A model-driven approach to studying dissociations between body size mental representations in anorexia nervosa

Body Image 20 (2017) 40–48

Contents lists available at ScienceDirect

Body Image journal homepage: www.elsevier.com/locate/bodyimage

A model-driven approach to studying dissociations between body size mental representations in anorexia nervosa Anne-Laure Moscone a,b,∗ , Michel-Ange Amorim a,b , Christine Le Scanff a,b , Pascale Leconte a,b,c a

CIAMS, Univ. Paris-Sud, Université Paris-Saclay, 91405 Orsay Cedex, France CIAMS, Université d’Orléans, 45067 Orléans, France c COMETE—UMR 1075, Université Caen Normandie, 14032 Caen Cedex, France b

a r t i c l e

i n f o

a b s t r a c t

Article history: Received 15 December 2015 Accepted 11 November 2016 Keywords: Anorexia nervosa Body size mental representations BMI norms Psychophysics Single channel model

This study compared dissociations between mental representations of current, ideal and normal body sizes (i.e., Current BS, Ideal BS and Normal BS) for women with anorexia nervosa (AN group, n = 56) and healthy women (control group, n = 56). Along the lines of the single channel model of Cornelissen et al. (2013), the discrepancy between Current BS and BMI for both groups was adequately described along a common linear continuum of Current BS (mis)perception. Body size mental representations were ranked similarly (Ideal BS < Current BS < Normal BS) in each group. Whilst the over-estimation of Current BS was much greater among the AN group than the control group, body dissatisfaction was better explained by Current BS for the AN group and by BMI for the control group. Dissociation between Current BS and participants’ BMI appears to be a key element when seeking to understand AN. © 2016 Elsevier Ltd. All rights reserved.

Introduction Body image disturbance, defined by the DSM-V as a distorted evaluation of personal body size (American Psychiatric Association, 2013), appears to be the main factor in anorexia nervosa (AN) (see Stice, 2002 for a review; Tury, Gülec¸, & Kohls, 2010). This disturbance comprises two components: a perceptual component (i.e., an inability to accurately estimate body size) and an attitudinal component (i.e., a subjective body dissatisfaction combined with negative attitudes to weight and shape; Cornelissen, Johns, & Tovée, 2013). Previous studies have investigated the perceptual component, highlighting the fact that women with AN tend to over-estimate their current body size (Current BS) compared with healthy women (Striegel-Moore et al., 2004; Tovée, Benson, Emery, Mason, & Cohen-Tovée, 2003). In turn, this over-estimation impacts on the attitudinal component, with greater body dissatisfaction found among women with AN than in healthy women (Hrabosky et al., 2009). Such dissatisfaction contributes to a concomitant drive for thinness, i.e., the desire for an ultra-thin ideal body size (Ideal BS) (American Psychiatric Association, 2013; Stice, 2002). Previous

∗ Corresponding author at: CIAMS, Univ. Paris-Sud, Université Paris-Saclay, 91405 Orsay Cedex, France. E-mail address: [email protected] (A.-L. Moscone). http://dx.doi.org/10.1016/j.bodyim.2016.11.003 1740-1445/© 2016 Elsevier Ltd. All rights reserved.

studies have mainly chosen to assess the attitudinal component of body dissatisfaction using, for example, the Body Shape Questionnaire (Cooper, Taylor, Cooper, & Fairburn, 1986) or the BD-Eating Disorder Inventory-2 (Garner, 1991). Conversely, we decided to use a Figure Rating Scale (FRS) in order to maintain a common reference for examining body size mental representations. Most existing studies of body image disturbance have assessed body image mental representations in two different ways, using either a FRS or video distortion techniques. A FRS offers an easy-to-administer self-reported measure of body image and is often used in clinical studies. In addition, there are attempts to link FRS responses to an objective measurement scale of body size such as the Body Mass Index (BMI) (Bulik et al., 2001; Francisco, Narciso, & Alarcao, 2012; see Swami, Salem, Furnham, & Tovée, 2008a for more details). With more sophisticated techniques, such as the video distortion technique (Collins, 1986), participants can modulate a photograph of their own silhouette displayed on a screen. Different versions of a participant’s body can be created by manipulating the photograph in order to measure how participants ‘see’ themselves (Mussap & Salton, 2006). The distortion is thus quantified in terms of a percentage variation. However, with no reference to the participant’s BMI, it is difficult to draw any definite conclusions about the over-estimation of Current BS. Likewise, in recent years, computer advances have helped create photorealistic 3D silhouettes of women, thus allowing participants to interactively vary

A.-L. Moscone et al. / Body Image 20 (2017) 40–48

shape characteristics (Crossley, Cornelissen, & Tovée, 2012). Using the methodology put forward by Tovée and Cornelissen (1999), it is possible for BMI to be inferred from the surface of a 2D body silhouette and from its volume when in 3D. In the initial 1999 study, the authors measured the path length around the perimeter of 2D body images and divided each by the area within its perimeter in order to measure the perimeter area ratio. This ratio appears to be highly correlated with the actual BMI of women to whom the images belong. Using this method, Cornelissen et al. (2013) studied the way in which women with and without AN estimate their BMI in terms of Current BS. They compared two models (i.e., the single channel model and the dual channel model) to examine whether participants’ BMI estimation varies as a linear function of their actual BMI (see Fig. 1). The single channel model predicts that the relationship between estimated BMI (i.e., Current BS) and actual BMI for women with and without AN is adequately described along a common linear continuum, with one intercept and one slope for all women. This model explains the difference in BMI estimation between women

41

with and without AN, using samples from the same unimodal distribution of a single perceptual process. Women with AN have a lower BMI than healthy women; thus, both groups stand along opposite ends of the same distribution (see Fig. 1). In contrast, the dual channel model predicts that the relationship between Current BS and participant’s BMI is better explained by distinct linear trends, with a different intercept for women with AN than for healthy women. Estimates of Current BS for the latter group would be accurately proportional to their BMI, whereas women with AN would show a systematic over-estimation bias of their Current BS (i.e., same slope but a higher intercept). Cornelissen et al. (2013) showed that Current BS overestimation of women with and without AN is not qualitatively different; indeed, it is consistent with a single channel model. Moreover, Current BS mental representation is better explained in relation to the participants’ BMI rather than their subjective attitude toward weight and shape (i.e., attitudinal component).

Fig. 1. Illustration of the single and dual channel models using simulated data; taken from Cornelissen et al. (2013). The top row of figures shows, for each model, the scatterplots of estimated Body Mass index (BMI) as a function of actual BMI for the AN group (filled dots) and the control group (empty dots). In each case, the dotted line represents veridical performance. The bottom row of figures shows the distributions of the difference between estimated and actual BMI (under/over estimation) for each model. In the first case, the distribution is unimodal, which is consistent with a single psychophysical process distributed continuously throughout the population. In the second case, there are two distributions, which is consistent with normal (white bars) vs. pathological (grey bars) psychophysical performance, under a dual-channel hypothesis.

42

A.-L. Moscone et al. / Body Image 20 (2017) 40–48

Our study builds on work by Cornelissen et al. (2013) in order to investigate the extent to which women with AN and healthy women rely on similar reasoning when they make use of body size mental representations. In order to further investigate the attitudinal component of body size mental representation, we examined ideal body size (Ideal BS), as well as normal body size (Normal BS), which corresponds to the mental representation of an acceptable body size (i.e., a silhouette that is neither too fat nor too thin). We theorized that the magnitude of body image disturbance, such as the drive for thinness, could be better understood by knowing the relationship between body size mental representations (i.e., Current BS, Ideal BS, and Normal BS). The literature is still poor about this kind of distinction. However, some authors have explored different BS, such as Woodman and Hemmings (2008), who investigated the effect of ought, ideal and feared body image representations among participants without an eating disorder. Brown and Slaughter (2011) examined the ‘thin ideal’ by evaluating discrepancies in the objective ratings of female ‘attractive’ body and ‘normal’ body sizes among 160 middle class male and female Australians. In this context, ‘attractive’ body reflects the desired body shape of participants and ‘normal’ body refers to the body size that participants believe is most common in their environment. The authors found that, for all age groups, ‘attractive’ body size was seen to be significantly thinner than ‘normal’ body size. They argued that social pressure leads males and females to consider thinner-than-normal adult female bodies as more attractive. To our knowledge, no study has yet investigated the mental representation of Normal BS among women with AN. In this paper, we provide quantified measures of Normal BS, Current BS and Ideal BS among women with and without AN in order to offer a qualitative (ordinal ranking) and quantitative comparison of body size mental representations of both groups. Instead of using a morphing procedure on participants’ photographs, as was the case in the study by Cornelissen et al. (2013), we tested the single channel model using a commonly used FRS, namely the ‘Contour Drawing Rating Scale’ (CDRS; Thompson & Gray, 1995). First, each CDRS silhouette number was converted into BMI units. Then, we identified possible dissociations between mental representations (using BMI scores) of current, ideal and normal body sizes among women with AN (the AN group) and compared them with those of healthy women (the control group). We theorized that these representations are thinner for women with AN than for healthy women. Finally, we investigated whether body dissatisfaction was also consistent with the single channel model. We theorized that the relationship between Current BS (i.e., estimated BMI) and participants’ BMI for women with and without AN can be adequately described along a common linear continuum. Thus, we sought to determine whether body dissatisfaction scores fit well with the single channel model; in other words, whether or not they lie along the same common linear continuum for both groups. Furthermore, we tested whether or not body dissatisfaction could be better explained by Current BS for one group and by participants’ BMI for the other group; such a discrepancy could reflect a possible dissociation between body size mental representations for the AN and control groups.

giving informed consent. In addition, 56 women without an eating disorder (Mage = 22.93 years, SD = 4.34, MBMI = 21.48, SD = 2.58) were age matched to form a control group. The women with AN had significantly smaller BMI measures than those in the control group (t(110) = 12.32, p < 0.001, Cohen’s d = 2.35, 95% CI = [4.40; 6.09]); however, they did not differ in terms of age (t(110) = 0.64, p = 0.53, d = 0.12). Materials We used a computerized version of the CDRS (Thompson & Gray, 1995), which is composed of nine drawings of whole-body size female silhouettes ranging from very thin to very fat. Throughout the experiment, participants were placed at a comfortable distance and height with respect to a computer screen, in a quiet room, away from noise and through traffic. Only the person running the experiment stood near to the participant. Measures BMI. Body mass index (BMI) is a simple weight-to-height index that is commonly used to classify adults who are underweight (BMI < 18.5 kg/m2 ), normal weight (18.5 kg/m2 < BMI < 25 kg/m2 ), (25 kg/m2 < BMI < 30 kg/m2 ) and obese overweight (BMI > 30 kg/m2 ) (WHO, 1995). Height and weight were selfreported by participants in the control group. For those in the AN group, these two variables were measured and reported by medical staff. According to the literature, self-reported and experimentally measured height and weight values are highly correlated (with 0.90 < r < 0.98) (Bulik et al., 2001; Kawada & Suzuki, 2005; see Dratva et al., 2016 for BMI). Each woman’s BMI was then calculated using Eq. (1): BMI (kg/m2 ) = mass (kg)/height (m)2

(1)

Current and ideal body sizes. Participants were shown an array of nine silhouettes from the CDRS. They were asked to select: (1) the silhouette that best represents what they look like (i.e., Current BS) and (2) the silhouette that best represents the body they would ideally like to have (i.e., Ideal BS). BMI scores of the CDRS silhouettes. Based on data collected by Francisco et al. (2012), we computed a second order polynomial equation function (R2 = 0.996), because this best fitted the BMI scores associated with each silhouette of the CDRS. In their study, which involved 848 female participants, for each CDRS silhouette they computed the mean BMI of participants, choosing the silhouette that best corresponded to their current appearance. The line of best fit equation was: y = 0.2262x2 − 0.5786x + 17.802

(2)

Eq. (2) was then used to compute the BMI scores (y) that corresponded with participants’ responses using the CDRS (x). For example, the BMI score of silhouette n◦ 4 would thus be 19.11 = (0.2262 × (42 )) − (0.5786 × 4) + 17.802.

Method Participants

Over-estimation. The over-estimation of Current BS was calculated as the difference between participants’ BMI and Current BS (i.e., O = Current BS − BMI), measured in BMI units.

Fifty-six women (Mage = 23.63 years, SD = 6.93; MBMI = 16.23, SD = 1.88) who met the DSM-IV criteria for anorexia nervosa (AN) were recruited from among the patients of five French hospitals. The advice and consent of the senior medical officers was sought in all cases. These inpatients took part in the study after

Body dissatisfaction. In accordance with the perceptual component of body image, body dissatisfaction is commonly assessed as the discrepancy between Current BD and Ideal BD with the result given in absolute values (Gardner, 1996). However, with this procedure we are given no indication as to whether there is a desire to

A.-L. Moscone et al. / Body Image 20 (2017) 40–48

43

Table 1 Mean and standard deviation (SD) of Current BS (CBS), Ideal BS (IBS) and Normal BS (NBS) body sizes, over-estimation (O), and body dissatisfaction (BD1, BD2) measures for the AN and control groups in BMI units (kg/m2 ).

CBS IBS NBS PSE NBS JND O = CBS − BMI BD1 = IBS − CBS BD2 = IBS − BMI

AN group (n = 56) Mean (SD)

Control group (n = 56) Mean (SD)

20.16 (2.76) 19.20 (1.90) 21.35 (2.32) 17.65 (0.11) 3.93 (2.75) −0.97 (3.26) 2.97 (2.72)

21.60 (2.07) 20.56 (2.16) 23.23 (1.95) 17.68 (0.09) 0.12 (2.31) −1.04 (2.53) −0.92 (3.46)

Results

Fig. 2. Graphical representation of the psychophysical measure of normal body size from the logistic function fitted to the mean responses of all the participants in each group. The computed point of subjective equality (PSE) corresponded to a silhouette that would be judged as ‘too fat’ or ‘too thin’ with equal probability.

lose or gain weight. For this reason, we first assessed body dissatisfaction (BD1) as the discrepancy between Ideal BS and Current BS (i.e., BD1 = Ideal BS − Current BS). In addition, we computed a second body dissatisfaction (BD2) variable as the discrepancy between Ideal BS and participants’ BMI (i.e., BD2 = Ideal BS − BMI). For BD1 and BD2, a negative score specified a desire to lose weight, whereas a positive score corresponded to a desire to gain weight. Both BD variables were computed in BMI units.

Normal body size. A psychophysical procedure was used to quantify Normal BS as a silhouette that is considered realistic and acceptable, i.e., a silhouette that is neither too fat nor too thin. First, each of the nine silhouettes was randomly displayed for 500 ms on a computer screen in order for the participants to familiarize themselves with the stimuli and to avoid subsequent range effects in their responses. Then, each silhouette was displayed until participants answered the following question: ‘if this silhouette was you, would you find yourself ‘too fat’. If so, push the enter key, or if ‘too thin’, push the escape key’. If the reaction time exceeded 500 ms, a warning signal was displayed. Each of the nine silhouettes was presented randomly 10 times, for a total of 90 trials. The method of constant stimuli we used was inspired by Mussap, McCabe, and Ricciardelli (2008). A cumulative logistic function (y = 1/1 + EXP(−(x − PSE)/JND)) was fitted to the proportion of ‘too fat’ answers (y) as a function of silhouette stimuli in order to estimate the point of subjective equality (PSE) and the just noticeable difference (JND) (see Fig. 2 for examples). The function was fitted to the scores of each participant, i.e., one curve per participant (except for Fig. 2, which shows where the function fitted to the mean responses of all the participants). PSE corresponds to the silhouette stimulus to which the participants responded ‘too fat’ and ‘too thin’ in equal proportion. It represents each participant’s judgment of what should be a Normal BS. JND is a measure of each participant’s sensitivity in detecting a change between two successive silhouettes. The JND value reflects the interquartile–range slope of the psychometric function. More precisely, it is the amount of change in x needed for an increase of about 0.25 in y. The flatter the logistic curve, the greater is the JND value (see Fig. 2). We theorized that PSE represents an attitudinal aspect of body size estimation, whereas JND reflects the perceptual ability to detect change in body size.

The following measures of effect size were reported: Cohen’s d followed by a 95% Confidence Interval of mean significances (between brackets) for significant paired comparisons tested with a Student’s t-test; 2 values were reported for the ANOVA results with significant effects. The groups (AN group vs. control group) were treated as a between-subjects factor, and body size mental representations (Current BS, Ideal BS, Normal BS) were treated as a within-subject factor. Post-hoc comparisons were performed using Tukey’s HSD (honest significant difference) test. Body Size Mental Representations (Current BS, Ideal BS, and Normal BS) Results showed a significant group effect (F(1, 110) = 35.36, p < .0001, ␩2 = .24, [1.04; 2.08]), revealing that, on average, the AN group chose thinner body size silhouettes than the control group. The results also revealed a significant effect of body size mental representations (F(2, 220) = 36.88, p < .0001, ␩2 = .25), (see Table 1 for means and SD). Post-hoc tests revealed that mean BMI values for the different BS variables varied significantly between each other (p < .002); and were ordered as follows: Ideal BS (M = 19.88) < Current BS (M = 20.88) < Normal BS (M = 22.29). However, we found no interaction between body size mental representations and groups (F(2, 220) = 0.50, p = .61), suggesting that both groups followed a similar rationale to conclude that Ideal BS < Current BS < Normal BS (Fig. 3), and that body size mental representations were all thinner (by a constant amount) among the AN group compared with the control group. Finally, analysis of the JND for normal BS showed an equivalent sensitivity in body size estimation for women with AN and for the controls, t(110) = 1.76, p = .08, d = .33 (see Table 1). Current BS Over-estimation The results showed that the Current BS of the AN group was significantly more important than their BMI (t(55) = 10.71, p < .0001, d = 1.43, [3.20; 4.67]), whereas these values did not differ among the control group (t(55) = 0.39, p > .69, d = .05). Moreover, it appears that the over-estimation made by the women in the AN group was significantly more important than that of the control group (t(110) = 7.94, p < .0001, d = 1.51, [2.86; 4.76], see the left-hand panel in Fig. 4). Indeed, the Current BS of the women in the AN group was over-estimated by about 3.93 BMI units relative to their BMI (16.23 kg/m2 ), whereas for the control group, this over-estimation was negligible (0.12 kg/m2 ). Body Bissatisfaction In contrast with the literature, which has indicated that the body dissatisfaction of women with AN is greater than that of healthy

44

A.-L. Moscone et al. / Body Image 20 (2017) 40–48

Fig. 3. Top graph: BMI scores associated with each CDRS silhouette (Thompson and Gray, 1995) distributed according to the BMI categories of the World Health Organization (WHO, 1995). Bottom graph: distribution of the body size mental representations and BMI of women from both the AN group and the control group.

Fig. 4. Current BS overestimation and scatterplots for both groups. The left-hand graph shows overlapping distributions of the mean overestimation (i.e., Current BS minus actual BMI). The right-hand graph shows scatterplots of participants’ Current BS estimates, plotted as a function of their actual BMI.

women (Cash & Deagle, 1997; Stice, 2002), our results showed that the mean BD1 (i.e., BD1 = Ideal BS − Current BS) did not differ between groups (t(110) = 0.132, p = .89, d = 0.03). However, when computing BD2 (i.e., BD2 = Ideal BS − BMI), our results showed that women in the AN group had a significantly greater BD2 than those in the control group (t(110) = 6.60, p < .0001, d = 1.26, [2.72; 5.05]).

Testing Cornelissen et al.’s (2013) single-channel model We sought to determine whether or not Current BS overestimation was qualitatively different between groups or resulted from a perceptual bias along a continuum in the magnitude representation of body size. First, we examined the distributions of the difference between BMI and Current BS in BMI units. The

A.-L. Moscone et al. / Body Image 20 (2017) 40–48

45

Fig. 5. Body dissatisfaction regressions for both groups. These figures show scatterplots of participants’ difference between Ideal BS and Current BS (left graph refers to BD1 measure) plotted as a function of their Current BS, and between Ideal BS and their actual BMI (right graph refers to BD2 measure) as a function of their actual BMI.

overlapping of both distributions, as illustrated in the left-hand panel in Fig. 4, does not provide a clear-cut account of whether or not a single or a dual-channel model better accounts for the results. To further examine our results in the light of the approach taken by Cornelissen et al. (2013), we performed regression analyses to compare the regression lines of participants’ Current BS as a function of their BMI (see the right-hand panel in Fig. 4). Our results showed that neither the regression slopes (t(108) = 0.46, p = .64) nor the regression intercepts (t(108) = 0.20, p = .84) differed significantly between the group of women with AN (respectively: ˇ = 0.35, SE = 0.13 and M = 11.88, SE = 3.06) and the control group (respectively: ˇ = 0.52, SE = 0.12 and M = 12.61, SE = 2.01). In order to check our regression model for influential cases we ran the regression analyses without two ‘outlier’ data points (see Fig. 4, right-hand graph: one AN participant with BMI = 13.33 and current BS = 30.92; and one control participant with BMI = 29.87 and current BS = 19.11). Again, the analyses showed that both groups did not differ, either in terms of slope (t(106) = 0.67, p = .50) or intercept (t(106) = 0.16, p = .87). To go one step further in the testing of the single channel model put forward by Cornelissen et al. (2013), we also examined body dissatisfaction with respect to the participants’ Current BS or their BMI. We regressed both measures of body dissatisfaction on the Current BS estimates and on participants’ actual BMI (Fig. 5). Our results for BD1 (BD1 = Ideal BS − Current BS), which were regressed on Current BS, showed that the regression slopes for the group of women with AN (ˇ = −0.81, SE = 0.08) and the control group (ˇ = −0.58, SE = 0.11) did not differ significantly, t(108) = 1.53, p = .13. Moreover, the regression intercepts for the group of women with AN (M = 18.35, SE = 1.90) and the control group (M = 14.24, SE = 2.96) also did not differ significantly, t(108) = 1.19, p = .24. A lack of difference between the groups in terms of both the intercepts and the slopes is consistent with a very similar rationale among both groups with respect to Current BS. In addition, we compared the regression of BD2 (BD2 = Ideal BS − BMI) on BMI for both groups. Our results showed that the regression slopes for the AN group (ˇ = −0.72, SE = 0.09) and the control group (ˇ = −0.78, SE = 0.08) did not differ significantly, t(108) = 0.07, p = .94. In the same way, no significant difference was found for regression intercepts between the women with AN (M = 19.81, SE = 2.25) and the control group (M = 21.66, SE = 2.46), t(108) = 0.55, p = .58. These results are consistent with a single channel model. However, a visual inspection of the R2 values of the linear regressions shows that body dissatisfaction was better explained by Current BS for the AN group (66% of explained variance, vs. 33% for

the control group; see Fig. 5 left), and by actual BMI for the control group (61% vs. 51% for the AN group; see Fig. 5 right). Finally, in order to check our regression models of body dissatisfaction for influential cases we ran the regression analyses without the ‘outlier’ data points (see Fig. 5). There were two outlying values for the regression of BD1 on Current BS (one AN participant with BD1 = −12.81 and current BS = 30.92; and one control participant with BD1 = 12.81 and current BS = 18.10), and also for the regression of BD2 on BMI (one control participant with BMI = 29.87 and BD2 = −10.76; and another control participant with BMI = 17.96 and BD2 = 12.96). Again, these additional analyses confirmed the results for the regression of BD2 on BMI: both groups still did not differ, either in terms of slope (t(106) = 0.85, p = .40) or intercept (t(106) = 0.52, p = .60). However, regressions that were run without the outlying values changed the conclusions when regressing BD1 on Current BS. The results showed that the regression slope for the AN group (ˇ = −0.75, SE = 0.09) was significantly greater than that for the control group (ˇ = −0.62, SE = 0.11), t(106) = 2.57, p = .01. In the same way, the regression intercept for women with AN (M = 17.26, SE = 2.22) was significantly greater than that for the control group (M = 10.09, SE = 1.97), t(106) = 2.32, p = .02. These results are suggestive of an increase in psychological concern about body size among women with AN (Cornelissen, Bester, Cairns, Tovée, & Cornelissen, 2015). General Discussion The aim of this research was two-fold: (a) to identify possible dissociations between mental representations of current, ideal and normal body sizes among women with and without AN, and (b) to test the single channel model of Cornelissen et al. (2013) with regard to the components of disturbance in body size mental representation, within the framework of BMI scores, and in terms of perceptual and attitudinal components. The perceptual component is related to Current BS as a function of BMI, whereas the attitudinal component is related to body dissatisfaction (computed with respect to subjective value: BD1 = Ideal BS − Current BS, and with respect to an objective reference: BD2 = Ideal BS − BMI). In order to be able to discuss our results in reference to the WHO classification of BMI, we established BMI scores for the CDRS silhouette stimuli. For this purpose, we used data gathered by Francisco et al. (2012) to report BMI scores associated with the CDRS silhouettes used by 848 female participants to indicate their Current BS. Although the participants’ height and weight measurements in their study were self-reported (as was the case for our controls), it

46

A.-L. Moscone et al. / Body Image 20 (2017) 40–48

has been shown that self-reported and experimentally measured height and weight values are highly correlated (with .90 < r < .98) (Bulik et al., 2001; Kawada & Suzuki, 2005). Furthermore, mean variations between self-reported and measured height and weight in women are small (Engstrom, Paterson, Doherty, Trabulsi, & Speer, 2003; see also Niedhammer, Bugel, Bonenfant, Goldberg, & Leclerc, 2000, who reported a difference of only 0.44 kg/m2 among 1905 French women). However, future research should also involve third parties to choose the silhouette stimuli that best correspond to participants’ body size in order to obtain more objective measures. Even so, this approach would not necessarily be immune to perceptual bias. The relationship between silhouette BS and objective anthropometric measurements is studied with more and more sophisticated methods, ranging from silhouette drawings (e.g., Bulik et al., 2001; Francisco et al., 2012), and photographs (e.g., Swami, Salem, Furnham, & Tovée, 2008b), to real-time morphing of 3D virtual characters (e.g., Crossley et al., 2012). Indeed, objective anthropometric measurements can provide additional information when examining participants’ body size mental representations, even when using 2D silhouettes. For example, Tovée and Cornelissen (1999), investigated men’s perceptions of female attractiveness using 2D silhouettes belonging to five BMI categories (Bray, 1978): emaciated (below 15 kg/m2 ), under-weight (15–19 kg/m2 ), normal (20–24 kg/m2 ), overweight (25–30 kg/m2 ) and obese (above 30 kg/m2 ). These authors concluded that female silhouettes perceived to be attractive by men corresponded to a BMI of 18–19 kg/m2 . Furthermore, they argued that this category of BMI is optimal for health and fertility according to medical data (Katz & Vollenhoven, 2000). Without BMI information, it would have been impossible to show that men’s perceptions of female attractiveness correspond to the optimal category for fertility. Our comparison of body size mental representations among participants with and without AN showed that, although all the body size mental representations (Current BS, Ideal BS and Normal BS) for both groups were within the normal weight category, the AN group chose thinner body size mental representations than the control group. Furthermore, both groups perceived their Current BS to be thinner than what they considered to be a normal body size. Indeed, even if they wanted to lose weight to achieve their Ideal BS (in view of their Current BS), the latter fell within the normal weight category for both groups. These body size mental representations among women with AN suggest a greater drive for thinness (American Psychiatric Association, 2013; Stice, 2002; Striegel-Moore et al., 2004) compared with the control group. On the other hand, our results showed that the Ideal BS of the AN group (19.20 kg/m2 ) belonged to the normal weight category (from 18.5 kg/m2 to 24.99 kg/m2 ). In relation to the literature, this new finding provides quantitative data that agrees with previous studies; namely, that even though women with AN have a thinner Ideal BS compared with healthy women, they do not necessarily aspire to excessive thinness (Probst, 2006; Vitousek & Hollon, 1990). Surprisingly, in our sample, the Ideal BS of women with AN was not extremely thin, as was assumed in the literature (Thiel, Broocks, Ohlmeier, Jacoby, & Schübler, 1995). Furthermore, the results suggest that if women with AN were made aware of their BMI they would look to gain weight to achieve their Ideal BS. Moreover, although the Normal BS of the AN group (21.35 kg/m2 ) was thinner than those of the control group (23.23 kg/m2 ), they were still within the normal weight category put forward by the WHO. This suggests that they have a realistic and healthy representation of what might be a normal and acceptable body size, the value of which is actually very close to both the Current BS and the BMI of our age-matched control group (see Fig. 3). However, for women with AN, the stronger desire to be thin, combined with a greater fear of gaining weight, makes it difficult for them to move toward

a Normal BS (Cooper & Turner, 2000; Vartanian, Herman, & Polivy, 2005). We cannot know objectively how much women with AN feel able to regain weight. The hospitalization bias means that they may provide an answer that is expected by the institution without really adopting the same view. Thus, we chose a two-alternative forced choice procedure, which was previously validated in a similar context (Mussap et al., 2008), and which does not require an explicit indication of what a Normal BS would look like on the CDRS. Our goal was to know which body size would likely correspond to a body that is ‘normally fat’; in other words, one that they would consider neither too thin (in terms of pathological underweight), nor too fat (inconceivable for it to be achieved), but rather a reasonable BS for them to move toward. By answering dichotomously, ‘if I had this body size, I would consider myself thin’ or ‘if I had this body size, I would consider myself fat’, participants provided responses that allowed a BS value to be computed without them having to explicitly choose the silhouette they would consider “normal” (i.e., neither too thin, nor too fat). Clinically, this BS mental representation may serve as a reference to help patients realize that, even though weight gain is required to be in good health, there is an achievable body size for which they would consider themselves as being neither too fat nor too thin. Nevertheless, this conclusion must be tempered by the fact that, when our experiment took place, our sample of female participants with AN was hospitalized. Therefore, they may have been influenced in their responses by the therapeutic treatment given them to encourage weight gain in order to recover their well-being. Building on the study by Cornelissen et al. (2013), we used BMI scores to investigate the extent to which women with AN and healthy women rely on similar reasoning in order to make mental representations of body size. Regression analyses suggested that a single channel model can better account for responses in both groups. Indeed, our results are in line with the single channel model, which shows that over-estimation of body size in women with AN is not qualitatively different from that found among healthy women and that, as weight increases, this value decreases for both groups. Although current BS over-estimation (i.e., the difference between BMI and Current BS) was much greater for the AN group than for the control group, they are located along the same continuum (see Fig. 4). This reveals that women with AN perceive their body size to be dramatically larger than it is in reality. These results are in agreement with previous findings (Striegel-Moore et al., 2004; Tovée et al., 2003) and provide additional evidence that women with AN are excessively concerned about physical anomalies and may suffer from body dysmorphic disorders (Grant & Phillips, 2004; Hartmann, Greenberg, & Wilhelm, 2013). According to these authors, both anorexia nervosa and body dysmorphia are highly co-morbid, sharing common features. When regressing current BS data on actual BMI (while excluding outliers), we have also provided evidence for increased psychological concern about body size among women with AN compared with healthy women. Some authors have suggested that body size representation may reflect an attitudinal rather than a perceptual component (Ben-Tovim, Walker, Murray, & Chin, 1990). However, perceptual problems in AN are not unitary. Troubles may result from just one specific perceptual process, such as aberrant sensory and proprioceptive impressions (Keizer et al., 2011; Keizer, Smeets, Dijkerman, van Elburg, & Postma, 2012), while sparing visual judgements of body silhouettes. Along these lines, both groups in our sample showed equivalent sensitivity (JND measures) in body size estimation. This suggests that the Current BS response of women with AN is due to a deficit in the judgment of their own body size rather than in their perception of the silhouette stimuli (Cornelissen et al., 2015; Gardner & Bokenkamp, 1996; Mussap et al., 2008). This judgment error reflects a “regression towards the mean” phenomenon

A.-L. Moscone et al. / Body Image 20 (2017) 40–48

in the population, known as contraction bias (Poulton, 1989). As a result, the slope that relates estimated BMI (Current BS) to actual BMI is less than one, as predicted by the single channel model (see Fig. 1), and illustrated in our data (see Fig. 4). The mean is a standard reference, based on the average size of bodies seen in real life; in other words, women with a low BMI would have a tendency to overestimate their own body, while women with a high BMI would tend to under-estimate their own body size compared with this standard reference (Winkler & Rhodes, 2005). Thus, the combination of an over-estimated Current BS representation with the “disconnection” between their BMI and their Current BS would lead women with AN to constantly believe that they need to lose weight. The discrepancy between their erroneous Current BS representation and their Ideal BS is an aggravating factor in their mental distress, increasing body dissatisfaction and feeding the vicious circle of anorexia nervosa. This finding underlines the necessity to deal with the impact of a drive for thinness in anorexia nervosa. From an attitudinal standpoint, our results revealed that body dissatisfaction could be better explained by Current BS for the AN group and by BMI for the control group. According to Slade and Russell (1973), the extent to which women with AN overestimate their body size tends to reduce as their weight increases. Re-establishing a more truthful Current BS could help women with AN to moderate their desire to lose weight and so reduce their sense of body dissatisfaction. Cognitive-behavioral theories (Cash, 2002) have assumed that the degree to which individuals invest in their physical appearance depends greatly on any related self-schemas (Cash, Melnyk, & Hrabosky, 2004). We found that for women with AN, the mode of reasoning with regard to mental representations of body size is qualitatively similar to that of healthy women: Ideal BS is smaller than Current BS, and Current BS is smaller than Normal BS. Dissociation between their Current BS and their BMI in the former population may originate from low-level deficits (rather than representational deficits), such as aberrant sensory and proprioceptive impressions that need to be treated (Keizer et al., 2011, 2012). Recent literature (Moscone, Leconte, & Le Scanff, 2014; Rizk, Kern, Godart, & Melchior, 2014) has indicated that treatment which includes adapted physical activities could act as a mediating tool to achieve body mindfulness. Low-intensity activities such as yoga, Tai Chi and stretching may help women with AN to improve body acceptance (Clance, Mitchell, & Engelman, 1980), to feel comfortable with their own body and to promote healthy eating behaviors (Daubenmier, 2005). The interesting point about adapted physical activities is that they encourage patients with AN to accept their body regardless of any thin-is-ideal messages and images conveyed by society and the media. Conclusion Body size dissociation in the mental representations of women with and without AN were interpreted within the framework of the WHO classification of the BMI. In line with the single channel model put forward by Cornelissen et al. (2013), we showed that women with AN and healthy women adopt similar reasoning with regard to the ordering of body size mental representations. Indeed, in our study, both groups indicated that, on average, they wished to be thinner than their current body size representation. At the same time, the latter was thinner than what they considered to be a normal body size. However, some dissociation remains. In our study, Current BS over-estimation with respect to BMI was much greater among the AN group than the control group. Moreover, body dissatisfaction was better explained by Current BS for the AN group and by BMI for the control group. Thus, although therapeutic intervention may help women with AN to accept their chosen Normal BS and try to reach it, so long as they continue to over-estimate their Current BS and use it as a reference instead of their BMI, the symp-

47

tomatology of AN will remain. The association of physical activities that focus on the body in motion with existing therapies may help women with AN to re-appropriate their body and teach them to reconstruct a CBS that is closer to their actual body size.

Acknowledgments We wish to thank Cornelissen et al. (2013) for having allowed us to reproduce their illustration of the single- and dual-channel models. We thank all the patients and staff whose help and participation made this work possible.

References American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Washington, DC: Author. Ben-Tovim, D. A., Walker, M. K., Murray, H., & Chin, G. (1990). Body size estimates: Body image or body attitude measures? International Journal of Eating Disorders, 9, 57–67. http://dx.doi.org/10.1002/1098-108X(199001)9:1<57::AIDEAT2260090107>3.0.CO;2-S Bray, G. A. (1978). Definition, measurement, and classification of the syndromes of obesity. International Journal of Obesity, 2, 99–112 (PMID: 711370). Brown, F. L., & Slaughter, V. (2011). Normal body, beautiful body: Discrepant perceptions reveal a pervasive thin-ideal from childhood to adulthood. Body Image, 8, 119–125. http://dx.doi.org/10.1016/j.bodyim.2011.02.002 Bulik, C. M., Wade, T. D., Heath, A. C., Martin, N. G., Stunkard, A. J., & Eaves, L. J. (2001). Relating body mass index to figural stimuli: Population-based normative data for Caucasians. International Journal of Obesity, 25, 1517–1524. http://dx.doi.org/ 10.1038/sj.ijo.0801742 Cash, T. F. (2002). The Situational Inventory of Body-Image Dysphoria: Psychometric evidence and development of a short form. International Journal of Eating Disorders, 32, 362–366. http://dx.doi.org/10.1002/eat.10100 Cash, T. F., & Deagle, E. A. (1997). The nature and extent of body-image disturbances in anorexia nervosa and bulimia nervosa: A meta-analysis. International Journal of Eating Disorders, 22, 107–125. http://dx.doi.org/10.1002/(SICI)1098108X(199709)22:2<107::AID-EAT1>3.0.CO;2-J Cash, T. F., Melnyk, S. E., & Hrabosky, J. I. (2004). The assessment of body investment: An extensive revision of the Appearance Schemas Inventory. International Journal of Eating Disorders, 35, 305–316. http://dx.doi.org/10.1002/eat.10264 Clance, P. R., Mitchell, M., & Engelman, S. R. (1980). Body cathexis in children as a function of awareness training and yoga. Journal of Clinical Child Psychology, 9, 82–85. http://dx.doi.org/10.1080/15374418009532956 Collins, J. K. (1986). The objective measurement of body image using a video technique: Reliability and validity studies. British Journal of Psychology, 77, 199–205. http://dx.doi.org/10.1111/j.2044-8295.1986.tb01994.x Cooper, P. J., Taylor, M. J., Cooper, Z., & Fairburn, C. G. (1986). The development and validation of the Body Shape Questionnaire. International Journal of Eating Disorders, 6, 485–494. http://dx.doi.org/10.1002/1098-108X(198707)6:4<485::AIDEAT2260060405>3.0.CO;2-O Cooper, M., & Turner, H. (2000). Underlying assumptions and core beliefs in anorexia nervosa and dieting. British Journal of Clinical Psychology, 39, 215–218. http://dx. doi.org/10.1348/014466500163239 Cornelissen, K. K., Bester, A., Cairns, P., Tovée, M. J., & Cornelissen, P. L. (2015). The influence of personal BMI on body size estimations and sensitivity to body size change in anorexia spectrum disorders. Body Image, 13, 75–85. http://dx.doi.org/ 10.1016/j.bodyim.2015.01.001 Cornelissen, P. L., Johns, A., & Tovée, M. J. (2013). Body size over-estimation in women with anorexia nervosa is not qualitatively different from female controls. Body Image, 10, 103–111. http://dx.doi.org/10.1016/j.bodyim.2012.09.003 Crossley, K. L., Cornelissen, P. L., & Tovée, M. J. (2012). What is an attractive body? Using an interactive 3D program to create the ideal body for you and your partner. PLoS One, 7, e50601. http://dx.doi.org/10.1371/journal.pone.0050601 Daubenmier, J. (2005). The relationship of yoga, body awareness, and body responsiveness to self-objectification and disordered eating. Psychology of Women Quarterly, 29, 207–219. http://dx.doi.org/10.1111/j.1471-6402.2005.00183.x Dratva, J., Bertelsen, R., Janson, C., Johannessen, A., Benediktsdóttir, B., Bråbäck, L., . . . & Real, F. G. (2016). Validation of self-reported figural drawing scales against anthropometric measurements in adults. Public Health Nutrition, 19, 1944–1951. http://dx.doi.org/10.1017/S136898001600015X Engstrom, J. L., Paterson, S. A., Doherty, A., Trabulsi, M., & Speer, K. L. (2003). Accuracy of self-reported height and weight in women: An integrative review of the literature. The Journal of Midwifery & Women’s Health, 48, 338–345. http://dx. doi.org/10.1016/S1526-9523(03)00281-2 Francisco, R., Narciso, I., & Alarcão, M. (2012). Body image (dis)satisfaction among Portuguese adolescents and adults: Contribution to the validation process of the Contour Drawing Rating Scale. A Revista Iberoamericana de Diagnóstico e Avaliac¸ão Psicológica, 34, 61–88. http://refhub.elsevier.com/S01401971(15)00032-9/sref21

48

A.-L. Moscone et al. / Body Image 20 (2017) 40–48

Gardner, R. M. (1996). Methodological issues in assessment of the perceptual component of body image disturbance. British Journal of Psychology, 87, 321–337 (PMID: 8673361). Gardner, R. M., & Bokenkamp, E. D. (1996). The role of sensory and nonsensory factors in body size estimations of eating disorder subjects. Journal of Clinical Psychology, 52, 3–15. http://dx.doi.org/10.1002/(SICI)10974679(199601)52:1<3::AID-JCLP1>3.0.CO;2-X Garner, D. M. (1991). EDI-2. Eating Disorder Inventory-2. Professional Manual. Odessa, FL: Psychological Assessment Resources, Inc. Grant, J. E., & Phillips, K. A. (2004). Is anorexia nervosa a subtype of body dysmorphic disorder? Probably not, but read on . . .. Harvard Review of Psychiatry, 12, 123–126. http://dx.doi.org/10.1080/10673220490447236 Hartmann, A. S., Greenberg, J. L., & Wilhelm, S. (2013). The relationship between anorexia nervosa and body dysmorphic disorder. Clinical Psychology Review, 33, 675–685. http://dx.doi.org/10.1016/j.cpr.2013.04.002 Hrabosky, J. I., Cash, T. F., Veale, D., Neziroglu, F., Soll, E. A., Garner, D. M., . . . & Phillips, K. A. (2009). Multidimensional body image comparisons among patients with eating disorders, body dysmorphic disorder, and clinical controls: A multisite study. Body Image, 6, 155–163. http://dx.doi.org/10.1016/j.bodyim.2009.03. 001 Katz, M. G., & Vollenhoven, B. (2000). The reproductive endocrine consequences of anorexia nervosa. British Journal of Obstetrics and Gynecology, 107, 707–713. http://dx.doi.org/10.1111/j.1471-0528.2000.tb13329.x Kawada, T., & Suzuki, S. (2005). Validation study on self-reported height, weight, and blood pressure. Perceptual and Motor Skills, 101, 187–191. http://dx.doi.org/ 10.2466/pms.101.1.187-191 Keizer, A., Smeets, M. A., Dijkerman, H. C., van den Hout, M., Klugkist, I., van Elburg, A., & Postma, A. (2011). Tactile body image disturbance in anorexia nervosa. Psychiatry Research, 190, 115–120. http://dx.doi.org/10.1016/j.psychres.2011.04.031 Keizer, A., Smeets, M. A., Dijkerman, H. C., van Elburg, A., & Postma, A. (2012). Aberrant somatosensory perception in anorexia nervosa. Psychiatry Research, 200, 530–537. http://dx.doi.org/10.1016/j.psychres.2012.05.001 Moscone, A. L., Leconte, P., & Le Scanff, C. (2014). L’anorexie et l’activité physique, une relation ambiguë. Movement & Sport Sciences, 84, 51–59. http://dx.doi.org/ 10.1051/sm/2013084 Mussap, M. P., McCabe, M. P., & Ricciardelli, M. P. (2008). Implications of accuracy, sensitivity, and variability of body size estimations to disordered eating. Body Image, 5, 80–90. http://dx.doi.org/10.1016/j.bodyim.2007.07.003 Mussap, A. J., & Salton, N. (2006). A ‘rubber-hand’ illusion reveals a relationship between perceptual body image and unhealthy body change. Journal of Health Psychology, 11, 627–639. http://dx.doi.org/10.1177/1359105306065022 Niedhammer, I., Bugel, I., Bonenfant, S., Goldberg, M., & Leclerc, A. (2000). Validity of self-reported weight and height in the French GAZEL cohort. International Journal of Obesity, 24, 1111–1118. http://dx.doi.org/10.1038/sj.ijo.0801375 Poulton, E. C. (1989). Bias in quantifying judgements. Hove: Erlbaum. Probst, M. (2006). The body experience in eating disorders: Research & theory. European Bulletin of Adaptated Physical Activity [on-line],. Retrieved from. http:// www.eufapa.upol.cz Rizk, M., Kern, L., Godart, N., & Melchior, J. C. (2014). Anorexie mentale, activité physique et nutrition: quelles potentialisations? Nutrition clinique et métabolisme, 28, 287–293. http://dx.doi.org/10.1016/j.nupar.2014.10.008

Slade, P. D., & Russell, G. F. (1973). Awareness of body dimensions in anorexia nervosa: Cross-sectional and longitudinal studies. Psychological Medicine, 3, 188–199. http://dx.doi.org/10.1017/ S0033291700048510 Stice, E. (2002). Risk and maintenance factors for eating disorder pathology: A metaanalytic review. Psychological Bulletin, 128, 825–848. http://dx.doi.org/10.1037// 0033-2909.128.5.825 Striegel-Moore, R. H., Franko, D. L., Thompson, D., Barton, B., Schreiber, G. B., & Daniels, S. R. (2004). Changes in weight and body image over time in women with eating disorders. International Journal of Eating Disorders, 36, 315–327. http://dx. doi.org/10.1002/eat.20053 Swami, V., Salem, N., Furnham, A., & Tovée, M. J. (2008). Initial examination of the validity and reliability of the female Photographic Figure Rating Scale for body image assessment. Personality and Individual Differences, 44, 1752–1761. http:// dx.doi.org/10.1016/j.paid.2008.02.002 Swami, V., Salem, N., Furnham, A., & Tovée, M. J. (2008). The influence of feminist ascription on judgements of women’s physical attractiveness. Body Image, 5, 224–229. http://dx.doi.org/10.1016/j.bodyim.2007.10.003 Thiel, A., Broocks, A., Ohlmeier, M., Jacoby, G. E., & Schübler, G. (1995). Obsessive–compulsive disorder among patients with anorexia nervosa and bulimia nervosa. American Journal of Psychiatry, 152, 72–75. http://dx.doi.org/ 10.1176/ajp.152.1.72 Thompson, M. A., & Gray, J. J. (1995). Development and validation of a new bodyimage assessment scale. Journal of Personality Assessment, 64, 258–269. http:// dx.doi.org/10.1207/s15327752jpa6402 6 Tovée, M. J., Benson, P. J., Emery, J. L., Mason, S. M., & Cohen-Tovée, E. M. (2003). Measurement of body size and shape perception in eating-disordered and control observers using body-shape software. British Journal of Psychology, 94, 501–516. http://dx.doi.org/10.1348/000712603322503060 Tovée, M. J., & Cornelissen, P. L. (1999). Visual cues to female physical attractiveness. Proceedings of the Royal Society of London, Series B: Biological Sciences, 266, 211–218. http://dx.doi.org/10.1098/rspb.1999.0624 Tury, F., Gülec¸, H., & Kohls, E. (2010). Assessment methods for eating disorders and body image disorders. Journal of Psychosomatic Research, 69, 601–611. http://dx. doi.org/10.1016/j.jpsychores.2009.05.012 Vartanian, L. R., Herman, P. C., & Polivy, J. (2005). Implicit and explicit attitudes toward fatness and thinness: The role of the internalization of societal standards. Body Image, 2, 373–381. http://dx.doi.org/10.1016/j.bodyim.2005.08.002 Vitousek, K. B., & Hollon, S. D. (1990). The investigation of schematic content and processing in eating disorders. Cognitive Therapy and Research, 14, 191–214. http:// dx.doi.org/10.1007/BF01176209 Winkler, C., & Rhodes, G. (2005). Perceptual adaptation affects attractiveness of female bodies. British Journal of Psychology, 96, 141–154. http://dx.doi.org/10. 1348/000712605X36343 Woodman, T., & Hemmings, S. (2008). Body image self-discrepancies and affect: Exploring the feared body self. Self and Identity, 7, 413–429. http://dx.doi.org/ 10.1080/15298860701800225 World Health Organization. (1995). Physical status: The use and interpretation of anthropometry. Report of a WHO Expert Committee. Technical Report Series No. 854. Geneva: WHO.