Compulsive exercise: The role of personality, psychological morbidity, and disordered eating


  • Huw Goodwin BSc (Hons), MRes,

    1. Loughborough University Centre for Research into Eating Disorders, Loughborough University, Leicestershire, United Kingdom
    Search for more papers by this author
  • Emma Haycraft PhD,

    Corresponding author
    1. Loughborough University Centre for Research into Eating Disorders, Loughborough University, Leicestershire, United Kingdom
    • Loughborough University Centre for Research into Eating Disorders, School of Sport, Exercise and Health Sciences, Loughborough University, Leicestershire LE11 3TU, United Kingdom
    Search for more papers by this author
  • Anne-Marie Willis RMN,

    1. Cotswold Spa, Huntercombe Hospitals, Worcestershire, United Kingdom
    Search for more papers by this author
  • Caroline Meyer PhD

    1. Loughborough University Centre for Research into Eating Disorders, Loughborough University, Leicestershire, United Kingdom
    Search for more papers by this author



Compulsive exercise has been closely linked with eating disorders, and has been widely reported in both clinical and nonclinical settings. It has been shown to have a negative impact on eating disorder treatment and outcome. However, the risk factors for compulsive exercise have not been examined. This study aimed to provide a first step in identifying potential cross-sectional predictors of compulsive exercise.


The sample consisted of 1,488 male and female adolescents, aged 12–14 years old, recruited from schools in the United Kingdom. Participants completed measures of compulsive exercise, personality, psychological morbidity, and disordered eating attitudes during a school class period.


Multiple stepwise regressions showed that the strongest cross-sectional predictors of compulsive exercise were a drive for thinness, perfectionism, and obsessive-compulsiveness.


These results are discussed in terms of the role that personal factors may play in the development of compulsive exercise. © 2011 by Wiley Periodicals, Inc.


Eating Disorders (ED) comprise a variety of problematic behaviors, including bingeing, purging, and restricting food consumption.1 Another such problematic behavior that has been widely reported in both clinical and nonclinical settings is that of compulsive exercise. Compulsive exercise has been defined as an intense drive to be active, often in a rigid, routine-like fashion that is predominantly performed to manage weight and shape, as well as alleviating negative emotions.2, 3 It has been found in as many as 39% of anorexia nervosa (AN) patients and 23% of Bulimia Nervosa (BN) patients at admission to an ED clinic4 and has been linked with greater treatment time, poorer outcome, and increased chance of relapse.5, 6

Additionally, compulsive exercise has been found in community samples,3 and has been implicated in the etiology of ED.7 The development of ED predominantly occurs around early adolescence,8 and yet little to no research on compulsive exercise has been conducted among adolescent samples. This age-group represents an important population to study in risk factor research into ED and compulsive exercise and, as such, investigations are required to identify whether a compulsive drive to exercise is directly linked to disordered eating attitudes at this early age.

Importantly, not all individuals with ED pathology display a compulsive drive to exercise. Therefore, it is likely that there are specific psychological and personality differences that render an individual at risk of specifically developing a compulsivity towards exercise. However, the risk factors for compulsive exercise are poorly understood.

One possible personality trait linked to the development of compulsive exercise is perfectionism, which has been reported as a risk factor for both AN9 and BN.10 Perfectionism has already been found to be an antecedent to obligatory exercise,11 as well as exercise dependence;12 terms which encompass similar, if not the same, constructs as compulsive exercising. Another possible risk factor is obsessive-compulsiveness, which has also been strongly linked with compulsive exercise in both clinical and nonclinical ED samples. For example, compulsive exercise has been related to obsessive-compulsiveness in a sample of health club exercisers (Wyatt, Unpublished Doctoral Dissertation), and to excessive exercise in clinical ED groups.13, 14

A key feature of compulsive exercise is a negative mood, such as, experiencing feelings of anxiety, depression, and guilt, when deprived of exercising.15 However, other studies have found negative affect, in its various forms (i.e., anxiety and depression), to be related to continued exercise and not simply a resultant state of exercise deprivation. For example, Coen and Ogles11 found that compulsive exercisers reported higher levels of anxiety than non-compulsive exercisers. Similarly, greater levels of depression have been related to compulsive exercise in both clinical and nonclinical ED samples.16, 17

In addition to general anxiety, social anxiety has been associated with ED, with women with eating disordered behavioral tendencies being more likely to be fearful of negative evaluation and have a greater sensitivity to the impressions of others.18 Specific to exercise, a tendency towards social comparisons has been shown to have a stronger impact on adolescent boys' exercising than on their eating.19 Previous research has also found that social physique anxiety, a body specific form of social anxiety, is positively related to greater exercise frequency.20 However, no study has focused on the role of social physique anxiety in the development of compulsive exercise.

In summary, compulsive exercise is a problematic behavior that affects many individuals with ED. A thorough review of previous research has identified several personality traits and psychological states as potential correlates of compulsive exercise,21 but relationships with compulsive exercise have yet to be tested in adolescents. It is also not clear which of these potential predictors of compulsive exercise are the most potent risk factors for its development.

Using a nonclinical adolescent sample, this study aims to identify which personality, psychological, and disordered eating factors are the best cross-sectional predictors of compulsive exercise. It is hypothesized that all significant predictors will be positively associated with compulsive exercise. Given the paucity of previous research, no predictions were made regarding which variable would be the best predictor of compulsive exercise.



This research was conducted in nine schools across the United Kingdom as part of an ongoing larger scale research project. This study reports on a sample of 1,488 participants, aged 12–14 years old (mean age of 12.98 years; SD = 0.73), with gender being equally distributed (girls = 54.1%; boys = 45.9%). The sample predominantly (95.3%) classified their ethnicity as “White British,” and all the schools were from areas of average to low levels of economic deprivation.22 Self-reported height and weight information was converted into body mass index (BMI) for each participant, which was then converted into a z score, so that they were standardized for both age and gender.23 The mean values for BMI z scores were 0.32 (SD = 1.39) for boys and 0.08 (SD = 1.34) for girls.

Measures and Procedure

Institutional Review Board ethical approval was granted before questionnaire packs were sent to the participating schools. Questionnaire packs were completed in a school class period by pupils aged between 12 and 14 years old. Following informed consent, the participants provided background information on nationality, ethnicity, age, gender, height and weight, and then completed the following validated measures:

Compulsive Exercise Test.3 The compulsive exercise test (CET) is a 24-item measure that assesses the level of compulsive exercise. It has five subscales that represent the five core features of the behavior, namely: Avoidance and Rule-Driven Behavior; Weight Control Exercise; Mood Improvement; Lack of Exercise Enjoyment; and Exercise Rigidity. Responses are scored on a six-point Likert scale, anchored with “0—never true” and “5—always true.” The level of compulsive exercise is then identified by creating a total CET score, which is calculated by summing the mean item score for each of the five subscales. Higher scores represent greater levels of compulsive exercise. A psychometric evaluation of the CET in these adolescents has already supported its use.2 In the current sample, the total CET had a Cronbach's alpha of 0.88.

Eating Disorder Inventory-2.24 The drive for thinness, bulimia, and body dissatisfaction subscales of the eating disorder inventory-2 (EDI-2) were administered to assess disordered eating attitudes. The EDI-2 measures the attitudes underpinning AN and BN and has been used reliably among adolescents.25 The Cronbach's alpha values found in this study were 0.84 (drive for thinness), 0.72 (bulimia), and 0.90 (body dissatisfaction).

Child and Adolescent Perfectionism Scale.26 The child and adolescent perfectionism scale (CAPS) is a 22-item, two-scaled measure of perfectionism specifically worded for use within child and adolescent samples. The two scales are self-orientated perfectionism (CAPS-self) and socially prescribed perfectionism (CAPS-social). The former subscale assesses the degree to which an individual imposes self-directed levels of perfectionistic standards and behaviors on to his or her self (e.g., “I try to be perfect in everything I do”). The latter subscale assesses the degree to which an individual feels that their perfectionism is imposed on them by others (e.g., “There are people in my life who expect me to be perfect”). The CAPS has been previously used with adolescents and demonstrated good reliability.27 Cronbach's alpha values for this sample were 0.81 for CAPS-Self and 0.87 for CAPS-Social.

Spence Child Anxiety Scale.28, 29 The Spence Child Anxiety Scale (SCAS) is a measure of anxiety symptoms among children. Only the obsessive-compulsive subscale was used in this study (SCAS-OC). It comprises six items assessing levels of obsessive-compulsiveness, which includes items such as “I can't seem to get bad or silly thoughts out of my head” and “I have to do some things in just the right way to stop bad things happening.” The SCAS has previously been reliably used with adolescents.29 The internal reliability figure for the obsessive-compulsiveness subscale for the current sample was 0.79.

Hospital Anxiety and Depression Scale.30 The Hospital Anxiety and Depression Scale (HADS) is a 14-item measure of anxiety and depression that is widely used in clinical and nonclinical research and practice. The HADS has been validated for use with adolescents,31 and provided Cronbach's alpha values of 0.73 for anxiety and 0.56 for depression for the current sample.

Social Physique Anxiety Scale.32 The Social Physique Anxiety Scale (SPAS) is a 9-item scale measuring a respondent's level of social physique anxiety, using such questions as: “In the presence of others, I feel apprehensive about my physique/figure.” The SPAS has displayed good psychometrics among adolescents,33 and obtained a Cronbach's alpha level of 0.87 for the sample in this current investigation.

Data Analysis

Data were initially screened for normality. As expected, a series of Kolmogorov-Smirnov tests showed that all variables were non-normally distributed, with the exception of the dependent variable, the CET Total score. Importantly, the residuals were normally distributed and, therefore, no transformations were made for the following regression analyses. Non-parametric tests were used, where appropriate. Preliminary analysis also demonstrated that age appropriate body mass index (BMI z-scores) was not significantly correlated with CET Total score for either boys (r = .02, p >.05) or girls (r = .06, p >.05). Therefore, BMI z scores were not included in any subsequent analysis as a control variable.

Analyses were conducted separately for boys and girls due to significant differences on many study variables (Table 1). For each gender, a multiple stepwise regression was conducted to examine relationships between compulsive exercise and the predictor variables (self-orientated perfectionism, socially prescribed perfectionism, obsessive-compulsiveness, anxiety, depression, social physique anxiety, drive for thinness, bulimic attitudes, and body dissatisfaction). Significance was set at p < .001 due to the large sample size.

Table 1. Means and standard deviations for study variables by gender
VariablesMean (SD)Test of Difference z
  • Note: Samples sizes differed between tests due to missing data;

  • *

    p< .001 (two-tailed); CET, compulsive exercise test; EDI, eating disorder inventory; CAPS, child and adolescent perfectionism scale; SCAS-OC, spence child anxiety scale obsessive-compulsiveness subscale; HADS, hospital anxiety and depression scale; SPAS, social physique anxiety scale.

CET total10.01 (4.08)9.94 (3.66)0.32
EDI-drive for thinness2.82 (3.98)6.13 (5.90)10.40*
EDI-bulimia1.99 (3.18)2.41 (3.49)3.06*
EDI-body dissatisfaction4.97 (5.67)10.56 (8.13)13.02*
CAPS-self34.05 (7.66)32.91 (8.35)2.81*
CAPS-social25.15 (8.17)23.70 (8.33)3.51*
SCAS-OC0.91 (0.63)1.00 (0.67)2.21
HADS-anxiety7.48 (3.65)8.67 (3.65)6.23*
HADS-depression4.24 (2.91)3.92 (2.74)3.64*
SPAS21.61 (6.98)28.01 (7.98)15.11*


Characteristics of the Sample

Descriptive statistics can be seen in Table1. The mean CET total score represents a mid-point scoring average for boys and girls, and is noticeably less than has been reported in clinical samples. The EDI subscale scores also represent average to low levels of disordered eating attitudes. The HADS subscale scores demonstrate normal levels of depression (i.e., none), while the anxiety subscale mean indicates mild levels of anxiety, according to suggested norms.34 The SPAS scores, CAPS-self and CAPS-social, and SCAS-OC scores all represent normal levels for this age group.27, 29, 33

Regression Analysis

The final model of the multiple stepwise regression for boys and girls can be seen in Table2. For boys, the final multiple stepwise regression model was significant, accounting for 39% of the variance of CET total score. In the final model, EDI-drive for thinness, CAPS-self, SCAS-OC, and CAPS-social were the significant predictors. The CET total score was not statistically predicted by HADS-anxiety, HADS-depression, SPAS, EDI-bulimia, or EDI-body dissatisfaction.

Table 2. Final model for multiple stepwise regression of personality, psychological, and disordered eating variables (predictor variables) on to CET total score (outcome) for boys and for girls
PredictorsF (df)Adj r2BetaT
  • *

    Note: *p < .001; CET, compulsive exercise test; Adj, adjusted; EDI, eating disorder inventory; CAPS, child and adolescent perfectionism scale; SCAS-OC, spence child anxiety scale obsessive-compulsiveness subscale.

 Model80.99 (4, 508)*.39  
 EDI-drive for thinness  0.297.48*
 CAPS-self  0.276.64*
 SCAS-OC  0.174.29*
 CAPS-social  0.143.45*
 Model108.47 (3, 621)*.34  
 EDI-drive for thinness  0.349.69*
 CAPS-Self  0.318.77*
 SCAS-OC  0.153.97*

The final model of the multiple stepwise regression for the girls was significant, and it accounted for 34% of the variance of CET total score. The final model produced three unique significant predictors. These were EDI-drive for thinness, CAPS-self, and SCAS-OC. For girls, CET total was not statistically predicted by CAPS-social, HADS-anxiety, HADS-Depression, SPAS, EDI-bulimia, or EDI-body dissatisfaction.


This study aimed to examine the best cross-sectional predictors of compulsive exercise among a sample of adolescents. The results indicate that for both boys and girls a drive for thinness was the best predictor, along with self-perfectionism, and then obsessive-compulsiveness. For boys only, social perfectionism was also an additional predictor, although it did not explain as much variance as the other significant variables. The hypothesis that all significant predictors would be positively associated with compulsive exercise was supported.

The significant drive for thinness finding supports the existing literature linking compulsive exercise closely with ED.4 A key finding in this sample, though, was that only a drive for thinness, rather than bulimic attitudes and body dissatisfaction, significantly predicted compulsive exercise, for both boys and girls. This drive for thinness is analogous to symptoms of AN, and so this finding concurs with previous literature identifying compulsive exercise as more prevalent among AN patients than other ED diagnoses.4, 7 This compulsivity towards exercise could be a key marker in the development of AN, particularly as this association with a drive for thinness has been found in a generally healthy, nonclinical group of young adolescents. Further research is needed to identify how compulsive exercise interacts with a drive for thinness over time, and whether the exercise compulsivity puts individuals at increased risk of subsequently developing AN.

High levels of perfectionism and obsessive-compulsiveness were also linked to compulsive exercise. The finding that perfectionism was among the best predictors of compulsive exercise is consistent with previous investigations.12, 14 Although both subscales of perfectionism were significant for the boys, there was a greater association with the self-orientated form of perfectionism. Likewise, for the girls, it was self-perfectionism and not social perfectionism that was found to be a significant predictor of compulsive exercise. Castro and her colleagues27 found that self-orientated perfectionism was more strongly associated with ED than socially prescribed perfectionism. This is also in accordance with a previous study that had shown AN patients to experience their perfectionism as self-imposed.35 Therefore, this would suggest that self-perfectionism could be influential in the development of ED, and specifically AN, and that it could be operating through compulsive exercise. The results from the current study demonstrate that this association between self-perfectionism and compulsive exercise occurs even in an adolescent school-based population, where levels of disordered eating symptoms were relatively low. Therefore, if replicated longitudinally, this finding could represent a key area for early intervention and/or prevention work of compulsive exercise attitudes, whereby the individual's self-imposed high standards could be targeted with the aim of reducing the compulsivity towards exercise.

The close link between compulsive exercise and obsessive-compulsiveness has been widely established in previous research (Wyatt, Unpublished Doctoral Dissertation).13 The findings from this investigation demonstrate that a compulsivity towards exercise is associated with obsessive-compulsive symptoms even in a community sample of adolescent boys and girls. This close and direct association could indicate another possible area for prevention work of compulsive exercise; work that could target certain individuals with greater levels of obsessive-compulsiveness. However, it is uncertain whether the compulsivity towards exercise actually develops into a wider obsessive-compulsiveness, or whether the causal direction is in fact the reverse, with individuals with obsessive-compulsive symptoms being at greater risk for developing compulsive exercise.

An interesting result from the present study was the lack of association between compulsive exercise and the psychological factors of anxiety and depression, as well as social physique anxiety. This is contrary to previous literature,16 although it is possible that in a nonclinical adolescent sample, the level of psychological morbidity may not yet be closely linked with exercising. Further research would be needed to identify whether the association between compulsive exercise and these psychological factors only occurs in specific samples, such as in a clinical ED or adult population.

This study has certain limitations that need to be highlighted. First, the cross-sectional nature of the design prevents causal attributions. Future research needs to replicate these findings using a longitudinal and/or experimental design to further establish causality. Second, the self-report nature of the measures could have been susceptible to reporter bias, as well as response error, particularly given the relatively young age of the sample. This is particularly true of the self-report nature of height and weight information. The use of self-report BMI data has been previously used in this age group,36 although it is accepted that self-report BMI in adolescents must be viewed with caution and objective BMI measurement would be preferable in future research.

Overall, the current study findings support a model where drive for thinness, perfectionism, and obsessive-compulsiveness all predict compulsive exercise in adolescents. The amount of variance accounted for by these personality predictors was large, with almost 40% of compulsive exercise in boys being explained by these variables. This suggests that compulsive exercise is largely a self-driven behavior that is affected by personal attributes and, as such, any potential prevention work needs to target the individual's existing personality motivations and general beliefs. Future research is also required to test the longitudinal associations between these personality traits and compulsive exercise.

Earn CE credit for this article!

Visit: for additional information. There may be a delay in the posting of the article, so continue to check back and look for the section on Eating Disorders. Additional information about the program is available at