Josie Geller, ¹² ²* Sarah J. Cockell, ² and Elliot M. Goldner ¹² ²

¹Department of Psychiatry, University of British Colombia,
Vancouver, British Columbia
²St. Paul’s Hospital Eating Disorders Clinic, Vancouver, British Columbia

Abstract: Objective: This study examined inhibited expression of negative feelings and interpersonal orientation in women with anorexia nervosa. Method: Twenty-one women meeting DSM-IV criteria for anorexia nervosa were compared with 21 psychiatric and 21 normal control women matched on education. Two measures were used to assess inhibited expression of negative feelings and interpersonal orientation: the State-Trait Anger Expression Inventory assesses the suppression and expression of anger and the Silencing the Self Scale assesses four cognitive schemas involving the repression of needs and feelings to protect interpersonal relationships. Results: Women with anorexia nervosa reported significantly higher scores on the four Silencing the Self Schemas and on suppressed anger after controlling for age [note or rage? Ds] These group differences were maintained for two of the cognitive schemas (Care and Silence) after controlling for depression, self-esteem, and global assessment of functioning. Inhibited expression of negative emotion and interpersonal orientation scores were also significantly related to cognitive and affective components of body image dissatisfaction and to trait and self-presentational dimensions of perfectionism.. Discussion: These findings are reviewed in context of health psychology, as well as feminist and temperament theories. Implications for treatment are addressed.. ©2000 by John Wiley & Sons, Inc. Int Eat Disord 28: 8-19, 2000

Key words: inhibited expression; anger; body image; perfectionism; anorexia nervosa

INTRODUCTION

Anorexia nervosa is a multidetermined disorder that has been approached from a number of perspectives, including feminist theory, temperament theory and health psychology. In this paper, theory and research findings from each of these areas are addressed in an attempt to develop a broader and more integrated understanding of interpersonal functioning in anorexia nervosa.
Current models of female socialization and moral development suggest that despite being open communicators, women are socialized to suppress negative feelings and needs in order to preserve close relationships (Gilligan, Rogers & Tolman, 1991). These models suggest that women focus on others’ feelings more than their own, inhibit their negative feelings, and fail to recognize and explore their own personal experiences. Four cognitive schemas have been described to result from this socialization (jack & Dill, 1992). They include the inhibition of self-expression to avoid conflict, the securing of attachments by putting the needs of others before the self; judging the self by external standards, and presenting an outer compliant self while the inner self grows angry and hostile. Feminist theories emphasize that the suppression of feelings and loss of voice involve assuming a position of diminished power in relationships. This may extend to other aspects of personal experience and have important implications to feelings of self-worth (Brown & Jasper, 1993; Gutwill, 1994).
There is growing evidence in the health psychology literature that inhibited self-expression increases risk for a number of health problems. Several studies have shown an association between failing to talk about traumatic events and poor performance on a number of health indicators ( see Pennebaker & Hoover, 1986, for a review). Conversely, writing and talking about difficult emotional experiences are associated with improved physical health, enhanced immune functioning, and fewer medical visits (Pennebaker & Beall,1986; Pennebaker, Colder & Sharp, 1990; Pennebaker & Hoover, 1986). A number of links have also been drawn between a lack of emotional expression and psychological disorders. For example, the cognitive schemas described earlier have been shown to correlate significantly with depression scores in several samples of women (Jack & Dill, 1992). In sum, the failure to express difficult feelings may be particularly common in women, and there is accumulating evidence that inhibited expression of negative emotions is associated with a variety of health problems.
Women with anorexia nervosa have been described as particularly prone to silencing negative affect. They have been depicted as prone to extreme harm avoidance, reward dependence, and low novelty seeking (Strober, 1991). From this temperamental perspective, the tendency to avoid expressing negative emotions may be accounted for by an overanticipation of distress and discomfort, an unusual sensitivity to the feelings and needs of others, and a preference for stable, invariant, and emotionally temperate environments. Because of this temperamental presentation, women with anorexia have been described as inclined to cope with the challenges of life using repetitive, reward-seeking behaviors, and to avoid uncomfortable communication involving negative affect (Sohlberg & Strober, 1994; Strober, 1991). Clinical description and controlled studies examining communication patterns in families of individuals with anorexia nervosa support this temperamental presentation (Dog & Vanderycken, 1985; Lieberman, 1995; Minuchin Rosman, & Baker, 1978; Shugar & Krueger, 1995). The extent to which women with anorexia nervosa demonstrate higher inhibited expression of negative emotions where compared with other psychiatric groups, however, has not been established.
The first purpose of this study was to determine the specificity of inhibited self-expression and externally focused interpersonal orientation to anorexia nervosa. Women with anorexia were compared with normal and psychiatric control women on these measures. Inclusion of a psychiatric group allowed for control of psychiatric feature comorbid with eating disorder symptomatology, including depressive symptoms. Demographic variables were also controlled through covariance. It was hypothesized that scores on inhibited expression of negative emotions and on measures of externally focused interpersonal orientation would be higher in the anorexia nervosa group than in the normal control group. No hypotheses were made with regard to differences between the anorexia and the psychiatric control groups.
Body image dissatisfaction is a commonly observed feature of anorexia. Negative thoughts and feelings about the body have been related to reports of general distress, depression, and self-esteem deficits (Geller, Johnston, & Madsen, 1997; Geller et al, 1998;Heilburn & Witt, 1990; Noles and Cash, 1985). Although no empirical studies of the relationship between inhibited expression of negative emotions and body dissatisfaction were found, it has been hypothesized that anorexia nervosa may occur as a result of displaced negative self-feelings into the body (Bruch, 1973, 1978). That is, rather than experiencing anger toward an external target, the individual with anorexia nervosa instead feels fat. Despite a number of clinical descriptions of this phenomenon, the relationship between body dissatisfaction and inhibited expression of negative emotions has not been empirically examined.
The second purpose of this study was to determine the relationship among measures of inhibited self-expression, externally focused interpersonal orientation, and negative thoughts and feelings about the body. It was hypothesized that inhibited expression of negative emotions and externally focused interpersonal orientation would be significantly and positively related to negative thoughts and feelings about the body.
Finally, given the health benefits of expressing emotions, understanding the conditions that contribute to the inhibition of feelings and the adoption of an externally focused interpersonal orientation may be useful in planning treatment interventions. One possible mechanism is the need to be perfect or to appear to be perfect. Trait and self-presentational dimensions of perfectionism have been established to be key features of anorexic symptomatology (Cockell, Hewitt, Goldner, Srikameswaran, & Flett, 1997; Goldner, Srikameswaran, & Cockell, in press). Possibly, the individual with a strong need to avoid exposing imperfections (and who holds the belief that expressing negative feelings and needs reflects a character flaw) avoids discussing negative experiences at all costs.
The final purpose of this study was to conduct exploratory analyses examining the relationship between perfectionism and inhibited self-expression in women with anorexia nervosa. It was hypothesized that perfectionism would be positively correlated with measures of inhibited expression of negative feelings and with an externally focused interpersonal orientation.

PARTICIPANTS

Anorexia Nervosa Group

Thirty-five women at various stages of treatment for anorexia nervosa were recruited from inpatient and outpatient programs at a Canadian metropolitan eating disorder clinic. At the time of assessment, 21 met criteria for anorexia nervosa specified in the Eating Disorder Examination (EDE; Cooper & Fairburn, 1987). The average age and years of formal education in this sample wee 29.0 (SD = 9.21; range = 18-47 years) and 14.0 (SD = 2.69; range - 11 - 20 years), respectively. The average age of symptom onset was 15.4 (SD = 7.1; range - 3-38 years). The average length of illness was 5.2 years (SD - 4.65; range - 1- 17 years). The average body mass index (BMI) was 15.3 (SD - 1.95; range = 12.11 - 18.62).

Psychiatric Control Group

Thirty-five women with a variety of psychiatric disorders were recruited from the same Canadian hospital. The diagnostic items of the EDE were used to screen for eating disorder symptomatology. Individuals were also asked if they had received a diagnosis or treatment for an eating disorder in the past. Using these exclusion criteria, 8 women were excluded from the study. Of the remaining 31 women, 21 were matched as closely as possible to the anorexic group on age and education. The mean age and years of formal education in this reduced sample were 38.7 (SD = 6.63; range = 24 - 48 years) and 15.4 (SD = 3.01; range - 9 - 20), respectively. According to clinical diagnoses provided by the therapist, 11 (52.4%) of the women received a diagnosis of major depressive episode, 9 (42.9%) bipolar disorder, and 1 (4.8%) dysthymic disorder. The average age of onset was 30.2 (SD = 7.81; range = 15 - 43 years). The average length of illness was 7.8 years (SD = 7.57; range = 1 -26 years). The average BMI in
this group was 24.4 (SD = 4.66; range = 18.6 - 36.8).

Normal Control Group

Thirty-four women were recruited from hospital staff, restaurant staff, and recreation center staff. Based on the same screening procedures used for the psychiatric control group 1 woman was excluded from the study. Of the remaining participants, 21 were matched by age and education to the anorexic group. The mean age and years of education in this reduced sample were 28.7 (SD = 8.36;range = 19 - 44 years) and 14.4 (SD = 1.73; range = 12 - 18 years) respectively. The average BMI was 22.1 (SD = 3.05; range = 17.7 - 29.4).

MEASURES

EDE

The EDE (Cooper & Fairburn, 1987) is a standardized investigator-based interview that elicits information on attitudes, feelings, and behaviors associated with eating, shape and weight. The interviewer rates the frequency of key behaviors (e.g., eating meals and snacks, restricting, overeating, purging) and the severity of eating disorder features. The EDE is internally consistent (Cooper, Cooper, & Fairburn, 1989) and has demonstrated good concurrent (Rosen, Vara, Wended, & Leitenberg, 1990( and discriminant (Cooper et al., 1989) validity.

The Silencing the Self Scale

The STSS (Jack & Dill, 1992) is a 24-item self-report measure of cognitive schemas pertaining to securing intimate relationships shown to be associated with depression in women. The scale consists of four rationally derived subscales: Externalized Self-Perception or the tendency to judge the self by external standards (e.g., I tend to judge myself by how I think other people see me); Care as Self-Sacrifice or the tendency to secure attachments by putting the needs of others before the self (e.g., caring means putting the other person’s needs in front of my own); Silencing the Self or inhibiting one’s self-expression and action to avoid conflict and possible loss of relationship (e.g., I don’t speak of my feelings in an intimate relationship when I know they will cause disagreement); and Divided Self or the experience of presenting an outer compliant self while the inner self grows angry and hostile (e.g., I find it is harder to be myself when I am in a close relationship than when I am on my own). Items are rated on 5-point scales, ranging from strongly disagree to strongly agree, with higher scores reflecting greater pressure to fulfill the norms of the stereotyped good woman. The STSS was demonstrated good test-retest reliability, internal consistency, and construct validity (Jack & Dill, 1992).

The State-Trial Anger Expression Inventory

The STAXI (Spielberger et al., 1985) consists of 44 items forming six scales, two of which were used for the present research. Anger Out (AX/Out) is an eight-item scale that measures how often anger is expressed toward other people or objects in the environment (e.g., I argue with others). Anger In (AX/In) measures the frequency with which angry feelings are held in or suppressed (e.g., I keep things in). The two scales have demonstrated good internal consistency as well as convergent and divergent validity.

Perceived Body Image Scale

The PBIS (Manley & Le Page, 1988) is a measure of body image dissatisfaction. It consists of 11 cards (5 in. x 7 ½ in.) Containing profile and full frontal outlines of female figures, ranging from emaciated to obese. The PBIS is conducted in the form of a card sort. Participants are asked to make to the following judgements: (1) Which body best represents the way you see yourself when you look in the mirror? (2) Which body best represents the way think you look? (3) Which body best represents the way you feel you are in your body? And (4) Which body best represents the way you would like to look? Participants select one of the PBIS plates in response to each question. A number from 1 to 11 appears on the back of each plate, and this is recorded by the investigator. Difference scores were calculated on the basis of mirror-ideal, think-ideal, and feel-ideal, reflected in the perceptual, cognitive, and affective dimensions of body image, respectively. The PBIS has been shown to discriminate eating groups from normal controls. It also possesses good reliability and validity (Manley & Le Page, 1988; Manley, Tonkin & Hammond, 1988).

Beck Depression Inventory

The BDI (Beck, Ward, Mendelson, Mock, & Erlbaugh, 1961) is a 21-item measure of depression symptoms. Items are rated on 4-point rating scales, with scores ranging from 0 to 63. The BDI demonstrates good internal consistency and concurrent and discriminant validity in clinical and nonclinical samples (Beck, Steer, & Garbin, 1988). It has also been shown to be an adequate measure of depression in eating disorder patients (Pulos 1996).

Rosenberg Self-Esteem Scale

The RSES (Rosenberg, 1965) is a 10-item scale measuring the general self-esteem. Respondents report feelings about the self on a 4-point rating scale. Scores can range from10 40, with higher scores representing lower self-esteem. The reliability and validity of the RSES have been documented (Blasovich & Tomaka, 1991).

Global Assessment Scale

The GAS (Endicott, Spitzer, Fleiss, & Cohen, 1976) is a 100-point scale that provides a summary measure of psychiatric severity. This scale has demonstrated good reliability and validity (Sohlberg, 1989).

Multidimensional Perfectionism Scale

The PMS (Hewitt & Flett, 1991) is a self-report measure of perfectionism traits, composed of three 15-item subscales measuring self-oriented, other-oriented, and socially prescribed perfectionism. Participants rate the extent to which they agree with the statements: One of my goals is to be perfect in everything I do (Self-Oriented Perfectionism). I have high expectations for the people who are important to me (Other-Oriented Perfectionism), and I feel that people are to demanding of me (Socially Prescribed Perfectionism). The MPS has been show to be internally consistent and to possess good test/retest reliability (Hewitt & Flett, 1991).

Perfectionism Self-Presentation Scale

The PSPS (Hewit, Flett, & Ediger, 1996) is a 27-item measure of three dimensions of perfectionistic, self-presentation. The Perfectionistic Self-Promotion subscale (10 items) measures the desire to present oneself as perfect to others (e.g., It is very important that I always appear to be on top of things). The Nondisplay of Imperfection subscale (10 items) measures he need to avoid being seen as less than perfect to others (e.g., I do not want people to see me do something unless I am very good at it). The Nondisclose of Imperfection subscale (seven items) measures the need to avoid admitting failures to others (e.g., ( I try to keep my faults to myself). The PSPS possesses good internal consistency, test-retest reliability, and adequate convergent and discriminant validity.

PROCEDURE

The EDE was administered to women in the anorexia nervosa group. Psychiatric and normal control group women were only administered the diagnostic item of the EDE. All participants completed a questionnaire packet.

RESULTS

Between-Group Comparisons on Demographic Values

In order to determine the extent to which the matching procedure was successful, two analysis of variance (ANOVA) procedures were conducted, comparing age and education across the three groups. The three groups did not differ significantly on education, but did differ on age, F (2.60( = 9.79, p < .001. Tukey’s HSD comparison revealed that psychiatric control women were significantly older than both women with anorexia nervosa and normal controls, who did not differ from one another. As a result, age was controlled through covariance in subsequent between-group analyses.

Between-Group Analyses on Measures of Psychological Distress

To compare overall levels of psychological distress and to ensure the diagnostic distinctiveness of the groups, the three groups were compared on BDI, RSES and GAS scores. Between-group differences were examined using the multivariate analyses of covariance (MANCOVA), with age as the covariate, and followed-up with analyses of covariance (ANCOVAs) and Tukey’s HSD comparisons. Unadjusted means and standard deviations for the three measures are reported on Table 1. The initial MANCOVA was significant, F (t,114) = 2.76, y < .001, and follow-up univariate tests revealed significant group differences on all three variables: BDI, F (2.59) = 90.35, p < .001, RSES, F (2.59) = 43.16, p < .001, and GAS, F (2.59) = 66.56, p < .001. Post-hoc comparisons revealed that women with anorexia nervosa reported significantly more depression than psychiatric control women, with no difference between the two control groups. On the self-esteem score, women in the anorexia group reported lower self-esteem than psychiatric control women, who reported lower self-esteem than normal control women. On the GAS, the same pattern emerged; the anorexic women reported more impairment due to psychiatric disturbance than psychiatric control women, who reported more impairment than normal control women.

Between-Group Comparisons on Inhibited Expression of Negative Emotions and
Interpersonal Orientation

Group differences on the STSS were also examined using MANCOVA, with group status as the independent variable, schema scores as the dependent variables, and age as the covariate. Unadjusted means and standard deviations for the schemas are reported in Table 1. The MANCOVA was significant F (8,108 = 8.23, p .001, and follow-up ANCOVAs were significant for all four schemas: F 2.55) = 21.4, p < .001 for Silence; F (2.55) = 16.0, p < .001 for Care; F (2.55) = 26.8, p < .001 for Divided; and F (2.55) = 24.4, p < .001 for External. Tukey’s HSD tests revealed that on all subscales, women with anorexia nervosa had higher scores than did
psychiatric and normal control women, who did not differ from one another.
A second MANTOVA examining group differences on the STAXI was performed, with age as the covariate. Unadjusted means and standard deviations for AX/In and AX/Out scales are also reported in Table 1. The MANCOVA was significant, F (4,108) = 6.76, p < .001. Tukey’s HSD test revealed that women with anorexia nervosa had higher scores than did psychiatric and normal control women, who did not differ from one another. The follow-up ANCOVA for AX/Out was not significant.
Because depression, self-esteem, and global assessment of functioning were significantly different across group, two additional MANCOVAs were performed, this time controlling for less variables in addition to age on the STS and STAXI scales. Only the STSS MANCOVA was significant, F (8,102) = 2.22, p < .05. Women with anorexia nervosa had higher scores than did psychiatric and normal control women on the Care, F (2.54) = 6.77, p <.05), and Silence F (2.54) =6.77, p < .05), and Silence F (2.54) = 4.10, p < .05 schemas, with no differences between the two control groups.

Body Image Dissatisfaction and Inhibited Expression of Negative Emotions and
Interpersonal Orientation

Correlations between STSS and Anger Expression scales, and thoughts and feelings about the body in the entire sample are shown on Table 2. To ensure that correlations were not falsely inflated due to the diagnostic distinctiveness of the three groups, scatter plots for each pair of correlations were first examined. This revealed no significant deviations from normality. Scatter plots for the entire sample revealed a single cluster (as opposed to two or three distinct clusters) for each pair of variables. Consequently, data from the three groups were pooled together without concern of an inflated correlation due to the three different groups. For each family of correlations, Bonferroni corrections were applied to control for the number of correlations performed (alpha of .05/12 = .004 was used).
As shown in Table 2, AX/In, Care, Silence, and Divided were significantly related to cognitive and affective components of body image dissatisfaction. To determine whether the pattern of relationships in the total sample was the same in women with anorexia, the correlations were performed a second time in the 21 women comprising that group.

Table 2. Correlations between perceptual,cognitive, and affective components of body image dissatisfaction and measures of inhibited expression of negative feelings and interpersonal orientation.

Perceptual

Cognitive

Affective

Anger Expression

Anger In .30 .48 .50
Anger Out .07 .34 .14

Silencing the Self
External .09 .25 .26
Care .26 .42 .39
Silence .18 .43 .39

Divided .27 .49 .46

p < .008

p < .004

Although fewer correlations reached statistical significance, a pattern of results emerged. Suppresion of Negative Emotion, Interpersonal Orientation, and Perfectionism among the entire sample are shown in Table 3. As in the previous set of correlation analyses, distributions of the perfectionism scores and scatter plots for each pair of correlations were examined. The distributions did not deviate significantly from normality, and single, as opposed to grouped, clusters were achieved. The Bonferroni correction procedure was applied to each family of correlations (.05/24 = .002 and .05/12 = .004 for STSS and STAXI correlations, respectively) to avoid Type I errors. AX/In and all four cognitive schemas were significantly related to Self-Oriented and Socially Prescribed Perfectionism, and to all three dimensions of perfectionistic self-presentation. Other-Oriented Perfectionism was unrelated to the STAXI scales or to the STSS schemas. As in the previous set of analyses, a similar pattern of correlations emerged in the anorexia nervosa group.

DISCUSSION

This study examined inhibited expression of negative feelings and interpersonal orientation among women with anorexia nervosa, women with other psychiatric disorders, and normal control women. The significantly higher scores in women with anorexia nervosa, when compared with the two control groups on anger suppression (and not anger expression) and on all the STSS schemas, support the hypothesis that women with anorexia nervosa are prone to suppress negative feelings and minimize their own needs in order to preserve close relationships. Two of the schemas, Care as Self Sacrifice and Silenced Self, remained significantly higher in the anorexia nervosa group than in the two control groups after controlling for depression, self-esteem, and global assessment of functioning. These latter findings suggest that women with anorexia nervosa are particularly inclined to avoid expressing thoughts and feelings when they conflict with those of others, and to give priority to others’ feelings over their own. In sum, these findings indicate that women with anorexia nervosa expend considerable energy in silencing thoughts and feelings.
Inhibited expression of negative feelings and interpersonal orientation were also related to negative feelings and thoughts about the body. A number of explanations for these findings exist. Consistent with the view that the eating disorders are associated with a lack of interoceptive awareness (Garner & Bemis, 1985), body dissatisfaction may reflect a difficulty to clearly identify feelings, and possibly a tendency to blur “pure affect” with “body affect.” Alternately, displacement hypothesis suggest that body image dissatisfaction stems from the avoidance of expressing threatening impulses or feelings toward appropriate targets, and the redirection of such feelings to a less threatening target, in this case, the body. This second hypothesis suggests that body dissatisfaction may be an outlet for unexpressed hostility and sadness. A third explanation is that suppression of negative feelings, outwardly focused interpersonal orientation, and body dissatisfaction all stem from a third variable, low self-esteem. Follow-up analysis did indeed indicate a strong relationship between low self-esteem and each of these constructs. The correlational design of this study only allows for speculation of such mechanisms and does not allow for causality questions to be answered. Future research

Table 3. Correlations between perfectionism scores and measures of inhibited expressions of negative feelings and interpersonal orientation.

MPS Other

MPS Self

MPS Social PSP Disclose PSP Display PSP Promote

Silencing the self
Silence ?.14 .55* .68* .65* .59* .57*
Divided ?.14 .61 .77 .68 .68 .74
Care ?.15 .65 .63 .63 .54 .66
External ?.14 .58 .65 .61 .60 .68

Anger Expression
Anger in ?.13 .58 .70 .64 .63 .62
Anger Out ?.15 .14 .06 .01 .09 .10

Note: MPS Other = Other-Oriented; Perfectionism; MP3 Self = Self-Oriented Perfectionism; MPS Social = Socially-Prescribed Perfectionism; PSP Disclose = Non disclose of Imperfection; PSP Display = Nondisplay of Imperfection; PSP Promote = Perfectionistc Self-Promotion

*p < .002
**p < .004

could follow-up on these findings using longitudinal and/or experimental designs in which the variables of interest are manipulated.

Results from this research are consistent with findings from the health psychology literature, in which a relationship between poor health indicators and suppression of difficult thoughts and feelings has been widely documented. A number of models have been proposed to explain the adverse health effects of inhibited self-expression. Most of these, however, remain largely untested. One model posits that active inhibition or suppression of feelings is a physiologically stressful process that takes a physical toll on the body over time. This model is supported by research linking active inhibition to increased brain activity and to electrodermal and cardiovascular responses (Gray, 1975; Fowles, 1980). Conversely, psychological models emphasize the health-enhancing role of social support, which the individual who inhibits expression of negative feelings may be less likely to receive. For instance, failing to disclose difficult feelings, the ability to make sense of personal experiences, and the opportunity to resolve areas of discordance. Failing to give voice to feelings may also deny the individual the opportunity to receive support and care during difficult times, possibly leaving the individual feeling stigmatized and ashamed for thoughts and feelings perceived to be unique to herself and unacceptable to others.
Given that individuals with anorexia nervosa were particularly reluctant to share difficult thoughts and feelings with others, rather than seeing the aforementioned benefits of self-disclosure, this group may hold the belief that confiding in others is likely to result in a negative outcome. Exploratory analyses in this study with trait and self-presentational measures of perfectionism demonstrated a relationship between the need to be perfect and need to appear to be perfect and inhibited expression and externally focused orientation. Possibly, women with anorexia nervosa view the expression of difficult thoughts and feelings to reflect personal imperfections or character flaws. Given that perfectionism is a central feature of anorexic symptomatology (Cockell et al., 1997; Goldner et al., in press), the greater inhibited self-expression and externally focused interpersonal orientation observed in the anorexia group are not surprising. Future research might investigate whether individuals with anorexia nervosa do indeed believe that the expression of negative feelings and personal needs reflect unattractive and/or undesirable character traits.