Body Image and School Engagement in Early Adolescence: Longitudinal Gender Differences

Body Image as a Potential Risk Factor for School Disengagement

School engagement is typically defined using the integrative model proposed by Fredricks and colleagues (2004). In this model, school engagement consists of three components: affective (e.g., valuing school, enjoying school), behavioural (e.g., paying attention, following rules), and cognitive (e.g., using effective strategies to plan and monitor learning). The extent to which children are engaged at school predicts short-term and long-term academic success (Pagani et al., 2012), as well as health promoting behaviours (e.g., balanced diet, adequate physical activity; Carter et al., 2007). Alternatively, children who are less engaged at school tend to struggle with both externalized (e.g., hyperactivity, opposition, defiance) and internalized problems (e.g., anxiety and depression; Olivier et al., 2020). Importantly, the process of becoming disengaged at school is best viewed as a long and gradual one, with certain adverse academic and social experiences in as early as the first grade, predicting eventual high school dropout (Finn, 1989). In other words, children’s experiences in elementary school appear to set them on subsequent school engagement trajectories.

Body image is a multi-dimensional construct comprised of a perceptual component as well as an affective component (Cash & Smolak, 2011). The perceptual component consists of a mental representation of your body that can be accurate or inaccurate (e.g., “I think I am fat”), whereas the affective component refers to your satisfaction with these perceptions (e.g., “I hate that I am fat”). Most children begin forming opinions about their body around the age of 7 or 8 years through comparisons to others (Tatangelo et al., 2016). When these opinions are negative, they tend to involve a distorted perception of their body (e.g., believing they are overweight when they are not) and/or being dissatisfied with their body. In cases where children develop a negative opinion about their body, this can be accompanied by unrealistic expectations for how their body should look (Myers & Crowther, 2009) as well as the development of maladaptive eating behaviors (Davison et al., 2003).

Despite being two distinct constructs, school engagement and body image both broadly impact children’s feelings, thoughts and behaviours. With that said, mechanisms linking body image and school engagement are likely indirect and multifaceted. Children who are dissatisfied with their bodies may become increasingly preoccupied with their appearance, engage in social comparison, or worry about being negatively evaluated by peers. These appearance-related concerns may tax cognitive resources needed for learning and self-regulation or contribute to avoidance-oriented coping strategies in school contexts (e.g., minimizing participation or avoiding attention from peers). In some cases, these processes may be amplified by experiences such as weight- or body-related teasing (Puhl & Luedicke, 2012). This perspective is generally consistent with self-objectification theory, which suggests that habitual monitoring of one’s appearance can interfere with cognitive functioning by diverting limited attentional resources (Fredrickson & Roberts, 1997; Winn & Cornelius, 2020). Importantly, body image concerns likely interact with other salient domains of functioning, such as academic self-concept, peer belonging, and social competence. Thus, body dissatisfaction may function as a psychosocial risk factor that increases vulnerability to disengagement under certain conditions.

Despite the potential interplay between these variables, cross-sectional studies examining the links between body image and school engagement have been scarce, as most research has instead focused on how students with a heavier weight status (i.e., high body mass index; BMI) tend to show poorer behavioural engagement and achievement (e.g., Lv et al., 2020). A common interpretation of this association is that students avoid attending or participating in class due to weight-based bullying and stigma (Finn et al., 2018). While this interpretation is plausible, it ignores the fact that many students experience body dissatisfaction who are not actually overweight (van den Berg et al., 2010). In fact, the link between BMI and school achievement has been shown to be better explained by a student’s weight perception (i.e., whether they view themselves as overweight, rather than being overweight; Florin et al., 2011).

To our knowledge, the only study that has investigated longitudinal associations between body image and school engagement was conducted by Guimond and Laursen (2022). Specifically, they investigated whether children’s body dissatisfaction could predict changes in school engagement over an eight-week period. Results indicated that greater body dissatisfaction predicted lower behavioural engagement (i.e., referred to as “academic” engagement) eight weeks later, and that these associations were strongest among children with emotional problems and children who felt rejected by their peers. Importantly, the link between body dissatisfaction and behavioural engagement still emerged after controlling for variables known to be associated with school engagement, such as school achievement, popularity, and disruptiveness. These findings serve as further evidence that body image may exert an influence on school engagement.

Creating Integrated Models for Predicting School Engagement

The study by Guimond and Laursen (2022) was the first, to our knowledge, to establish temporal precedence in the relationship between body image and school engagement (i.e., that body dissatisfaction predicts subsequent decreases in behavioral school engagement). However, their measures of both school engagement and body image were unidimensional. Specifically, only behavioural engagement and body dissatisfaction were measured. By including measures of affective and cognitive engagement, as well as body perception, we stand to obtain a more thorough understanding of the longitudinal associations between these constructs.

While body perception can be measured in relation to various aspects of one’s appearance (e.g., weight, hair), weight perception (i.e., whether you view your weight as normal, overweight, or underweight) is a commonly measured perceptual component in children (Wang et al., 2018) and has been linked to body dissatisfaction (Dion et al., 2016) and school-related outcomes (Florin et al., 2011). These factors make it a worthwhile addition to a study designed to investigate the associations between body image and school engagement in children. Moreover, although children’s behavioural and cognitive engagement levels tend to be stable across classes, their affective engagement can vary between class subject (e.g., languages, math; Archambault & Vandenbossche-Makombo, 2014), suggesting that a measure of affective engagement should be subject-specific.

The Current Study

The current study aimed to examine longitudinal associations between body image and school engagement in pre-adolescent children. Our first objective was to investigate the longitudinal associations between body dissatisfaction and weight perception with the three types of school engagement. We hypothesized that children who were more dissatisfied with their body or who perceived themselves as heavier at the start of the school year would show decreases in affective, behavioural, and cognitive engagement by the end of the school year, even after accounting for children’s age, prosocial behaviour, and externalized behavioural problems. As a secondary objective, we examined whether these longitudinal associations differed by gender. Given evidence that gender differences in body dissatisfaction emerge in nuanced ways during childhood and early adolescence, we treated the moderating role of gender as an exploratory research question rather than advancing a directional hypothesis.

Participants

Our initial sample consisted of 1,400 elementary school students from 9 to 12 years old. After excluding participants who disclosed a diagnosis of a mental disorder or learning disability, our final sample included 1,070 students. In Fall 2021, these participants were recruited from 47 French public schools in semi-urban areas of Montreal, Canada. We collected data at two time points: which included near the beginning (Time 1) and end (Time 2) of the school year. Specifically, our Time 1 data collection occurred between November-December 2021 (n = 1,070 students; boys = 472, girls = 588, not specified = 10; M _age = 11.13 years), and our Time 2 data collection occurred between May-June 2022 (n = 707 students; boys = 317, girls = 382, not specified = 8; M _age = 11.09 years).

Participants were 78.9% White and 21.1% identified as racialized (2.5% Black, 2.3% Latinx, 1.4% Indigenous, 1.4% West Asian, 1.3% East/Southeast Asian, and 12.2% other). Only 6.5% of children were born outside of Canada. French was the majority language spoken at home (98.6%), followed by English (1.4%). Some children also specified that they spoke other languages at home (5.8%). Most children reported that their mother (77.3%) and father (58.9%) had at least a college or a bachelor’s degree and were employed (93.6% and 96.5%, respectively). As children might be unaware of their household annual income, the Family Affluence Scale (FAS; Boyce et al., 2006) was used to calculate an index of family wealth. The FAS is moderately associated to parental earned income and has been shown to distinguish between low- and high-income households (Corell et al., 2021). Scores are calculated by summing the number of cars, computers, tablets, and cellphones owned, as well as whether the child has their own bedroom. Results suggested that our sample was homogenous in terms of family affluence and consisted of children from relatively high-income households (M _{total =} 14.33, SD = 1.70, range 9–18). Almost all children reported that their family had at least one car (99.9%), a computer (98.1%), a tablet (87.3%), a cellphone (100%), and that they had their own bedroom (89.6%).

Procedure

This study was approved by the research ethics board at the University of Ottawa and at the University of Sherbrooke and was part of a larger longitudinal study on student academic adjustment and outcomes. We initially contacted schoolboards in the province of Quebec, Canada to seek permission to recruit from schools in their respective districts. If we received approval, our request was then forwarded to principals of those schools within each district. If school principals approved our request, we then contacted teachers in those schools about our research project. Finally, if teachers approved of our study, we began our informed consent process with children, their parents, and their teacher. Parents and teachers provided their consent, whereas assent was obtained from all children who participated.

Upon obtaining consent and assent, data collection dates were scheduled with teachers at the various schools. Data collection took place virtually, as surveys at both time points were completed by children on tablets during class time. Surveys took approximately 30–45 min to complete. All measures were administered in French at both time points.

Measures

Plan of Analysis

As preliminary analyses, bivariate associations between Time 1 and Time 2 study variables were conducted. These analyses were conducted using generalized estimating equations (GEE), due to some children being in the same class, which violated the assumption of data independence (i.e., children in the same class/school may be more similar than those who are not). To examine concurrent and longitudinal associations between body dissatisfaction, weight perception, and school engagement, path analyses were conducted in a structural equation modeling framework using Mplus (version 8.2), and the COMPLEX function was applied to address potential classroom-level differences in school engagement due to the nested nature of our data.

Our first objective was to investigate the longitudinal associations between body dissatisfaction and weight perception with school engagement. To do so, we created four models—one for each type of school engagement that we measured (i.e., affective French, and affective math, behavioural, and cognitive). Each model included body dissatisfaction, weight perception, and school engagement at both Time 1 and Time 2, as well as three control variables at Time 1 (i.e., age, prosocial behaviours, and externalized behaviour problems). Including each construct at both time points allowed us to estimate autoregressive paths (i.e., stability effects) and cross-lagged paths, such that significant longitudinal associations were interpreted as predicting change in the outcome variable over time after accounting for its prior level. Moreover, by including control variables at Time 1, any significant longitudinal associations between body dissatisfaction, weight perception, and school engagement were interpreted as above and beyond the effects of these other variables.

As a secondary objective, we examined whether the longitudinal associations between body dissatisfaction, weight perception, and each type of school engagement differed by gender. We began by calculating the Sattora-Bentler Chi-Square difference (TRd) for each of our four models. This statistic was calculated by comparing a constrained version of the model (i.e., all paths were set to be equal between boys and girls) to an unconstrained version. A significant TRd suggested that a model had significant gender differences, which we followed up on by comparing the pattern of results of boys and girls and identifying differences in their statistically significant longitudinal paths.

To assess model fit, several standard indices were assessed. This included the Chi-square index (should be non-significant), the root-mean-square error of approximation (RMSEA; should be .06 or lower), as well as the comparative fit index (CFI) and Tucker-Lewis index (TLI; both should exceed 0.95; Kline, 2016). Of the 1,070 children who took the survey at Time 1, 707 took it again at Time 2 (33.93% of attrition). For those who completed the survey at both time points, an average of 2.31% (range = 0%–5%) of the data was missing for the study variables. A Little’s MCAR test indicated that the data were missing completely at random, χ ² (176) = 191.57, p = .20. Missing data were handled with the full information maximum likelihood (FIML) estimation method.

Preliminary Analyses

Table 1 presents the bivariate associations among all study variables, estimated using generalized estimating equations to account for the clustering of students within classrooms. Notably, there were significant negative associations, both concurrent and longitudinal, between children’s body dissatisfaction and nearly all types of engagement. Specifically, children who were more dissatisfied with their body were less engaged at school (i.e., affectively in math, behaviourally, and cognitively), whereas affective French engagement was not longitudinally associated with body dissatisfaction. Alternatively, children’s weight perception was not concurrently nor longitudinally associated with any type of engagement. All three of our control variables (i.e., age, prosocial behaviours, and externalized behaviour problems) showed concurrent and longitudinal associations with all types of engagement, except for age and cognitive engagement. Specifically, prosocial behaviour showed positive concurrent and longitudinal associations with all types of engagement, such that children who were more prosocial were more engaged. Alternatively, age and externalized behaviour problems showed negative concurrent and longitudinal associations with engagement, such that older children and those with more externalized behaviour problems were less engaged.

OPEN IN VIEWER

Table 1.

Bivariate Associations Among Study Variables: Generalized Estimating Equation Standardized Beta Coefficients

Variable	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
1. Body dis. T1	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–
2. Weight perception T1	.04 [−.02, .10]	–	–	–	–	–	–	–	–	–	–	–	–	–	–
3. Affective engagement (French) T1	−.10** [−.17, −.04]	.01 [−.04, .08]	–	–	–	–	–	–	–	–	–	–	–	–	–
4. Affective engagement (math) T1	−.25*** [−.31, −.17]	.00 [−.06, .06]	.40*** [.33, .46]	–	–	–	–	–	–	–	–	–	–	–	–
5. Behavioural engagement T1	−.20*** [−.26, −.13]	.04 [−.01, .10]	.51*** [.45, .57]	.49*** [.43, .55]	–	–	–	–	–	–	–	–	–	–	–
6. Cognitive engagement T1	−.18*** [−.24, −.11]	.02 [−.04, .08]	.45*** [.38, .51]	.49*** [.43, .55]	.74*** [.70, .79]	–	–	–	–	–	–	–	–	–	–
7. Body dis. T2	.71*** [.63, .77]	−.00 [−.09, .08]	−.06 [−.13, .01]	−.22*** [−.31, −.13]	−.12*** [−.19, −.05]	−.11** [−.18, −.04]	–	–	–	–	–	–	–	–	–
8. Weight perception T2	.14** [.04, .22]	.65*** [.56, .74]	−.01 [−.08, .06]	−.05 [−.14, .04]	−.02 [−.10, .04]	−.06 [−.13, .01]	.10* [.01, .19]	–	–	–	–	–	–	–	–
9. Affective engagement (French) T2	−.08 [−.15, .00]	.03 [−.04, .10]	.70*** [.65, .75]	.26*** [.19, .33]	.37*** [.30, .44]	.31*** [.24, .38]	−.09* [−.17, −.02]	.00 [−.07, .07]	–	–	–	–	–	–	–
10. Affective engagement (math) T2	−.23*** [−.29, −.15]	.01 [−.07, .09]	.31*** [.23, .39]	.73*** [.67, .78]	.36*** [.29, .43]	.38*** [.30, .45]	−.22*** [−.29, −.14]	−.01 [−.10, .08]	.37*** [.28, .45]	–	–	–	–	–	–
11. Behavioural engagement T2	−.15*** [−.21, −.07]	.01 [−.06, .10]	.46*** [.38, .54]	.45*** [.37, .54]	.58*** [.51, .64]	.47*** [.40, .55]	−.17*** [−.26, −.08]	−.02 [−.10, .05]	.51*** [.44, .59]	.56*** [.49, .64]	–	–	–	–	–
12. Cognitive engagement T2	−.17*** [−.24, −.09]	−.00 [−.08, .06]	.40*** [.34, .47]	.45*** [.37, .53]	.52*** [.45, .59]	.63*** [.56, .70]	−.16*** [−.24, −.07]	−.02 [−.10, .05]	.47*** [.40, .54]	.52*** [.43, .60]	.74*** [.69, .79]	–	–	–	–
13. Externalized behaviours T1	.31*** [.25, .36]	−.08* [−.15, −.01]	−.44*** [−.49, −.38]	−.37*** [−.43, −.32]	−.56*** [−.61, −.51]	−.51*** [−.57, −.46]	.23*** [.15, .30]	−.04 [−.12, .04]	−.32*** [−.38, −.26]	−.31*** [−.39, −.22]	−.41*** [−.50, −.33]	−.43*** [−.51, −.36]	–	–	–
14. Prosocial behaviours T1	−.06 [−.12, .00]	.09* [.02, .16]	.35*** [.29, .42]	.22*** [.15, .28]	.41*** [.33, .49]	.41*** [.34, .48]	−.00 [−.08, .07]	.02 [−.05, .10]	.28*** [.20, .35]	.16*** [.07, .25]	.26*** [.17, .36]	.28*** [.19, .38]	−.35*** [−.42, −.29]	–	–
15. Age T1	.07** [.02, .12]	−.06* [−.10, −.00]	−.09*** [−.14, −.03]	−.15*** [−.20, −.09]	−.07*** [−.12, −.03]	−.00 [−.06, .04]	.10*** [.04, .16]	−.03 [−.09, .03]	−.11*** [−.18, −.04]	−.16*** [−.23, −.09]	−.10*** [−.15, −.04]	−.05 [−.11, .01]	−.01 [−.06, .03]	−.00 [−.04, .03]	–

Main Analyses

Gender Differences

The behavioural and cognitive engagement models contained significant gender differences, as evidenced by a significant Sattora-Bentler Chi-Square difference (TRd). Alternatively, the affective engagement models did not contain any significant gender differences in our hypothesized paths (see Figures 1 and 2).OPEN IN VIEWER

Figure 1.

Concurrent and longitudinal associations between affective French engagement and body image variables (boys and girls). Note. This figure shows the concurrent and longitudinal paths between our body image variables and affective French engagement for boys and girls. Significant paths are illustrated with solid arrows (solid = overall sample; square dot = boys only; rectangle dot = girls only. Non-significant paths are illustrated with transparent dashed lines. Absent from this figure are three control variables (i.e., age, prosocial behaviours, and externalized behaviours), all of which were controlled for at Time 1

OPEN IN VIEWER

Figure 2.

Concurrent and longitudinal associations between affective math engagement and body image variables (boys and girls). Note. This figure shows the concurrent and longitudinal paths between our body image variables and affective math engagement for boys and girls. Significant paths are illustrated with solid arrows (solid = overall sample; square dot = boys only; rectangle dot = girls only). Non-significant paths are illustrated with transparent dashed lines. Absent from this figure are three control variables (i.e., age, prosocial behaviours, and externalized behaviours), all of which were controlled for at Time 1

Behavioural Engagement

In our behavioural engagement model (TRd = 55.71, p = .008), there was one notable gender difference. First, boys showed a negative association between behavioural engagement at Time 1 and body dissatisfaction at Time 2 (b = −.08, SE = .05, p = .035), whereas girls did not (b = −.04, SE = .04, p = .289). Specifically, boys who were less behaviourally engaged at Time 1 were more dissatisfied with their body at Time 2. This gender difference is illustrated in Figure 3.OPEN IN VIEWER

Figure 3.

Concurrent and longitudinal associations between behavioural engagement and body image variables (boys and girls). Note. This figure shows the concurrent and longitudinal paths between our body image variables and behavioural engagement for boys and girls. Significant paths are illustrated with solid arrows (solid = overall sample; square dot = boys only; rectangle dot = girls only). Non-significant paths are illustrated with transparent dashed lines. Absent from this figure are three control variables (i.e., age, prosocial behaviours, and externalized behaviours), all of which were controlled for at Time 1

Cognitive Engagement

In our cognitive engagement model (TRd = 55.39, p = .008), two gender differences emerged. First, girls showed a negative association between body dissatisfaction at Time 1 and cognitive engagement at Time 2 (b = −.08, SE = .04, p = .047), whereas boys did not (b = −.08, SE = .06, p = .156). Specifically, girls who were more dissatisfied with their body at Time 1 were less cognitively engaged at Time 2. Second, girls showed a negative association between cognitive engagement at Time 1 and weight perception at Time 2 (b = −.07, SE = .03, p = .031), whereas boys did not (b = −.07, SE = .04, p = .131). Specifically, girls who were less cognitively engaged at Time 1 perceived themselves as heavier at Time 2. These gender differences are illustrated in Figure 4.OPEN IN VIEWER

Figure 4.

Significant concurrent and longitudinal associations between cognitive engagement and body image variables (boys and girls). Note. This figure shows the concurrent and longitudinal paths between our body image variables and cognitive engagement for boys and girls. Significant paths are illustrated with solid arrows (solid = overall sample; square dot = boys only; rectangle dot = girls only). Non-significant paths are illustrated with transparent dashed lines. Absent from this figure are three control variables (i.e., age, prosocial behaviours, and externalized behaviours), all of which were controlled for at Time 1

Body Dissatisfaction and School Engagement

Weight Perception and School Engagement

Contrary to our hypotheses, weight perception at Time 1 did not predict engagement at Time 2 in any model. This lack of associations may support the notion that a child’s weight is less related to school engagement than their level of satisfaction with their weight. However, without an objective measure of children’s weight status (i.e., BMI), we are unable to determine whether children’s weight perception is accurate or distorted (i.e., weighing in the normal range but perceiving yourself as overweight), which prevents us from drawing this conclusion. With that said, it did not seem to matter whether children perceived themselves as heavier or slimmer—what mattered more with regards to engagement was the extent to which they were dissatisfied with this perception of their weight. This is contrary to research by Florin and colleagues (2011), who found that children’s academic achievement was predicted by their weight perception (i.e., students who perceived themselves as heavier performed worse at school). Although school achievement and engagement are distinct constructs, they are highly related to one another (Hughes et al., 2008). A possible explanation for this discrepancy in findings is that including more meaningful predictors in our models (i.e., body dissatisfaction, age, prosocial behaviours, and externalized behavioural problems) may make weight perception a nonsignificant predictor of engagement.

Also contrary to our hypotheses, lower behavioural and cognitive engagement at Time 1 predicted heavier weight perception at Time 2 in our overall sample. Again, this finding is difficult to interpret as we were unable to conclude whether children’s weight perception was accurate or distorted. For instance, if children’s weight perception was accurate, this finding could suggest that being less behaviourally and cognitively engaged puts children at risk of becoming overweight. This would align with research that has concurrently linked overweight status with a range of academic difficulties in childhood (Lv et al., 2020) and adolescence (Finn et al., 2018). On the other hand, if children’s weight perception was distorted, this might be best explained by underlying self esteem issues that initially influence engagement before later manifesting as weight-based concerns (i.e., vulnerability model of self-esteem; Beck, 1967).

Strengths, Limitations, and Future Directions

Most notably, our study has identified novel findings regarding the long-term interplay between body image and school engagement in children. In particular, gender differences suggest directionally specific effects, such that boys and girls experience distinct risk factors for becoming dissatisfied with their body and disengaged in school. The primary strengths of our study are its large sample size and longitudinal design—both of which represent marked improvements in comparison to prior research. Moreover, our approach of including multi-dimensional measures of body image and school engagement is a notable improvement compared to prior research that has typically measured only a single dimension of each. Moreover, our inclusion of known predictors of engagement as control variables in our models strengthens our confidence in the findings, as all our significant findings were found to be above and beyond the influence of these well-established risk factors (i.e., prosocial behaviours, externalized behavioural issues, age).

Despite these strengths, there are limitations worth noting when interpreting our findings. First, weight perception was measured using a single item, which prevented us from assessing its internal consistency. Beyond reliability concerns, the primary issue with our measure of weight perception is that, without an objective measure of weight status, we were unable to determine whether children perceived their weight accurately or in a distorted manner (e.g., reporting that they perceive themselves as overweight when they are in the normal range). Ultimately, this makes it difficult to interpret the lack of directional effects from weight status to school engagement. Moreover, although weight perception is commonly assessed as a perceptual dimension of body image in youth research (Wang et al., 2018), body image perceptions can also encompass broader aspects of appearance, such as body shape or specific physical features. Future studies using more comprehensive measures of body perception may reveal stronger or more nuanced associations with school engagement.

Second, while our sample was considerably larger than most prior studies, we did experience notable attrition from Time 1 to Time 2 (i.e., 34%). This sample size reduction led to lower statistical power, which increased our risk of Type II errors—ultimately making it more challenging to detect real effects. While our findings suggest that children’s body image and engagement influence one another over time, additional research is needed to clarify the nature of these associations. For instance, future studies should include an objective measure of children’s weight status (e.g., age- and sex-specific BMI percentile) as this would allow us to tease apart the relative influences of body image (i.e., satisfaction and perception) and weight on engagement. Additionally, although family wealth was measured, our sample was relatively affluent and socioeconomically homogeneous, which limited our ability to meaningfully examine the potential role of socioeconomic factors in these associations. Future studies should replicate these findings in more socioeconomically diverse samples, given that findings suggest that lower socioeconomic status is associated with lower school engagement, likely due to differences in access to educational resources, parental support, and increased exposure to chronic stressors (Martins et al., 2022).

Conclusions

Our findings provide new insights into how children’s body image and school engagement influence each other over time. Should our findings continue to be replicated, this could inform gender-specific intervention approaches to address body image issues and school disengagement in children. Specifically, boys appear to become disengaged prior to becoming dissatisfied with their body, whereas the reverse seems true for girls. This could suggest that young boys may benefit more from preventative interventions designed to promote school engagement, whereas young girls may benefit more from early interventions designed to foster a positive body image.