Introduction

Cognitive processes seem to play a vital role in the construct of body image, suggesting they are also responsible for the development of specific eating attitudes. The term body image refers to a psychological representation of the body’s size, shape and form as well as of our own feelings toward it [1]. It is commonly assumed that negative automatic thoughts and dysfunctional assumptions relating to body size, shape and weight are key etiological and maintenance factors in eating disorder psychopathology [2, 3].

According to several studies, fashion models [4, 5], along with dancers [6, 7], gymnasts, and certain athletes [8–10] represent professions in which the prevalence of full- or partial syndrome eating disorders seems to be higher than for the general population. Research suggests that exposure to media images depicting the thin-ideal body may be linked to body image distortion in women [11], and an increasingly thin ideal is dominating the press, television, and internet communities. Indeed, women presented in public media today are thinner than media images of women in the past [12], thinner than the actual female population [13], and often, thinner than would normally meet the diagnostic criteria for AN [14]. This ideal is pervasive with fashion models, cartoon characters, movie and television actresses, all having become increasingly slimmer over the past few decades [15–18]. Thus, media aimed at girls, adolescents, and young women are replete with extremely skinny models that portray an ideal which is unattainable to most [11]. What happens to the models who work for an industry that sets such high standards for beauty and body shape? Why is it that so many young models are dissatisfied with their bodies, regardless of their size?

Both ‘body image’ and ‘body weight’ are strongly connected to an individual’s sense of self-worth. Body image as a theoretical construct is thought to be multidimensional, consisting of perceptual, cognitive, affective, and behavioral components [2]. A dysfunction in one or more of those components leads to specific consequences. The perceptual component of body image refers to an overestimation of body dimensions [19]. The cognitive-affective component of a disordered body image is marked by dysfunctional thoughts and negative feelings toward one’s own body [20]. With regard to the behavioral component of a distorted body image, a distinction between body avoidance and checking behavior is made [21]. Body-related avoidance and checking behavior is primarily motivated by a fear of gaining weight. Various studies have established the importance of the part a distorted body image plays in the development [22] as well as the maintenance of eating disorders. Changes in dysfunctional cognitions regarding the body have been associated with a reduction in eating disorder psychopathology in the course of body image therapy [23, 24].

A vast body of research attempting to create a compilation of personality styles that may render a person more susceptible to eating disorders has been produced. The research [25] shows that eating disorders, particularly the restrictive type of AN, are related to obsessive–compulsive personality disorder (OCPD) traits. In the meta-analysis conducted by Young et al. [26], certain relationships between obsessive compulsive personality style/disorder and AN have been established. Seven out of the ten studies reviewed demonstrated a positive relationship between OCPD and/or obsessive compulsive disorder (OCD) in AN patients who exercise excessively. Anorexia nervosa patients demonstrate personality traits which are highly concordant with OCPD-perfectionism, rigidity, higher impulse control, and emotional restraint. Anorexia nervosa patients show minimal changes in obsessive personality characteristics following weight restoration, suggesting that such pre-morbid personality traits play a role in the pathogenesis of AN. People suffering from AN (the restrictive type) are perceived as being anxious, as well as avoidant [27]. Individuals suffering from the binging/purging type of AN are more social. They strive for the attention of others and have a tendency to exaggerate their feelings and thoughts. It has also been pointed out that bulimic individuals manifest certain borderline personality traits, whereas anorexics tend, most notably, to manifest obsessive–compulsive traits.

Among the research related to dissatisfaction with body image, it is hard to find studies directly aimed at assessing this problem in professional models. In addition, there are no satisfactory results comparing professional models with a control group in terms of eating attitudes and body image. Similarly, no research examining dysfunctional cognitive processes in relation to eating behaviors, in the context of personality styles has been conducted for models. The aim of this study was to investigate possible differences between professional models and a control group of women in terms of their attitudes toward eating, dysfunctional thoughts about eating and body image, and personality styles. Based on the findings reviewed in the previous paragraphs, our first hypothesis predicts that models will show higher levels of dysfunctional thoughts about eating and body image than the controls.

Personality styles should play a critical role in the risk of developing eating disorders. Eating disorders are associated with dysfunctional cognitions concerning (a) the body and self-image, (b) dietary behavior, and (c) loss of control in eating. These three factors are assessed by the “Eating Disorder Cognition Questionnaire” [28]. To date, there has been no research relating the EDCQ scales to personality styles. Here, we measure personality with the “Personality Styles and Disorders Inventory” [29]. This questionnaire assesses 14 personality styles in the non-pathological range, which are related to well-known personality disorders, as described in the DSM-IV (American Psychiatric Association, 2000).

The most relevant scales for this study are (a) ambitious/narcissistic, (b) conscientious/compulsive, and (c) spontaneous/borderline. Because these scales are of particular relevance for successful models and based on the research reviewed previously, we hypothesize that these personality styles will moderate dysfunctional thoughts about eating as assessed by the EDCQ scales, when comparing models and controls. That is, we expect that models with higher scores in these personality styles will show significantly higher scores for dysfunctional thoughts about eating and body image than women in the control group.

Specifically, our second hypothesis predicts that the ambitious/narcissistic style will moderate “negative body and self-esteem” (E-Body). Our third hypothesis predicts that the conscientious/compulsive style will moderate “diet restriction” (E-Diet). Our fourth hypothesis predicts that the spontaneous/borderline style will moderate “loss of control in eating” (E-Control-Loss), which is closely related to bulimic disorder.

Finally, we hypothesize that personality styles will moderate dysfunctional thoughts about eating as assessed by the EDCQ scales, when comparing models and controls. That is, we expect that models with higher scores in these personality styles will show significantly higher scores for dysfunctional thoughts about eating and body image than women in the control group.

Method

Participants

This study included 43 female professional models and, as a control group, 43 women who had never been involved in professional modeling and who did not follow a specific diet regimen. Of 100 models contacted, 43 agreed to complete the survey, and their participation was voluntary. 25 were Polish, five came from the USA, five came from the UK, three were Italian and five were German. The women from the control group came exclusively from Poland and answered the questionnaires where the instructions were in Polish. They had never been involved in professional modeling and were matched with the models in terms of age, socioeconomic status, and educational level. The mean age for the models was 25.39 (SD = 2.76; Range 17–34 years), and for the control participants: M = 24.79 (SD = 2.36; Range 17–35 years). For models, 48 % had an upper secondary education level, 35 % vocational education, and 17 % had completed higher education. For controls, 49 % had an upper secondary education level, 31 % vocational education, and 20 % had completed higher education. Most of the models came from a city, as did most of the control women.

Self-report measures

Dysfunctional assumptions in eating disorders

To assess the level of dysfunctional assumptions in eating disorders, we used the “Eating Disorder Cognition Questionnaire” [28] (in Polish and English versions). Responses are given on a 4-point Likert scale [“In the last month I have had this thought: not at all (0), sometimes (1), frequently (2), or constantly (3)”]. The EDCQ consists of 28 items and three scales: (1) “Negative assumptions about body and self-esteem”. This encompasses the desire for a perfect figure, dissatisfaction with one’s body, negative thoughts regarding one’s body, approval and dependency on others as well as a sense of self-worth in relation to body satisfaction. A sample item is: “My thoughts circle around my body weight”. This scale correlates with several types of eating disorders [28], (α = 0.92 and α = 0.91, for the Polish and English versions, respectively). (2) “Restriction and dietary restraint”. This scale refers to thoughts about dieting, restrictions, dietary rules aimed at regulating body weight and it shows positive correlations with anorexic behaviors. A sample item is: “I have to get a grip on myself and not eat anything” (α = 0.87 and α = 0.92, for the Polish and English versions, respectively). (3) “Eating and loss of control”. The scale consists of statements about food cravings as well as strategies employed to control food intake and is associated with bulimic behaviors. A sample item is: “I will lose control and have a binge episode any moment” (α = 0.79 and α = 0.92, for the Polish and English versions, respectively).

Attitudes toward eating

The Eating Attitudes Test (EAT) [30] was administered. This instrument contains 26 statements representing attitudes and behaviors about eating, and it is associated with AN. The EAT-26 contains three factors: (1) “Dieting”, drive for thinness and dieting behaviors, (2) “Bulimia and Food Preoccupation”, food thoughts, and bulimic behaviors, and (3) “Oral Control”, pressure from others to gain weight and control eating. The EAT-26 questionnaire is the most widely used tool in the diagnosis of AN. The questionnaire consists of items related to eating behaviors, rated on a 4-point Likert scale, ranging from never (0) to always (3). Sample items: “I cut my food into small pieces”; “I avoid eating when I am hungry”. The Polish version of the scale has a Cronbach’s α of 0.88.

Personality styles

We applied the Personality Styles and Disorders Inventory [28]. The PSDI is a standardized and normalized measure designed to assess personality styles in the non-pathological range, which are analogous to the personality disorders described in psychiatric diagnostic manuals, such as the DSM-IV (American Psychiatric Association, 2000). It was constructed on the basis of Personality Systems Interactions (PSI) theory [29], and on the symptoms described in the DSM-IV Axis II related to personality disorders. The scales of the PSDI are (Style/disorder): Ambitious/narcissistic, Assertive/antisocial, Charming/histrionic, Conscientious/compulsive, Critical/passive aggressive, Intuitive/schizotypal, Loyal/dependent, Optimistic/rhapsodic, Passive/depressive, Reserved/schizoid, Self-Critical/avoidant, Spontaneous/borderline, Unselfish/self-sacrificing, and Willful/paranoid. The scales correspond to the 11 DSM-IV personality disorders, with the exception of the Optimistic/rhapsodic, Passive/depressive, and Unselfish/self-sacrificing scales, which are additionally included in the PSDI. Responses are given using a 4-point Likert scale [“This statement applies to me: not at all (0), somewhat (1), much (2), or completely (3)”]. The most relevant PSDI-S scales for this research were: Ambitious/narcissistic (Sample item: “Being admired is especially appealing to me.”), Conscientious/compulsive (Sample item: “I have firm principles that I always adhere to.”), and Spontaneous/borderline (Sample item: “My feelings about something or someone frequently change very abruptly.”). The reliability of the scales (Cronbach’s α) for the English version lies between α = 0.74 and α = 0.86 [30]. For the Polish translation, the reliability lies between α = 0.61 and α = 0.83 [31].

Procedure

The study was conducted online. One of the authors is a professional model and she recruited the models through well-known modeling agencies. All were currently employed. Her picture appeared on an Internet page with an invitation to participate in the survey, which consisted of four questionnaires (in English or in Polish). They also completed the survey through the Internet. The website with questionnaires was written in following computer languages: HTML, Java Script, Ajax, CSS3, PHP5 and MySQL. The website was built to protect the anonymity of participants and store in a database all the information necessary for this research and then to calculate the final results. There was no information that could have been used to identify the respondents. The website does not restore sensitive information such as: name, surname or address or other information that could lead to the identification of the participant. For better security protection, no information was stored about the device that a participant used to answer the questionnaires. All participants provided their informed consent. The study was approved by the Institutional Review Board. Completion of the survey took about 30 min.

Results

Differences between models and controls

Preliminary analyses indicated that most of the variables were not normally distributed, which prevented us from using t tests. We therefore carried out non-parametric Mann–Whitney U tests for independent groups to find out whether Fashion Models differed significantly from Controls in terms of the defined variables: Dysfunctional thoughts in eating disorders, eating attitudes, and personality styles. The analyses encompassed the following variables: the general EAT score, the EDCQ factors ‘negative body and self-esteem’ (E-Body), ‘dietary restraint’ (E-Diet), and ‘loss of control in eating’ (E-Control-Loss), and personality styles measured with the PSDI-S. To control for the family-wise error rate due to multiple comparisons, we applied the Holm–Bonferroni correction [32]. The results are shown in Table 1.

Table 1 Mann–Whitney U tests contrasting models versus controls on Eating Attitudes (EAT), Dysfunctional Thoughts Related to Eating (EDCQ), and Personality Styles (PSDI-S). Family-wise error rate was controlled with the Holm–Bonferroni correction (α = 0.05)
Full size table

Confirming our first hypothesis, fashion models had a significantly higher EAT general score than controls, measuring the level of attitudes and behaviors associated with eating. The models also had significantly higher scores for dysfunctional thoughts on eating, as measured by the EDCQ subscales (see Table 1). Concerning personality styles, the models had higher scores than controls on nine PSDI-S scales, including ambitious/Narcissistic and spontaneous/Borderline. Unexpectedly, on the scale conscientious/Compulsive, the controls had numerically higher scores than the models (Mean = 6.98 vs. Mean = 5.23, respectively), although the contrast using the Mann–Whitney test failed to reach the level of significance (see Table 1).

Relationships between EDCQ vs. and EAT or personality styles

As a next step, the relationships between the three indicators of dysfunctional thoughts in eating disorders (EDCQ) or personality styles (PSDI-S) and the general EAT score were examined with non-parametric Spearman Rho correlations. The results are shown in Table 2, presented separately for models (n = 43) and for controls (n = 43). All of the EDCQ scales—Negative body and self-esteem, dietary restraint, and loss of control in eating—correlated significantly with the EAT general score in the control group. For models, only the negative body and self-image scales were significant. Concerning the PSDI-S, most scales correlated with the EDCQ scales, especially for the group of models. It is noteworthy that for models the scales Ambitious/narcissistic and Spontaneous/borderline correlated positively with all three EDCQ scales, which was not the case for participants from the control group (see Table 2). The scale Conscientious/compulsive correlated significantly with each EDCQ scale for both groups of participants (see Table 2).

Table 2 Spearman Rho correlations between the Eating Attitudes Test (EAT) or the Personality Styles and Disorders Inventory (PSDI-S) and the three subscales of the Eating Disorder Cognition Questionnaire (EDCQ)
Full size table

Moderation of personality styles on EDCQ scales

Ambitious/narcissistic style moderates negative body and self-esteem

To test our second hypothesis that the ambitious/narcissistic style moderates negative body and self-esteem (E-Body), we carried out a hierarchical regression analysis. Because negative body and self-esteem was not normally distributed, we applied a square root transformation. In addition, we centered the predictors.

Before doing the analysis, we checked if the assumptions of the regression analysis were met. In a preliminary analysis, we calculated the distance of Mahalanobis (D2) on the data, which identified two participants as multivariate outliers. After removing them, in the final analysis, an inspection of graphical distribution of D2 on Q–Q plots indicated a normal multivariate distribution. An inspection of Cook’s distance showed that no participant’s data would change the regression analysis coefficients significantly (neither maximum nor minimum Cook’s distance exceeded 1). The data also met the assumption of non-zero variances (Body and Self-esteem, Variance = 0.98; Ambitious Personality, Variance = 7.13).

We examined the absence of multicollinearity using the tolerance index (Ti) and the variance inflation factor (VIF). A Ti of more than 0.02 and a value less than 5.0 for VIF are considered reliable cutoff points for the absence of multicollinearity [33]. The results showed that multicollinearity was not a concern either at step 1 (Ambitious, Ti = 0.78, VIF = 1.29) or at step 2 (Ambitious, Ti = 0.64, VIF = 1.57; Group, Tolerance = 0.75, VIF = 1.33; Group × Ambitious, Ti = 0.82, VIF = 1.23) of the analysis. The data also met the assumption of independent errors: Durbin–Watson value = 1.73 (the value was close to 2). Finally, the analysis of the standardized residuals indicated that the data contained approximately normally distributed errors (the results of the Kolmogorov–Smirnov and the Wilks–Shapiro tests were both non-significant, p > 0.05) and the correlation between the standardized residuals and each independent variable was non-significant (close to 0.0).

In order to carry out the hierarchical regression analysis, we entered the centered ambitious personality and centered group in the first step (as dummy variable: 0 = controls; 1 = models). In the second step, we entered the interaction of these variables. The outcome was the square root transformation of negative body and self-esteem. The results yielded a significant effect of ambitious personality, β = 0.58 (t = 4.77, p < 0.001) and a significant interaction, β = 0.30 (t = 2.78, p = 0.007). The variables entered in the first step accounted for 17 % of the variance (R 2), and the model was significant, F(2, 80) = 9.10, p < 0.001. The interaction entered in the second step accounted for 7.7 % of additional variance (R 2), and the model was significant, F(3, 79) = 7.68, p < 0.001. To explore the interaction, we used restructured simple regression equations and plotted regression lines at 1 SD below and 1 SD above the mean values of the moderator [34]. The results are depicted as standard scores in Fig. 1. The simple-slope analysis for models was significant: t = 4.09, p < 0.001, whereas the simple-slope analysis for the control group was not: t = 1.77, ns. These results show that models characterized by a more prevalent ambitious/narcissistic personality style display a higher amount of negative thoughts about their body and self-esteem than models with lower scores on this style and individuals from the control group.

Fig. 1
figure 1

Negative body and self-esteem (E-Body; standard scores) as a function of group and personality (low vs. high ambitious/narcissistic). Low and high values on personality correspond to one standard deviation below or above the mean, respectively. n = 84

Full size image

Conscientious/compulsive style moderates dietary restraint

To test our third hypothesis that the conscientious/compulsive style moderates dietary restraint (E-Diet), we carried out an analogous hierarchical regression analysis. Because dietary restraint was not normally distributed, we applied the square root transformation. In addition, we centered the predictors.

Before doing the analysis, we checked if the assumptions of the regression analysis were met. In a preliminary analysis, we calculated the distance of Mahalanobis (D2) on the data, which identified four participants as multivariate outliers. After removing them, in the final analysis, an inspection of graphical distribution of D2 on Q–Q plots indicated a normal multivariate distribution. An inspection of Cook’s distance showed that no participant’s data would change the regression analysis coefficients significantly (neither maximum nor minimum Cook’s distance exceeded 1). The data also met the assumption of non-zero variances (Diet restraint, Variance = 0.52; Obsessive Personality, Variance = 8.30).

Results showed that multicollinearity was not concern either at step 1 (Obsessive, Ti = 0.80, VIF = 1.26) or at step 2 (Obsessive, Ti = 0.84, VIF = 1.19; Group, Tolerance = 0.75, VIF = 1.33; Group × Obsessive, Ti = 0.92, VIF = 1.10) of the analysis. The data also met the assumption of independent errors: Durbin–Watson value = 1.59 (the value was higher than 1 and closer to 2). Finally, analysis of the standardized residuals indicated that the data contained normally distributed errors (the results of the Kolmogorov–Smirnov and the Wilks–Shapiro tests were both non-significant, p > 0.20 in each case) and the correlation between the standardized residuals and each independent variable was non-significant (close to 0.0).

In order to carry out the hierarchical regression analysis, in the first step, we entered centered the obsessive personality and centered group (as dummy variable: 0 = controls; 1 = models). In the second step, we entered their interaction. The outcome was the square root transformation of dietary restraint. The results yielded the significant effect of obsessive personality, β = 0.29 (t = 2.70, p = 0.008), of group, β = 0.48 (t = 4.49, p < 0.001), and a significant interaction, β = 0.20 (t = 1.99, p < 0.05). The variables entered in the first step accounted for 21 % of the variance (R 2), and the model was significant, F(2, 81) = 10.72, p < 0.001. The interaction entered in the second step accounted for 4.0 % of additional variance (R 2), and the model was significant, F(3, 81) = 8.73, p < 0.001. To explore the interaction, we used restructured simple regression equations and plotted regression lines at 1 SD below and 1 SD above the mean values of the moderator [32]. The results are presented as standard scores in Fig. 2. The simple-slope analysis for models was highly significant: t = 3.73, p < 0.001, whereas the simple-slope analysis for the control group was not significant: t = 1.03, ns. These results show that models characterized by a more prevalent conscientious/compulsive personality style display a higher amount of dietary restraint than models with lower scores for this style and individuals from the control group.

Fig. 2
figure 2

Dietary restraint (E-Diet, standard scores) as a function of group and personality (low vs. high conscientious/compulsive). Low and high values on personality correspond to one standard deviation below or above the mean, respectively. n = 82

Full size image

Spontaneous/borderline style moderates loss of control in eating

To test our fourth hypothesis that the spontaneous/borderline style moderates loss of control in eating (E-Control-Loss), we carried out an analogous hierarchical regression analysis. Because loss of control in eating was not normally distributed, we applied the square root transformation. In addition, we centered the predictors.

Before doing the analysis, we checked if the assumptions of the regression analysis were met. In a preliminary analysis, we calculated the distance of Mahalanobis (D2) on the data, which identified four participants as multivariate outliers. After removing them, in the final analysis, an inspection of graphical distribution of D2 on Q–Q plots indicated a normal multivariate distribution. An inspection of Cook’s distance showed that no participant’s data would change the regression analysis coefficients significantly (neither maximum nor minimum Cook’s distance exceeded 1). The data also met the assumption of non-zero variances (Diet restraint, Variance = 0.45; Obsessive Personality, Variance = 5.59).

The results showed that multicollinearity was not concern either at step 1 (Obsessive, Ti = 0.87, VIF = 1.15) or at step 2 (Spontaneous, Ti = 0.64, VIF = 1.57; Group, Tolerance = 0.86, VIF = 1.17; Group × Spontaneous, Ti = 0.73, VIF = 1.37) of the analysis. The data also met the assumption of independent errors: Durbin–Watson value = 2.08 (the value was very close to 2). Finally, the analysis of the standardized residuals indicated that the data contained normally distributed errors (the results of the Kolmogorov–Smirnov and the Wilks–Shapiro tests were both non-significant, p > 0.20 in each case) and the correlation between the standardized residuals and each independent variable was non-significant (close to 0.0).

To carry out the hierarchical regression analysis, we entered the centered spontaneous personality and centered group in the first step (as dummy variable: 0 = controls; 1 = models). In the second step, we entered their interaction. The outcome was the standardized score for loss of control in eating. The results yielded the significant effect of a spontaneous personality, β = 0.54 (t = 5.37, p < 0.001), and a significant interaction, β = 0.25 (t = 2.27, p = 0.026). The variables entered in the first step accounted for 31 % of the variance (R 2), and the model was significant, F(2, 80) = 17.10, p < 0.001. The interaction entered in the second step accounted for 4.4 % of additional variance (R 2) and the model was significant, F(3, 80) = 13.73, p < 0.001. To explore the interaction, we used restructured simple regression equations and plotted regression lines at 1 SD below and 1 SD above the mean values of the moderator [33]. The results are presented as standard scores in Fig. 3. The simple-slope analysis for models was highly significant: t = 5.35, p < 0.0001, and the simple-slope analysis for the control group was also significant: t = 3.59, p < 0.001. These results show that models characterized by a more prevalent spontaneous/borderline personality style display a higher amount of loss of control in eating (related to bulimia) than models with lower scores for this style and individuals from the control group. Notice that women in the control group with higher scores in terms of spontaneous personality also show less control in eating than those with lower scores.

Fig. 3
figure 3

Loss of control in eating (E-Control-Loss, standard scores) as a function of group and personality (low vs. high spontaneous/borderline). Low and high values on personality correspond to one standard deviation below or above the mean, respectively. n = 82

Full size image

Discussion

The findings suggest that professional female models displayed a statistically higher level of attitudes and behaviors associated with AN (EAT), as well as a higher level of dysfunctional thoughts related to ‘negative body and self-esteem’, ‘dietary restraint’, and ‘loss of control in eating’ than the control group. These results confirm our first hypothesis and suggest that models may be at risk of developing serious eating disorders, such as anorexia or bulimia nervosa [2, 3]. The research suggests that direct contact with “thinness ideals”, as presented in the media, may result in a distorted body image in women [11], which models would be especially prone to.

The results regarding the importance of personality styles (PSDI-S), as shown in Table 1, indicate that models achieved higher scores on the ambitious/narcissistic, spontaneous/borderline, independent/schizoid, critical/passive aggressive, passive/depressive, and self-critical/avoidant (as a tendency) scales than the controls. All of those personality styles also demonstrated significant relationships with dysfunctional thoughts about eating, as measured by the EDCQ (Table 2). The results have been partially confirmed by other researchers [35], who have established that, for example, borderline personality traits co-occur with eating disorders. The models scored higher in terms of passive/depressive and self-critical/avoidant (tendency) personality styles. These styles characterize individuals with higher levels of anxiety (including social anxiety), who are in need of social approval, have a low sensitivity to positive affect, and are susceptible to criticism and afraid of rejection. Diagnoses of personality disorders from the C cluster of DSM-IV, accompanied by high levels of anxiety, including social anxiety, are more common among AN patients [27].

The moderator analyses carried out were consistent with our hypotheses #2, #3, and #4, respectively: The ambitious/narcissistic style moderates “negative body and self-esteem” (Fig. 1), the conscientious/compulsive style moderates “diet restriction” (Fig. 2), and the spontaneous/borderline style moderates “loss of control in eating”, which is closely related to bulimic disorder (Fig. 3). In each case, models with higher scores on the respective personality style had higher scores on the respective dysfunctional thought concerning body and self-image or eating problems, as assessed by the EDCQ scales.

The models had higher scores for the ambitious/narcissistic and spontaneous/borderline personality styles (Table 1) than the controls. These personality traits lie closer to the B cluster of the DSM-IV, and form the so-called “emotional” personality disorders. The ambitious/narcissistic style is characterized by an intense sensitivity to stimuli related to positive affectivity; for example, deep levels of commitment, striving for success, and extraordinary achievements [30]. Uncertain self-esteem, dependent on external reinforcement and a chronic need for admiration and appreciation are typical for the modeling industry, where external appearance is crucial. For models, external admiration of their body is not only associated with a successful career, but may also be related to one’s self-esteem, especially for models with high scores on ambitious/narcissistic personality, leading to diverse eating disorders. It has been found that people suffering from eating disorders also display high levels of narcissism [35, 36].

Individuals displaying the spontaneous/borderline personality are characterized by impulsivity and frequent mood swings. These people have a tendency to initiate intensive, but unstable relationships. A chronic feeling of emptiness and a lack of the ability to identify with their self-image often follow [29]. It is assumed that borderline personality disorder is a far more common comorbid disorder of the bulimic (purging) subtype of AN, rather than the restrictive type. Davis et al. [37] established that features of the spontaneous/borderline personality make for a significant risk factor for developing maladaptive beliefs concerning body weight and may lead to the development of an eating disorder.

A relationship between obsessive/compulsive personality and restrictive behaviors in individuals prone to AN is reported in the meta-analysis of Young et al. [26]. For example, AN patients with obsessive/compulsive personality, who exercised excessively, showed high levels of depression, low self-esteem and dietary restraint [26]. Notably, there were no differences between women in the control group and models on the conscientious/compulsive style (Table 1), which was unexpected. On the other hand, only models, and not the controls, showed significant relationships between this style and each EDCQ scale (Table 2). Moreover, as shown in Fig. 2, only models with higher scores in terms of conscientiousness (and not controls) showed higher levels of “diet restriction”.

To explain the above pattern of results, we suggest that what is important to predict dysfunctional thoughts about eating and maladaptive eating behavior is the relative level of the conscientious/compulsive personality within a particular group known to be at risk of developing an eating disorder (e.g. models). Personality-specific traits within the group of compulsive models display the striving for a beauty ideal and perfectionism with regard to their body image. A never-ending striving to attain and maintain a beautiful and perfect body through the drive for thinness and restrictive dietary behaviors is an important feature of compulsive models.

More specific treatments should focus on self-regulatory processes. For example, for models displaying a high level of the spontaneous/borderline style, the emphasis should be placed on self-relaxation processes, which have proven to be effective in controlling eating, as well as anorexic and bulimic behavior. Activating self-regulatory mechanisms translates not only into a decrease in eating disorder symptoms, but also reduces psychosomatic complaints, depressive symptoms as well as anxiety. Moreover, extreme levels of personality traits are only maladaptive if they are associated with low self-regulatory abilities [38]. Although changing these personality styles is not easy, there are different therapeutic approaches that can help to diminish their influence on maladaptive cognitions and behavior, such as body image therapy [23, 24], or approaches focusing on the development of self-regulatory competences.

Some limitations of this study should be noted. There was no longitudinal design and no follow-up measurement. The correlational design of this study does not allow conclusions to be drawn on causal relationships between self-reported personality factors and eating disorder levels. This will have be to investigated in an experimental or longitudinal study design. We cannot substantiate any statement regarding the risk of developing pathology, as this is a cross-sectional study. The prevalence of eating disorders seems to be higher in professional fashion models than it is for the general population; however, this prevalence may be exactly measured in a longitudinal study with a very large sample size. The strengths of this study are the very rare and interesting sample size and the results showing that not all models are susceptible to dysfunctional cognitions about eating and body images and the importance of the moderating role of their personality styles. The personality styles: conscientious compulsive-like, spontaneous borderline-like and ambitious narcissistic-like personality relate to dysfunctional cognitive processes in eating and weight. In a future research project, we need to design more studies directly aimed at assessing the relationship between these personality styles and eating attitudes in professional models.