- Email: [email protected]
Thoughts about eating, weight and shape in anorexia nervosa and bulimia nervosa
00057967/92 $5.00 + 0.00 Copyright 0 1992 Pergamon Press Ltd
Behaa. Res. The-r. Vol. 30, No. 5, pp. 501-51 I, 1992 Printed in Great Britain. All rights reserved
WEIGHT AND SHAPE THOUGHTS ABOUT EATING, ANOREXIA NERVOSA AND BULIMIA NERVOSA MYRA Department
of Psychiatry,
J. COOPER* Oxford
and
University,
G.
CHRISTOPHER Warneford
Hospital,
IN
FAIRBURN Oxford
OX3 7JX, England
(Received 24 October 1991) Summary-Concurrent verbalisation and a self-report questionnaire were used to investigateself-statements in patients with anorexia nervosa, patients with bulimia nervosa, two groups of dieters and non-dieting controls. Thoughts were collected while subjects performed three behavioural tasks, looking at themselves in a full-length mirror, weighing themselves and eating a chocolate covered mint. Both groups of patients had more negative thoughts related to eating, weight and shape than those in the three control groups. In addition, patients with anorexia nervosa showed a greater concern with eating while patients with bulimia nervosa showed a greater concern with weight and appearance. Differences were found between the patients and non-dieting controls using both methods but the self-report questionnaire was less sensitive than concurrent verbalisation to differences between the patients and dieters. Implications of the findings for cognitive-behavioural treatments of the two disorders are discussed.
Cognitive theories assign thoughts about eating, weight and shape a causal role in the aetiology of anorexia nervosa (Garner & Bemis, 1982) and of bulimia nervosa (Fairburn, Cooper & Cooper, 1986). To date few studies have investigated whether these disorders are indeed characterised by the type of cognitive disturbance implied by the theories. Clinical and anecdotal reports (e.g. Bruch, 1973; Russell, 1979) are consistent with the theories but are unsatisfactory because they lack control groups and are usually retrospective. In so far as cognitive-behavioural treatments are effective (for a review see Fairburn, Agras & Wilson, 1992) they provide some support for a cognitive disturbance, but they do not tell us precisely what this is. With the exception of four semi-structured interview-based studies of the cognitive characteristics of these patients (Cooper & Fairburn, 1987; Cooper, Cooper & Fairburn, 1989; Wilson & Smith, 1989; Rosen, Vara, Wendt & Leitenburg, 1990) all existing research studies (e.g. Clark, Feldman & Channon, 1989; Phelan, 1987) have employed self-report questionnaires. These require endorsement of predetermined items and so are only able to capture those thoughts which the investigator has decided are of interest. Most studies have used normal young women as their control group. However, those who diet are an important comparison group since they may show some of the same characteristics (Wilson, 1989). Finally, with the exception of one study (Clark et al., 1989), existing studies have only measured frequency of thoughts and not other response dimensions, for example, duration and belief, and have investigated only those with bulimia nervosa and not also those with anorexia nervosa. The present study was designed to address these issues and extend the work on the cognitive disturbance characteristic of eating disorders. To investigate specificity, two groups of dieters were included in addition to Ss with anorexia nervosa and Ss with bulimia nervosa. To detect thoughts not included on self-report questionnaires, concurrent verbalisation was used. It was predicted that patients with eating disorders would have more thoughts about eating, weight and shape and, in particular, more negative thoughts about these issues than a group of non-dieting controls, and that dieters would occupy an intermediate position, sharing some of the same thoughts as the patients and some with the non-dieting controls. METHOD Design Patients with anorexia nervosa, patients with bulimia nervosa, two groups of dieters and a non-dieting control group were investigated. All Ss were female. Each S completed three tasks *Author
for correspondence. 501
502
MYRA J. C&PER and CHRISTOPHER G. FAIRBURN
devised to elicit thoughts relevant to eating, weight and shape. These were weighing eating a chocolate mint and looking at themselves in a full-length mirror.
themselves,
Subjects Patients with anorexia nervosa. These were 12 patients who fulfilled DSM-III-R criteria for 1987). Nine were recruited through their anorexia nervosa (American Psychiatric Association, therapist and three were recruited through a local self-help group for eating disorders. Patients with bulimia nervosa. These were 12 patients who fulfilled DSM-III-R criteria for bulimia nervosa (American Psychiatric Association, 1987). All patients also fulfilled a strict operational definition of bulimia nervosa based on these criteria (Fairburn, 1987). All were recruited through their therapist. Normal dieters. These Ss were 12 volunteers who fulfilled a strict operational definition of dieting. They were recruited by placing posters in two local Universities asking for female volunteers, under the age of 35, who had been making an attempt to lose weight for at least 4 weeks. They were invited to take part in a study investigating ‘thoughts about eating, weight and shape’. No reference was made to eating disorders. The age limit was specified in order to obtain a group similar in age to the two patient groups. To be included the volunteers had to have been making a serious attempt to lose weight over the preceding 4 weeks. The diet followed and their success in adhering to it were irrelevant. A serious attempt was defined as following a standard reducing diet and/or the setting of predetermined and rigid rules such as a definite calorie limit, preset quantities of food or rules about what should be eaten. Ss were excluded if they had a psychiatric history, if they had ever met DSM-III-R criteria for anorexia nervosa or bulimia nervosa, or if they had shown any of the core behavioural features of either eating disorder, i.e. objective bulimic episodes, actual or attempted self-induced vomiting to lose weight, weight maintenance below 85% mean population matched weight. Symptomatic dieters. These were 12 volunteers, recruited in exactly the same way as the normal dieters, who fulfilled the same criteria for inclusion and exclusion as this group but who, in addition, had either currently or in the past experienced some of the behavioural features of anorexia nervosa or bulimia nervosa as previously described above. Reliability of allocating the dieters to the two groups was determined by an independent judge who was provided with details of Ss’ eating habits and behaviours. Complete agreement was reached in all 24 cases. Normal controls. These were 12 volunteers recruited in the same way as the normal dieters and who fulfilled the same exclusion criteria but who had not been attempting to lose weight over the previous 4 weeks. Measures Information was obtained on demographic and background features, including age, years in full-time education, weight and height. Each S also completed the Eating Attitudes Test (EAT; Garner & Garfinkel, 1979) and the Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock & Erbaugh, 1961), self-report measures of the features of eating disorders and depression respectively. Ss’ self-statements were recorded during the three separate behavioural tasks using two techniques, concurrent verbalisation (or ‘thinking aloud’) and a self-report questionnaire (or Thoughts Checklist) devised specifically for this study. Materials The Thoughts Checklist was made up of self-statements chosen to be typical of the thoughts that patients with eating disorders might be expected to have while carrying out each of the three behavioural tasks. Items were taken from a pool of 133 thoughts (44 for the weighing task, 48 for the mirror task, 41 for the eating task) generated by asking two clinicians (one clinical psychologist and one psychiatrist), as well as the experimenter, to make three lists, one for each task, “of the thoughts which you think might occur to someone with an eating disorder while doing each of the three tasks”. The thoughts obtained were reduced to 15 negative thoughts for the eating task and 16 each for the weighing and mirror tasks. In addition, a task-specific positive thought was added to the weighing and mirror tasks. Examples of thoughts are “I’m getting fatter and fatter” (for the weighing task) and “I’ve no self-control” (for the eating
Thoughts
about
eating,
weight
and shape
in AN and BN
503
task).* The Thoughts Checklist was used to yield several scores for the negative thoughts and, where relevant, for the positive thoughts. These were an overall score (made up of absolute frequency x duration of the thoughts), absolute frequency of thoughts, mean duration of thoughts, and a score of mean degree of belief in the thoughts. Concurrent oerbalisation Breaking transcripts into units. Thoughts obtained using the ‘thinking aloud’ method were transcribed and broken up into units for classification using an approach similar to that employed by Davison, Robins and Johnson (1983). An independent rater (rater 2) broke ten randomly selected transcripts (17% of the total) into units. Reliability of this method was calculated in two different ways. In the first overall agreement on position of end boundaries was calculated by noting the number of end boundaries on which both raters agreed and dividing this by the larger number of boundaries used by either rater. In the second, since rater 1 divided the rest of the data into units, rater 2s agreement with rater 1 was calculated by dividing the number of end boundaries agreed on by the total number used by rater 2. Mean percentage agreement between the two raters was 91.7 and 93.5% respectively for the two methods. Categorising idea units. A manual was written with detailed instructions and examples of how to code the idea units. These were coded using two different methods.
(1) Immediately
relevant negative and positive thoughts. In this coding each task was taken individually and the number of negative and positive thoughts directly and immediately related to each task was calculated (i.e. number of negative and positive thoughts concerned with weighing in the weighing task, number of negative and positive thoughts concerned with eating in the eating task, and number of negative and positive thoughts concerned with looking in the mirror in the mirror task). Past associations and memories were excluded. (4 Content, reference and valence. Using this method, thoughts in each of the three tasks were coded along three dimensions. (4 Content, in which thoughts were divided into those relating to food, eating, weight and shape (target thoughts) and those relating to other issues (non-target thoughts). Unlike the first method, thoughts could include memories, associations and thoughts about all food, eating, weight and shape issues. (b) Reference, in which thoughts were divided into those relating to the self (self-referent) and those relating to other issues (other-referent). Valence, in which thoughts were classified as positive, negative or neutral. (cl
Reliability of coding was calculated separately for each category using the two methods above. Mean percentage agreement between two raters ranged from 84.4 to 98.7%.
outlined
Procedure A standard procedure was used. On arrival the procedure was explained to Ss. Criteria for inclusion and exclusion were established, demographic data was collected and Ss completed the self-report questionnaires. Instructions were then given to verbalise all thoughts that came to mind during each of the three behavioural tasks. Ss first completed a practice task, sorting pictures of faces into order of age, to familiarise them with ‘thinking aloud’ and with the instructions that were to be given for each of the three main tasks. The three behavioural tasks, weighing, looking in a full-length mirror and eating an ‘After Eight’ (a chocolate covered mint), were then carried out. There were six possible orders in which the three tasks could be presented. To control for possible order effects two Ss in each group completed the tasks in each order. Ss were left alone to complete each task with signals being given at two points, the first to indicate that the S could begin the task and the second to indicate that the S should stop doing the task. Thoughts were thus obtained for three periods, before, during and after the task. The procedure was exactly as it had been for the practice task except that Ss were given twice as long (5 min) to say their thoughts out loud, i.e. 14min before the task, 2 min during the task and lf min after the task.
*Copies
of the Thoughts
Checklist
can be obtained
from the authors.
MYRA f. CAPER and CHRISTOPHER G. FAIRBURN
504
In the weighing task Ss were asked to stand on a set of weighing scales, find out exactly how much they weighed, and remain on the scales until a signal came from the experimenter to stop. In the mirror task Ss were asked to look carefully at themselves in a full-length mirror, starting with a side view, and turning slowly round to look at themselves from different angles. The experimenter demonstrated what was to be done. In the eating task Ss were asked to eat a chocolate covered mint, a ‘forbidden food’ for many patients with eating disorders and for many dieters. After carrying out the three tasks Ss completed the Thoughts Checklist and rated how typical their thoughts had been of the thoughts that they might expect to have when doing each of the tasks in everyday life. A O-10 scale was used with 0 being ‘not at all typical’ and 10 being ‘completely typical’. Finally weight and height were measured and Ss were debriefed. RESULTS
Demographic features
Information on age, years of full-time education, Body Mass Index (BMI) and scores on the EAT and BDI are presented in Table 1. (BMI = weight in kg/height in m2.) Ss in the different groups were of a similar age. They differed in number of years in full-time education with the two patient groups having received significantly fewer years of full-time education than the other three groups. All were of a similar BMI, except for the patients with anorexia nervosa whose BMI was, as expected, lower than that of the other four groups. On the EAT the normal control group had a significantly lower score than the symptomatic dieters, patients with bulimia nervosa and patients with anorexia nervosa, while the normal dieters had a si~ificantly lower score than the two patient groups. Ss in both patient groups had higher scores on the BDI than those in the other three groups. Both dieting groups had been attempting to lose weight for equivalent lengths of time [11.6 + 10.1 weeks in the normal dieter group and 12.7 f 14.5 weeks in the symptomatic dieter group, t(21) = 0.6, NS]. In the symptomatic dieter group seven Ss reported objective bulimic episodes, nine reported attempted or actual self-induced vomiting to lose weight and one had briefly reached a weight below 85% mean population matched weight by dieting. In the anorexia nervosa group most patients were suffering from ‘restricting’ anorexia nervosa, only one reported episodes of overeating followed by self-induced vomiting. There were no differences between the five groups in the typicality ratings made for each task (range 6.0-8.5). This indicated that the thoughts obtained were equally typical of thoughts that might be expected in real life situations. Immediately reletlant negatitte and positiue thoughts
A three-way analysis of variance (group x task x type of thought) showed no main effect of group [F(4,55) = 1.42, NS], indicating that the groups did not differ in overall number of thoughts produced, or of task [F&l 10) = 1.04, NS]. There was a main effect of type of thought [F(8,110) = 4.3, P < 0.0002] but this was modified by a significant group x task x type of thought interaction [r”( 16,220) = 2.10, P < 0.021. To locate the source of the difference found, two separate two-way analyses of variance (group x task) were performed, one with the negative thoughts as the dependent variable and the other with the positive thoughts as the dependent variable. Since Table 1. Demographic NC 4s
( yr)
Education (yr) BMI EAT BDI
X
SD X SD X SD X SD X SD
22.0 2.3 16.la I.5 21.3’ 2.1 6.8”” 4.0 5.7” 4.4
ND 23.5 4.6 15.4 1.2 24.4” 1.8 17.6abc 9.4 7.2’ 5.9
features of the five groups
SD
BN
AN
F/H ratio
P
23.0 3.1 15.8’ 2.4 23.3” 3.0 30.8bc 12.0 9.38 6.5
24.0 4.3 13.3b 1.7 24.8* 5.9 42.3M 11.0 Zl.lb 10.6
21.9 3.2 14.0b 1.2 I5.0b 1.5 53.lM 20.4 25.c 10.4
F = 0.8
NS
F = 6.4
<0.0003
H = 34.1
H = 41.6
<0.0001
I== 12.8
to.0001
All d.f. = 4.55. n = 12 for each group. NC = Normal controls, ND = Normal dieters, SD = Symptomatic dieters, BN = Bulimia nervosa, AN = Anorexia nervosa. BMI = Body mass Index = weight in kg/height in m*. EAT = Eating Attitudes Test. BDI = Beck Depression Inventory. Within each measure means with different superscripts are significantly different from each other.
Table
2. Mean
Task
about
percentage verbalisation
scores for negative and positive thoughts obtained using concurrent for each of the five groups in each of the three tasks
Valence
eating,
NC
weight
and shape
ND Thoughts
Weighing Eating Mirror
Weighing Eating Mirror
505
Thoughts
Negative Positive Negative Positive Negative Positive
10.9’ II.2 8.2” 12.8 16.7” 13.1
19.8’ 3.4 8.61 6.6 21.6’ 6.4
Negative Positive Negative Positive Negative Positive
19.7” 8.4 19.9” II.3 24.0* 10.7
29.0Pd 5.2 18.5” 5.3 29.7” 8.8
in AN and BN
BN
SD immediately
relevant
AN
to task
25.7”’ 3.1 18.4sb Il.3 32.0’ 8.5 All thoughts
46.9b 3.0 33.lk 6.1 50.Sb 4.0
39.Ok 4.0 45. IC 1.9 31.9” 3.6
31.5M 1.1 27.9” II.0 38.6”’ 10.3
56.7’ 4.1 44.Sbc 6.0 61.3b 4.8
46.1cd 4.8 56.3b 5.1 41.4c 9.9
n = I2 for each group. NC = Normal controls, ND = Normal dieters, SD = Symptomatic dieters, BN = Bulimia nervosa, AN = Anorexia nervosa. Within each measure means with different superscripts are significantly different from each other.
the groups did not differ in overall number of thoughts for the three tasks, to assess the extent to which they differed in emotional content of the thoughts produced, the percent negative and the percent positive scores were compared. This took into account the large differences between groups in the number of neutral thoughts reported in each task. For percentage of negative thoughts, the two-way analysis of variance showed a main effect of group [F(4,55) = 13.06, P < O.OOOl]and of task [F(2,110) = 7.08, P < 0.0021. However, these main effects were modified by a group x task interaction [F(8,110) = 3.39, P c 0.0021. Mean percentage scores of immediately relevant negative thoughts for each of the five groups in each of the three taks are shown in Table 2. To locate the source of the interaction post hoc Tukey Honestly Significant Difference tests were performed. As predicted, in the between group comparisons, taking each task separately, the two patient groups generally had more negative thoughts directly and immediately related to each of the three tasks than the non-dieting female controls. Patients in both groups had a greater percentage of negative thoughts in the weighing and eating tasks than the non-dieting controls (all P values ~0.01). In addition, patients with bulimia nervosa, but not those with anorexia nervosa, had a greater percentage of negative thoughts in the mirror task than the normal controls (P < 0.01). There were no significant differences between the two dieting groups and the normal controls in any of the three tasks and several of the comparisons between the two dieting groups and the two patient groups were also not significant. These were comparisons between the symptomatic dieters and the patients with anorexia nervosa in the weighing task, between the symptomatic dieters and the patients with bulimia nervosa in the eating task, and between the two dieting groups and the patients with anorexia nervosa in the mirror task. This suggested that, as predicted, the two dieting groups occupied a position intermediate between that of the normal controls and the two patient groups. When the two dieting groups were compared with the two patient groups it emerged that those with bulimia nervosa had a greater percentage of negative thoughts immediately relevant to the weighing and mirror tasks while those with anorexia nervosa had a greater percentage of negative thoughts immediately relevant to the eating task (all P values ~0.01). Patients with bulimia nervosa also had a greater percentage of negative thoughts immediately relevant to the eating task than the normal dieters and a greater percentage of negative thoughts in the mirror task than those with anorexia nervosa (both P values ~0.01). Similar findings emerged from the within group comparisons. Patients with bulimia nervosa had a greater percentage of negative thoughts in the mirror (P < 0.01) and weighing (P < 0.05) tasks than in the eating task. In this they showed a pattern similar to that of the two dieting groups, both of which had a greater percentage of negative thoughts in the mirror task than in the eating task (both P values <0.05). However, in the patients with anorexia nervosa the reverse pattern was observed. This group had a greater percentage of negative thoughts in the eating task than in the mirror task (P < 0.05). There were no differences between tasks in the normal control group.
MYRA J. CARPERand CHRISTOPHER G. FAIRBURN
506
Percentage of positive thoughts was also subjected to a two-way analysis of variance (group x task). There was a main effect of group [F(4,55) = 4.47, P < 0.004)] and of task [F(2,110) = 8.71, P < 0.0004] but no significant group x task interaction [F(8,110) = 1.53, NS]. When subjected to post hoc tests the main effect of group indicated that the normal control Ss did not report a greater percentage of positive thoughts immediately relevant to the three tasks than the two patient groups. Mean percentage scores for the five groups in each of the three tasks can be seen in Table 2. Content,
reference
and valence
a three-way analysis of variance On the target vs non-target thoughts dimension (group x task x target vs non-target thoughts) showed that there was no main effect of group [F(4,55) = 1.25, NS]. This indicated that the groups did not differ in overall number of thoughts reported. There was no main effect of task [F(2,110) < l] but there was a main effect of type of target thought [F(4,55) = 155.45, P < O.OOOl]. There was no interaction between group, task and type of thought [F(8,110) < I]. The main effect of type of target thought indicated that all groups reported more thoughts about eating, weight and shape related issues than about other issues. Mean scores for the five groups in each of the three tasks can be seen in Table 3. While the groups did not differ in type of thought reported in the three tasks, as we shall see below, the groups did differ in valence or emotional content of the thoughts produced. With regard to reference, a three-way analysis of variance (group x task x self-referent vs other referent) showed no main effect of group [F(4,55) = 1.54, NS], again indicating that the groups did not differ in overall number of thoughts produced. There was no main effect of task [F(2,110) = 1.19, NS] but there was a main effect of reference [F(4,55) = 74.20, P < O.OOOl]. There was no group x task x type of thought interaction [F(8,110) = 1.88, NS]. The main effect of reference indicated that all groups had more self-referent thoughts than other-referent thoughts. Mean scores for the five groups in each of the three tasks can be seen in Table 3. With regard to valence, a three-way analysis of variance (group x task x negative vs positive vs neutral thoughts), showed no main effect of group [F(4,55) = 1.33, NS]. As before, this indicated that the groups did not differ in overall number of thoughts produced in the three tasks. There was no main effect of task [F(2,110) < 1] but there was a main effect of valence [F(8,110) = 74.84, P < O.OOOl] and a valence x group interaction [F(8,110) = 6.34, P < O.OOl]. However, these findings were modified by a group x task x type of thought interaction [F(l6,220) = 1.73, P < 0.051. To locate the source of the interaction, two separate two way analyses of variance (group x task) were carried out, one using the negative thoughts as the dependent variable and one using the positive thoughts as the dependent variable. Since there were no differences between groups in overall number of thoughts the extent to which the groups differed in emotional content of the thoughts was assessed by comparing the percent negative and the percent positive scores of each group. With negative thoughts as the dependent variable, the two way analysis of variance yielded a main effect of group [F(4,55) = 16.84) P < O.OOOl] and of task [F(2,110) = 3.44, P < 0.041. However, these main effects were modified by an interaction between group and task Table 3. Mean raw scores for target and self-referent tmn for each of the five groups Task
Thought
Weighing
Target Non-target Target Non-target Target Non-target Self-referent Other-referent Self-referent Other-referent Self-referent Other-referent
Eating Mirror Weighing Eating Mirror
type
thoughts obtained using concurrent in each of the three tasks
verbalisa-
NC
ND
SD
BN
AN
37.3 13.1 32.8 17.1 37.7 14.1 30.4 19.9 22.3 27.7 35.6 16.2
37.7 6.2 32.7 II.3 40.3 7.0 30. I 13.8 22.3 18.3 34.4 12.9
49.6 10.2 50.3 9.8 50.4 1.2 38.0 21.9 31.7 22.8 45.6 12.8
38.6 4.7 31.2 8.4 35.3 4.6 30.4 6.8 30.3 9.3 33.5 6.4
42.9 6.8 39.3 6.6 48.6 5.8 33.3 15.8 33.8 12.2 44.9 7.9
n = I2 for each group. NC = Normal controls, ND = Normal dieters, SD = Symptomatic dieters, EN = Bulimia nervosa, AN = Anorexia nervosa. Within each measure means with different superscripts are significantly different from each other.
Thoughts about eating, weight and shape in AN and BN
507
[F(8,110) = 3.42, P < 0.0021. Mean percentage scores for negative thoughts for the five groups in each of the three tasks are shown in Table 2. Post hoc Tukey Honestly Significant Difference tests were used to locate the source of the interaction. As predicted, between group comparisons, taking each task separately, showed that the two patient groups had more negative thoughts in each of the three tasks thaz the non-dieting female controls (all P values ~0.01). There were no significant differences between the two dieting groups and the normal controls in most tasks, the one exception being the comparison between the normal controls and the symptomatic diet group in the weighing task (P < 0.05), and several comparisons between the two dieting groups and the two patient groups and the patients with anorexia nervosa in the weighing and mirror tasks and between the symptomatic dieters and the patients with bulimia nervosa in the eating task. This suggested that, as predicted, the two dieting groups occupied a position intermediate between the normal controls and the two patient groups. When the two dieting groups were compared with the two patient groups a pattern of results similar to that for immediately relevant negative thoughts was found, i.e. patient with bulimia nervosa had more negative thoughts in the weighing and mirror task while those with anorexia nervosa had more negative thoughts in the eating task (all P values co.01 except for the comparison between patients with bulimia nervosa and symptomatic dieters on the weighing task P < 0.05). In addition patients with bulimia nervosa had more negative thoughts in the eating task than the normal dieters and more negative thoughts in the mirror task than those with anorexia nervosa (both P values
Checklist
When overall negative score was subjected to a two-way (group x task) analysis of variance there was a main effect of group [F(4,55) = 8.57, P < O.OOl] and of task [F(2,110) = 28.37, P < O.OOl]. However, these main effects were modified by a significant group x task interaction [F(8,110) = 3.53, P < 0.0021. Mean overall negative scores on the Thoughts Checklist for the five groups in each of the three tasks are shown in Table 4. As predicted, between group comparisons, taking each task separately, showed that the two patient groups had a higher overall negative score in most of the three tasks than the non-dieting controls. The one exception was for the comparison between the normal controls and the patients with anorexia nervosa in the mirror task. All comparisons were significant at the level P < 0.01, except for that between the normal controls and the anorexia nervosa patients in the weighing task (P < 0.05). There were few significant differences between the two dieting groups and the normal controls in most tasks, the exceptions being the comparisons between the symptomatic dieter group and the normal controls in the weighing task (P < 0.01) and in the mirror task (P < 0.05). Several of the comparisons between the two dieting groups and the two patient groups were also not significant. These were comparisons between the patients with anorexia nervosa and the two diet groups in the weighing and mirror tasks, the comparison between the patients with anorexia nervosa and the symptomatic dieters in the eating task and between the patients with bulimia nervosa and the symptomatic dieters in all three tasks. This again suggested that, as predicted, the two dieting groups appeared to occupy a position between that of the normal controls and the two patient groups. When the two dieting groups were compared with the two patient groups the
508
MYRA J. OXPER and CHRISTOPHER G. FAIRBURN Table 4. Mean scores on the Thouehts Task
NC
Checklist
for each of the five crams
ND
SD Frequency
Weighing Eating Mirror
20.0” 16.4” 18.9”
Weighmg Eating Mirror
2.5 0.9 2.1
33.3”’ 20.1” 26.7” Absolute
Duration Weighing Eating Mirror
Weighing Eating Mirror
2.7” 2.5’ 2.4”
78.5 89.3 66.2
3.9b 2.6” 3.4”b
Weighing Mirror
4.2 4.3
69.2 65.4 59.4
81.0” 80.0”
2.6 4.0
32.0b 60.0”
9.2 6.6 8.0 of negative
58.9 49.0 54.1 of positive
2.3 3.2 Belief m posltlve 51.7dh 68.3”
AN
thoughts 49.Q 36.7b 40.9b thoughts
I I.2
36.3k 35.9b 27.Bab
8.8 9.4
7.5 7.0 6.2
4.0b 3.2ab 3.5b
3.P 4.0b 3.0ab
thoughts
3.gh 3.0db 3.2” Belief in negative thoughts*
Frequency Weighing Mirror
of negative
40.9k 28.7*b 34.5h number of negative
6.2 3.2 4.9
in each of the three tasks BN
69.7 58.9 59.6
71.5 76.7 64.8
3.7 3.0
3.8 3.4
63.36b 40.0”
46.0db 48.0”
thoughtst
thought&
n = 12 for each group. *n = 7 for NC, n = I I for ND, SD. BN and n = IO for AN. tn = IO for NC, n = 5 for ND, AN, n = 6 for SD and n = 3 for BN. NC = Normal controls, ND = Normal dieters. SD = Symptomatic dieters, BN = Bulimia nervosa, AN = Anorexia nervosa. Within measures means with different superscripts are significantly different from each other.
pattern of differences between groups was less clear than that observed for immediately relevant negative thoughts. As before, patients with bulimia nervosa had a greater frequency of negative thoughts than the normal dieters in all three taks (all P values ~0.05) while patients with anorexia nervosa had a greater frequency of negative thoughts than the normal dieters only in the eating task (P < 0.05). Thus, as in the analysis of immediately relevant thoughts, overall negative score successfully distinguished the normal dieters from the two patient groups. However, comparisons between the two patient groups and the symptomatic dieters showed no significant results. Thus, unlike the measure of immediately relevant negative thoughts, overall negative score on the Thoughts Questionnaire did not discriminate between the two patient groups and those who showed some symptoms, but not a diagnosis, of an eating disorder. Comparisons within groups produced results that were similar to those observed for immediately relevant negative thoughts. Patients with bulimia nervosa had a greater score in the weighing than in the eating task (P < 0.01) and in the weighing than in the mirror task (P < 0.01). Patients with anorexia nervosa, however, had greater scores in the eating than in the mirror task (P < 0.05) and in the weighing than in the mirror task (P < 0.05). As found for immediately relevant negative thoughts, the two dieting groups showed a pattern similar to that of the patients with bulimia nervosa. They had a greater score in the weighing than in the eating task (both P values ~0.01) as well as a greater score in the mirror than in the eating task (both P values <0.05). Normal controls again showed no differences in scores obtained in each task. When absolute number of negative thoughts was subjected to a two-way (group x task) analysis of variance there was a main effect of group [F(4,55) = 13.57, P < O.OOOl] and of task [F(2,110) = 17.49, P < O.OOOl] but no group x task interaction [F@,llO) = 1.25, NS]. In order to test the prediction that the two patient groups would have a greater absolute number of negative thoughts than the non-dieting controls post hoc Tukey Honestly Significant Difference tests were carried out on the main effect of group. As predicted, the two patient groups had a greater absolute number of negative thoughts than the non-dieting controls (both P values c 0.01). The prediction that the two dieting groups would occupy a position intermediate between the normal controls and the two patient groups was partially supported. There were no significant differences between the normal dieting group and the non-dieting controls or patients with anorexia nervosa which
Thoughts about eating, weight and shape in AN and BN
509
suggested that they occupied a position intermediate between the normal controls and these patients. Scores of the symptomatic dieters were similar to those of the two patient groups and different from those of the non-dieting controls (P < O.Ol), suggesting that they were most like the two patient groups. Mean scores for absolute number of negative thoughts on the Thoughts Checklist for each of the five groups on the three tasks can be seen in Table 4. With mean duration of negative thoughts as the dependent variable in a two-way (group x task) analysis of variance there was a main effect of group [F(4,45) = 3.11, P < 0.021 and of task [F(2,90) = 10.67, P < O.OOOl]. These effects were modified by a group x task interaction [F(8,90) = 3.61, P < 0.0031. Mean scores for mean duration of negative thoughts on the Thoughts Checklist in each of the five groups on the three tasks can be seen in Table 4. The prediction that negative thoughts would last longer in the two patient groups than in the non-dieting controls (i.e. that the two patient groups would spend longer thinking about the negative thoughts than the non-dieting controls) was partially supported. Between group comparisons showed that, in the weighing and mirror tasks, negative thoughts lasted longer in the patients with bulimia nervosa than they did in the normal controls. In the eating task, negative thoughts lasted longer in the patients with anorexia nervosa than in the normal controls (all P values ~0.05). There were few differences between the two dieting groups and the normal controls, the exceptions being the comparison between the normal controls and the two diet groups in the weighing task (both P values <0.05), and most of the comparisons between the dieting groups and the patients groups were also not significant. These were comparisons between the two patient groups and the two diet groups in the weighing and mirror tasks and between the patients and the normal dieters and between the patients with anorexia nervosa and the symptomatic dieters in the eating task. This suggested that, as predicted, the scores of the two dieting groups fell between those of the normal controls and the two patient groups. When the two dieting groups were compared with the patient groups only one significant finding was observed, the patients with anorexia nervosa spent longer thinking about the negative thoughts in the eating task than the normal dieters (P < 0.05). Thus, although duration of negative thoughts in the three tasks distinguished the normal controls from the patients to some extent, it did not generally distinguish the patients from the two groups of dieters. As with previous measures, the results from the within group comparisons suggested greater concern with weight in the patients with bulimia nervosa and greater concern with eating in the patients with anorexia nervosa. The normal controls spent similar amounts of time thinking about negative thoughts in all three tasks. However, in the patients with bulimia nervosa, negative thoughts in the weighing task lasted longer than they did in the eating task (P < 0.05). The two dieting groups again showed a pattern similar to the patients with bulimia nervosa. In the normal dieters negative thoughts lasted longer in the weighing (P < 0.01) and mirror tasks (P < 0.05) than in the eating task. In the symptomatic dieters negative thoughts in the weighing task lasted longer than negative thoughts in the eating task (P < 0.05). The reverse pattern was observed in the patients with anorexia nervosa where negative thoughts lasted longer in the eating task than in the mirror task (P < 0.01). When mean degree of belief in the negative thoughts was subjected to a two-way (group x task) analysis of variance there was a main effect of group [F(4,45) = 3.00, P < 0.031 and of task [F(2,90) = 4.87, P < 0.011. There was no group x task interaction [F(8,90) = 1.52, NS]. In order to test the prediction that the patient groups would believe more strongly in the negative thoughts reported on the questionnaire than the normal controls post hoc Tukey Honestly Significant Difference tests were performed on the main effect of group. No evidence was found to support this prediction. This suggested that the patients did not believe the negative thoughts to a greater degree than the normal controls. Two measures were obtained from the positive thoughts on the weighing and mirror tasks. These were overall positive score (which was equivalent to mean duration of each positive thought because there was only one positive thought in each of the two tasks) and mean degree of belief in each positive thought. When overall positive score was subjected to a two way (group x task) analysis of variance there was no main effect of group [F(4,24) = 2.43, NS] or of task [F( 1,24) < 11. There was no group x task interaction [F(4,24) = 1.lO, NS]. Thus, as for immediately relevant positive thoughts the normal controls did not have a greater overall positive score than
510
MYKA J. CARPERand CHRISTOPHERG. FAIRBURN
the two patient groups. When mean degree of belief in positive thoughts was subjected to a twoway (group x task) analysis of variance there was a main effect of group [F(4,24) = 3.86, P < 0.021 but no main effect of task [F( 1,24) = 1.29, NS]. However, the main effect of group was modified by a group x task interaction [F(4,24) = 4.06, P < 0.021. Post hoc Tukey Honestly Significant Difference tests were carried out to locate the source of the interaction. When between group comparisons were made the normal controls did not believe the positive thoughts to a greater degree than the two patient groups. Only one finding was significant. This indicated that the normal controls believed the positive weighing thought more than the normal controls (P < 0.01). The within group comparisons showed that all groups, except the normal dieter group, believed equally in the positive thoughts in the weighing and mirror task. The one exception was that the normal dieter group believed the positive mirror thought to a greater degree than the positive weighing thought (P < 0.01). DISCUSSION Using the ‘thinking aloud’ method, patients with eating disorders were found to have more negative thoughts related to eating, weight and shape in each of the three tasks than the non-dieting controls, while the two dieting groups appeared to occupy an intermediate position. Patients with bulimia nervosa were distinguished from dieters, including those with some symptoms of an eating disorder, by a greater number of negative thoughts about weight and appearance. Patients with anorexia nervosa, however, were distinguished from the two dieting groups by a greater number of negative thoughts about eating. Within group comparisons showed a similar concern with weight and appearance in the bulimia nervosa patients and with eating in the anorexia nervosa patients. The two dieting groups showed a pattern similar to that of the patients with bulimia nervosa while normal controls showed no difference in prominence of negative thoughts across the three tasks. Positive thoughts did not distinguish the groups. No differences were found between groups in number of target thoughts, i.e. total number of thoughts related to food, eating, weight and shape. This result may have been due to a ceiling effect since most of the thoughts in all groups were related to these issues. However, an alternative explanation, which receives some support here, is that while all Ss experienced such thoughts when placed in situations concerned with eating, weight and shape what distinguished the groups most clearly was the emotional content of the thoughts experienced. No differences were found between groups for the reference dimension, i.e. self-referent vs other-referent thoughts. This may also have been due to a ceiling effect since the majority of thoughts in all groups were self-referent rather than other-referent. On the Thoughts Checklist findings on all measures, negative and positive, were generally similar to the findings for immediately relevant thoughts. There was, however, one important exception. Unlike the measure of immediately relevant negative thoughts, none of the measures obtained from the questionnaire distinguished the symptomatic dieters from the two patient groups. This suggests that the self-report questionnaire is a less sensitive measure than the concurrent verbalisation method. This could be because retrospective recall of thoughts is less accurate than immediate report. The results also suggest that response dimensions other than frequency, particularly duration of thoughts, are relevant. The results of this study suggest that negative self-statements may be an important focus for cognitive interventions in patients with eating disorders. In particular, the finding that patients with bulimia nervosa are particularly concerned with weight and appearance while patients with anorexia nervosa are more concerned with eating suggests that in treatment it may be desirable to focus on different content when attempting to alter self-statements in these two groups. There has been a tendency amongst clinicians and researchers to emphasise the similarity of the cognitive concerns in the two disorders and in particular to stress the importance of concerns about weight 1986). This emphasis may help to explain why and appearance (e.g. Fairburn & Garner, cognitive-behaviour therapy has been an effective treatment for bulimia nervosa. However, for cognitive-behavioural treatment to be as successful in anorexia nervosa it may be necessary to focus particularly on concerns about eating. The finding that only negative thoughts and not positive thoughts distinguish the groups supports what Kendall (e.g. Kendall & Hollon, 1981) has called It suggests that therapeutic interventions which focus the power of ‘non negative thinking’.
Thoughts
about
eating,
weight
and shape in AN and BN
511
primarily on negative and dysfunctional thinking (e.g. those developed by Ellis, 1962, and Beck, 1976) may be more successful in treating patients with eating disorders than interventions which focus on encouraging positive coping responses (e.g. that developed by Meichenbaum, 1977). Ackmnvledgements-Myra J. Cooper was supported by a Studentship from the Medical Research Council while this research was completed. Christopher G. Fairburn holds a Wellcome Trust Senior Lectureship. We would like to thank David Clark and Paul Salkovskis for their helpful comments on a draft of this manuscript. Correspondence and requests for reprints should be sent to Myra Cooper, University Department of Psychiatry, Warneford Hospital, Oxford OX3 7JX, England.
REFERENCES American Psychiatric Association (1987). DSM-III-R. Washington, D.C.: American Psychiatric Association. Beck, A. T. (1976). Cognit& therapy and the emotional disorders. Madison, Conn.: International Univerisities Press. Beck, A. T., Ward, C. H., Mendelson, M., Mock, J. & Erbaugh, J. (1961). An inventory for measuring depression. Archiues of General Psychiatry, 4, 53363. Bruch, H. (1973). Eating disorders: Obesity, anorexia nercosa and /he person within. New York: Basic Books, Clark, D. A., Feldman, J. & Channon, S. (1989). Dysfunctional thinking in anorexia and bulimia nervosa. Cognitive Therapy and Research, 13, 377-387. Cooper, Z. & Fairburn, C. G. (1987). The Eating Disorder Examination: A semi-structured interview for the assessment of the specific psychopathology of eating disorders. International Journal of Eating Disorders, 6, l-8. Cooper, Z., Cooper, P. J. & Fairburn, C. G. (1989). The validity of the Eating Disorder Examination and its subscales. British Journal of Psychiatry. 154. 80778 12. Davison, G. C., Robins, C. & Johnson, M. K. (1983). Articulated thoughts during simulated situations: A paradigm for studying cognition in emotion and behaviour. Cognifive Therapy and Research, 7, 17-40. Ellis, A. (1962). Reason and emofion in ps.ychotherapy. Secaucus, N.J.: Citadel Press, Fairburn, C. G. (1987). The definition of bulimia nervosa. Annals of Behapioural Medicine, 9, 337. Fairburn, C. G. & Garner, D. M. (1986). The diagnosis of bulimia nervosa. International Journal of Baling Disorders, 5, 403-420. Fairburn, C. G., Agras. W. S. & Wilson, G. T. (1992). The research on the treatment of bulimia nervosa: Practical and theoretical implications. In Anderson, G. H. & Kennedy, S. H. (Eds). The biology offeasr andfamine: Relevance to eating disorders. New York: Academic Press, Fairburn, C. G., Cooper, Z. & Cooper, P. J. (1986). The clinical features and maintenance of bulimia nervosa. In Brownell, K. D. & Foreyt, J. P. (Eds). Physiology, psychology and treatment of the eating disorders. New York: Basic books. Garner, D. M. & Bemis, K. M. (1982). A cognitive-behavioural approach to anorexia nervosa. Cognitiue Therapy and Research, 6, 1233150. Garner, D. M. & Garfinkel, P. E. (1979). The Eating Attitudes Test: An index of the symptoms of anorexia nervosa. Psychological Medicine, 9, 2733279. Kendall, P. C. & Hollon, S. D. (1981). Assessing self-referent speech: Methods in the measurement of self-statements. In Kendall, P. C. & Hollon. S. D. (Eds). Assessment Strategies for Cognitive-Beharioural Intert~entions (pp. 85-117). London: Academic Press, Meichenbaum, D. (1977). Cognitioe-behaoiour modification: An integrative approach. New York: Plenum Press. Phelan, P. W. (1987). Cognitive correlates of bulimia: The bulimic thoughts questionnaire. Infernarional Journal of Eating Disorders, 6, 5933607. Rosen, J. C., Vara, L., Wendt, S. & Leitenberg, H. (1990). Validity studies of the Eating Disorder Examination. International Journal of Eating Disorders, 9. 519-528. Russell, G. F. M. (1979). Bulimia nervosa: An ominous variant of anorexia nervosa. Psychological Medicine, 9, 429-488. Wilson, G. T. (1989). Bulimia nervosa: Models, assessment and treatment. Current Opinion in Psychiatry, 2, 790-794. Wilson, G. T. & Smith, D. (1989). Assessment of bulimia nervosa: An evaluation of the Eating Disorder Examination, International Journal of Euring Disorders, 8, 173-l 79.
Behaa. Res. The-r. Vol. 30, No. 5, pp. 501-51 I, 1992 Printed in Great Britain. All rights reserved
WEIGHT AND SHAPE THOUGHTS ABOUT EATING, ANOREXIA NERVOSA AND BULIMIA NERVOSA MYRA Department
of Psychiatry,
J. COOPER* Oxford
and
University,
G.
CHRISTOPHER Warneford
Hospital,
IN
FAIRBURN Oxford
OX3 7JX, England
(Received 24 October 1991) Summary-Concurrent verbalisation and a self-report questionnaire were used to investigateself-statements in patients with anorexia nervosa, patients with bulimia nervosa, two groups of dieters and non-dieting controls. Thoughts were collected while subjects performed three behavioural tasks, looking at themselves in a full-length mirror, weighing themselves and eating a chocolate covered mint. Both groups of patients had more negative thoughts related to eating, weight and shape than those in the three control groups. In addition, patients with anorexia nervosa showed a greater concern with eating while patients with bulimia nervosa showed a greater concern with weight and appearance. Differences were found between the patients and non-dieting controls using both methods but the self-report questionnaire was less sensitive than concurrent verbalisation to differences between the patients and dieters. Implications of the findings for cognitive-behavioural treatments of the two disorders are discussed.
Cognitive theories assign thoughts about eating, weight and shape a causal role in the aetiology of anorexia nervosa (Garner & Bemis, 1982) and of bulimia nervosa (Fairburn, Cooper & Cooper, 1986). To date few studies have investigated whether these disorders are indeed characterised by the type of cognitive disturbance implied by the theories. Clinical and anecdotal reports (e.g. Bruch, 1973; Russell, 1979) are consistent with the theories but are unsatisfactory because they lack control groups and are usually retrospective. In so far as cognitive-behavioural treatments are effective (for a review see Fairburn, Agras & Wilson, 1992) they provide some support for a cognitive disturbance, but they do not tell us precisely what this is. With the exception of four semi-structured interview-based studies of the cognitive characteristics of these patients (Cooper & Fairburn, 1987; Cooper, Cooper & Fairburn, 1989; Wilson & Smith, 1989; Rosen, Vara, Wendt & Leitenburg, 1990) all existing research studies (e.g. Clark, Feldman & Channon, 1989; Phelan, 1987) have employed self-report questionnaires. These require endorsement of predetermined items and so are only able to capture those thoughts which the investigator has decided are of interest. Most studies have used normal young women as their control group. However, those who diet are an important comparison group since they may show some of the same characteristics (Wilson, 1989). Finally, with the exception of one study (Clark et al., 1989), existing studies have only measured frequency of thoughts and not other response dimensions, for example, duration and belief, and have investigated only those with bulimia nervosa and not also those with anorexia nervosa. The present study was designed to address these issues and extend the work on the cognitive disturbance characteristic of eating disorders. To investigate specificity, two groups of dieters were included in addition to Ss with anorexia nervosa and Ss with bulimia nervosa. To detect thoughts not included on self-report questionnaires, concurrent verbalisation was used. It was predicted that patients with eating disorders would have more thoughts about eating, weight and shape and, in particular, more negative thoughts about these issues than a group of non-dieting controls, and that dieters would occupy an intermediate position, sharing some of the same thoughts as the patients and some with the non-dieting controls. METHOD Design Patients with anorexia nervosa, patients with bulimia nervosa, two groups of dieters and a non-dieting control group were investigated. All Ss were female. Each S completed three tasks *Author
for correspondence. 501
502
MYRA J. C&PER and CHRISTOPHER G. FAIRBURN
devised to elicit thoughts relevant to eating, weight and shape. These were weighing eating a chocolate mint and looking at themselves in a full-length mirror.
themselves,
Subjects Patients with anorexia nervosa. These were 12 patients who fulfilled DSM-III-R criteria for 1987). Nine were recruited through their anorexia nervosa (American Psychiatric Association, therapist and three were recruited through a local self-help group for eating disorders. Patients with bulimia nervosa. These were 12 patients who fulfilled DSM-III-R criteria for bulimia nervosa (American Psychiatric Association, 1987). All patients also fulfilled a strict operational definition of bulimia nervosa based on these criteria (Fairburn, 1987). All were recruited through their therapist. Normal dieters. These Ss were 12 volunteers who fulfilled a strict operational definition of dieting. They were recruited by placing posters in two local Universities asking for female volunteers, under the age of 35, who had been making an attempt to lose weight for at least 4 weeks. They were invited to take part in a study investigating ‘thoughts about eating, weight and shape’. No reference was made to eating disorders. The age limit was specified in order to obtain a group similar in age to the two patient groups. To be included the volunteers had to have been making a serious attempt to lose weight over the preceding 4 weeks. The diet followed and their success in adhering to it were irrelevant. A serious attempt was defined as following a standard reducing diet and/or the setting of predetermined and rigid rules such as a definite calorie limit, preset quantities of food or rules about what should be eaten. Ss were excluded if they had a psychiatric history, if they had ever met DSM-III-R criteria for anorexia nervosa or bulimia nervosa, or if they had shown any of the core behavioural features of either eating disorder, i.e. objective bulimic episodes, actual or attempted self-induced vomiting to lose weight, weight maintenance below 85% mean population matched weight. Symptomatic dieters. These were 12 volunteers, recruited in exactly the same way as the normal dieters, who fulfilled the same criteria for inclusion and exclusion as this group but who, in addition, had either currently or in the past experienced some of the behavioural features of anorexia nervosa or bulimia nervosa as previously described above. Reliability of allocating the dieters to the two groups was determined by an independent judge who was provided with details of Ss’ eating habits and behaviours. Complete agreement was reached in all 24 cases. Normal controls. These were 12 volunteers recruited in the same way as the normal dieters and who fulfilled the same exclusion criteria but who had not been attempting to lose weight over the previous 4 weeks. Measures Information was obtained on demographic and background features, including age, years in full-time education, weight and height. Each S also completed the Eating Attitudes Test (EAT; Garner & Garfinkel, 1979) and the Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock & Erbaugh, 1961), self-report measures of the features of eating disorders and depression respectively. Ss’ self-statements were recorded during the three separate behavioural tasks using two techniques, concurrent verbalisation (or ‘thinking aloud’) and a self-report questionnaire (or Thoughts Checklist) devised specifically for this study. Materials The Thoughts Checklist was made up of self-statements chosen to be typical of the thoughts that patients with eating disorders might be expected to have while carrying out each of the three behavioural tasks. Items were taken from a pool of 133 thoughts (44 for the weighing task, 48 for the mirror task, 41 for the eating task) generated by asking two clinicians (one clinical psychologist and one psychiatrist), as well as the experimenter, to make three lists, one for each task, “of the thoughts which you think might occur to someone with an eating disorder while doing each of the three tasks”. The thoughts obtained were reduced to 15 negative thoughts for the eating task and 16 each for the weighing and mirror tasks. In addition, a task-specific positive thought was added to the weighing and mirror tasks. Examples of thoughts are “I’m getting fatter and fatter” (for the weighing task) and “I’ve no self-control” (for the eating
Thoughts
about
eating,
weight
and shape
in AN and BN
503
task).* The Thoughts Checklist was used to yield several scores for the negative thoughts and, where relevant, for the positive thoughts. These were an overall score (made up of absolute frequency x duration of the thoughts), absolute frequency of thoughts, mean duration of thoughts, and a score of mean degree of belief in the thoughts. Concurrent oerbalisation Breaking transcripts into units. Thoughts obtained using the ‘thinking aloud’ method were transcribed and broken up into units for classification using an approach similar to that employed by Davison, Robins and Johnson (1983). An independent rater (rater 2) broke ten randomly selected transcripts (17% of the total) into units. Reliability of this method was calculated in two different ways. In the first overall agreement on position of end boundaries was calculated by noting the number of end boundaries on which both raters agreed and dividing this by the larger number of boundaries used by either rater. In the second, since rater 1 divided the rest of the data into units, rater 2s agreement with rater 1 was calculated by dividing the number of end boundaries agreed on by the total number used by rater 2. Mean percentage agreement between the two raters was 91.7 and 93.5% respectively for the two methods. Categorising idea units. A manual was written with detailed instructions and examples of how to code the idea units. These were coded using two different methods.
(1) Immediately
relevant negative and positive thoughts. In this coding each task was taken individually and the number of negative and positive thoughts directly and immediately related to each task was calculated (i.e. number of negative and positive thoughts concerned with weighing in the weighing task, number of negative and positive thoughts concerned with eating in the eating task, and number of negative and positive thoughts concerned with looking in the mirror in the mirror task). Past associations and memories were excluded. (4 Content, reference and valence. Using this method, thoughts in each of the three tasks were coded along three dimensions. (4 Content, in which thoughts were divided into those relating to food, eating, weight and shape (target thoughts) and those relating to other issues (non-target thoughts). Unlike the first method, thoughts could include memories, associations and thoughts about all food, eating, weight and shape issues. (b) Reference, in which thoughts were divided into those relating to the self (self-referent) and those relating to other issues (other-referent). Valence, in which thoughts were classified as positive, negative or neutral. (cl
Reliability of coding was calculated separately for each category using the two methods above. Mean percentage agreement between two raters ranged from 84.4 to 98.7%.
outlined
Procedure A standard procedure was used. On arrival the procedure was explained to Ss. Criteria for inclusion and exclusion were established, demographic data was collected and Ss completed the self-report questionnaires. Instructions were then given to verbalise all thoughts that came to mind during each of the three behavioural tasks. Ss first completed a practice task, sorting pictures of faces into order of age, to familiarise them with ‘thinking aloud’ and with the instructions that were to be given for each of the three main tasks. The three behavioural tasks, weighing, looking in a full-length mirror and eating an ‘After Eight’ (a chocolate covered mint), were then carried out. There were six possible orders in which the three tasks could be presented. To control for possible order effects two Ss in each group completed the tasks in each order. Ss were left alone to complete each task with signals being given at two points, the first to indicate that the S could begin the task and the second to indicate that the S should stop doing the task. Thoughts were thus obtained for three periods, before, during and after the task. The procedure was exactly as it had been for the practice task except that Ss were given twice as long (5 min) to say their thoughts out loud, i.e. 14min before the task, 2 min during the task and lf min after the task.
*Copies
of the Thoughts
Checklist
can be obtained
from the authors.
MYRA f. CAPER and CHRISTOPHER G. FAIRBURN
504
In the weighing task Ss were asked to stand on a set of weighing scales, find out exactly how much they weighed, and remain on the scales until a signal came from the experimenter to stop. In the mirror task Ss were asked to look carefully at themselves in a full-length mirror, starting with a side view, and turning slowly round to look at themselves from different angles. The experimenter demonstrated what was to be done. In the eating task Ss were asked to eat a chocolate covered mint, a ‘forbidden food’ for many patients with eating disorders and for many dieters. After carrying out the three tasks Ss completed the Thoughts Checklist and rated how typical their thoughts had been of the thoughts that they might expect to have when doing each of the tasks in everyday life. A O-10 scale was used with 0 being ‘not at all typical’ and 10 being ‘completely typical’. Finally weight and height were measured and Ss were debriefed. RESULTS
Demographic features
Information on age, years of full-time education, Body Mass Index (BMI) and scores on the EAT and BDI are presented in Table 1. (BMI = weight in kg/height in m2.) Ss in the different groups were of a similar age. They differed in number of years in full-time education with the two patient groups having received significantly fewer years of full-time education than the other three groups. All were of a similar BMI, except for the patients with anorexia nervosa whose BMI was, as expected, lower than that of the other four groups. On the EAT the normal control group had a significantly lower score than the symptomatic dieters, patients with bulimia nervosa and patients with anorexia nervosa, while the normal dieters had a si~ificantly lower score than the two patient groups. Ss in both patient groups had higher scores on the BDI than those in the other three groups. Both dieting groups had been attempting to lose weight for equivalent lengths of time [11.6 + 10.1 weeks in the normal dieter group and 12.7 f 14.5 weeks in the symptomatic dieter group, t(21) = 0.6, NS]. In the symptomatic dieter group seven Ss reported objective bulimic episodes, nine reported attempted or actual self-induced vomiting to lose weight and one had briefly reached a weight below 85% mean population matched weight by dieting. In the anorexia nervosa group most patients were suffering from ‘restricting’ anorexia nervosa, only one reported episodes of overeating followed by self-induced vomiting. There were no differences between the five groups in the typicality ratings made for each task (range 6.0-8.5). This indicated that the thoughts obtained were equally typical of thoughts that might be expected in real life situations. Immediately reletlant negatitte and positiue thoughts
A three-way analysis of variance (group x task x type of thought) showed no main effect of group [F(4,55) = 1.42, NS], indicating that the groups did not differ in overall number of thoughts produced, or of task [F&l 10) = 1.04, NS]. There was a main effect of type of thought [F(8,110) = 4.3, P < 0.0002] but this was modified by a significant group x task x type of thought interaction [r”( 16,220) = 2.10, P < 0.021. To locate the source of the difference found, two separate two-way analyses of variance (group x task) were performed, one with the negative thoughts as the dependent variable and the other with the positive thoughts as the dependent variable. Since Table 1. Demographic NC 4s
( yr)
Education (yr) BMI EAT BDI
X
SD X SD X SD X SD X SD
22.0 2.3 16.la I.5 21.3’ 2.1 6.8”” 4.0 5.7” 4.4
ND 23.5 4.6 15.4 1.2 24.4” 1.8 17.6abc 9.4 7.2’ 5.9
features of the five groups
SD
BN
AN
F/H ratio
P
23.0 3.1 15.8’ 2.4 23.3” 3.0 30.8bc 12.0 9.38 6.5
24.0 4.3 13.3b 1.7 24.8* 5.9 42.3M 11.0 Zl.lb 10.6
21.9 3.2 14.0b 1.2 I5.0b 1.5 53.lM 20.4 25.c 10.4
F = 0.8
NS
F = 6.4
<0.0003
H = 34.1
H = 41.6
<0.0001
I== 12.8
to.0001
All d.f. = 4.55. n = 12 for each group. NC = Normal controls, ND = Normal dieters, SD = Symptomatic dieters, BN = Bulimia nervosa, AN = Anorexia nervosa. BMI = Body mass Index = weight in kg/height in m*. EAT = Eating Attitudes Test. BDI = Beck Depression Inventory. Within each measure means with different superscripts are significantly different from each other.
Table
2. Mean
Task
about
percentage verbalisation
scores for negative and positive thoughts obtained using concurrent for each of the five groups in each of the three tasks
Valence
eating,
NC
weight
and shape
ND Thoughts
Weighing Eating Mirror
Weighing Eating Mirror
505
Thoughts
Negative Positive Negative Positive Negative Positive
10.9’ II.2 8.2” 12.8 16.7” 13.1
19.8’ 3.4 8.61 6.6 21.6’ 6.4
Negative Positive Negative Positive Negative Positive
19.7” 8.4 19.9” II.3 24.0* 10.7
29.0Pd 5.2 18.5” 5.3 29.7” 8.8
in AN and BN
BN
SD immediately
relevant
AN
to task
25.7”’ 3.1 18.4sb Il.3 32.0’ 8.5 All thoughts
46.9b 3.0 33.lk 6.1 50.Sb 4.0
39.Ok 4.0 45. IC 1.9 31.9” 3.6
31.5M 1.1 27.9” II.0 38.6”’ 10.3
56.7’ 4.1 44.Sbc 6.0 61.3b 4.8
46.1cd 4.8 56.3b 5.1 41.4c 9.9
n = I2 for each group. NC = Normal controls, ND = Normal dieters, SD = Symptomatic dieters, BN = Bulimia nervosa, AN = Anorexia nervosa. Within each measure means with different superscripts are significantly different from each other.
the groups did not differ in overall number of thoughts for the three tasks, to assess the extent to which they differed in emotional content of the thoughts produced, the percent negative and the percent positive scores were compared. This took into account the large differences between groups in the number of neutral thoughts reported in each task. For percentage of negative thoughts, the two-way analysis of variance showed a main effect of group [F(4,55) = 13.06, P < O.OOOl]and of task [F(2,110) = 7.08, P < 0.0021. However, these main effects were modified by a group x task interaction [F(8,110) = 3.39, P c 0.0021. Mean percentage scores of immediately relevant negative thoughts for each of the five groups in each of the three taks are shown in Table 2. To locate the source of the interaction post hoc Tukey Honestly Significant Difference tests were performed. As predicted, in the between group comparisons, taking each task separately, the two patient groups generally had more negative thoughts directly and immediately related to each of the three tasks than the non-dieting female controls. Patients in both groups had a greater percentage of negative thoughts in the weighing and eating tasks than the non-dieting controls (all P values ~0.01). In addition, patients with bulimia nervosa, but not those with anorexia nervosa, had a greater percentage of negative thoughts in the mirror task than the normal controls (P < 0.01). There were no significant differences between the two dieting groups and the normal controls in any of the three tasks and several of the comparisons between the two dieting groups and the two patient groups were also not significant. These were comparisons between the symptomatic dieters and the patients with anorexia nervosa in the weighing task, between the symptomatic dieters and the patients with bulimia nervosa in the eating task, and between the two dieting groups and the patients with anorexia nervosa in the mirror task. This suggested that, as predicted, the two dieting groups occupied a position intermediate between that of the normal controls and the two patient groups. When the two dieting groups were compared with the two patient groups it emerged that those with bulimia nervosa had a greater percentage of negative thoughts immediately relevant to the weighing and mirror tasks while those with anorexia nervosa had a greater percentage of negative thoughts immediately relevant to the eating task (all P values ~0.01). Patients with bulimia nervosa also had a greater percentage of negative thoughts immediately relevant to the eating task than the normal dieters and a greater percentage of negative thoughts in the mirror task than those with anorexia nervosa (both P values ~0.01). Similar findings emerged from the within group comparisons. Patients with bulimia nervosa had a greater percentage of negative thoughts in the mirror (P < 0.01) and weighing (P < 0.05) tasks than in the eating task. In this they showed a pattern similar to that of the two dieting groups, both of which had a greater percentage of negative thoughts in the mirror task than in the eating task (both P values <0.05). However, in the patients with anorexia nervosa the reverse pattern was observed. This group had a greater percentage of negative thoughts in the eating task than in the mirror task (P < 0.05). There were no differences between tasks in the normal control group.
MYRA J. CARPERand CHRISTOPHER G. FAIRBURN
506
Percentage of positive thoughts was also subjected to a two-way analysis of variance (group x task). There was a main effect of group [F(4,55) = 4.47, P < 0.004)] and of task [F(2,110) = 8.71, P < 0.0004] but no significant group x task interaction [F(8,110) = 1.53, NS]. When subjected to post hoc tests the main effect of group indicated that the normal control Ss did not report a greater percentage of positive thoughts immediately relevant to the three tasks than the two patient groups. Mean percentage scores for the five groups in each of the three tasks can be seen in Table 2. Content,
reference
and valence
a three-way analysis of variance On the target vs non-target thoughts dimension (group x task x target vs non-target thoughts) showed that there was no main effect of group [F(4,55) = 1.25, NS]. This indicated that the groups did not differ in overall number of thoughts reported. There was no main effect of task [F(2,110) < l] but there was a main effect of type of target thought [F(4,55) = 155.45, P < O.OOOl]. There was no interaction between group, task and type of thought [F(8,110) < I]. The main effect of type of target thought indicated that all groups reported more thoughts about eating, weight and shape related issues than about other issues. Mean scores for the five groups in each of the three tasks can be seen in Table 3. While the groups did not differ in type of thought reported in the three tasks, as we shall see below, the groups did differ in valence or emotional content of the thoughts produced. With regard to reference, a three-way analysis of variance (group x task x self-referent vs other referent) showed no main effect of group [F(4,55) = 1.54, NS], again indicating that the groups did not differ in overall number of thoughts produced. There was no main effect of task [F(2,110) = 1.19, NS] but there was a main effect of reference [F(4,55) = 74.20, P < O.OOOl]. There was no group x task x type of thought interaction [F(8,110) = 1.88, NS]. The main effect of reference indicated that all groups had more self-referent thoughts than other-referent thoughts. Mean scores for the five groups in each of the three tasks can be seen in Table 3. With regard to valence, a three-way analysis of variance (group x task x negative vs positive vs neutral thoughts), showed no main effect of group [F(4,55) = 1.33, NS]. As before, this indicated that the groups did not differ in overall number of thoughts produced in the three tasks. There was no main effect of task [F(2,110) < 1] but there was a main effect of valence [F(8,110) = 74.84, P < O.OOOl] and a valence x group interaction [F(8,110) = 6.34, P < O.OOl]. However, these findings were modified by a group x task x type of thought interaction [F(l6,220) = 1.73, P < 0.051. To locate the source of the interaction, two separate two way analyses of variance (group x task) were carried out, one using the negative thoughts as the dependent variable and one using the positive thoughts as the dependent variable. Since there were no differences between groups in overall number of thoughts the extent to which the groups differed in emotional content of the thoughts was assessed by comparing the percent negative and the percent positive scores of each group. With negative thoughts as the dependent variable, the two way analysis of variance yielded a main effect of group [F(4,55) = 16.84) P < O.OOOl] and of task [F(2,110) = 3.44, P < 0.041. However, these main effects were modified by an interaction between group and task Table 3. Mean raw scores for target and self-referent tmn for each of the five groups Task
Thought
Weighing
Target Non-target Target Non-target Target Non-target Self-referent Other-referent Self-referent Other-referent Self-referent Other-referent
Eating Mirror Weighing Eating Mirror
type
thoughts obtained using concurrent in each of the three tasks
verbalisa-
NC
ND
SD
BN
AN
37.3 13.1 32.8 17.1 37.7 14.1 30.4 19.9 22.3 27.7 35.6 16.2
37.7 6.2 32.7 II.3 40.3 7.0 30. I 13.8 22.3 18.3 34.4 12.9
49.6 10.2 50.3 9.8 50.4 1.2 38.0 21.9 31.7 22.8 45.6 12.8
38.6 4.7 31.2 8.4 35.3 4.6 30.4 6.8 30.3 9.3 33.5 6.4
42.9 6.8 39.3 6.6 48.6 5.8 33.3 15.8 33.8 12.2 44.9 7.9
n = I2 for each group. NC = Normal controls, ND = Normal dieters, SD = Symptomatic dieters, EN = Bulimia nervosa, AN = Anorexia nervosa. Within each measure means with different superscripts are significantly different from each other.
Thoughts about eating, weight and shape in AN and BN
507
[F(8,110) = 3.42, P < 0.0021. Mean percentage scores for negative thoughts for the five groups in each of the three tasks are shown in Table 2. Post hoc Tukey Honestly Significant Difference tests were used to locate the source of the interaction. As predicted, between group comparisons, taking each task separately, showed that the two patient groups had more negative thoughts in each of the three tasks thaz the non-dieting female controls (all P values ~0.01). There were no significant differences between the two dieting groups and the normal controls in most tasks, the one exception being the comparison between the normal controls and the symptomatic diet group in the weighing task (P < 0.05), and several comparisons between the two dieting groups and the two patient groups and the patients with anorexia nervosa in the weighing and mirror tasks and between the symptomatic dieters and the patients with bulimia nervosa in the eating task. This suggested that, as predicted, the two dieting groups occupied a position intermediate between the normal controls and the two patient groups. When the two dieting groups were compared with the two patient groups a pattern of results similar to that for immediately relevant negative thoughts was found, i.e. patient with bulimia nervosa had more negative thoughts in the weighing and mirror task while those with anorexia nervosa had more negative thoughts in the eating task (all P values co.01 except for the comparison between patients with bulimia nervosa and symptomatic dieters on the weighing task P < 0.05). In addition patients with bulimia nervosa had more negative thoughts in the eating task than the normal dieters and more negative thoughts in the mirror task than those with anorexia nervosa (both P values
Checklist
When overall negative score was subjected to a two-way (group x task) analysis of variance there was a main effect of group [F(4,55) = 8.57, P < O.OOl] and of task [F(2,110) = 28.37, P < O.OOl]. However, these main effects were modified by a significant group x task interaction [F(8,110) = 3.53, P < 0.0021. Mean overall negative scores on the Thoughts Checklist for the five groups in each of the three tasks are shown in Table 4. As predicted, between group comparisons, taking each task separately, showed that the two patient groups had a higher overall negative score in most of the three tasks than the non-dieting controls. The one exception was for the comparison between the normal controls and the patients with anorexia nervosa in the mirror task. All comparisons were significant at the level P < 0.01, except for that between the normal controls and the anorexia nervosa patients in the weighing task (P < 0.05). There were few significant differences between the two dieting groups and the normal controls in most tasks, the exceptions being the comparisons between the symptomatic dieter group and the normal controls in the weighing task (P < 0.01) and in the mirror task (P < 0.05). Several of the comparisons between the two dieting groups and the two patient groups were also not significant. These were comparisons between the patients with anorexia nervosa and the two diet groups in the weighing and mirror tasks, the comparison between the patients with anorexia nervosa and the symptomatic dieters in the eating task and between the patients with bulimia nervosa and the symptomatic dieters in all three tasks. This again suggested that, as predicted, the two dieting groups appeared to occupy a position between that of the normal controls and the two patient groups. When the two dieting groups were compared with the two patient groups the
508
MYRA J. OXPER and CHRISTOPHER G. FAIRBURN Table 4. Mean scores on the Thouehts Task
NC
Checklist
for each of the five crams
ND
SD Frequency
Weighing Eating Mirror
20.0” 16.4” 18.9”
Weighmg Eating Mirror
2.5 0.9 2.1
33.3”’ 20.1” 26.7” Absolute
Duration Weighing Eating Mirror
Weighing Eating Mirror
2.7” 2.5’ 2.4”
78.5 89.3 66.2
3.9b 2.6” 3.4”b
Weighing Mirror
4.2 4.3
69.2 65.4 59.4
81.0” 80.0”
2.6 4.0
32.0b 60.0”
9.2 6.6 8.0 of negative
58.9 49.0 54.1 of positive
2.3 3.2 Belief m posltlve 51.7dh 68.3”
AN
thoughts 49.Q 36.7b 40.9b thoughts
I I.2
36.3k 35.9b 27.Bab
8.8 9.4
7.5 7.0 6.2
4.0b 3.2ab 3.5b
3.P 4.0b 3.0ab
thoughts
3.gh 3.0db 3.2” Belief in negative thoughts*
Frequency Weighing Mirror
of negative
40.9k 28.7*b 34.5h number of negative
6.2 3.2 4.9
in each of the three tasks BN
69.7 58.9 59.6
71.5 76.7 64.8
3.7 3.0
3.8 3.4
63.36b 40.0”
46.0db 48.0”
thoughtst
thought&
n = 12 for each group. *n = 7 for NC, n = I I for ND, SD. BN and n = IO for AN. tn = IO for NC, n = 5 for ND, AN, n = 6 for SD and n = 3 for BN. NC = Normal controls, ND = Normal dieters. SD = Symptomatic dieters, BN = Bulimia nervosa, AN = Anorexia nervosa. Within measures means with different superscripts are significantly different from each other.
pattern of differences between groups was less clear than that observed for immediately relevant negative thoughts. As before, patients with bulimia nervosa had a greater frequency of negative thoughts than the normal dieters in all three taks (all P values ~0.05) while patients with anorexia nervosa had a greater frequency of negative thoughts than the normal dieters only in the eating task (P < 0.05). Thus, as in the analysis of immediately relevant thoughts, overall negative score successfully distinguished the normal dieters from the two patient groups. However, comparisons between the two patient groups and the symptomatic dieters showed no significant results. Thus, unlike the measure of immediately relevant negative thoughts, overall negative score on the Thoughts Questionnaire did not discriminate between the two patient groups and those who showed some symptoms, but not a diagnosis, of an eating disorder. Comparisons within groups produced results that were similar to those observed for immediately relevant negative thoughts. Patients with bulimia nervosa had a greater score in the weighing than in the eating task (P < 0.01) and in the weighing than in the mirror task (P < 0.01). Patients with anorexia nervosa, however, had greater scores in the eating than in the mirror task (P < 0.05) and in the weighing than in the mirror task (P < 0.05). As found for immediately relevant negative thoughts, the two dieting groups showed a pattern similar to that of the patients with bulimia nervosa. They had a greater score in the weighing than in the eating task (both P values ~0.01) as well as a greater score in the mirror than in the eating task (both P values <0.05). Normal controls again showed no differences in scores obtained in each task. When absolute number of negative thoughts was subjected to a two-way (group x task) analysis of variance there was a main effect of group [F(4,55) = 13.57, P < O.OOOl] and of task [F(2,110) = 17.49, P < O.OOOl] but no group x task interaction [F@,llO) = 1.25, NS]. In order to test the prediction that the two patient groups would have a greater absolute number of negative thoughts than the non-dieting controls post hoc Tukey Honestly Significant Difference tests were carried out on the main effect of group. As predicted, the two patient groups had a greater absolute number of negative thoughts than the non-dieting controls (both P values c 0.01). The prediction that the two dieting groups would occupy a position intermediate between the normal controls and the two patient groups was partially supported. There were no significant differences between the normal dieting group and the non-dieting controls or patients with anorexia nervosa which
Thoughts about eating, weight and shape in AN and BN
509
suggested that they occupied a position intermediate between the normal controls and these patients. Scores of the symptomatic dieters were similar to those of the two patient groups and different from those of the non-dieting controls (P < O.Ol), suggesting that they were most like the two patient groups. Mean scores for absolute number of negative thoughts on the Thoughts Checklist for each of the five groups on the three tasks can be seen in Table 4. With mean duration of negative thoughts as the dependent variable in a two-way (group x task) analysis of variance there was a main effect of group [F(4,45) = 3.11, P < 0.021 and of task [F(2,90) = 10.67, P < O.OOOl]. These effects were modified by a group x task interaction [F(8,90) = 3.61, P < 0.0031. Mean scores for mean duration of negative thoughts on the Thoughts Checklist in each of the five groups on the three tasks can be seen in Table 4. The prediction that negative thoughts would last longer in the two patient groups than in the non-dieting controls (i.e. that the two patient groups would spend longer thinking about the negative thoughts than the non-dieting controls) was partially supported. Between group comparisons showed that, in the weighing and mirror tasks, negative thoughts lasted longer in the patients with bulimia nervosa than they did in the normal controls. In the eating task, negative thoughts lasted longer in the patients with anorexia nervosa than in the normal controls (all P values ~0.05). There were few differences between the two dieting groups and the normal controls, the exceptions being the comparison between the normal controls and the two diet groups in the weighing task (both P values <0.05), and most of the comparisons between the dieting groups and the patients groups were also not significant. These were comparisons between the two patient groups and the two diet groups in the weighing and mirror tasks and between the patients and the normal dieters and between the patients with anorexia nervosa and the symptomatic dieters in the eating task. This suggested that, as predicted, the scores of the two dieting groups fell between those of the normal controls and the two patient groups. When the two dieting groups were compared with the patient groups only one significant finding was observed, the patients with anorexia nervosa spent longer thinking about the negative thoughts in the eating task than the normal dieters (P < 0.05). Thus, although duration of negative thoughts in the three tasks distinguished the normal controls from the patients to some extent, it did not generally distinguish the patients from the two groups of dieters. As with previous measures, the results from the within group comparisons suggested greater concern with weight in the patients with bulimia nervosa and greater concern with eating in the patients with anorexia nervosa. The normal controls spent similar amounts of time thinking about negative thoughts in all three tasks. However, in the patients with bulimia nervosa, negative thoughts in the weighing task lasted longer than they did in the eating task (P < 0.05). The two dieting groups again showed a pattern similar to the patients with bulimia nervosa. In the normal dieters negative thoughts lasted longer in the weighing (P < 0.01) and mirror tasks (P < 0.05) than in the eating task. In the symptomatic dieters negative thoughts in the weighing task lasted longer than negative thoughts in the eating task (P < 0.05). The reverse pattern was observed in the patients with anorexia nervosa where negative thoughts lasted longer in the eating task than in the mirror task (P < 0.01). When mean degree of belief in the negative thoughts was subjected to a two-way (group x task) analysis of variance there was a main effect of group [F(4,45) = 3.00, P < 0.031 and of task [F(2,90) = 4.87, P < 0.011. There was no group x task interaction [F(8,90) = 1.52, NS]. In order to test the prediction that the patient groups would believe more strongly in the negative thoughts reported on the questionnaire than the normal controls post hoc Tukey Honestly Significant Difference tests were performed on the main effect of group. No evidence was found to support this prediction. This suggested that the patients did not believe the negative thoughts to a greater degree than the normal controls. Two measures were obtained from the positive thoughts on the weighing and mirror tasks. These were overall positive score (which was equivalent to mean duration of each positive thought because there was only one positive thought in each of the two tasks) and mean degree of belief in each positive thought. When overall positive score was subjected to a two way (group x task) analysis of variance there was no main effect of group [F(4,24) = 2.43, NS] or of task [F( 1,24) < 11. There was no group x task interaction [F(4,24) = 1.lO, NS]. Thus, as for immediately relevant positive thoughts the normal controls did not have a greater overall positive score than
510
MYKA J. CARPERand CHRISTOPHERG. FAIRBURN
the two patient groups. When mean degree of belief in positive thoughts was subjected to a twoway (group x task) analysis of variance there was a main effect of group [F(4,24) = 3.86, P < 0.021 but no main effect of task [F( 1,24) = 1.29, NS]. However, the main effect of group was modified by a group x task interaction [F(4,24) = 4.06, P < 0.021. Post hoc Tukey Honestly Significant Difference tests were carried out to locate the source of the interaction. When between group comparisons were made the normal controls did not believe the positive thoughts to a greater degree than the two patient groups. Only one finding was significant. This indicated that the normal controls believed the positive weighing thought more than the normal controls (P < 0.01). The within group comparisons showed that all groups, except the normal dieter group, believed equally in the positive thoughts in the weighing and mirror task. The one exception was that the normal dieter group believed the positive mirror thought to a greater degree than the positive weighing thought (P < 0.01). DISCUSSION Using the ‘thinking aloud’ method, patients with eating disorders were found to have more negative thoughts related to eating, weight and shape in each of the three tasks than the non-dieting controls, while the two dieting groups appeared to occupy an intermediate position. Patients with bulimia nervosa were distinguished from dieters, including those with some symptoms of an eating disorder, by a greater number of negative thoughts about weight and appearance. Patients with anorexia nervosa, however, were distinguished from the two dieting groups by a greater number of negative thoughts about eating. Within group comparisons showed a similar concern with weight and appearance in the bulimia nervosa patients and with eating in the anorexia nervosa patients. The two dieting groups showed a pattern similar to that of the patients with bulimia nervosa while normal controls showed no difference in prominence of negative thoughts across the three tasks. Positive thoughts did not distinguish the groups. No differences were found between groups in number of target thoughts, i.e. total number of thoughts related to food, eating, weight and shape. This result may have been due to a ceiling effect since most of the thoughts in all groups were related to these issues. However, an alternative explanation, which receives some support here, is that while all Ss experienced such thoughts when placed in situations concerned with eating, weight and shape what distinguished the groups most clearly was the emotional content of the thoughts experienced. No differences were found between groups for the reference dimension, i.e. self-referent vs other-referent thoughts. This may also have been due to a ceiling effect since the majority of thoughts in all groups were self-referent rather than other-referent. On the Thoughts Checklist findings on all measures, negative and positive, were generally similar to the findings for immediately relevant thoughts. There was, however, one important exception. Unlike the measure of immediately relevant negative thoughts, none of the measures obtained from the questionnaire distinguished the symptomatic dieters from the two patient groups. This suggests that the self-report questionnaire is a less sensitive measure than the concurrent verbalisation method. This could be because retrospective recall of thoughts is less accurate than immediate report. The results also suggest that response dimensions other than frequency, particularly duration of thoughts, are relevant. The results of this study suggest that negative self-statements may be an important focus for cognitive interventions in patients with eating disorders. In particular, the finding that patients with bulimia nervosa are particularly concerned with weight and appearance while patients with anorexia nervosa are more concerned with eating suggests that in treatment it may be desirable to focus on different content when attempting to alter self-statements in these two groups. There has been a tendency amongst clinicians and researchers to emphasise the similarity of the cognitive concerns in the two disorders and in particular to stress the importance of concerns about weight 1986). This emphasis may help to explain why and appearance (e.g. Fairburn & Garner, cognitive-behaviour therapy has been an effective treatment for bulimia nervosa. However, for cognitive-behavioural treatment to be as successful in anorexia nervosa it may be necessary to focus particularly on concerns about eating. The finding that only negative thoughts and not positive thoughts distinguish the groups supports what Kendall (e.g. Kendall & Hollon, 1981) has called It suggests that therapeutic interventions which focus the power of ‘non negative thinking’.
Thoughts
about
eating,
weight
and shape in AN and BN
511
primarily on negative and dysfunctional thinking (e.g. those developed by Ellis, 1962, and Beck, 1976) may be more successful in treating patients with eating disorders than interventions which focus on encouraging positive coping responses (e.g. that developed by Meichenbaum, 1977). Ackmnvledgements-Myra J. Cooper was supported by a Studentship from the Medical Research Council while this research was completed. Christopher G. Fairburn holds a Wellcome Trust Senior Lectureship. We would like to thank David Clark and Paul Salkovskis for their helpful comments on a draft of this manuscript. Correspondence and requests for reprints should be sent to Myra Cooper, University Department of Psychiatry, Warneford Hospital, Oxford OX3 7JX, England.
REFERENCES American Psychiatric Association (1987). DSM-III-R. Washington, D.C.: American Psychiatric Association. Beck, A. T. (1976). Cognit& therapy and the emotional disorders. Madison, Conn.: International Univerisities Press. Beck, A. T., Ward, C. H., Mendelson, M., Mock, J. & Erbaugh, J. (1961). An inventory for measuring depression. Archiues of General Psychiatry, 4, 53363. Bruch, H. (1973). Eating disorders: Obesity, anorexia nercosa and /he person within. New York: Basic Books, Clark, D. A., Feldman, J. & Channon, S. (1989). Dysfunctional thinking in anorexia and bulimia nervosa. Cognitive Therapy and Research, 13, 377-387. Cooper, Z. & Fairburn, C. G. (1987). The Eating Disorder Examination: A semi-structured interview for the assessment of the specific psychopathology of eating disorders. International Journal of Eating Disorders, 6, l-8. Cooper, Z., Cooper, P. J. & Fairburn, C. G. (1989). The validity of the Eating Disorder Examination and its subscales. British Journal of Psychiatry. 154. 80778 12. Davison, G. C., Robins, C. & Johnson, M. K. (1983). Articulated thoughts during simulated situations: A paradigm for studying cognition in emotion and behaviour. Cognifive Therapy and Research, 7, 17-40. Ellis, A. (1962). Reason and emofion in ps.ychotherapy. Secaucus, N.J.: Citadel Press, Fairburn, C. G. (1987). The definition of bulimia nervosa. Annals of Behapioural Medicine, 9, 337. Fairburn, C. G. & Garner, D. M. (1986). The diagnosis of bulimia nervosa. International Journal of Baling Disorders, 5, 403-420. Fairburn, C. G., Agras. W. S. & Wilson, G. T. (1992). The research on the treatment of bulimia nervosa: Practical and theoretical implications. In Anderson, G. H. & Kennedy, S. H. (Eds). The biology offeasr andfamine: Relevance to eating disorders. New York: Academic Press, Fairburn, C. G., Cooper, Z. & Cooper, P. J. (1986). The clinical features and maintenance of bulimia nervosa. In Brownell, K. D. & Foreyt, J. P. (Eds). Physiology, psychology and treatment of the eating disorders. New York: Basic books. Garner, D. M. & Bemis, K. M. (1982). A cognitive-behavioural approach to anorexia nervosa. Cognitiue Therapy and Research, 6, 1233150. Garner, D. M. & Garfinkel, P. E. (1979). The Eating Attitudes Test: An index of the symptoms of anorexia nervosa. Psychological Medicine, 9, 2733279. Kendall, P. C. & Hollon, S. D. (1981). Assessing self-referent speech: Methods in the measurement of self-statements. In Kendall, P. C. & Hollon. S. D. (Eds). Assessment Strategies for Cognitive-Beharioural Intert~entions (pp. 85-117). London: Academic Press, Meichenbaum, D. (1977). Cognitioe-behaoiour modification: An integrative approach. New York: Plenum Press. Phelan, P. W. (1987). Cognitive correlates of bulimia: The bulimic thoughts questionnaire. Infernarional Journal of Eating Disorders, 6, 5933607. Rosen, J. C., Vara, L., Wendt, S. & Leitenberg, H. (1990). Validity studies of the Eating Disorder Examination. International Journal of Eating Disorders, 9. 519-528. Russell, G. F. M. (1979). Bulimia nervosa: An ominous variant of anorexia nervosa. Psychological Medicine, 9, 429-488. Wilson, G. T. (1989). Bulimia nervosa: Models, assessment and treatment. Current Opinion in Psychiatry, 2, 790-794. Wilson, G. T. & Smith, D. (1989). Assessment of bulimia nervosa: An evaluation of the Eating Disorder Examination, International Journal of Euring Disorders, 8, 173-l 79.