditions, and all participants received the average-weight control condition. A between-subjects design was selected over a within-subjects design for the obesity condition to avoid practice/carry-over effects and to reduce the cognitive burden for children that would be required with employing a within-subjects design, where participants would be required to read four vignettes (each obesity condition and average-weight condition) and answer the same questions on each vignette multiple times.

The vignettes presented across the three experimental obesity conditions were similar with the exception that, for the biological condition, the following explanation for the peer's overweight was also provided: Anna is a girl who is about the same age as you. She is average weight compared to most children her age. Anna likes to read and do puzzles and she likes to read every night before bed. For the overweight peers, all three vignettes stated: Suzy is a girl who is about the same age as you. She is overweight compared to most children her age. Suzy likes to read and do puzzles and she likes to read every night before bed. The vignettes presented across the three experimental obesity conditions were the same with the exception that, for the biological condition, the following explanation for the peer's overweight was also provided: Suzy's mum recently brought her to the doctor to discuss the weight problem and the doctor identified Suzy's difficulty to be because of a problem with her glands where she is able to store more food in her body than most children her age. In the environmental condition, a different explanation for the peer's overweight was provided: Suzy's mum recently brought her to the doctor to discuss the weight problem and the doctor identified Suzy's difficulty to be because of a poor diet and lack of exercise. All vignettes were uniform with the exception of the biological/environmental/no causal information manipulation and each vignette included the same introduction statement. Vignettes were specifically created for the purpose of this study and were assessed for reading and face validity by a panel of child developmental experts. Vignettes were also pilot tested with school-aged children to assess suitability and comprehension.

The AC [30] consisted of 32 adjectives (16 positive and 16 negative) that represented four domains: affective feelings (e.g., happy, sad); physical appearance (e.g., pretty, ugly); academic assessment (e.g., smart, dumb); and social behaviour (e.g., cheerful, bored). Participants were required to circle adjectives that they believed to be appropriate descriptions of the target girl. The checklist was scored by subtracting the total number of negative adjectives from the total number of positive adjectives and adding a constant of 20. Thus, scores range from 4-36, with scores above 20 indicating more positive attitudes towards the target girl. Factor analyses confirmed construct validity for positive or negative value of the adjectives, and a coefficient alpha of 0.81 indicated acceptable internal consistency [30]. Good construct and concurrent validity with the AC has been shown [31].

The SAQ-self [32] was employed to assess participant's willingness to engage in certain activities with the average-weight and overweight girls presented in the vignettes. It has three broad activity areas consisting of 8 items each: general social (e.g., ‘Eat lunch next to Suzy at school'); academic (e.g., ‘Work in the school library with Suzy'), and active recreational (e.g., ‘Pick Suzy to be on my soccer team'). As the SAQ was originally created for use with American children, some item wordings were appropriately revised for use with Irish children [33]. Below each item are line drawings of faces with corresponding responses: ‘yes' with a smile (score of 3), ‘maybe' with a neutral expression (score of 2), and ‘no' with a frown (score of 1). The participant circles the answer that shows how he or she feels about sharing each activity with the target child. A total score was obtained by summing responses (ranging from 24-72); and scores for the three individual activity areas area were also obtained (ranging from 8-24). Good concurrent validity of the SAQ-self has been demonstrated with children [32]. In the present study, alphas for the SAQ-self were 0.73, 0.86, and 0.65 (p < 0.001) for the girl presented as overweight in the no causal information, environmental and biological condition, respectively. For the girl presented as average weight, one item, ‘Do you think you would go to the cinema with Suzy?', was deleted from the active recreational subscale because its inclusion resulted in a marked reduction in alpha (α = 0.59); the resulting alpha was 0.92 (95% CI 0.90-0.93; p < 0.001).

A modified version of the SAQ-self, the SAQ-other, was administered to participants in order to detect socially desirable responding [see also [29]. This measure had the same format as the SAQ-self outlined above, with the exception that participants responded in accordance with how willing they thought their classmates would be to interact with the targets. For the overweight girl in the biological, environmental and no causal information conditions, alpha was 0.93, 0.94, and 0.90 respectively (p < 0.001). For the average-weight girl, one item ‘Do you think your friend would eat lunch next to Suzy at school?' was deleted from the general social subscale because its inclusion resulted in a marked reduction in alpha (α = 0.60); the resulting alpha was 0.91 (95% CI 0.88-0.92; p < 0.001).

Ethical approval for this study was granted by the School of Psychology Ethics Review Group at the host university. Participants were informed that they would read two short stories about children of their own age. All participants read the vignette depicting the average-weight girl. Participants at each school year were randomly assigned to read a vignette of a peer presented as overweight in one of three conditions: i) overweight girl with biological causal information, ii) overweight girl with environmental causal information and iii) overweight girl with no causal information. To control for order effects, the vignettes were counterbalanced so that half of the participants received the average-weight vignette first, followed by the overweight vignette, while the other half received the vignettes in the opposite order. Chi-square tests revealed no significant differences in sex distribution from condition to condition across school year. ANOVAs revealed no significant differences in age between conditions.

After reading each vignette, participants completed the AC followed by the SAQ-self and the SAQ-other. The dependent measures were administered to children in a class setting, and each item was read aloud to participants by the researcher. Six children were unable to participate where developmental delays precluded them from completing study measures or where English was not their first language. The researcher was available to answer any questions about the vignette or survey items in the event that a child needed assistance. Children with reading difficulties were assisted by the classroom teacher who read instructions but required the child to circle their own response.

Data were screened for normality and outliers. All survey items showed skewness and kurtosis within an acceptable range [34]. A series of paired samples t-tests were conducted to examine if there were differences in ratings on the AC and SAQ-self for the average-weight peer and the peer depicted as overweight in each of the three experimental conditions (biological, environmental, no causal information). A series of 3 × 2 between-subjects analysis of co-variances (ANCOVA) were conducted to examine the effect of experimental condition and sex on the AC and SAQ-self. A mixed analysis of variance (ANOVA) was conducted to examine the effect of experimental condition on social desirability responses (SAQ-other). Alpha was set to 0.01 to avoid type 1 errors with multiple tests.

Table 1 shows the means and standard deviations for scores on the AC and for subscale scores and total scores on the SAQ-self and SAQ-other.

A series of paired samples t-tests were conducted to determine if there were significant differences between children's ratings on the AC for the girl in the average-weight condition and the girls presented as overweight in each of the experimental conditions (biological, environmental and no causal information). Results showed that in each case ratings were significantly higher for the average-weight target than for the overweight targets (biological: t(62) = -9.43, p < 0.001; environmental: t (57) = -11.08, p < 0.001; no causal information: t (54) = -6.861, p < 0.001; see table 2).

A 3 (experimental condition: biological, environmental, no causal information) × 2 (sex) between-subjects analysis of covariance (ANCOVA) was employed to analyse total scores on the AC, while correcting for social desirability (scores on the SAQ-other). A significant main effect was found for experimental condition, F (2, 169) = 10.01, p < 0.001, partial η² = 0.11. Post-hoc tests indicated that children in the biological condition rated the girl presented as overweight more favourably than children in the environmental condition (p ≤ 0.01). Post-hoc tests also indicated that children in the no causal information condition rated the girl presented as overweight more favourably than children in the environmental condition (p < 0.001).

A series of paired samples t-tests were also conducted to determine if there were significant differences between children's ratings on the SAQ-self for the girl in the average-weight condition and the girls presented as overweight in each of the other three conditions (biological, environmental and no causal information). Significant differences were observed in each case (biological: t (62) = 3.64, p < 0.001; environmental: t (57) = 4.74, p < 0.001; no causal information: t (54) = 3.27, p < 0.001; see table 2).

A 3 (experimental condition) × 2 (sex) between-subjects ANCOVA was employed to analyse total scores on the SAQ, while correcting for social desirability (scores on the SAQ-other). A significant main effect was found for sex, F (1, 169) = 28.54, p < 0.001, partial η² = 0.14. Girls (58.94 ± 8.80) reported significantly more positive behavioural intentions toward a girl presented as overweight than boys (46.29 ± 11.17).

To determine the effect of socially desirable responding on SAQ scores, a mixed between-within subjects ANOVA was conducted to examine if there were significant differences between ratings on the two versions of the SAQ (SAQ-self and SAQ-other) across the three experimental conditions (biological, environmental, no causal information). There was no significant interaction between ratings on the SAQ-self versus SAQ-other and experimental condition (Wilks Lambda = 0.99, F (2, 173) = 0.54, p = 0.59, partial ε² = 0.006). There was a main effect for scores on SAQ-self versus SAQ-other (Wilks Lambda = 0.92, F (1, 173) = 14.19, p < 0.001) with all conditions showing higher scores on the SAQ-self versus SAQ-other. According to guidelines by Cohen [35], this effect was moderate (partial ε² = 0.08). The main effect comparing the three conditions was not significant (F (2, 173) = 1.26, p = 0.29, partial ε² = 0.01), suggesting no difference in the type of experimental condition on social desirability responses. Examination of the means in table 1 shows that participants in the three experimental conditions were more positive in their own ratings on the SAQ-self than they were in rating how their classmates would respond (SAQ-other).

Social desirability effects were also examined on SAQ scores for average-weight peer in each of three conditions. A mixed ANOVA was conducted and showed that there was no interaction between scores on the SAQ-self versus SAQ-other and experimental condition (Wilks Lambda = 0.98, F(2, 173) = 0.18, p = 0.179, partial ε² = 0.02. There was a significant social desirability effect for the three experimental conditions (Wilks Lambda = 0.96, F(1, 173) = 6.78, p < = 0.01). According to guidelines by Cohen [35], this effect was small (partial ε² = 0.04). The main effect for condition was not significant (F(2, 173) = 0.87, p < 0.87, ε² = 0.002).

This study examined children's attitudes towards peers presented as average weight and overweight as well as the impact of different types of causal information on children's responses toward a peer depicted as overweight. Consistent with previous research [5,8], the current findings demonstrated that children tended to rate the target depicted as overweight more negatively across all conditions than the average-weight target. In line with previous research conducted in the USA, children in the current study seem to have internalized the prevailing socio-cultural message that ‘fat is bad' and ‘thin is good' [36].

When examining the impact of causal information toward the peer who was overweight, different patterns of results emerged for attitudes and behavioural intentions. There is some support for the view that when children think the peer is responsible for their own condition (i.e., they have a poor diet and no exercise) this is associated with more negative attitudes; however, the same pattern is not observed for behavioural intentions. The latter finding is in line with previous research [7]. In addition, there is no evidence to suggest that the provision of medical information results in a more positive pattern of responses toward the peer who was overweight. The current findings therefore do not provide strong support for Weiner's attribution theory. However, a recent study of weight bias in medical students may be encouraging where it was found that genetic information or information about the behavioural mechanisms of obesity reduced anti-fat stereotypes relative to a control (i.e., chronic head pain) [37]. However, given that this type of study has been implemented in adults with mixed or little success [38,39], it is not surprising that there was little effect on behavioural intentions in the current study.

The findings suggest that researchers need to reconsider whether presenting causal information is an appropriate direction for future interventions. It has been suggested that attributions of responsibility may not be the primary reasons for bias in behavioural responses towards children who are overweight [38]. The effects of an educational film intervention to enhance students' understanding for obese adolescents' problems revealed that, although there was an increased understanding of the problems of obese people after the intervention, students showed stronger prejudice against them [40].

Social desirability effects were identified in the present study in that children provided more positive behavioural intentions in the SAQ-self in comparison to the SAQ-other would respond to an overweight peer and an average-weight peer in the three experimental conditions. However, the results showed that the social desirability effect was stronger for the overweight peer than for the average-weight peer based on Cohen's [35] guidelines. To the authors' knowledge, this is one of the first studies to consider socially desirable responding in weight bias research with children. It is recommended that social desirability is controlled for in future attitudinal studies on overweight. There is also scope for greater reliance on implicit attitude measures in order to tap into automatic associative processes. Research using the Implicit Association Test (IAT) has found higher levels of bias on the IAT than with explicit measures [41].

An important strength of this study was the experimental design employed to investigate the effects of different types of causal information on children's attitudes and behavioural intentions towards a peer depicted as overweight. This study extended Bell and Morgan's [7] research by examining the effect of environmental information on children's weight-based attitudes and intentions. An additional strength is that this research contributes to the literature by providing evidence of weight-based stigma in a non-US sample of youth.

There are several study limitations that must be considered for improving future research. Developmental factors such as age and pubertal status may be important constructs to consider when studying weight stereotypes among children, and these were not considered in the current work. Other variables that may also impact on children's judgements of a peer who is overweight include ethnicity, the amount of contact a child has with a peer who is overweight/obese, exposure to the media, and whether a child has a weight difficulty of his or her own. Recent studies have begun to examine some of these variables. Gray et al. [42] found that certain age groups (i.e., children compared to adolescents) and populations (i.e., White adolescents compared to African-American adolescents) had higher levels of weight-based stigmatization. The current study found that girls had more positive behavioural intentions toward an overweight female target than boys. However, a limitation of this study is that all children were presented with vignettes of female targets only, which may potentially bias the results and lead to better behavioural intentions among girls. Further research is warranted to examine whether there are gender differences in behavioural intentions towards an overweight child and whether this varies depending on the gender of the target child.

Another limitation is the extent to which children's responses on attitude and behavioural intention measures to a hypothetical overweight child can predict their actual behaviours with a peer who is overweight. Although attitudes have been described as good predictors of behaviours, in real-life situations there is often a lack of correspondence between them [43]. A wider range of assessments of social relationships and social responses could also be explored, such as peer and friendship nominations and observations of children's behaviours toward peers who are overweight/obese in the school setting. It would also be worthwhile to measure participants' BMI or perceived weight in future research. Research on a large community-based sample of adults found that implicit and explicit anti-fat biases were significantly weaker among people with high BMIs compared with those with low BMIs [44]. However, other studies with smaller samples found that overweight individuals did not hold more favourable attitudes toward overweight group members [45,46]. Further research is warranted to examine whether overweight children show more favourable attitudes and intentions towards an overweight peer when compared to normal-weight children. Finally, another consideration would be to employ a visual target as well as presenting a vignette with an explanation of the child's overweight condition, as visual cues may change children's ratings towards the obese peer. Presenting the child in action via video is a more realistic interaction and would have more external validity than that created by written vignettes.

The present research suggests a need to explore alternative approaches to improving attitudes towards children who are overweight and fostering integration of these children in the society. One possible approach might be to promote messages emphasizing both children's similarities and differences to others (e.g., ‘all children are the same in a way, but each child is also unique'). There is some evidence that this approach was effective among Black and Latino youth in improving desired social closeness to White youth [47]; however, it has yet to be tested as a viable approach for children with health conditions. Another potential strategy is the promotion of body size acceptance. A puppet program implemented among elementary school children to promote size acceptance was effective in reducing negative stereotypes about large body shapes [48]. Gray et al. [42] suggest that the limited effectiveness of interventions to date is due to their exclusive focus on children, as opposed to their broader social network. More complex multi-component interventions that address the broad social ecology, spanning peers, family, school and society, may be required [42].

In conclusion, the present study found that overweight peers were viewed negatively by their peers. Information explaining obesity had a minimal positive effect on children's attitudes and no effect on their behavioural intentions toward a peer presented as overweight. Further research needs to consider alternative ways of reducing stigma towards obese peers and how to promote acceptance among children who are overweight.

The authors declare that there are no conflicts of interest.