Introduction

According to Cash [2], body image “encompasses one’s body-related self-perceptions and self-attitudes, including thoughts, beliefs, feelings and behaviors” (pp 1–2). In other words, body image comprises a perceptual, a cognitive-affective and a behavioral component. Regarding the perceptual component, previous studies have shown that patients with Anorexia Nervosa (AN) and Bulimia Nervosa (BN) tend to overestimate the size of their bodies to a comparable amount ([24]; for a meta-analytic review, see [3]), while healthy controls tend to underestimate the size of their bodies [30], although results are mixed and depend on the methods being used [3, 9, 27]. Farrell et al. [9] conclude in a review that optical distortion techniques are closest to achieving construct and ecological validity and that they are superior to analogue scale procedures. Moreover, a review by Cash and Deagle [3] showed that larger effects occurred with whole body than with body-part size estimation techniques. It has been argued that body-size overestimation can largely be explained by the underweight of patients with AN [7], although Cash and Deagle [3] found that perceptual distortion indices do not differentiate between subject with AN and BN. The cognitive-affective component of body image comprises deviations in cognitive processing [6, 8, 16] and negative emotions that are elicited by being confronted with one’s own body [28, 31]. The behavioral component of body image includes avoidance behavior [26], i.e., the avoidance of situations that might provoke concerns about physical appearance like looking in the mirror, wearing tight clothes or going swimming [23]. However, patients with eating disorders differ in the amount of avoidance behavior and some patients tend to frequently go through body checking behavior [25].

Based on clinical observations, it is assumed that body-related avoidance behavior prevents individuals from disconfirming the distorted perception of their own body. This assumption is supported by a finding from Geissner and Bauer [13] who could show that confrontation with one’s body, which counteracts body-related avoidance behavior, can help to correct a distorted self image. To our knowledge, only one study [23] has investigated the association between body-related avoidance behavior and body-size overestimation to date. Rosen et al. [23] used a visual analogue scale and asked subjects to estimate the size of four different body areas (bust, waist, hips and abdomen). Body-size distortion was then calculated as the percentage of over- and underestimation of the average of these four regions (see [33]). In their nonclinical sample, a small but significant correlation between body-related avoidance behavior and overestimation of the size of four body areas emerged [23]. Although this study provides valuable hints that body-related avoidance behavior and body-size distortion are associated, the question remains open whether this association also holds true for clinical samples, i.e., in patients with eating disorders. Furthermore, it is not clear yet if this relationship still emerges when subjects have to evaluate whole body size instead of body areas as these are considered two methodologically different approaches [3]. The latter is especially important, as there is evidence that in patients with eating disorders, overestimation of one’s whole body is more pronounced as compared to overestimation of specific body areas [3]. Therefore, the main aim of the present study was to analyze the association between body-related avoidance behavior and overestimation of one’s body size based on an estimation of the size of one’s whole body instead of particular body areas like in the study by Rosen [23]. Moreover, we used a digital photo distortion technique instead of a visual analogue scale because the former is considered to have a higher ecological validity than the use of visual analogue scales [9]. In order to control for the possible influence of weight and shape concerns and Body Mass Index (BMI), we used a regression analysis with body-related avoidance behavior, weight and shape concerns and BMI as predictors for the overestimation of one’s body size. As it has been suggested that in AN, body-size overestimation might be a secondary effect of the observer’s low body size and might not indicate eating disorder pathology [7], we analyzed a sample of participants with eating disorders that did not include any patients with AN.

Method

Participants

Participants were female outpatients from the psychotherapy centers of the Ruhr University Bochum and the Johannes-Gutenberg University Mainz, Germany. The entire sample comprised N = 78 participants (N = 31 females with BN, N = 47 patients with an Eating Disorder Not Otherwise Specified, EDNOS). The diagnoses were assessed using a German version of the Structured Clinical Interview for DSM-IV Axis I disorders (SCID-I; [34]). We included participants who were at least 16 years old and suffered an eating disorder other than AN. The studies which provided data for the analyses reported here were approved by the responsible ethics committee and signed informed consent was obtained from all participants. Information on the sample characteristics can be derived from Table 1. Subjects with EDNOS were significantly older (t = 4.81, p < 0.001), had a higher BMI (t = 8.37, p < 0.001), showed more avoidance (t = 4.57, p < 0.001) and had less weight and shape concerns (t = 4.99, p < 0.001) as compared to subjects with BN.

Table 1 Sample characteristics
Full size table

Measures

Eating disorder examination-questionnaire

The Eating Disorder Examination-Questionnaire (EDE-Q, [10, 17]) is a self-report measure of symptoms and relevant characteristics of eating disorders that were experienced during the last 28 days. The items are rated on a scale from 0 to 6. For the present study, we used a brief version consisting of the Weight and Shape Concern items (8 items), as previous studies have shown that the two scales “Weight Concerns” and “Shape Concerns” largely combined into one factor [1, 15]. The EDE-Q composite score “Weight and Shape Concern” had an internal consistency of α = 0.92 for participants with an eating disorder [1].

Body image avoidance behavior

Avoidance behavior was assessed with the subscales “Clothing” and “Social Activities” from the Body Image Avoidance Behavior Questionnaire (BIAQ; [20, 23]. Participants indicated on a six-point Likert scale (“never” to “always”) how much a specific avoidance behavior applies to them. As only the subscales “Clothing” and “Social Activities” directly assess body-related avoidance behavior, we only used the 13-item general avoidance score that had also been used in previous research [32]. Cronbach’s Alpha for the general avoidance score was α = 0.77 [32].

Digital photo distortion technique

A digital photo distortion technique (for a detailed description see [30]) was used to explore the degree of body-size misperception. For this purpose, a photo of each participant was taken from a frontal perspective while wearing a standardized tight-fitting suit with short arms and legs. Participants were then shown photos that were initially distorted to 80 and 120 % of their real dimensions. The order of the starting points was counterbalanced. Responding to the question “What do you really look like?”, participants had to estimate their actual body dimensions by adjusting the width of the presented photo with the left and right arrow keys. The procedure was repeated four times, two times from each of the two starting points. The order of the starting points was counterbalanced. The four values from two different starting points were merged into an overall score. The degree of deviation was automatically assessed by a computer program that saved the answers as a percentage of the real body size. Scores of more than zero indicate a tendency towards overestimation, whereas scores of less than zero indicate a tendency towards underestimation. Internal consistency of the photo distortion technique was high, ranging between α = 0.90 and α = 0.92 [30, 31].

Procedure

In a first session, the SCID-I was administered to each participant by a trained clinical psychologist. Age, height and weight as well as the self-report questionnaire measures were assessed by the patients in a paper–pencil format. In a second session, the photo distortion technique was applied.

Data analysis

Pearson correlations were calculated to find out how BMI, body-related avoidance behavior, weight and shape concerns and overestimation of one’s body size were interrelated. In a multiple regression analysis (enter method) controlling for BMI, the EDE-Q composite score “Weight and Shape Concern” as well as the composite score of the BIAQ, we tested if body-related avoidance behavior significantly contributes to the explanation of body-size overestimation. In order to rule out multicollinearity, variance inflation factors and tolerance scores were computed for each predictor in the regression analysis. As recommended by Urban and Meyerl [29], none of the variance inflation factors was >5 and all tolerance scores were >0.25. In detail, the highest variance inflation factor was 1.5 and the lowest tolerance score was 0.67. An alpha level of 0.05 was used for all statistical tests.

Results

Accuracy of estimation

The mean average score for the deviation was M = 0.044 (SD = 0.13), which indicated a mean overestimation of 4.4 %. An additional t test indicated that the overestimation was significantly different from zero [t (1,77) = 3.12, p = 0.003], while zero would represent an accurate estimation of body size, a significant positive value represents an overestimation. Participants with BN (M = 0.047, SD = 0.14) and EDNOS (M = 0.42, SD = 0.12) did not differ in the amount of overestimation [t (76) = 0.169, p = 0.87].

Correlational analysis

The correlational analysis (Pearson) revealed a significant correlation between the composite score of the BIAQ and the photo distortion technique [r(76) = 0.25, p = 0.03]. Table 2 presents all Pearson correlation coefficients between the BIAQ, the BMI, EDE-Q score weight/shape concerns and the photo distortion technique.

Table 2 Correlations between photo distortion technique, body mass index, shape and weight concerns, and Body Image Avoidance Questionnaire
Full size table

Regression analysis

Whereas the variables BMI and EDE-Q score weight/shape concerns did not significantly explain variance in body-size overestimation, the regression analysis showed that the BIAQ exclusively contributed to the explanation of variance in body-size overestimation (see Table 3).

Table 3 Linear regression (enter method) to predict body-size overestimation
Full size table

Discussion

The main aim of the present study was to analyze the association between body-related avoidance behavior and body-size overestimation in a clinical sample of participants with BN and EDNOS. We were able to demonstrate that our participants significantly overestimated their body size by 4.4 % on average, which is comparable to the overestimation of 4.5 % that Tovée et al. [27] found in participants with AN. Therefore, we conclude that in our sample, a small but relevant overestimation occurred. This contradicts the assumption that overestimation is a result of low body weight instead of a consequence of the eating disorder symptoms [7], but it is in line with the results of the metaanalysis of Cash and Deagle [3] who found that the amount of overestimation is comparable in AN and BN. Our main result is that body-related avoidance behavior is positively associated with body-size overestimation over and above potential contributions of weight and shape concerns and BMI. This result provides preliminary support for the hypothesis that body-related avoidance behavior prevents individuals with eating disorders from disconfirming their distorted view of body size [13]. This result is also in line with a finding from a study on a nonclinical sample by Rosen et al. [23] who reported a similar association between body-related avoidance behavior and body-size overestimation with a different methodology. In our sample, we did not find the positive association between weight and shape concerns and overestimation of one’s body size reported several times before [14, 19, 22, 25]. Weight and shape concerns did not significantly contribute to the explanation of body-size overestimation in the regression model. Therefore, our study suggests that the positive association between body-related avoidance behavior and overestimation does not simply go back to weight and shape concerns.

Although our study was the first to find an association between perception of overall body-size and body-related avoidance behavior in a clinical sample of patients with eating disorders, some limitations warrant mentioning. Firstly, we did not carry out a prospective study and therefore cannot conclude whether body-related avoidance behavior is cause or consequence of the development and maintenance of a distorted perception of the size of one’s own body. Moreover, it would be premature to generalize the finding to all kinds of eating disorders as our sample is only compromised of subjects with BN and EDNOS, who had a relatively high BMI. However, assuming that comparable mechanisms are at work in the psychopathology of all kinds of eating disorders [11], it is not unlikely that similar findings emerge in other types of eating disorders in future studies. Moreover, our results strengthen the view that overestimation is associated with eating disorder psychopathology instead of simply going back to low body weight as it has been suggested for AN [7].

Overall, we found evidence that body-related avoidance behavior is positively associated with an overestimation of the size of one’s body and that this association seems to be independent from BMI or weight and shape concerns. This result provides empirical support for the assumption that the association is caused by the limited feedback concerning one’s body size which goes along with avoidance behavior. However, the amount of variance that can be explained by the regression is only 10 %, indicating that factors other than avoidance behavior also contribute to the overestimation of body size. One example might be body checking behavior, for which Shafran et al. [25] suggest that eating disorder psychopathology can be maintained via cognitive biases. Although at first glance, body checking behavior like mirror gazing seems to be a phenomenologically opposite behavior to avoidance behavior, both behaviors are behavioral manifestation of a disturbed body image [2] and women with eating disorders tend to alternate between these behaviors [25]. In our view, body checking behavior and avoidance behavior might both contribute to a distorted perception, maybe partly because they both lead to a preoccupation with the disliked aspects of one’s own body that are actively checked or avoided. Moreover, in the case of avoidance, corrective feedback is generally limited and in the case of checking, it might be limited to aspects of the own body that are disliked, leading to a distorted perception of the whole body. Yet, like in anxiety disorders and in obsessive–compulsive disorders, the complex relation between checking behavior and avoidance that might vary between individuals and points of time still needs further investigation. This issue and other factors than avoidance behavior that might contribute to the overestimation of body size need to be addressed in future research. Moreover, intervention studies need to investigate if a reduction in body-related avoidance behavior causes a reduction in body-size overestimation. From a cognitive perspective, our finding is in line with studies that have shown that body exposure therapy [30, 31] is effective in reducing the overestimation of body size (e.g., [12, 18, 21]) as cognitive theories suggest that exposure is useful when avoidance has previously maintained the disorder by preventing disconfirmations of maladaptive appraisals [4, 5]. In eating disorders, one of the maladaptive appraisals that can be changed by body exposure seems to be the overestimation of one’s body size.