- Email: [email protected]
Emotional face processing in women with high and low levels of eating disorder related symptoms
Eating Behaviors 9 (2008) 389–397
Contents lists available at ScienceDirect
Eating Behaviors
Emotional face processing in women with high and low levels of eating disorder related symptoms Lynne Jones, Catherine Harmer, Phil Cowen, Myra Cooper ⁎ Isis Education Centre and Department of Psychiatry University of Oxford, UK
a r t i c l e
i n f o
Article history: Received 7 October 2007 Received in revised form 16 February 2008 Accepted 7 March 2008 Keywords: Eating disorders Emotional processing Happy Anger Surprise
a b s t r a c t Objectives: Emotional processing has rarely been investigated in those “at risk” of developing an eating disorder. This study investigated the processing of six basic emotions depicted on faces in an “at risk” group, compared to a control group. Design: Participants were women with high (N = 29) and low (N = 23) levels of eating disorder symptoms who were not taking psychotropic medication. A well characterised computerised task (Facial Expression Emotion Task) was administered to all participants. Results: Women with high levels of eating disorder symptoms, compared to those with low levels, were less accurate at recognising happy and neutral faces, but showed no differences in their accuracy at recognising other emotions. They also showed a trend to be less good at discriminating anger, but better at discriminating surprise from other emotions. Depressive and anxious symptoms did not provide a complete explanation for the findings. Conclusions: The findings support the inclusion of emotional processing in models of eating disorders, and suggest that it may have a role in their development. Emotional processing warrants further investigation particularly in those “at risk” but also in those with eating disorders. © 2008 Elsevier Ltd. All rights reserved.
1. Introduction Clinical descriptions have long suggested that people with eating disorders have deficits in emotional processing, including in the context of interpersonal relationships (e.g. Bruch, 1973, p50). It has been suggested that these may help account for the fact that patients with eating disorders often have significant problems with social functioning (Kucharska-Pietura, Nikolaou, Masiak, & Treasure, 2004). In the context of interpersonal relationships, for example, misinterpretation of another person's emotional state might result in an unhelpful behaviour or response which may then have a negative consequence. Social functioning difficulties may persist even after behavioural symptoms have remitted (Yager, Landsverk, & Edelstein, 1987), and become chronic (Ratnasuriya, Eisler, Szmuckler, & Russell, 1991). Problems with social functioning may also help maintain the eating disorder symptoms (Fairburn & Harrison, 2003). Research suggests that social and interpersonal situations, for example, are frequent triggers for eating disorder related behaviour (Abraham & Beumont, 1982). Research has investigated ability to understand and process patients own internal emotions, and this ability has a role in cognitive models of eating disorders. However, little research attention has been given to the ability of those with eating disorders to understand/interpret the emotional states of other people. Recent developments in cognitive models of eating disorders (e.g. Cooper, Wells, & Todd, 2004; Fairburn, Cooper, & Shafran, 2003) conceptualise binge-eating, to a greater or less extent, as a means of coping with intense (internal) cognitive and affective distress, while lack of awareness of own emotional states has been identified as a significant risk factor in cognitive models of
⁎ Corresponding author. Isis Education Centre, Oxford, OX3 7JX, UK. E-mail address: [email protected] (M. Cooper). 1471-0153/$ – see front matter © 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.eatbeh.2008.03.001
390
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
anorexia nervosa (AN) (e.g. Connan, Campbell, Katzman, Lightman, & Treasure, 2003; Guidano and Liotti, 1983). Cognitive models, however, pay much less attention to patient's ability to comprehend other people's emotional states, and do not generally provide an explanation of whether and how this might play a role in the development or maintenance of eating disorders. Theoretically, in developmental models, understanding of one's own and of other's emotional states is often regarded as interdependent. Attachment theory outlines how deficits in one will affect the development of the other, and vice versa (e.g. Fonagy, Gergely, Jurist, & Target, 2002). Emotional processing difficulties also arise in the context of the early mother child relationship, with the mother and infant forming an affective relationship from the beginning of life (e.g. Bowlby, 1969). Historically, developmental factors have not been emphasised in cognitive models of EDs although there is recent work providing a preliminary integration of the two models (Cooper, 2005). There is also a cognitive–interpersonal model (Schmidt and Treasure, 2006) which includes difficulties in recognising others emotions as part of one process (avoidance) maintaining AN. Importantly, a deficit in understanding others emotions is likely to make cognitive therapy tasks difficult to complete successfully. A number of studies have used the self report Toronto Alexithymia Scale (TAS: Taylor, Ryan, & Bagby, 1985) to assess emotional processing. These have indicated that patients with eating disorders, including those with AN (Beales and Dolton, 2000) and bulimia nervosa (BN) (Bydlowski et al., 2005) have deficits in emotional processing. Two recent studies in non-clinical groups, for example, have also found high levels of eating disorder related symptoms are related to deficits on the TAS (Kiyotaki and Yokoyama, 2006; Quinton and Wagner, 2005), suggesting that emotional processing deficits may also be characteristic of those “at risk” or who have sub-clinical presentations. Although these findings are of interest, the TAS has been criticised for amalgamating several aspects of emotional processing (Cooper, 2003), including both awareness of own and other's emotional states. It has also been suggested that more specific and objective measures are needed (Gilboa-Schechtman, Avnon, Zubery, & Jeczmien, 2006). A small number of studies have used more objective means to assess emotional processing in clinical samples. These typically involve ability to identify emotion in others. To date, most studies have used emotional faces as stimuli, but have produced mixed findings. Zonnevylle-Bender, van Goozen, Cohen-Kettenis, van Elburg, and van Engeland (2002) found that adolescents with eating disorders were worse than controls at identifying emotion, using pictures presented on videotape. Kucharska-Pietura, Nikolaou, Masiak, and Treasure (2004), report similar findings in women with AN, using nine emotions presented on slides. However, Mendlewicz, Linkowski, Bazelmans, and Philippot (2005) used a computer based task, composed of facial stimuli taken from Ekman's series (Matsumoto and Ekman, 1998), but found no difference between those with AN and controls in accuracy or reaction time to five different emotions. Similarly, Kessler, Schwarze, Filipic, Traue, and von Wietersheim (2006), also using computerised presentation, found no difference between those with AN and controls in ability to recognise the six basic facial emotions, with the exception of a trend for the patients to recognise surprise less rapidly than controls. There appear to be no comparable studies in those with BN, and little information on which to base differential predictions in BN compared to AN. A wide range of tasks have been used in these studies, most of which have unknown ability to discriminate between different groups. In several studies the medication status of the patients tested is unclear, and some are reported to be taking selective serotonin reuptake inhibitors (SSRIs). This is likely to be significant given that several studies have found similar tasks to be highly sensitive to SSRI administration (e.g. Harmer et al., 2003). Despite the high prevalence of depression in those with eating disorders, and increasing recognition of the importance of anxiety disorders (e.g. Hinrichsen, Wright, Waller, & Meyer, 2003), only one study controlled for depressive symptoms (Kucharska-Pietura et al. (2004), and none for anxiety, thus some of the significant effects observed may be due to general distress, rather than any eating disorder specific features. This is particularly relevant given that both anxiety and depression are known to be associated with both emotional processing deficits and characteristic biases in the processing of, for example, negative information. Clinically, there is increased recognition of the prevalence and seriousness of sub-clinical eating disorders. While their extent and nature remains understudied, it has been suggested that this group (which includes those identified in the Diagnostic and Statistical Manual-IV (American Psychiatric Association, 1994) manual as Eating Disorder Not Otherwise Specified: ED-NOS) is by far the largest of those with an eating disorder (Fairburn & Harrison, 2003). Moreover, many will subsequently develop either AN or BN. To date, however, no investigation of emotional processing using facial stimuli appear to be have been conducted in a group “at risk” of developing an eating disorder, or who have sub-clinical symptoms. This is an important area for research given that relatively little is known about this group. Investigation of their emotional processing may also provide some insight into why and how those “at risk” may develop AN or BN, and how similar those “at risk” are to those with the full syndrome disorders. The aim of the current study was to investigate the processing of emotional facial stimuli in a group of women with high scores on a measure of eating disorder related symptoms (i.e. an “at risk” or sub-clinical group) and a group of women with low scores. A well characterised, computerised task of emotional facial stimuli, measuring the full range of basic emotions, was used. Potential participants were excluded if they were taking any centrally acting medication (e.g. SSRIs or other medication known to affect the central nervous system) at the time of study and depression and anxiety were taken into account statistically in conducting the data analysis. Given previously conflicting findings, the study was designed to be exploratory and no specific hypotheses were investigated. 2. Method 2.1. Participants A sample of two hundred and eighty-one women students enrolled in various undergraduate and post-graduate degrees at two universities completed the Eating Attitudes Test-26 (EAT; Garner, Olmsted, Bohr, & Garfinkel, 1982) and provided their contact
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
391
details. The range of scores was 0 to 52 for the entire sample (from a possible range of 0 to 78). Women were then contacted consecutively, from the highest scorers down and the lowest scorers up, until a minimum of 20 participants had potentially been recruited in two groups. These groups were labelled: low EAT (EAT score 3 or less) and high EAT (EAT score 16 or greater). They represented the top and bottom 15–20% of the sample. A power calculation was not possible, given the lack of previous research with this task in this group, and a minimum of 20 was chosen based on that needed for significance in studies using the same measure in different populations (e.g. Harmer et al., 2003). No potential participant with whom contact was made in this way declined to take part. Exclusion criteria included visual or motor impairments which prevented participants accessing visually presented material or using a keyboard. These were not evident in any of the women approached. As noted above, potential participants were also excluded if they were currently taking any centrally acting medication. This exclusion applied not only to recognised psychotropic drugs but also general medical drugs with central actions. Some of the latter agents, for example, propanolol, are known to modify emotional processing (Harmer, Perrett, Cowen, & Goodwin, 2001). 2.2. Demographic information Participants were asked to provide their age, self-described ethnic origin and number of years spent in education. Self-report of height and weight was requested and used to calculate participants' body mass indices (BMI).1 Participants were asked whether they were currently taking any medication and, if so, what the name of it was, and the reason they were taking it (in order to exclude those taking centrally acting medication). They were also asked about their history of mental health difficulties. 2.3. Eating Attitudes Test-26 (EAT-26: Garner et al., 1982) The EAT-26 measures symptoms of eating disorders. Higher scores suggest more disordered eating, with scores of 20 or above representing the clinical cut-off. It has high internal consistency (alpha = .90) and acceptable criterion-related validity, being highly accurate in classifying eating disordered and non-eating disordered individuals (Garner et al., 1982) and has been used to identify abnormal eating patterns among college students (Thompson & Schwartz, 1982). It is widely used to identify those at risk of developing an eating disorder (anorexia nervosa and bulimia nervosa), and in analogue studies in non-clinical populations designed to be relevant to eating disorders. 2.4. The Facial Expression Recognition Task (FERT; Harmer et al., 2003) The FERT is an interactive computer task containing facial expression stimuli (posed by actors) associated with the 6 basic emotions (anger, disgust, fear, happiness, sadness, surprise), taken from the Ekman and Friesen (1976) Pictures of Affect Series. Each emotion has been morphed (i.e. transformed) between 100% intensity (full emotion expression) and 0% intensity (neutral expression), in 10% increments to create a set of stimuli displaying each emotion at 10 different levels of intensity. The task includes four examples of each emotion at each intensity (10–100%) i.e. 40 examples of each emotion. The 10 actors' faces used are also presented once each in neutral (0%). This provides a total of 250 stimuli presentations. Compared to using stimuli that have not been morphed, morphing is useful as a systematic means of deriving large numbers of stimuli that potentially vary only in a limited number of ways (in this case, intensity of the emotion), and not along a range of dimensions which would then require matching or controlling in an experimental design such as that used here. Each example is presented on a computer screen for 500 ms then immediately replaced by a blank screen. Participants are asked to indicate the emotion seen by pressing a labelled key on the keyboard as soon as they identify it. The task takes approximately 10-minutes to complete. The task has proved sensitive to change in previous studies investigating depression in both non-clinical and “at risk” samples (Bhagwagar, Cowen, Goodwin, & Harmer, 2003; Hayward, Goodwin, Cowen, & Harmer, 2005). 2.5. Visual analogue scales (VAS) Visual analogue scales were used to assess participants' subjective affective state before and after administration of the FERT. These comprised a series of 10-centimetre lines drawn on paper, each with a different one of the following adjectives written above it: anger; disgust; fear; happiness; sadness; surprise; alertness; and anxiety, and ‘not at all’ and ‘extremely’ written at either endpoint. Participants were requested to mark the lines according to the extent each state described them at the time. The placement of the marks on the lines was measured using a ruler to quantify responses. 2.6. Hospital Anxiety and Depression Scale (HADS) (Zigmond & Snaith, 1983) This self-report questionnaire was designed to detect anxiety and depression in general medical out-patients and has been widely used in clinical practice and research. It comprises anxiety and depression subscales, each with 7-items. For both subscales scores between 8 and 10 identify mild cases, 11–15 moderate cases, and 16- or above, severe cases. It places less emphasis on somatic symptoms than other measures and therefore was considered most appropriate given the physical symptoms associated
1
Body Mass Index (BMI; weight (kg) / height (m2)).
392
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
with eating disorders. It has acceptable reliability (Cronbach's alphas of .82 and .77 for anxiety and depression respectively) and normative data is available (Crawford, Henry, Crombie, & Taylor, 2001). 2.7. Procedure For the screening phase, two methods of recruitment were utilised; one for each of two universities sampled. At one, various lectures were visited and after a verbal presentation of the study, specifically addressing women, students were provided with an information sheet, a copy of the EAT-26 questionnaire, a sheet requesting contact details and permission to contact for possible further involvement in the study. A pre-paid addressed envelope was attached for the return of completed questionnaires. At the other university, the University's women's welfare representative emailed the same information to women students, including a link to a website where the EAT-26 and contact details could be completed on-line. Completed questionnaires were scored and the high- and low EAT groups created (as described above). Women in each group were contacted by their preferred means (email or telephone) and invited to participate in the main study. Self report questionnaires were posted to participants for completion at home and returned at the meeting. Where impractical, these were distributed at meetings for completion after the task. At the meeting the first VAS was administered. Instructions for the FERT were then given to participants and they completed the task. The second VAS was administered. Finally, information regarding medication was sought. 2.8. Data analysis Two tailed tests were used throughout as these are more conservative and reduce the probability of making a Type I error. They were also appropriate given the relative lack of existing research in this area. Demographic, self report questionnaire and VAS ratings were normally distributed, and differences between the two groups were analysed using t tests, with the degrees of freedom appropriately adjusted if the homogeneity of variance assumption was not met. The face recognition data were not normally distributed, and an arcsine transformation (often recommended for percentage data) did not alter this. The parametric analyses planned (ANOVA and ANCOVA) do not have non-parametric equivalents, but both are relatively robust tests. Given that there were only minor departures from normality in most cases, analysis thus proceeded as planned. The emotional faces were analysed by accuracy (percentage of total number of stimuli correctly identified in each emotion category), reaction time (time in milliseconds to correctly identify the stimuli in each emotion category), misclassification (percentage classified as an example of an emotion category which it did not represent; for example, responding to a sad face with the surprised response would be an example of misclassification of surprise) and by discrimination (ability to identify each emotion correctly from amongst all other emotions). Each of the four measures was analysed using a mixed factorial analysis of variance (ANOVA), with EAT group as a between participants factor with two levels (high EAT and low EAT) and emotion type as a within participants factor with seven levels (i.e. anger, disgust, fear, happy, sad, surprised and neutral). To examine the extent to which the findings were due to differences between the groups in symptoms of depression or anxiety, as opposed to eating disorder related symptoms, all analyses were also conducted using HADS depression and HADS anxiety scores as covariates. 3. Results 3.1. Response rate In the screening phase, approximately 800 EAT-26 questionnaires were distributed of which 281 were completed and returned, providing an estimated response rate of 35%. Of those who were later contacted to take part in the main study, the response rate was approximately 80%, and was similar in the high and low EAT groups.
Table 1 Demographic and self report questionnaire scores for the high and low EAT groups
Age (years) Education (years) Body Mass Index Eating Attitudes Test HADS depression HADS anxiety
Low EAT (N = 29)
High EAT (N = 23)
Mean (SD)
Mean (SD)
27.0 (6.6) 17.5 (1.8) 23.2 (3.1) 0.3 (0.5) 5.16 (2.56) 1.72 (2.49)
23.4 (3.7) 17.2 (2.6) 22.0 (2.5) 26.7 (9.9) 10.72 (4.54) 5.84 (4.19)
EAT = Eating Attitudes Test. SD = Standard Deviation, shown in parentheses. HADS = Hospital Anxiety and Depression Scale.
T value
Significance
2.4 0.4 1.6 −5.3 −3.3 −6.1
.02 .67 .12 b .001 .001 b .001
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
393
Table 2 Visual analogue scale scores for the high and low EAT groups Visual analogue scale
Low EAT (N = 29)
Happy Sad Angry Frightened Anxious Alert Disgust Surprise
High EAT (N = 23)
Mean (SD)
Mean (SD)
65.6 (10.9) 12.0 (12.0) 13.1 (14.8) 11.5 (17.9) 27.9 (21.6) 61.7 (13.0) 3.1 (5.7) 11.9 (9.7)
59.2 (20.2) 29.0 (23.6) 11.9 (17.5) 18.0 (18.0) 46.9 (21.2) 62.2 (13.8) 8.5 (13.8) 18.9 (11.4)
T value
Significance
1.3 −3.3 0.2 1.2 −3.1 −0.1 1.7 0.5
.20 .005 .80 .22 .004 .90 .10 .64
3.2. Participants Twelve potential participants were excluded from the high EAT group and one participant from the low EAT group because they were taking either antidepressants (N = 6), or other centrally acting medicines (e.g. diazepam, steroids). In the final sample the high EAT group consisted of 29 women and the low EAT consisted of 23 women. 3.3. Demographic details Information on age, years in full time education, and Body Mass Index can be seen in Table 1. The majority of participants in each group described themselves as ‘white, British’. The remaining minority comprised three in the low EAT group who reported Irish, British-Asian and Middle Eastern origins and one in the high EAT group who stated they were ‘European’. T tests were performed to compare ages, years in education and BMI between the two groups. The only significant difference found between groups was in age (t (1,50) = 2.4, p = .02). High EAT participants were significantly younger than low EAT participants. Age was significantly correlated with HADS depression, and percentage of angry faces misclassified (low EAT group only), and was thus used as a covariate in subsequent analyses involving these variables (Tabachnick & Fidell, 1989). The results are not reported separately, as all the analyses involved that were significant remained highly significant when age was used as a covariate. Most of the women in the high EAT group reported a history of mental health difficulties (N = 21) compared to a minority in the low EAT group (N = 2). This included four cases of past AN and eight of BN in the high EAT group, and no cases of an ED in the low EAT group. Sixteen participants in the high EAT group scored 20 or above (the recommended cut off point for an eating disorder diagnosis) on the EAT. 3.4. Self report questionnaires Scores for the two groups on the EAT-26 and HADS can also be seen in Table 1. The high EAT group, compared to the low EAT group, had significantly higher scores on the HADS depression scale (t (50) = −3.3, p = .001) and on the HADS anxiety scale (t (50) = −6.1, p b .001). 3.5. VAS ratings The VAS scores for the two groups immediately before completing the computer task can be seen in Table 2. The high EAT group had significantly higher scores on two scales, as follows: sadness (t (27.7) = −3.3, p = .005); anxiety (t (47) = −3.1, p = .004).
Table 3 Percentage of emotional faces correctly identified in the two groups Emotion
Angry Disgust Fear Happy Sad Surprised Neutral Overall SD = Standard Deviation. EAT = Eating Attitudes Test.
Percentage of emotional faces correctly identified Low EAT (N = 29)
High EAT (N = 23)
Mean (SD)
Mean (SD)
52.2 (10.1) 52.5 (15.0) 48.8 (13.4) 72.3 (8.3) 53.1 (15.0) 54.4 (12.2) 64.5 (16.2) 57.6 (5.2)
53.3 50.9 53.6 65.3 57.8 56.8 53.5 57.5
(15.5) (15.8) (18.1) (15.4) (14.8) (10.8) (13.6) (4.9)
T value
Significance
0.3 0.4 1.1 3.1 1.1 0.8 2.4
.76 .71 .28 .003 .26 .45 .02
394
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
Table 4 Time to recognise emotions in the faces accurately in the two groups Emotion
Mean time in milliseconds to recognise emotions accurately in the faces Low EAT (N = 29) Mean SD
Angry Disgust Fear Happy Sad Surprised Neutral Overall
1917.8 1854.0 1997.5 1592.6 1893.0 1939.8 1818.4 1859.0
(590.0) (651.8) (388.8) (344.3) (557.1) (579.4) (492.8) (408.49)
High EAT (N = 23) Mean SD 1858.8 2102.1 2065.9 1576.5 1781.2 1986.9 1768.5 1877.1
(508.3) (740.1) (382.7) (615.1) (512.0) (568.5) (615.1) (426.9)
SD = Standard Deviation. EAT = Eating Attitudes Test.
3.6. Facial Expression Recognition Test 3.6.1. Accuracy The mean percentage of faces correctly identified by each group for each discrete emotion and overall can be seen in Table 3. The ANOVA revealed a significant main effect of emotion type (F (6, 300) = 98.9, p b .001), and a significant interaction between emotion type and EAT group (F (6, 300) = 2.1, p = .02). Analyses of covariance (ANCOVAs) were conducted to statistically examine the extent to which depression or anxiety scores contributed to this effect using HADS depression and HADS anxiety ratings as covariates. The interaction remained significant when HADS depression was controlled (p b .05), and when HADS anxiety was controlled (p = .04). Post hoc between group tests for each emotion, to locate the source of the interaction, indicated that the high EAT group accurately recognised fewer happy faces (t (1,50) = 3.1, p = .003) and fewer neutral faces (t (1,50) = 2.4, p = .02) than the low EAT group. 3.6.2. Reaction time The mean reaction times for each emotion in each group can be seen in Table 4. The ANOVA revealed a main effect of emotion (F (6, 300) = 8.64, p b .001) but no main effect of group and no significant interaction between type of emotion and group. 3.6.3. Misclassification Percentage of each type of emotion that was wrongly classified in each group (e.g. wrongly classified as sad when in fact it depicted a different emotion) can be seen in Table 5. The ANOVA revealed a significant main effect of emotion type (F (6, 300) = 151.8, p b .001), but no significant interaction between emotion type and EAT group. 3.6.4. Discrimination Discrimination accuracy was computed for each emotion using the signal detection formula described by Surguladze et al. (2004). ANOVA revealed an interaction between each emotion type and group (F (1,50) = 5.9, p = .02), which remained significant when HADS depression and HADS anxiety ratings were controlled (p = .03 and p = .02 respectively), using ANCOVAs. Post hoc between group analyses indicated that there was a trend for anger discrimination (p = .09) to be less accurate, and surprise
Table 5 Percentage of emotional faces misclassified in the two groups Emotion
Percentage of emotional faces misclassified Low EAT (N = 29) Mean SD
Angry Disgust Fear Happy Sad Surprised Neutral Overall SD = Standard Deviation. EAT = Eating Attitudes Test.
10.2 6.7 7.9 5.1 11.7 7.0 49.2 56.70
(6.1) (5.6) (5.4) (5.1) (9.0) (4.0) (16.0) (5.47)
High EAT(N = 23) Mean SD 14.7 (7.9) 7.1 (6.7) 8.7 (6.3) 3.6 (4.6) 14.4 (7.8) 9.3 (4.3) 44.6 (12.9) 56.48 (5.44)
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
395
discrimination (p = .04) to be more accurate in the high EAT group, i.e. these participants were less accurate in identifying anger and more accurate at identifying surprise from amongst all other emotions. 3.7. Error rates The accuracy analysis indicated that the high EAT group, compared to the low EAT group, correctly identified fewer happy and fewer neutral faces. The numbers involved in the relevant cells were small, but the distribution of inaccurate responses to happy and neutral faces was inspected visually in order to investigate whether there were differences between the groups in the labels given to these two types of emotional faces. No consistent pattern emerged for happy faces, with the exception of a tendency for the high EAT group, compared to the low EAT group to categorise more of these as neutral. For neutral faces, visual inspection of the means suggested that those in the high EAT group appeared more likely than those in the low EAT group to categorise neutral faces as angry or sad, and less likely to categorise them as happy or surprise. 4. Discussion The results indicate that those with high levels of eating disorder related symptoms, compared to those with low levels, were less accurate at recognising happy and neutral faces. There was also some evidence that they were less good at discriminating anger, but better at discriminating surprise, from other emotions. Interestingly, these differences appeared to be due to increased eating disorder symptoms in the “at risk” group rather than to any mood disturbance. The finding that the effects found remained even when depression was partialled out is consistent with the findings of Kucharsha-Pietura et al. (2003). It further extends their finding to anxiety. While the medication status of participants in previous studies has been unclear, or some participants have clearly been on medication, the current study excluded those on psychotropic and other centrally acting medicines; thus the findings are not affected by any increased use of medication in the high EAT group. It is intriguing that decreased recognition of happy faces was evident. Kucharsha-Pietura et al. (2003) did not find decreased recognition of positive emotion faces, perhaps because they grouped happiness together with interest and surprise in their analysis. Zonnevylle-Bender et al. (2002) found an overall deficit in emotional processing but did not separate different emotions and thus were unable to locate the source of any differential processing. It is also intriguing that decreased recognition of neutral faces was found in the high EAT group. This is not a finding that has been previously reported, or indeed examined in detail, in this group, but it is consistent with the finding that depressed patients are impaired in processing neutral faces (Leppänen, Milders, Bell, Terriere, & Hietanen, 2004). As Leppänen et al. (2004) suggest for depressed people, this finding indicates that people with eating disorder symptoms, like those with depression, are less likely to recognise emotionally neutral faces accurately. However, in the current sample, this difference does not appear to be dependent on mood, or as in Leppänen et al., 2004, status as a patient with depression. Error rates indicate that those with high EAT scores tend to categorise happy faces as neutral, whereas they tend to categorise neutral faces more frequently as either anger or sad and less frequently as happy or surprise. Error rates have not generally been examined in relation to eating disorders or eating disorder symptoms. However, the categorisation of neutral as sad is consistent with the findings of Leppänen et al. (2004) in depressed patients. The current findings also suggest that high levels of eating disorder symptoms are associated with both a bias to interpret happy faces as emotionally neutral as well as biases to interpret neutral faces as emotionally negative (anger and sad) and not as emotionally positive (happy and surprise). Again, this is consistent with the negative biases found to be associated with depression, but is apparently not related to mood in the current study. Relatively poor discrimination of anger but good discrimination of surprise was evident in those with high levels of eating disorder symptoms; again, this appears to be largely independent of depressed and anxious mood. The current findings suggest that those with high levels of eating disorder symptoms find it difficult to identify facial expressions that indicate anger, and relatively easy to identify those indicating surprise. Anger has not previously been investigated in this group as a discrete emotion in paradigms similar to that used here, but the current findings suggest that it is worthy of further attention. It has been suggested that anger is an understudied emotion in those with eating disorders (Waller et al., 2003). One possibility is that our finding reflects lack of attention to, or “neglect” of anger, perhaps due to fear of social rejection. Kessler et al. (2006) found a trend for surprise to be less rapidly recognised by patients with AN; one possibility is that increased time spent identifying surprise may result in better discrimination. Although we did not find increased reaction time for surprise identification here, surprise may also be an emotion worthy of further study in this group. Clearly, the finding for anger, indicative only of a trend here, needs to be interpreted with caution, and requires replication in a larger study, and further investigation. Theoretically, the findings support the inclusion of emotional processing difficulties in models of eating disorders. The use of an objective measure of emotional processing has facilitated a more detailed analysis of the specific deficits than would have been possible using the TAS. Interestingly, while high EAT scorers, compared to low EAT scorers show some “deficits”, they also show some strengths or enhanced processing abilities. These may, of course, also be related to the development of psychopathology. Specific focus on emotional processing per se is not typically part of cognitive behavioural therapy (CBT) for eating disorders. Nevertheless CBT may need to teach those with disturbed eating how to accurately decode facial expressions, before moving on to help them re-assess the meaning of what people say or do, including for example, by modifying biases in the processing of emotional information. The use of an objective measure, such as the FERT, when compared to existing questionnaires, enables us to
396
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
understand the precise nature of the difficulties that those with eating disorder related symptoms may have. This may enable treatment of those with eating disorders to target specific difficulties more precisely. The study has several limitations. We studied a non-clinical group, thus it is important to investigate in future work whether these patterns of emotional processing also characterise those with diagnosed eating disorders. Nevertheless, the mean score of our high EAT group was above the clinical cut off, and there is some evidence that our screening tool, the EAT, performs well in identifying partial syndrome eating disorders (Scheinberg et al., 1993), particularly those with sub-clinical AN. The sample size was relatively small, and it is important that the current findings are replicated. A larger sample might also permit statistical analysis of error rates. The sample was self selecting, i.e. compromised only of those who completed the screening tool, thus potentially not representative of all those in the “at risk” category. Reliance on self report, for example, for report of height and weight, is also a limitation, as self report of these data may not be accurate. The EAT has not often been investigated for it's ability to identify BN, or indeed those who present to services with an ED-NOS diagnosis, thus it is particularly important to determine whether the current findings also apply to these patient groups. The decision to investigate the role of mood by controlling statistically for HADS depression and HADS anxiety score was a difficult one. It is possible that, in the case of eating disorder related psychopathology, such an analysis may control for symptoms that are a key aspect of the eating related disturbance. Nevertheless, the key differences between the high and low groups largely remained significant when these analyses were conducted. While it is possible that the HADS is a relatively poor measure of mood in those with eating disorder symptoms, this seems unlikely given its extensive psychometric assessment. We did not have full information on any psychological treatment received by participants. While we would not expect this to create deficits in emotional processing it is possible that those in the high EAT group who had received some psychological treatment performed better than those who had not. This requires further investigation. Overall, the findings suggest that the role of mood in relation to eating disorder related psychopathology is complex, and requires further examination. One possibility, which is consistent with our clinical observations, is that the negative emotional experiences and negative sense of self seen in patients with eating disorders are an intrinsic part of the condition and not attributable solely to concomitant depressive symptoms. In particular, our impression is that “depressed mood” in those with eating disorders may reflect an extreme and intense emotional and cognitive state not readily captured by the qualitative experience of depression alone, and also not captured by current assessment measures. This may help explain a pattern of results in those “at risk” which is similar to that found in depression, but which appears not to be attributable to depressed mood as assessed on standardised measures. Clearly, this requires further investigation. Overall, the findings indicate that emotional processing may be an important topic for further investigation in relation to eating disorder symptoms. The FERT would appear to be a useful tool to study emotional processing, with particular promise in identifying disturbances in the processing of discrete emotions in those “at risk” or with sub-clinical eating disturbance. References Abraham, S. F., & Beumont, P. J. V. (1982). How patients describe bulimia or binge eating. Psychological Medicine, 12, 625−635. American Psychiatric Association (1994). Diagnostic and Statistical Manual of Mental Disorders, 4th edition revised. Washington, DC: Author. Beales, D. L., & Dolton, R. (2000). Eating disordered patients: Personality, alexithymia, and implications for primary care. British Journal of General Practice, 50, 21−26. Bhagwagar, Z., Cowen, P. J., Goodwin, G. M., & Harmer, C. J. (2003). Normalisation of enhanced fear recognition by acute SSRI treatment in subjects with a previous history of depression. American Journal of Psychiatry, 161, 166−168. Bowlby, J. (1969). Attachment and Loss, Vol 1: Attachment. New York: Basic Books. Bruch, H. (1973). Eating Disorders. New York: Basic Books. Bydlowski, S., Corcos, M., Jeammet, P., Paterniti, S., Berthoz, S., Laurier, C., et al. (2005). Emotion-processing deficits in eating disorders. International Journal of Eating Disorders, 37, 321−329. Connan, F., Campbell, I. C., Katzman, M., Lightman, S. L., & Treasure, J. (2003). A neurodevelopmental model for anorexia nervosa. Physiology and Behavior, 79, 13−24. Cooper, M. J. (2003). Bulimia nervosa: A cognitive perspective. Oxford: Oxford University Press. Cooper, M. J. (2005). A vulnerability stress model of eating disorders: A cognitive perspective. In B. L. Hankin, & J. Abela (Eds.), Development of Psychopathology: A Vulnerability Stress Perspective Thousand Oaks, CA: Sage. Cooper, M., Wells, A., & Todd, G. (2004). A cognitive model of BN. British Journal of Clinical Psychology, 43, 1−16. Crawford, J. R., Henry, J. D., Crombie, C., & Taylor, E. P. (2001). Normative data for the HADS from a large non-clinical sample. British Journal of Clinical Psychology, 40, 429−434. Ekman, P., & Friesen, W. V. (1976). Pictures of facial affect. Palo Alto, Calif: Consulting Psychologists Press. Fairburn, C. G., & Harrison, P. J. (2003). Eating disorders. The Lancet, 361, 407−416. Fairburn, C. G., Cooper, Z., & Shafran, R. (2003). Cognitive behaviour therapy for eating disorders: A “transdiagnostic” theory and treatment. Behaviour Research and Therapy, 41, 509−528. Fonagy, P., Gergely, G., Jurist, E. L., & Target, M. (2002). Affect regulation, mentalisation and the development of the self. London: Karnac. Garner, D. M., Olmsted, M. P., Bohr, Y., & Garfinkel, P. (1982). The Eating Attitudes Test: Psychometric features and clinical correlates. Psychological Medicine, 12, 871−878. Gilboa-Schechtman, E., Avnon, L., Zubery, E., & Jeczmien, P. (2006). Emotional processing in eating disorders: Specific impairment or general distress related deficiency? Depression and Anxiety, 23, 331−339. Guidano, V. F., & Liotti, G. (1983). Cognitive Processes and Emotional Disorders: A Structured Approach to Psychotherapy. New York: Guilford Press. Harmer, C. J., Bhagwagar, Z., Perrett, D. I., Vollm, B. A., Cowen, P. J., & Goodwin, G. M. (2003). Acute SSRI administration affects the processing of social cues in healthy volunteers. Neuropsychopharmacology, 28, 148−152. Harmer, C. J., Perrett, D. J., Cowen, P. J., & Goodwin, G. M. (2001). Administration of the beta-adrenoreceptor blocker propranolol impairs the processing of facial expressions of sadness. Psychopharmacology, 154, 383−389. Hayward, G., Goodwin, G., Cowen, P., & Harmer, C. (2005). Low-dose tryptophan depletion in recovered depressed patients induces changes in cognitive processing without depressive symptoms. Biological Psychiatry, 57, 517−524. Hinrichsen, H., Wright, F., Waller, G., & Meyer, C. (2003). Social anxiety and coping strategies in the eating disorders. Eating Behaviors, 4, 117−126. Kessler, H., Schwarze, M., Filipic, Traue, H. C., & von Wietersheim, J. (2006). Alexithymia and facial emotion recognition in patients with eating disorders. International Journal of Eating Disorders, 39, 245−251. Kiyotaki, Y., & Yokoyama, K. (2006). Relationships of eating disturbances to alexithymia, need for social approval, and gender identity among Japanese female undergraduate students. Personality and Individual Differences, 41, 609−618.
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
397
Kucharska-Pietura, K., Nikolaou, V., Masiak, M., & Treasure, J. (2004). The recognition of emotion in the faces and voice of anorexia nervosa. International Journal of Eating Disorders, 35, 42−47. Leppänen, J., Milders, M., Bell, J., Terriere, E., & Hietanen, J. (2004). Depression biases the recognition of emotionally neutral faces. Psychiatry Research, 128, 123−133. Matsumoto, D., & Ekman, P. (1998). Japanese and Caucasian Facial Expressions of Emotion and Neutral Faces. San Francisco: San Francisco State University. Mendlewicz, L., Linkowski, P., Bazelmans, C., & Philippot, P. (2005). Decoding emotional facial expressions in depressed and anorexic patients. Journal of Affective Disorders, 89, 195−199. Quinton, S., & Wagner, H. L. (2005). Alexithymia, ambivalence over emotional expression, and eating attitudes. Personality and Individual Differences, 38, 1163−1173. Ratnasuriya, R. H., Eisler, I., Szmuckler, G. I., & Russell, G. F. M. (1991). Anorexia nervosa: Outcome and prognostic factors after 20 years. British Journal of Psychiatry, 158, 495−502. Scheinberg, Z., Koslowsky, M., Bleich, A., Mark, M., Apter, A., Danon, Y., et al. (1993). Sensitivity, specificity, and positive predictive value as measures of prediction accuracy: The case of the EAT-26. Educational and Psychological Measurement, 53, 831−839. Schmidt, U., & Treasure, J. (2006). Anorexia nervosa: Valuable and visible. A cognitive–interpersonal maintenance model and its implications for research and practice. British Journal of Clinical Psychology, 45, 343−366. Surguladze, S. A., Young, A. W., Senior, C., Brebion, G., Travis, M. J., & Phillips, M. L. (2004). Recognition accuracy and response bias to happy and sad facial expressions in patients with major depression. Neuropsychology, 18, 212−218. Tabachnick, B. G., & Fidell, L. S. (1989). Using Multivariate Statistics. New York: Harper Collins. Taylor, G. J., Ryan, D., & Bagby, R. M. (1985). Toward the development of a new self-report alexithymia scale. Psychotherapy and Psychosomatics, 44, 191−199. Thompson, M. G., & Schwartz, D. M. (1982). Life-adjustment of women with anorexia nervosa and anorexic-like behaviour. International Journal of Eating Disorders, 1, 47−60. Waller, G., Babbs, M., Milligan, R., Meyer, C., Ohanian, V., & Leung, N. (2003). Anger and core beliefs in the eating disorders. International Journal of Eating Disorders, 34, 118−124. Yager, J., Landsverk, J., & Edelstein, C. K. (1987). A 20 month follow up study of 628 women with eating disorders, I: Course and severity. American Journal of Psychiatry, 144, 1172−1177. Zigmond, A. S., & Snaith, R. P. (1983). The hospital anxiety and depression scale. Acta Psychiatrica Scandinavica, 67, 361−370. Zonnevylle-Bender, M. J. S., van Goozen, S. H. M., Cohen-Kettenis, P. T., van Elburg, A., & van Engeland, H. (2002). Do adolescent anorexia nervosa patients have deficits in emotional functioning? European Child and Adolescent Psychiatry, 11, 38−42.
Contents lists available at ScienceDirect
Eating Behaviors
Emotional face processing in women with high and low levels of eating disorder related symptoms Lynne Jones, Catherine Harmer, Phil Cowen, Myra Cooper ⁎ Isis Education Centre and Department of Psychiatry University of Oxford, UK
a r t i c l e
i n f o
Article history: Received 7 October 2007 Received in revised form 16 February 2008 Accepted 7 March 2008 Keywords: Eating disorders Emotional processing Happy Anger Surprise
a b s t r a c t Objectives: Emotional processing has rarely been investigated in those “at risk” of developing an eating disorder. This study investigated the processing of six basic emotions depicted on faces in an “at risk” group, compared to a control group. Design: Participants were women with high (N = 29) and low (N = 23) levels of eating disorder symptoms who were not taking psychotropic medication. A well characterised computerised task (Facial Expression Emotion Task) was administered to all participants. Results: Women with high levels of eating disorder symptoms, compared to those with low levels, were less accurate at recognising happy and neutral faces, but showed no differences in their accuracy at recognising other emotions. They also showed a trend to be less good at discriminating anger, but better at discriminating surprise from other emotions. Depressive and anxious symptoms did not provide a complete explanation for the findings. Conclusions: The findings support the inclusion of emotional processing in models of eating disorders, and suggest that it may have a role in their development. Emotional processing warrants further investigation particularly in those “at risk” but also in those with eating disorders. © 2008 Elsevier Ltd. All rights reserved.
1. Introduction Clinical descriptions have long suggested that people with eating disorders have deficits in emotional processing, including in the context of interpersonal relationships (e.g. Bruch, 1973, p50). It has been suggested that these may help account for the fact that patients with eating disorders often have significant problems with social functioning (Kucharska-Pietura, Nikolaou, Masiak, & Treasure, 2004). In the context of interpersonal relationships, for example, misinterpretation of another person's emotional state might result in an unhelpful behaviour or response which may then have a negative consequence. Social functioning difficulties may persist even after behavioural symptoms have remitted (Yager, Landsverk, & Edelstein, 1987), and become chronic (Ratnasuriya, Eisler, Szmuckler, & Russell, 1991). Problems with social functioning may also help maintain the eating disorder symptoms (Fairburn & Harrison, 2003). Research suggests that social and interpersonal situations, for example, are frequent triggers for eating disorder related behaviour (Abraham & Beumont, 1982). Research has investigated ability to understand and process patients own internal emotions, and this ability has a role in cognitive models of eating disorders. However, little research attention has been given to the ability of those with eating disorders to understand/interpret the emotional states of other people. Recent developments in cognitive models of eating disorders (e.g. Cooper, Wells, & Todd, 2004; Fairburn, Cooper, & Shafran, 2003) conceptualise binge-eating, to a greater or less extent, as a means of coping with intense (internal) cognitive and affective distress, while lack of awareness of own emotional states has been identified as a significant risk factor in cognitive models of
⁎ Corresponding author. Isis Education Centre, Oxford, OX3 7JX, UK. E-mail address: [email protected] (M. Cooper). 1471-0153/$ – see front matter © 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.eatbeh.2008.03.001
390
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
anorexia nervosa (AN) (e.g. Connan, Campbell, Katzman, Lightman, & Treasure, 2003; Guidano and Liotti, 1983). Cognitive models, however, pay much less attention to patient's ability to comprehend other people's emotional states, and do not generally provide an explanation of whether and how this might play a role in the development or maintenance of eating disorders. Theoretically, in developmental models, understanding of one's own and of other's emotional states is often regarded as interdependent. Attachment theory outlines how deficits in one will affect the development of the other, and vice versa (e.g. Fonagy, Gergely, Jurist, & Target, 2002). Emotional processing difficulties also arise in the context of the early mother child relationship, with the mother and infant forming an affective relationship from the beginning of life (e.g. Bowlby, 1969). Historically, developmental factors have not been emphasised in cognitive models of EDs although there is recent work providing a preliminary integration of the two models (Cooper, 2005). There is also a cognitive–interpersonal model (Schmidt and Treasure, 2006) which includes difficulties in recognising others emotions as part of one process (avoidance) maintaining AN. Importantly, a deficit in understanding others emotions is likely to make cognitive therapy tasks difficult to complete successfully. A number of studies have used the self report Toronto Alexithymia Scale (TAS: Taylor, Ryan, & Bagby, 1985) to assess emotional processing. These have indicated that patients with eating disorders, including those with AN (Beales and Dolton, 2000) and bulimia nervosa (BN) (Bydlowski et al., 2005) have deficits in emotional processing. Two recent studies in non-clinical groups, for example, have also found high levels of eating disorder related symptoms are related to deficits on the TAS (Kiyotaki and Yokoyama, 2006; Quinton and Wagner, 2005), suggesting that emotional processing deficits may also be characteristic of those “at risk” or who have sub-clinical presentations. Although these findings are of interest, the TAS has been criticised for amalgamating several aspects of emotional processing (Cooper, 2003), including both awareness of own and other's emotional states. It has also been suggested that more specific and objective measures are needed (Gilboa-Schechtman, Avnon, Zubery, & Jeczmien, 2006). A small number of studies have used more objective means to assess emotional processing in clinical samples. These typically involve ability to identify emotion in others. To date, most studies have used emotional faces as stimuli, but have produced mixed findings. Zonnevylle-Bender, van Goozen, Cohen-Kettenis, van Elburg, and van Engeland (2002) found that adolescents with eating disorders were worse than controls at identifying emotion, using pictures presented on videotape. Kucharska-Pietura, Nikolaou, Masiak, and Treasure (2004), report similar findings in women with AN, using nine emotions presented on slides. However, Mendlewicz, Linkowski, Bazelmans, and Philippot (2005) used a computer based task, composed of facial stimuli taken from Ekman's series (Matsumoto and Ekman, 1998), but found no difference between those with AN and controls in accuracy or reaction time to five different emotions. Similarly, Kessler, Schwarze, Filipic, Traue, and von Wietersheim (2006), also using computerised presentation, found no difference between those with AN and controls in ability to recognise the six basic facial emotions, with the exception of a trend for the patients to recognise surprise less rapidly than controls. There appear to be no comparable studies in those with BN, and little information on which to base differential predictions in BN compared to AN. A wide range of tasks have been used in these studies, most of which have unknown ability to discriminate between different groups. In several studies the medication status of the patients tested is unclear, and some are reported to be taking selective serotonin reuptake inhibitors (SSRIs). This is likely to be significant given that several studies have found similar tasks to be highly sensitive to SSRI administration (e.g. Harmer et al., 2003). Despite the high prevalence of depression in those with eating disorders, and increasing recognition of the importance of anxiety disorders (e.g. Hinrichsen, Wright, Waller, & Meyer, 2003), only one study controlled for depressive symptoms (Kucharska-Pietura et al. (2004), and none for anxiety, thus some of the significant effects observed may be due to general distress, rather than any eating disorder specific features. This is particularly relevant given that both anxiety and depression are known to be associated with both emotional processing deficits and characteristic biases in the processing of, for example, negative information. Clinically, there is increased recognition of the prevalence and seriousness of sub-clinical eating disorders. While their extent and nature remains understudied, it has been suggested that this group (which includes those identified in the Diagnostic and Statistical Manual-IV (American Psychiatric Association, 1994) manual as Eating Disorder Not Otherwise Specified: ED-NOS) is by far the largest of those with an eating disorder (Fairburn & Harrison, 2003). Moreover, many will subsequently develop either AN or BN. To date, however, no investigation of emotional processing using facial stimuli appear to be have been conducted in a group “at risk” of developing an eating disorder, or who have sub-clinical symptoms. This is an important area for research given that relatively little is known about this group. Investigation of their emotional processing may also provide some insight into why and how those “at risk” may develop AN or BN, and how similar those “at risk” are to those with the full syndrome disorders. The aim of the current study was to investigate the processing of emotional facial stimuli in a group of women with high scores on a measure of eating disorder related symptoms (i.e. an “at risk” or sub-clinical group) and a group of women with low scores. A well characterised, computerised task of emotional facial stimuli, measuring the full range of basic emotions, was used. Potential participants were excluded if they were taking any centrally acting medication (e.g. SSRIs or other medication known to affect the central nervous system) at the time of study and depression and anxiety were taken into account statistically in conducting the data analysis. Given previously conflicting findings, the study was designed to be exploratory and no specific hypotheses were investigated. 2. Method 2.1. Participants A sample of two hundred and eighty-one women students enrolled in various undergraduate and post-graduate degrees at two universities completed the Eating Attitudes Test-26 (EAT; Garner, Olmsted, Bohr, & Garfinkel, 1982) and provided their contact
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
391
details. The range of scores was 0 to 52 for the entire sample (from a possible range of 0 to 78). Women were then contacted consecutively, from the highest scorers down and the lowest scorers up, until a minimum of 20 participants had potentially been recruited in two groups. These groups were labelled: low EAT (EAT score 3 or less) and high EAT (EAT score 16 or greater). They represented the top and bottom 15–20% of the sample. A power calculation was not possible, given the lack of previous research with this task in this group, and a minimum of 20 was chosen based on that needed for significance in studies using the same measure in different populations (e.g. Harmer et al., 2003). No potential participant with whom contact was made in this way declined to take part. Exclusion criteria included visual or motor impairments which prevented participants accessing visually presented material or using a keyboard. These were not evident in any of the women approached. As noted above, potential participants were also excluded if they were currently taking any centrally acting medication. This exclusion applied not only to recognised psychotropic drugs but also general medical drugs with central actions. Some of the latter agents, for example, propanolol, are known to modify emotional processing (Harmer, Perrett, Cowen, & Goodwin, 2001). 2.2. Demographic information Participants were asked to provide their age, self-described ethnic origin and number of years spent in education. Self-report of height and weight was requested and used to calculate participants' body mass indices (BMI).1 Participants were asked whether they were currently taking any medication and, if so, what the name of it was, and the reason they were taking it (in order to exclude those taking centrally acting medication). They were also asked about their history of mental health difficulties. 2.3. Eating Attitudes Test-26 (EAT-26: Garner et al., 1982) The EAT-26 measures symptoms of eating disorders. Higher scores suggest more disordered eating, with scores of 20 or above representing the clinical cut-off. It has high internal consistency (alpha = .90) and acceptable criterion-related validity, being highly accurate in classifying eating disordered and non-eating disordered individuals (Garner et al., 1982) and has been used to identify abnormal eating patterns among college students (Thompson & Schwartz, 1982). It is widely used to identify those at risk of developing an eating disorder (anorexia nervosa and bulimia nervosa), and in analogue studies in non-clinical populations designed to be relevant to eating disorders. 2.4. The Facial Expression Recognition Task (FERT; Harmer et al., 2003) The FERT is an interactive computer task containing facial expression stimuli (posed by actors) associated with the 6 basic emotions (anger, disgust, fear, happiness, sadness, surprise), taken from the Ekman and Friesen (1976) Pictures of Affect Series. Each emotion has been morphed (i.e. transformed) between 100% intensity (full emotion expression) and 0% intensity (neutral expression), in 10% increments to create a set of stimuli displaying each emotion at 10 different levels of intensity. The task includes four examples of each emotion at each intensity (10–100%) i.e. 40 examples of each emotion. The 10 actors' faces used are also presented once each in neutral (0%). This provides a total of 250 stimuli presentations. Compared to using stimuli that have not been morphed, morphing is useful as a systematic means of deriving large numbers of stimuli that potentially vary only in a limited number of ways (in this case, intensity of the emotion), and not along a range of dimensions which would then require matching or controlling in an experimental design such as that used here. Each example is presented on a computer screen for 500 ms then immediately replaced by a blank screen. Participants are asked to indicate the emotion seen by pressing a labelled key on the keyboard as soon as they identify it. The task takes approximately 10-minutes to complete. The task has proved sensitive to change in previous studies investigating depression in both non-clinical and “at risk” samples (Bhagwagar, Cowen, Goodwin, & Harmer, 2003; Hayward, Goodwin, Cowen, & Harmer, 2005). 2.5. Visual analogue scales (VAS) Visual analogue scales were used to assess participants' subjective affective state before and after administration of the FERT. These comprised a series of 10-centimetre lines drawn on paper, each with a different one of the following adjectives written above it: anger; disgust; fear; happiness; sadness; surprise; alertness; and anxiety, and ‘not at all’ and ‘extremely’ written at either endpoint. Participants were requested to mark the lines according to the extent each state described them at the time. The placement of the marks on the lines was measured using a ruler to quantify responses. 2.6. Hospital Anxiety and Depression Scale (HADS) (Zigmond & Snaith, 1983) This self-report questionnaire was designed to detect anxiety and depression in general medical out-patients and has been widely used in clinical practice and research. It comprises anxiety and depression subscales, each with 7-items. For both subscales scores between 8 and 10 identify mild cases, 11–15 moderate cases, and 16- or above, severe cases. It places less emphasis on somatic symptoms than other measures and therefore was considered most appropriate given the physical symptoms associated
1
Body Mass Index (BMI; weight (kg) / height (m2)).
392
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
with eating disorders. It has acceptable reliability (Cronbach's alphas of .82 and .77 for anxiety and depression respectively) and normative data is available (Crawford, Henry, Crombie, & Taylor, 2001). 2.7. Procedure For the screening phase, two methods of recruitment were utilised; one for each of two universities sampled. At one, various lectures were visited and after a verbal presentation of the study, specifically addressing women, students were provided with an information sheet, a copy of the EAT-26 questionnaire, a sheet requesting contact details and permission to contact for possible further involvement in the study. A pre-paid addressed envelope was attached for the return of completed questionnaires. At the other university, the University's women's welfare representative emailed the same information to women students, including a link to a website where the EAT-26 and contact details could be completed on-line. Completed questionnaires were scored and the high- and low EAT groups created (as described above). Women in each group were contacted by their preferred means (email or telephone) and invited to participate in the main study. Self report questionnaires were posted to participants for completion at home and returned at the meeting. Where impractical, these were distributed at meetings for completion after the task. At the meeting the first VAS was administered. Instructions for the FERT were then given to participants and they completed the task. The second VAS was administered. Finally, information regarding medication was sought. 2.8. Data analysis Two tailed tests were used throughout as these are more conservative and reduce the probability of making a Type I error. They were also appropriate given the relative lack of existing research in this area. Demographic, self report questionnaire and VAS ratings were normally distributed, and differences between the two groups were analysed using t tests, with the degrees of freedom appropriately adjusted if the homogeneity of variance assumption was not met. The face recognition data were not normally distributed, and an arcsine transformation (often recommended for percentage data) did not alter this. The parametric analyses planned (ANOVA and ANCOVA) do not have non-parametric equivalents, but both are relatively robust tests. Given that there were only minor departures from normality in most cases, analysis thus proceeded as planned. The emotional faces were analysed by accuracy (percentage of total number of stimuli correctly identified in each emotion category), reaction time (time in milliseconds to correctly identify the stimuli in each emotion category), misclassification (percentage classified as an example of an emotion category which it did not represent; for example, responding to a sad face with the surprised response would be an example of misclassification of surprise) and by discrimination (ability to identify each emotion correctly from amongst all other emotions). Each of the four measures was analysed using a mixed factorial analysis of variance (ANOVA), with EAT group as a between participants factor with two levels (high EAT and low EAT) and emotion type as a within participants factor with seven levels (i.e. anger, disgust, fear, happy, sad, surprised and neutral). To examine the extent to which the findings were due to differences between the groups in symptoms of depression or anxiety, as opposed to eating disorder related symptoms, all analyses were also conducted using HADS depression and HADS anxiety scores as covariates. 3. Results 3.1. Response rate In the screening phase, approximately 800 EAT-26 questionnaires were distributed of which 281 were completed and returned, providing an estimated response rate of 35%. Of those who were later contacted to take part in the main study, the response rate was approximately 80%, and was similar in the high and low EAT groups.
Table 1 Demographic and self report questionnaire scores for the high and low EAT groups
Age (years) Education (years) Body Mass Index Eating Attitudes Test HADS depression HADS anxiety
Low EAT (N = 29)
High EAT (N = 23)
Mean (SD)
Mean (SD)
27.0 (6.6) 17.5 (1.8) 23.2 (3.1) 0.3 (0.5) 5.16 (2.56) 1.72 (2.49)
23.4 (3.7) 17.2 (2.6) 22.0 (2.5) 26.7 (9.9) 10.72 (4.54) 5.84 (4.19)
EAT = Eating Attitudes Test. SD = Standard Deviation, shown in parentheses. HADS = Hospital Anxiety and Depression Scale.
T value
Significance
2.4 0.4 1.6 −5.3 −3.3 −6.1
.02 .67 .12 b .001 .001 b .001
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
393
Table 2 Visual analogue scale scores for the high and low EAT groups Visual analogue scale
Low EAT (N = 29)
Happy Sad Angry Frightened Anxious Alert Disgust Surprise
High EAT (N = 23)
Mean (SD)
Mean (SD)
65.6 (10.9) 12.0 (12.0) 13.1 (14.8) 11.5 (17.9) 27.9 (21.6) 61.7 (13.0) 3.1 (5.7) 11.9 (9.7)
59.2 (20.2) 29.0 (23.6) 11.9 (17.5) 18.0 (18.0) 46.9 (21.2) 62.2 (13.8) 8.5 (13.8) 18.9 (11.4)
T value
Significance
1.3 −3.3 0.2 1.2 −3.1 −0.1 1.7 0.5
.20 .005 .80 .22 .004 .90 .10 .64
3.2. Participants Twelve potential participants were excluded from the high EAT group and one participant from the low EAT group because they were taking either antidepressants (N = 6), or other centrally acting medicines (e.g. diazepam, steroids). In the final sample the high EAT group consisted of 29 women and the low EAT consisted of 23 women. 3.3. Demographic details Information on age, years in full time education, and Body Mass Index can be seen in Table 1. The majority of participants in each group described themselves as ‘white, British’. The remaining minority comprised three in the low EAT group who reported Irish, British-Asian and Middle Eastern origins and one in the high EAT group who stated they were ‘European’. T tests were performed to compare ages, years in education and BMI between the two groups. The only significant difference found between groups was in age (t (1,50) = 2.4, p = .02). High EAT participants were significantly younger than low EAT participants. Age was significantly correlated with HADS depression, and percentage of angry faces misclassified (low EAT group only), and was thus used as a covariate in subsequent analyses involving these variables (Tabachnick & Fidell, 1989). The results are not reported separately, as all the analyses involved that were significant remained highly significant when age was used as a covariate. Most of the women in the high EAT group reported a history of mental health difficulties (N = 21) compared to a minority in the low EAT group (N = 2). This included four cases of past AN and eight of BN in the high EAT group, and no cases of an ED in the low EAT group. Sixteen participants in the high EAT group scored 20 or above (the recommended cut off point for an eating disorder diagnosis) on the EAT. 3.4. Self report questionnaires Scores for the two groups on the EAT-26 and HADS can also be seen in Table 1. The high EAT group, compared to the low EAT group, had significantly higher scores on the HADS depression scale (t (50) = −3.3, p = .001) and on the HADS anxiety scale (t (50) = −6.1, p b .001). 3.5. VAS ratings The VAS scores for the two groups immediately before completing the computer task can be seen in Table 2. The high EAT group had significantly higher scores on two scales, as follows: sadness (t (27.7) = −3.3, p = .005); anxiety (t (47) = −3.1, p = .004).
Table 3 Percentage of emotional faces correctly identified in the two groups Emotion
Angry Disgust Fear Happy Sad Surprised Neutral Overall SD = Standard Deviation. EAT = Eating Attitudes Test.
Percentage of emotional faces correctly identified Low EAT (N = 29)
High EAT (N = 23)
Mean (SD)
Mean (SD)
52.2 (10.1) 52.5 (15.0) 48.8 (13.4) 72.3 (8.3) 53.1 (15.0) 54.4 (12.2) 64.5 (16.2) 57.6 (5.2)
53.3 50.9 53.6 65.3 57.8 56.8 53.5 57.5
(15.5) (15.8) (18.1) (15.4) (14.8) (10.8) (13.6) (4.9)
T value
Significance
0.3 0.4 1.1 3.1 1.1 0.8 2.4
.76 .71 .28 .003 .26 .45 .02
394
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
Table 4 Time to recognise emotions in the faces accurately in the two groups Emotion
Mean time in milliseconds to recognise emotions accurately in the faces Low EAT (N = 29) Mean SD
Angry Disgust Fear Happy Sad Surprised Neutral Overall
1917.8 1854.0 1997.5 1592.6 1893.0 1939.8 1818.4 1859.0
(590.0) (651.8) (388.8) (344.3) (557.1) (579.4) (492.8) (408.49)
High EAT (N = 23) Mean SD 1858.8 2102.1 2065.9 1576.5 1781.2 1986.9 1768.5 1877.1
(508.3) (740.1) (382.7) (615.1) (512.0) (568.5) (615.1) (426.9)
SD = Standard Deviation. EAT = Eating Attitudes Test.
3.6. Facial Expression Recognition Test 3.6.1. Accuracy The mean percentage of faces correctly identified by each group for each discrete emotion and overall can be seen in Table 3. The ANOVA revealed a significant main effect of emotion type (F (6, 300) = 98.9, p b .001), and a significant interaction between emotion type and EAT group (F (6, 300) = 2.1, p = .02). Analyses of covariance (ANCOVAs) were conducted to statistically examine the extent to which depression or anxiety scores contributed to this effect using HADS depression and HADS anxiety ratings as covariates. The interaction remained significant when HADS depression was controlled (p b .05), and when HADS anxiety was controlled (p = .04). Post hoc between group tests for each emotion, to locate the source of the interaction, indicated that the high EAT group accurately recognised fewer happy faces (t (1,50) = 3.1, p = .003) and fewer neutral faces (t (1,50) = 2.4, p = .02) than the low EAT group. 3.6.2. Reaction time The mean reaction times for each emotion in each group can be seen in Table 4. The ANOVA revealed a main effect of emotion (F (6, 300) = 8.64, p b .001) but no main effect of group and no significant interaction between type of emotion and group. 3.6.3. Misclassification Percentage of each type of emotion that was wrongly classified in each group (e.g. wrongly classified as sad when in fact it depicted a different emotion) can be seen in Table 5. The ANOVA revealed a significant main effect of emotion type (F (6, 300) = 151.8, p b .001), but no significant interaction between emotion type and EAT group. 3.6.4. Discrimination Discrimination accuracy was computed for each emotion using the signal detection formula described by Surguladze et al. (2004). ANOVA revealed an interaction between each emotion type and group (F (1,50) = 5.9, p = .02), which remained significant when HADS depression and HADS anxiety ratings were controlled (p = .03 and p = .02 respectively), using ANCOVAs. Post hoc between group analyses indicated that there was a trend for anger discrimination (p = .09) to be less accurate, and surprise
Table 5 Percentage of emotional faces misclassified in the two groups Emotion
Percentage of emotional faces misclassified Low EAT (N = 29) Mean SD
Angry Disgust Fear Happy Sad Surprised Neutral Overall SD = Standard Deviation. EAT = Eating Attitudes Test.
10.2 6.7 7.9 5.1 11.7 7.0 49.2 56.70
(6.1) (5.6) (5.4) (5.1) (9.0) (4.0) (16.0) (5.47)
High EAT(N = 23) Mean SD 14.7 (7.9) 7.1 (6.7) 8.7 (6.3) 3.6 (4.6) 14.4 (7.8) 9.3 (4.3) 44.6 (12.9) 56.48 (5.44)
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
395
discrimination (p = .04) to be more accurate in the high EAT group, i.e. these participants were less accurate in identifying anger and more accurate at identifying surprise from amongst all other emotions. 3.7. Error rates The accuracy analysis indicated that the high EAT group, compared to the low EAT group, correctly identified fewer happy and fewer neutral faces. The numbers involved in the relevant cells were small, but the distribution of inaccurate responses to happy and neutral faces was inspected visually in order to investigate whether there were differences between the groups in the labels given to these two types of emotional faces. No consistent pattern emerged for happy faces, with the exception of a tendency for the high EAT group, compared to the low EAT group to categorise more of these as neutral. For neutral faces, visual inspection of the means suggested that those in the high EAT group appeared more likely than those in the low EAT group to categorise neutral faces as angry or sad, and less likely to categorise them as happy or surprise. 4. Discussion The results indicate that those with high levels of eating disorder related symptoms, compared to those with low levels, were less accurate at recognising happy and neutral faces. There was also some evidence that they were less good at discriminating anger, but better at discriminating surprise, from other emotions. Interestingly, these differences appeared to be due to increased eating disorder symptoms in the “at risk” group rather than to any mood disturbance. The finding that the effects found remained even when depression was partialled out is consistent with the findings of Kucharsha-Pietura et al. (2003). It further extends their finding to anxiety. While the medication status of participants in previous studies has been unclear, or some participants have clearly been on medication, the current study excluded those on psychotropic and other centrally acting medicines; thus the findings are not affected by any increased use of medication in the high EAT group. It is intriguing that decreased recognition of happy faces was evident. Kucharsha-Pietura et al. (2003) did not find decreased recognition of positive emotion faces, perhaps because they grouped happiness together with interest and surprise in their analysis. Zonnevylle-Bender et al. (2002) found an overall deficit in emotional processing but did not separate different emotions and thus were unable to locate the source of any differential processing. It is also intriguing that decreased recognition of neutral faces was found in the high EAT group. This is not a finding that has been previously reported, or indeed examined in detail, in this group, but it is consistent with the finding that depressed patients are impaired in processing neutral faces (Leppänen, Milders, Bell, Terriere, & Hietanen, 2004). As Leppänen et al. (2004) suggest for depressed people, this finding indicates that people with eating disorder symptoms, like those with depression, are less likely to recognise emotionally neutral faces accurately. However, in the current sample, this difference does not appear to be dependent on mood, or as in Leppänen et al., 2004, status as a patient with depression. Error rates indicate that those with high EAT scores tend to categorise happy faces as neutral, whereas they tend to categorise neutral faces more frequently as either anger or sad and less frequently as happy or surprise. Error rates have not generally been examined in relation to eating disorders or eating disorder symptoms. However, the categorisation of neutral as sad is consistent with the findings of Leppänen et al. (2004) in depressed patients. The current findings also suggest that high levels of eating disorder symptoms are associated with both a bias to interpret happy faces as emotionally neutral as well as biases to interpret neutral faces as emotionally negative (anger and sad) and not as emotionally positive (happy and surprise). Again, this is consistent with the negative biases found to be associated with depression, but is apparently not related to mood in the current study. Relatively poor discrimination of anger but good discrimination of surprise was evident in those with high levels of eating disorder symptoms; again, this appears to be largely independent of depressed and anxious mood. The current findings suggest that those with high levels of eating disorder symptoms find it difficult to identify facial expressions that indicate anger, and relatively easy to identify those indicating surprise. Anger has not previously been investigated in this group as a discrete emotion in paradigms similar to that used here, but the current findings suggest that it is worthy of further attention. It has been suggested that anger is an understudied emotion in those with eating disorders (Waller et al., 2003). One possibility is that our finding reflects lack of attention to, or “neglect” of anger, perhaps due to fear of social rejection. Kessler et al. (2006) found a trend for surprise to be less rapidly recognised by patients with AN; one possibility is that increased time spent identifying surprise may result in better discrimination. Although we did not find increased reaction time for surprise identification here, surprise may also be an emotion worthy of further study in this group. Clearly, the finding for anger, indicative only of a trend here, needs to be interpreted with caution, and requires replication in a larger study, and further investigation. Theoretically, the findings support the inclusion of emotional processing difficulties in models of eating disorders. The use of an objective measure of emotional processing has facilitated a more detailed analysis of the specific deficits than would have been possible using the TAS. Interestingly, while high EAT scorers, compared to low EAT scorers show some “deficits”, they also show some strengths or enhanced processing abilities. These may, of course, also be related to the development of psychopathology. Specific focus on emotional processing per se is not typically part of cognitive behavioural therapy (CBT) for eating disorders. Nevertheless CBT may need to teach those with disturbed eating how to accurately decode facial expressions, before moving on to help them re-assess the meaning of what people say or do, including for example, by modifying biases in the processing of emotional information. The use of an objective measure, such as the FERT, when compared to existing questionnaires, enables us to
396
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
understand the precise nature of the difficulties that those with eating disorder related symptoms may have. This may enable treatment of those with eating disorders to target specific difficulties more precisely. The study has several limitations. We studied a non-clinical group, thus it is important to investigate in future work whether these patterns of emotional processing also characterise those with diagnosed eating disorders. Nevertheless, the mean score of our high EAT group was above the clinical cut off, and there is some evidence that our screening tool, the EAT, performs well in identifying partial syndrome eating disorders (Scheinberg et al., 1993), particularly those with sub-clinical AN. The sample size was relatively small, and it is important that the current findings are replicated. A larger sample might also permit statistical analysis of error rates. The sample was self selecting, i.e. compromised only of those who completed the screening tool, thus potentially not representative of all those in the “at risk” category. Reliance on self report, for example, for report of height and weight, is also a limitation, as self report of these data may not be accurate. The EAT has not often been investigated for it's ability to identify BN, or indeed those who present to services with an ED-NOS diagnosis, thus it is particularly important to determine whether the current findings also apply to these patient groups. The decision to investigate the role of mood by controlling statistically for HADS depression and HADS anxiety score was a difficult one. It is possible that, in the case of eating disorder related psychopathology, such an analysis may control for symptoms that are a key aspect of the eating related disturbance. Nevertheless, the key differences between the high and low groups largely remained significant when these analyses were conducted. While it is possible that the HADS is a relatively poor measure of mood in those with eating disorder symptoms, this seems unlikely given its extensive psychometric assessment. We did not have full information on any psychological treatment received by participants. While we would not expect this to create deficits in emotional processing it is possible that those in the high EAT group who had received some psychological treatment performed better than those who had not. This requires further investigation. Overall, the findings suggest that the role of mood in relation to eating disorder related psychopathology is complex, and requires further examination. One possibility, which is consistent with our clinical observations, is that the negative emotional experiences and negative sense of self seen in patients with eating disorders are an intrinsic part of the condition and not attributable solely to concomitant depressive symptoms. In particular, our impression is that “depressed mood” in those with eating disorders may reflect an extreme and intense emotional and cognitive state not readily captured by the qualitative experience of depression alone, and also not captured by current assessment measures. This may help explain a pattern of results in those “at risk” which is similar to that found in depression, but which appears not to be attributable to depressed mood as assessed on standardised measures. Clearly, this requires further investigation. Overall, the findings indicate that emotional processing may be an important topic for further investigation in relation to eating disorder symptoms. The FERT would appear to be a useful tool to study emotional processing, with particular promise in identifying disturbances in the processing of discrete emotions in those “at risk” or with sub-clinical eating disturbance. References Abraham, S. F., & Beumont, P. J. V. (1982). How patients describe bulimia or binge eating. Psychological Medicine, 12, 625−635. American Psychiatric Association (1994). Diagnostic and Statistical Manual of Mental Disorders, 4th edition revised. Washington, DC: Author. Beales, D. L., & Dolton, R. (2000). Eating disordered patients: Personality, alexithymia, and implications for primary care. British Journal of General Practice, 50, 21−26. Bhagwagar, Z., Cowen, P. J., Goodwin, G. M., & Harmer, C. J. (2003). Normalisation of enhanced fear recognition by acute SSRI treatment in subjects with a previous history of depression. American Journal of Psychiatry, 161, 166−168. Bowlby, J. (1969). Attachment and Loss, Vol 1: Attachment. New York: Basic Books. Bruch, H. (1973). Eating Disorders. New York: Basic Books. Bydlowski, S., Corcos, M., Jeammet, P., Paterniti, S., Berthoz, S., Laurier, C., et al. (2005). Emotion-processing deficits in eating disorders. International Journal of Eating Disorders, 37, 321−329. Connan, F., Campbell, I. C., Katzman, M., Lightman, S. L., & Treasure, J. (2003). A neurodevelopmental model for anorexia nervosa. Physiology and Behavior, 79, 13−24. Cooper, M. J. (2003). Bulimia nervosa: A cognitive perspective. Oxford: Oxford University Press. Cooper, M. J. (2005). A vulnerability stress model of eating disorders: A cognitive perspective. In B. L. Hankin, & J. Abela (Eds.), Development of Psychopathology: A Vulnerability Stress Perspective Thousand Oaks, CA: Sage. Cooper, M., Wells, A., & Todd, G. (2004). A cognitive model of BN. British Journal of Clinical Psychology, 43, 1−16. Crawford, J. R., Henry, J. D., Crombie, C., & Taylor, E. P. (2001). Normative data for the HADS from a large non-clinical sample. British Journal of Clinical Psychology, 40, 429−434. Ekman, P., & Friesen, W. V. (1976). Pictures of facial affect. Palo Alto, Calif: Consulting Psychologists Press. Fairburn, C. G., & Harrison, P. J. (2003). Eating disorders. The Lancet, 361, 407−416. Fairburn, C. G., Cooper, Z., & Shafran, R. (2003). Cognitive behaviour therapy for eating disorders: A “transdiagnostic” theory and treatment. Behaviour Research and Therapy, 41, 509−528. Fonagy, P., Gergely, G., Jurist, E. L., & Target, M. (2002). Affect regulation, mentalisation and the development of the self. London: Karnac. Garner, D. M., Olmsted, M. P., Bohr, Y., & Garfinkel, P. (1982). The Eating Attitudes Test: Psychometric features and clinical correlates. Psychological Medicine, 12, 871−878. Gilboa-Schechtman, E., Avnon, L., Zubery, E., & Jeczmien, P. (2006). Emotional processing in eating disorders: Specific impairment or general distress related deficiency? Depression and Anxiety, 23, 331−339. Guidano, V. F., & Liotti, G. (1983). Cognitive Processes and Emotional Disorders: A Structured Approach to Psychotherapy. New York: Guilford Press. Harmer, C. J., Bhagwagar, Z., Perrett, D. I., Vollm, B. A., Cowen, P. J., & Goodwin, G. M. (2003). Acute SSRI administration affects the processing of social cues in healthy volunteers. Neuropsychopharmacology, 28, 148−152. Harmer, C. J., Perrett, D. J., Cowen, P. J., & Goodwin, G. M. (2001). Administration of the beta-adrenoreceptor blocker propranolol impairs the processing of facial expressions of sadness. Psychopharmacology, 154, 383−389. Hayward, G., Goodwin, G., Cowen, P., & Harmer, C. (2005). Low-dose tryptophan depletion in recovered depressed patients induces changes in cognitive processing without depressive symptoms. Biological Psychiatry, 57, 517−524. Hinrichsen, H., Wright, F., Waller, G., & Meyer, C. (2003). Social anxiety and coping strategies in the eating disorders. Eating Behaviors, 4, 117−126. Kessler, H., Schwarze, M., Filipic, Traue, H. C., & von Wietersheim, J. (2006). Alexithymia and facial emotion recognition in patients with eating disorders. International Journal of Eating Disorders, 39, 245−251. Kiyotaki, Y., & Yokoyama, K. (2006). Relationships of eating disturbances to alexithymia, need for social approval, and gender identity among Japanese female undergraduate students. Personality and Individual Differences, 41, 609−618.
L. Jones et al. / Eating Behaviors 9 (2008) 389–397
397
Kucharska-Pietura, K., Nikolaou, V., Masiak, M., & Treasure, J. (2004). The recognition of emotion in the faces and voice of anorexia nervosa. International Journal of Eating Disorders, 35, 42−47. Leppänen, J., Milders, M., Bell, J., Terriere, E., & Hietanen, J. (2004). Depression biases the recognition of emotionally neutral faces. Psychiatry Research, 128, 123−133. Matsumoto, D., & Ekman, P. (1998). Japanese and Caucasian Facial Expressions of Emotion and Neutral Faces. San Francisco: San Francisco State University. Mendlewicz, L., Linkowski, P., Bazelmans, C., & Philippot, P. (2005). Decoding emotional facial expressions in depressed and anorexic patients. Journal of Affective Disorders, 89, 195−199. Quinton, S., & Wagner, H. L. (2005). Alexithymia, ambivalence over emotional expression, and eating attitudes. Personality and Individual Differences, 38, 1163−1173. Ratnasuriya, R. H., Eisler, I., Szmuckler, G. I., & Russell, G. F. M. (1991). Anorexia nervosa: Outcome and prognostic factors after 20 years. British Journal of Psychiatry, 158, 495−502. Scheinberg, Z., Koslowsky, M., Bleich, A., Mark, M., Apter, A., Danon, Y., et al. (1993). Sensitivity, specificity, and positive predictive value as measures of prediction accuracy: The case of the EAT-26. Educational and Psychological Measurement, 53, 831−839. Schmidt, U., & Treasure, J. (2006). Anorexia nervosa: Valuable and visible. A cognitive–interpersonal maintenance model and its implications for research and practice. British Journal of Clinical Psychology, 45, 343−366. Surguladze, S. A., Young, A. W., Senior, C., Brebion, G., Travis, M. J., & Phillips, M. L. (2004). Recognition accuracy and response bias to happy and sad facial expressions in patients with major depression. Neuropsychology, 18, 212−218. Tabachnick, B. G., & Fidell, L. S. (1989). Using Multivariate Statistics. New York: Harper Collins. Taylor, G. J., Ryan, D., & Bagby, R. M. (1985). Toward the development of a new self-report alexithymia scale. Psychotherapy and Psychosomatics, 44, 191−199. Thompson, M. G., & Schwartz, D. M. (1982). Life-adjustment of women with anorexia nervosa and anorexic-like behaviour. International Journal of Eating Disorders, 1, 47−60. Waller, G., Babbs, M., Milligan, R., Meyer, C., Ohanian, V., & Leung, N. (2003). Anger and core beliefs in the eating disorders. International Journal of Eating Disorders, 34, 118−124. Yager, J., Landsverk, J., & Edelstein, C. K. (1987). A 20 month follow up study of 628 women with eating disorders, I: Course and severity. American Journal of Psychiatry, 144, 1172−1177. Zigmond, A. S., & Snaith, R. P. (1983). The hospital anxiety and depression scale. Acta Psychiatrica Scandinavica, 67, 361−370. Zonnevylle-Bender, M. J. S., van Goozen, S. H. M., Cohen-Kettenis, P. T., van Elburg, A., & van Engeland, H. (2002). Do adolescent anorexia nervosa patients have deficits in emotional functioning? European Child and Adolescent Psychiatry, 11, 38−42.