Evoked facial emotional expression and emotional experience in people with anorexia nervosa

Preview:

Citation preview

Evoked Facial Emotional Expression and EmotionalExperience in People with Anorexia Nervosa

Helen Davies, BA, BSc1*Ulrike Schmidt, MD, PhD,FRCPsych1

Daniel Stahl, PhD2

Kate Tchanturia, PhD1

ABSTRACT

Objective: To use an experimental

paradigm to assess facial expression, sub-

jective experience of emotion and the

relationship between them in people with

anorexia nervosa (AN).

Method: Film clips are used to elicit

emotion and participants’ facial expres-

sion and subjective experience are

recorded. Thirty inpatients with AN and

34 healthy control (HC) women are

included in the study.

Results: People with AN are less facially

expressive than HC while watching posi-

tive and negative film clips and report

feeling less positive emotion than HC but

not less negative emotion. People with

AN look away significantly more than HC

during the negative film clip. Duration of

illness and depression relate to attenu-

ated positive facial expression and eating

pathology to attenuated negative facial

expression.

Discussion: This experimental study

supports self report studies showing peo-

ple with AN attenuate emotional expres-

sion and avoid negative affect. Such

behavior may affect social interaction

and contribute to the maintenance of

the disorder. VVC 2010 by Wiley Periodi-

cals, Inc.

Keywords: emotion; expression; anorexia

nervosa; eating disorders

(Int J Eat Disord 2011; 44:531–539)

Introduction

Unlike verbal language, where meanings can bearbitrarily conveyed,1 facial expressions containuniversal messages in their performance and intheir perception.2 This is supported by six basicexpression categories being recognized across cul-tures3 as well as in people born deaf and blind.4,5 Inaddition, facial displays such as the eyebrow flash,yawning, startle response, embarrassment, andshame are also universal.4 To this end, it can beunderstood that we have the same facial muscula-

ture and move it in the same way under similar cir-cumstances, denoting facial expression as a behav-ioral phenotype.6 Within this phenotype, there isindividual variation in our ability and tendency toproduce facial expressions based on factors such asgender, culture, and psychopathology.6 For exam-ple, it has been shown that women smile morethan men7,8 and generally ‘‘specialize’’ in expres-sions of happiness.9

Since the nineteenth century, scientists havebeen interested in the function of human facialexpressions, chiefly because of the informationthey convey about emotional experience10,11 andalso because of their critical role in social interac-tion as a visible signal of others’ intentions.6 It islikely that the original function of facial expressionwas a sensory response to stimuli in the environ-ment (e.g., facial expression of fear facilitatesincreased sensory exposure), but has evolved tobecome related to internal experiences.2,10 There-fore, facial expression can be understood as onepart of a coordinated system of emotional respond-ing, which also includes physiological and experi-ential elements.11,12 Emotional responses canbecome dysfunctional when these elements areincongruent with each other,13,14 for example,when experiential responding is inconsistent withexpressivity.

As mentioned above, facial expression has beenco-opted for efficient social communication2 used

This work was supported by the NIHR Biomedical Research

Centre for Mental Health award to the South London and Maudsley

NHS Foundation Trust and the Institute of Psychiatry, King’s College

London and the ARIADNE programme (Applied Research into Ano-

rexia Nervosa and Not Otherwise Specified Eating Disorders)

funded by a Department of Health NIHR Programme Grant for

Applied Research (Reference number RP-PG-0606-1043).

*Correspondence to: Helen Davies, PO47, Institute of Psychiatry,

De Crespigny Park, London, SE5 8AF, United Kingdom.

E-mail: helen.davies@kcl.ac.uk

Accepted 27 June 2010

1 Section of Eating Disorders, Division of Psychological Medicine

and Psychiatry, Institute of Psychiatry, King’s College London,

United Kingdom2 Section of Biostatistics, Division of Psychological Medicine and

Psychiatry, Institute of Psychiatry, King’s College London, United

Kingdom

Published online 18 October 2010 in Wiley Online Library

(wileyonlinelibrary.com). DOI: 10.1002/eat.20852

VVC 2010 Wiley Periodicals, Inc.

International Journal of Eating Disorders 44:6 531–539 2011 531

REGULAR ARTICLE

as a signaling system that benefits signaler and re-ceiver.15 Thus, for example, people associate smil-ing with positive intentions directed toward them(liking)16 and with increased sociability that canbring about positive responses.17 Also, smiling canhave a positive feedback for the signaler, by elicit-ing or reinforcing positive feelings,18 therefore hav-ing an active role in regulating emotion. Theexpression of negative emotion, such as sadness, isjust as important as it provides others with infor-mation about one’s needs thereby signaling trustthat elicits responsiveness (e.g., Ref. 19). Of note,while facial expression can bring about advantagesin social interaction, nonexpression can be activelydisadvantageous (although under some circum-stances the reverse is true). Developmentalresearch shows that depressed mothers’ reducedexpression of positive emotion is associated withincreased anxiety, distress, and disengagement inthe child.20 In a clinical context, it has beenproposed that attenuated emotion expression inpeople with schizophrenia,21 depression,22 or facialparalysis23 may contribute to social difficulties.

People with anorexia nervosa (AN) have signifi-cant and wide ranging impairments in the socio-emotional domain.24–26 As yet, very few studieshave focused on emotion expression in eating dis-orders (EDs), and those that do have used mainlyself-report measures.27,28

Existing studies have focused on the beliefs asso-ciated with the expression of emotions and havefound that while women with AN had similar levelsof negative affect as controls, they reported inhibi-ting the expression of negative feelings in order topreserve relationships and reduce conflict.27,28

Similarly, Lawson et al.29 found that in AN difficultyin describing one’s feelings to others were associ-ated with emotional inhibition cognitions (i.e., thebelief that emotional expression or experience willhave aversive consequences). Other studies haveshown that the expression of positive emotion isalso inhibited in EDs.30–32 In several studies higherlevels of emotional inhibition were found to beassociated with greater ED symptomatology,including weight and shape concern, dietaryrestraint, and binge eating.33,34 Only one study to-date has assessed facial expression in AN andhealthy controls (HC) objectively in response toexplicit food pictures and subliminal affect picturesfrom the International Affective Pictures System(IAPs).35 The only significant difference regardingfacial expressiveness was increased smiles in theHC group in response to food pictures (with noaffective prime). However, the IAPs images havethe limitation of being static images that perhaps

does not allow for naturally occurring emotion toevolve, moreover the affect pictures were shownsubliminally.

In sum, investigating expression of emotions inAN has relied, in the main, on self-report measureswith one study using static images from IAPs toelicit and measure facial expression. Given the im-portance of facial expression in relating to both theself and the social arena, the present study aims toaddress some of these limitations by using a cross-sectional experimental design to measure bothpositive and negative facial emotion expressionand experience of participants during the showingof emotionally salient film clips.

We hypothesize that people with AN will showless facial expression during both positive and neg-ative films clips. With regard to subjective response,there will be no difference in ratings of negativeaffect between AN and HC in response to the nega-tive film clip; however, we expect positive affect tobe attenuated in people with AN in response to thepositive film clip. This is in accordance with find-ings from a recent study,36 which found subjectiveratings to positive stimuli (happiness) attenuatedin people with AN.

Method

Participants

Thirty consecutively referred patients with AN fulfilling

DSM-IV diagnostic criteria37 and 34 HC with no personal

or family history of psychiatric illness took part in the

study. Exclusion criteria for both groups were poor liter-

acy, nonfluent English, or a history of head injury. Inclu-

sion age was 18–60 as the ED unit where recruitment

took place treats people in this age range.

Because no previous study in EDs has used this para-

digm, the sample size power calculation was based on

previously published work in a different clinical group.21

The power calculation indicates that a sample size of 30

participants in each group would have 80% power to

detect significant differences (d 5 0.74) between groups

with a 0.05 two-tailed significance level.

AN patients were recruited from the South London

and Maudsley NHS Foundation Trust’s (SLaM) Eating

Disorders Inpatient Unit. For inclusion in this group, par-

ticipants had to have a Body Mass Index (BMI) below

17.5 kg/m2 and be diagnosed by experienced ED clini-

cians as fulfilling DSM-IV criteria for AN.37 Sixteen par-

ticipants were of the restrictive, six of the binge-purge,

and eight of the purging subtype of AN.

HC participants were recruited via advertisements in

the local community and from University students and

DAVIES ET AL.

532 International Journal of Eating Disorders 44:6 531–539 2011

employees in the host institution. Inclusion criteria

included a BMI of between 19 and 25 (kg/m2), no personal

or family history of psychiatric illness or EDs. Two HC par-

ticipants were excluded from the original 36 screened.

Measures and Procedure

Ethical approval was obtained through the local NHS

ethics committee (ref. 08/H0606/58). All participants

gave written, informed consent to take part. Self-report

questionnaires were completed prior to meeting with the

researcher.

Demographic and Clinical Information

Participants recorded their age, ethnicity, and years in

education. In addition, duration of illness was obtained

from the clinical notes. At the beginning of the assess-

ment, the Structured Clinical Interview Diagnostic tool38

research version was administered by a trained researcher.

This highlighted current and past comorbidity in patients

and was used as a screening tool in HC. The following

self-report measures were completed by all participants.

Eating Disorder Examination Questionnaire

(EDE-Q)39

This 36-item self-report questionnaire is a measure of

psychopathological and behavioral indicators of disor-

dered eating. It is derived from and scored in the same

way as the Eating Disorder Examination interview sched-

ule (EDE).40 The EDE-Q provides a global score and has

four subscales measuring dietary restraint, eating con-

cern, weight concern, and shape concern. Subscale and

global scores range from zero to six, with higher scores

representing greater pathology. The EDE-Q also assesses

frequency of key behavioral indicators of disordered eat-

ing: The EDE-Q has acceptable case detection and con-

current validity in community samples.41 In the current

study, the overall Cronbach a coefficient was .93.

Depression, Anxiety, and Stress Scale (DASS).42 This is

a 21 item measure with 7 items in each of the 3

subscalesa Each subscale has a maximum score of

21. Internal consistency for the current study using

Cronbach a coefficient was .87 for the total score.

Positive and Negative Affect Scale (PANAS).43 This self-

report measure has 22 adjectives designed to

measure current positive and negative affect In

completing the PANAS, participants are instructed

to indicate [using a 0–4 Likert scale (05 very slightly

or not at all, 4 5 extremely)] how they feel at the

present moment for each adjective. In addition to

the 22 items, 2 further items were added asking

whether the film clip had been seen before and

whether the participant felt any other emotion while

watching the clip. Internal consistency for the cur-

rent study using Cronbach a coefficient was .81.

Experimental Stimuli

Film clips have previously been used in other clinical

groups as a means to elicit facial emotion expression (e.g.,

Refs. 21, 44). Films were chosen as an ecologically valid

method for eliciting emotion and do not rely on partici-

pants’ ability to recall past experiences and unlike slides

or photographs provide a more realistic context in which

emotional experiences typically develop over time. The 2

film clips chosen were selected from a total of 11 after

being viewed by 20 people who met HC criteria. The film

clips were selected based on eliciting the highest positive

and negative scores on the PANAS. Specifically, the

positive film clip was taken from ‘‘Four Weddings and a

Funeral’’ and shows a humorous depiction of a wedding

ceremony and the negative film clip was taken from

‘‘Shadowlands’’ and depicts a dying woman saying fare-

well to her son and husband. Each clip lasts �2 min. A

third film clip, of simulated waves, was chosen as a neutral

stimulus. Films were presented in a fixed order of positive,

neutral, and negative. This is based on the premise that

negative affect has a more lasting carry over effect. Stimuli

were presented on a 15-inch computer screen.

Procedure

Subsequent to completing clinical questionnaires and

on a different day, participants took part in the current

paradigm as part of a larger study examining emotion

processing. Participants were instructed to ‘‘allow them-

selves to get into the story’’ as much as possible. This line

of instruction was intended to allow participants to attend

to the film without revealing the true nature of the study

in order to reduce demand characteristics.21 While viewing

each film clip, participants’ faces were recorded with their

knowledge and agreement, using a small camera fixed

onto the computer screen. Following the positive and neg-

ative film clip, participants completed the PANAS.43

Because the neutral film clip was primarily included to

minimize affect and thus reduce carry over effects

between the positive and negative film clips, PANAS rat-

ings were not obtained after the showing of this clip.

Coding the Film Clips

Recordings of participants were coded using the Facial

Expression Coding System (FACES, Kring and Sloan,

Unpublished). Unlike systems based on a discrete emo-

tions theory (e.g., Ref. 45), FACES reflects an alternative

perspective that focuses on global dimensions of positive

and negative affect. In FACES, an expression is defined as

any change in the face from a neutral display (i.e.,

no expression) to a non-neutral display and back to a

neutral display. If the participant shifts from neutral to

aThe DASS 21 is a short form of the 42 item measure. Therefore,

when the final score for each subscale is calculated it is multiplied

by 2.

EMOTIONAL EXPRESSION AND EXPERIENCE IN ANOREXIA NERVOSA

International Journal of Eating Disorders 44:6 531–539 2011 533

non-neutral but does not return to a neutral display but

evidences another clear display of affective expression,

this is coded as an additional expression. Each time a sep-

arate expression occurs coders’ rate whether it is a positive

or negative expression (valence) and a frequency count of

expression is initiated. The intensity of the expression is

also assessed using a 4-point Likert scale (1 5 low, 4 5

very high) as is duration (in seconds). In addition, the fre-

quency of ‘‘looking away’’ is also counted. At the end of

each clip, coders compute the following information (1)

the total number (frequency) of positive expressions, their

mean intensity, and mean duration; (2) the total number

of negative expressions, their mean intensity, and mean

duration; (3) the total number of times looking away.

Two researchers rated the participants’ facial expres-

sions. One researcher was blind to the hypotheses of the

study and to the nature and names of the film clips.

Inter-rater agreement between coders was medium size

(j5 .63), which is regarded as substantial agreement.46

Data Analysis

All analyses were carried out using SPSS version 15.

Repeated measures ANOVAs were used to assess facial

expression and PANAS for main effect of film type and

diagnoses. For significant interactions, planned pairwise

tests were performed for between-group comparison.

Data that were non-normally distributed were trans-

formed using the square root method. For between-

group analyses involving clinical (relevant only to AN)

and demographic variables, Mann–Whitney U tests were

used. Correlation analyses investigated relationships

between facial expression, age, length of illness, BMI,

and comorbidity in each group separately. Spearman’s

rho correlations were used as a nonparametric measure

of association. a was set at p\ .05.

Results

Medians, upper, and lower quartiles and the resultsof Mann–Whitney U tests for demographic, illness,comorbidity characteristics (participant age, BMI,years of education, duration of illness, depression,and anxiety) are presented in Table 1. As expected,the AN group had a significantly lower BMI thanthe HC group but were comparable on years ofeducation and age. The oldest participants in theAN and HC groups were 53 years and 55 years,respectively. Significant group differences emergedfor anxiety and depression. The AN group weremore anxious and depressed than the HC group. Asexpected, a significant difference was observed onthe EDE-Q global score with AN showing signifi-cantly higher scores.

FACES Variables

Correlations between the individual FACES varia-bles (frequency, intensity, and duration), computedseparately for AN and HC are reported in Table 2.Within each of the three film clips, positive expres-sion frequency, positive expression intensity, andpositive expression duration were highly correlated.Similarly, negative expression frequency, negativeexpression intensity, and negative expressionduration were highly correlated (overall averagespearman rho correlation for the AN group: r 5 .93;

TABLE 1. Between group comparisons on demographic and clinical symptomatology

AN, n5 30 HC, n5 34 Test Statistic MWa p-value

Age 24.5 (19–33.5) 23 (19–31.5) 510 .81BMI 14.6 (12.9–15.6) 21 (20.1–22) 0.00 \.001*Duration of illness 9.5 (4.9–18.0) N/A N/A N/ANumber of years in education 15 (12–16) 16 (15–17) 582.00 .071DASS anxiety 13.0 (12–18) 1 (0–1) 6.00 \.001*DASS depression 17 (12–18) 1 (0–3) 5.50 \.001*Total global EDE-Q score 4.9 (3.2–5.8) 0.3 (0.2–0.7) 10.00 \.001*

a MW-test statistics for Mann–Whitney U non-normally distributed data.Medians with upper and lower quartiles are presented.* p\.001.

TABLE 2. Intercorrelations of FACES variables

Positive Expressions Negative Expressions

1 2 3 1 2 3

Happy Film1. Frequency .69** .80** .92** .92**

2. Intensity .57** .80** 1.00** 1.00**

3. Duration .30 .33* 1.00** 1.00**

Sad Film1. Frequency .44* .44* .97** .97**

2. Intensity .98** 1.00** .97** .97**

3. Duration .99** .98** .80** .78**

Neutral Film1. Frequency 1.00** 1.00** 1.00** 1.00**

2. Intensity 1.00** 1.00** 1.00** 1.00**

3. Duration 1.00** 1.00** 1.0** 1.00**

Notes: FACES, Facial Expression Coding System.Correlations for ANs are in bold, above the diagonal.* p\ .05;** p\ .01.1 5 Frequency.2 5 Intensity.3 5 Duration.

DAVIES ET AL.

534 International Journal of Eating Disorders 44:6 531–539 2011

overall average spearman rho correlation for HC:r 5 .87). This shows that there is overall high corre-lation between each of the three dependent varia-bles for positive and negative expression. Becauseof this, and like other studies that have used theFACES system (e.g., Refs. 44, 47) we used ‘‘fre-quency’’ as the prime index of emotional expressiv-ity in order to reduce the number of highly corre-lated dependent variables. Therefore, in the analy-sis ‘‘frequency’’ of positive and negative expressionis the dependent variable. FACES also includes avariable that codes for frequency of looking awayfrom the stimuli. Therefore, we have a separateanalysis for ‘‘frequency of looking away’’ inresponse to the positive, negative, and neutral filmsand this is reported separately in the results.

Frequency of Emotional Expressivity

We conducted a 2 (AN vs. HC) 3 3 (happy, sad,neutral) repeated measures ANOVA separately forfrequency of positive and negative expressiveness,with group as a between subjects factor and filmtype as a within subjects factor.

For positive expressivity there was a significantmain effect for film type, [F(2, 124) 5 146.36, p \.001, partial e2 5 .702]. A significant main effect wasalso found for group, [F(1, 62)5 46.48, p\ .001, par-tial e2 5 .43]. The response pattern to film type dif-fered between groups as the significant film type 3group interaction revealed [F(2, 124, 5 39.41, p \.001, partial e2 5 .39]. A planned pairwise compari-son showed there was a significant differencebetween the groups in the positive film (with amean difference in frequency in the number of posi-tive expressions of 6.5, 95% CI 4.7–8.3, d 5 21.8)with the HC showing more expressivity while therewere no significant differences between groups inthe negative or neutral film type (means and p-val-ues presented in Table 3). Including anxiety as acovariate did not change the main results; however,including depression as a covariate did change the

main findings and the results for the interactionbetween groups became nonsignificant p5 .14.

For negative expressivity, data were transformeddue to non-normal distribution (and unequal var-iances between groups). There were significantmain effects for film type [F(2, 124) 5 41.33, p 5\.001, partial e2 5 .40] and group [F(1, 62) 5 73.33,p \ .001, partial e2 5 .54] and film type 3 groupinteraction [F(2, 124) 5 21.13, p 5 .001, partial e2 5.25]. A further planned pairwise comparisonshowed there was a significant difference betweenthe groups in the negative film (mean difference infrequency of negative expressions: 0.81 trans-formed, 95% CI 0.45–1.20, d 5 0.9) with the HCshowing more expressivity while there was no dif-ference in negative expressivity for the positive orneutral film clip (means and p-values presented inTable 3). Including anxiety and depression as cova-riates, separately, did not change the main results.Therefore, the observed differences between groupscannot be explained by depression or anxiety.

Because of the small number of events and littlevariance in the ‘‘neutral’’ film group, the repeatedmeasures ANOVAS without the neutral film werererun to assess the robustness of the results. Theinteraction between film type and diagnosisremained and the results of the pairwise compari-sons did not change.

Frequency of Looking Away

The FACES coding system also includes a variableto count for frequency of looking away. Data weretransformed due to non-normative distribution(and unequal variances between groups). The varia-bles were entered into a 2 (AN vs. HC) 3 3 (happy,sad, neutral) repeated measures ANOVA with groupas a between subjects factor and film type as awithin subjects factor. A significant main effect wasfound for film type [F(2, 124) 11.28, p\ .001, partiale2 5 .15] and group [F(1, 62) 5 60.16, p\ .001, par-tial e2 5 .49]. The film type 3 group interactionwas significant [F(2, 124) 5 3.81, p \ .025, partial

TABLE 3. FACES variables, PANAS descriptive data, and planned pairwise comparisons

Positive Film Negative Film Neutral Film

AN Mean HC Mean p AN Mean HC Mean p AN Mean HC Mean p

Positive expression* 3.2 (3.2) 9.8 (4.1) \.001** 0.3 (1.4) 0.5 (2.0) .80 0.0 (0.0) 0.0 (0.0) \1Negative expression* 0.3 (0.7) 0.0 (0.2) .06 0.4 (0.7) 2.0 (2.1) \.001** 0.1 (0.4) 0.0 (0.0) .30Looking away* 1.2 (2.8) 0.1 (0.5) .07 4.3 (5.0) 0.4 (0.6) \.001** 3.6 (5.8) 4.7 (12.0) .80PANAS—positive subscale 11.3 (7.2) 18.2 (6.2) \.001** 5.0 (4.7) 7.6 (6.0) .06PANAS—negative subscale 3.2 (4.3) 1.2 (2.8) .06 7.7 (7.9) 6.7 (5.2) .54

Notes: FACES, Facial Expression Coding System; PANAS, Positive and Negative Affect Scale.Means and standard deviations presented of untransformed data.* Frequency of expression.** p\ .05.

EMOTIONAL EXPRESSION AND EXPERIENCE IN ANOREXIA NERVOSA

International Journal of Eating Disorders 44:6 531–539 2011 535

e2 5 .06]. A planned pairwise comparison showedthere was a significant difference between thegroups in the negative film (mean difference in thefrequency of looking away 1.05 transformed, 95% CI.66–1.4, d 5 1.4) with the AN group looking awaysignificantly more. No significant difference wasfound for the positive or neutral films (means andp-values are presented in Table 3). Including anxietyand depression as covariates, separately, did notchange the main results. Therefore, the observeddifferences between groups cannot be explained bydepression or anxiety.

Emotional Experience

Subjective response to the film clips wasascertained for the positive and negative, but notneutral, film clips. Similar to the analyses forexpressivity, to assess group differences in subjec-tive experience the PANAS ratings served asdependent variables in a 2 (AN vs. HC) 3 2 (happyor sad) repeated measures ANOVA conducted sepa-rately for positive affect and negative affect. Forpositive affect, a significant main effect was foundfor film type [F(1, 62) 5 156.46, p5\.001, partial e2

5 .72], for group [F (1,62) 5 239.35, p\ .001, partiale2 5 .79] and for film type 3 group interaction [F(1, 62) 5 9.9, p \ .001, partial e2 5 .13]. Plannedpairwise comparisons reveal significantly higherpositive affect for the HC group in response to thepositive film clip (mean difference 6.9, 95% CI 3.5–10.2, d 5 1.0) with no significant difference in thenegative film clip. Including anxiety as a covariatedid not change the main results; however, includ-ing depression as a covariate did change the mainfindings and the results for the interaction betweengroups became nonsignificant (p5 .84).

For negative affect, there was a significant maineffect for film type [F (1, 62) 5 53.52, p 5 \.001,partial e2 5 .46] but not for group [F(1, 62) 5 1.78,p 5 .18, partial e2 5 .02] and no significant interac-tion for film type 3 group was found [F (1, 62) 5.48, p 5 .49, partial e2 5 .008]. As there was no sig-nificant interaction, further planned pairwise com-parisons were not carried out. However, as shownin Table 3, there was a trend for higher negativeaffect scores to the positive and negative film clipfor the AN group.

The Impact of Demographic, Clinical, and

Comorbid Variables on Congruent Positive and

Negative Facial Expression, Emotion

Experience, and Looking Away

Correlation analyses were run on each groupseparately. Only significant results are presented.These were found only in the AN group. Variables

included age, BMI, length of illness, EDE-Q globalscore and the DASS variables, anxiety and depres-sion. Spearman correlations were used due to non-normal distribution of some of the data.

Age, length of illness, and depression correlatednegatively with congruent positive facial expression(r 5 2.43, p \ .05, r 5 2.49, p \ .05, r 5 2.47,p\ .05, respectively). This means that the older thepatient, the longer the duration of illness and thehigher the score on depression the less positiveexpression is shown to positive stimuli.

The EDE-Q global score correlated negativelywith negative facial expression to negative stimuli(r 5 2.43, p \ .05), suggesting that the higher thelevel of eating pathology the less negative expres-sion to congruent stimuli.

Discussion

The hypotheses for this study were confirmed as itwas observed that during both positive and negativefilm clips, the AN group were less facially expressivethan the HC group while reporting feeling less posi-tive emotion but the same level of negative emotionas HC. There were no significant differences in facialresponse to the neutral film clip, suggesting theemotionally eliciting stimuli evoked a differentiatedresponse. Our findings are consistent with previousresearch using self-report measures, which showthat people with AN do not express themselves.27

They are also consistent with studies using differentmethodologies that found differences in how posi-tive and negative affect is reported by people withAN with reduced positive affect but similar levels ofnegative affect to control participants.36,48

An additional finding, was that the AN partici-pants looked away significantly more during thenegative film clip. One possible explanation couldbe that looking away may be the result of lack ofattention or boredom. However, if this was the caseone would expect the highest levels of looking awayto be in response to the neutral film clip rather thaneither of the other two clips. Interestingly, the HCslooked away mostly in response to the neutral filmclip, although this was not a significant effect. Giventhe AN participants looked away significantly morein response to the negative film clip, this could beconstrued as an attempt to avoid any negative feel-ings the stimulus was evoking. This is supported byprevious studies, which suggest that avoidance inAN is used as a means of reducing affectivestates.31,49 Alternatively, looking away could beused as a way of hiding negative expression. As the

DAVIES ET AL.

536 International Journal of Eating Disorders 44:6 531–539 2011

self-report literature demonstrates,27 showing nega-tive expression is unacceptable to people with ANdue to feared negative consequences. In our studyparticipants were on their own while watching thefilm clips, but would have been aware of theirexpressions being filmed. Fridlund (1994) has sug-gested that even emotional expression that takesplace in solitude involves implicit or imagined audi-ences whereby solitary expression is a means of con-trolling images projected during imagined socialinteractions.15

Although we can only speculate at this stage as tothe reason, or reasons, why looking away may arisein this context in people with AN, it is of note that inother clinical groups (using similar paradigms) thephenomenon has not been reported.21,47 In schizo-phrenia, looking away has been examined with anegative outcome (Kring, personal communication)and in depression has not been examined (Sloan,personal communication). It is worth noting thatthere is evidence from functional magnetic reso-nance imaging studies that people with anxietydemonstrate prefrontal/limbic activation patternsthat indicate compensatory attentional strategiesinvolving avoidance of emotionally evocative stimuli(e.g., Ref. 50). Therefore, using paradigms similar tothe one in this study may give us more clarity as towhether similar behavioral patterns are present inanxiety disorders.

These main findings are supported by correla-tional analyses that inspected the associationsbetween clinical factors (e.g., weight status, illnessduration, ED symptomatology, depression, and anx-iety) and experimental variables. It was observedthat the greater the ED symptomatology the lessnegative facial expression was produced. Withregard to positive expression, it was observed thatlength of illness and age along with depressionrelated to attenuated positive expression. Depres-sion was also found to explain the significant groupdifference with regard to positive expressivity andpositive experience. A similar pattern was found inthe study by Joos (2009) with attenuated positiveaffect in response to positive pictures from the IAPsexplained by depression.36 The association betweendepression and lack of positive facial expression andemotion experience in AN suggests that comorbiddepressive symptomatology may play a greater rolein influencing positive affect than it does negativeexpression that seemingly has more associationwith eating pathology, in this current study.

Although we do not know the degree to whichattenuation of expression is an effortful act, ourfindings do relate to theories of facial expression asa means of regulating emotion.51 Studies in non-

clinical groups have demonstrated that suppres-sion of positive facial expression can be used toattenuate positive feelings,13,51 whereas suppress-ing negative expressions does not alter negativefeelings.13,52 Speculatively, in the present study it isconceivable that people with AN who perceive thatthe experience and expression of emotion is dan-gerous and may have unwanted consequences(Hambrook et al., submitted) have learnt to usenonexpression and looking away as a way to regu-late their emotions.

Limitations and Strengths

The main limitation of this study is lack of a psy-chiatric control group. As attenuation of expres-sion has been observed in other clinical groups,for example, depression, inclusion of such a groupwill be important in order to demonstrate theextent to which these effects are found in AN, EDsglobally or more generally related to psychiatricdysfunction.

As this is an exploratory study and we aimed, atthis stage, to investigate differences in facialexpression and experience between positive andnegative domains, future studies could use stimulithat will elicit discrete emotions such as fear, anger,compassion, or contentment and might alsoexplain deliberate suppression of emotion expres-sion and how this relates to subjective experience.Current findings are also limited to inpatients, thatis, a very severely underweight group. It will be im-portant to see whether change in emotional facialexpressiveness occurs as people’s level of nutritionimproves (however, it is of note that BMI did notshow any correlation with expressiveness in the ANgroup) and to test the paradigm in people recov-ered from AN and before and after treatment. Inaddition, it will be fruitful to cover the other majorcomponent of emotion processing, that of physiol-ogy. Assessment of skin conductance would give ameasure of interoceptive response thus comple-menting the data gathered on subjective experien-tial and expression response.

A strength of this study is that it is using an eco-logically valid approach for evoking emotionalresponses, something which, to our knowledge, hasnot been used to assess emotional response in peo-ple with ED. Film clips allow for naturally occurringemotion to evolve and thus a greater opportunityto identify objective expression and subjective feel-ings. Our findings do not support the only otherstudy, to our knowledge, which has recorded facialbehavior in ED.35 However, they used IAPS picturesand affective pictures were shown subliminally that

EMOTIONAL EXPRESSION AND EXPERIENCE IN ANOREXIA NERVOSA

International Journal of Eating Disorders 44:6 531–539 2011 537

may not be sufficient to obtain explicit measures offacial expression.

Conclusion

Expressing signs of positive and negative emotionis advantageous in social interaction, allows peo-ple to respond, and potentially helps in gettingneeds met.53,54 Using an experimental method,this study has found that people with AN haveattenuated facial expression in response to posi-tive and negative stimuli, supporting previousself-report studies in ED. We have also found thatwhile positive affect is generally attenuated, anincongruence between facial expression and feel-ing is demonstrated for negative affect. Attenuatedemotion expression and looking away could relateto negative beliefs associated with expression, as ameans to reduce affect, or both. Future studiescan answer some of these questions. Whatever thereason, or reasons, for this behavior, avoidance ofemotion is proposed as one of the maintainingfactors in AN.55

The views expressed here are not necessarily those ofthe NIHR.

References

1. Sassure FD. Course in General Linguistics. New York: McGraw-

Hill Book Company, 1966.

2. Susskind JM, Lee DH, Cusi A, Feiman R, Grabski W, Anderson

AK. Expressing fear enhances sensory acquisition. Nat Neurosci

2008;11:843–850.

3. Ekman P, Friesen WV. Constants across cultures in the face and

emotion. J Pers Soc Psychol 1971;17:124–129.

4. Eibl-Eibesfeldt I. Human Ethology. New York: Aldine de Gruyter,

1989.

5. Izard CE. Human Emotions. New York: Plenum, 1977.

6. Schmidt KL, Cohn JF. Human facial expressions as adaptations:

Evolutionary questions in facial expression research. Yearb

Phys Anthropol 2001;44:3–24.

7. Briton NJ, Hall JA. Gender-based expectancies and observer

judgments of smiling. J Nonverbal Behav 1995;19:49–65.

8. Barr CL, Kleck RE. Self-other perception of the intensity of facial

expressions of emotion: Do we know what we show? J Pers Soc

Psychol 1995;68:608–618.

9. Coats EJ, Feldman RS. Gender differences in nonverbal corre-

lates of social status. Pers Soc Psychol Bull 1996;22:1014–

1022.

10. Darwin C. The Expression of the Emotions in Man and Animals.

New York: Oxford University Press, 1872–1998.

11. Ekman P. Facial expressions of emotion: New findings, new

questions. Psychol Sci 1992;3:34–38.

12. Levenson RW. Human emotions: A functional view. In: Ekman P,

Davidson RJ, editors. The Nature of Emotion: Fundamental

Questions. New York: Oxford University Press, 1994, pp. 123–126.

13. Gross JJ, Levenson RW. Hiding feelings: The acute effects of in-

hibiting negative and positive emotion. J Abnorm Psychol

1997;106:95–103.

14. Rosenberg EL, Ekman P. Coherence between expressive and

experiential systems in emotion. Cogn Emot 1994;8:201–229.

15. Fridlund AJ. Human Facial Expression: An Evolutionary View.

San Diego, CA: Academic Press, 1994.

16. Floyd K, Burgoon JK. Reacting to nonverbal expressions of lik-

ing: A test of interaction adaptation theory. Commun Monogr

1999;66:219–239.

17. Sullivan DG, Masters RD. Facial displays and political leader-

ship: Some experimental findings. In: Schubert G, Masters RD,

editors. Primate Politics. Carbondale: Southern Illinois Univer-

sity Press, 1991, pp.188–206.

18. Merkel KE, Schmidt KL, Levenstein RM, VanSwearingen JM,

Bentley BC. Positive affect predicts improved lip movement in

facial movement disorder. Otolaryngol Head Neck Surg

2007;137:100–104.

19. Clark MS, Brissette I. Relationship beliefs and emotion: Recipro-

cal effects. In: Frijda N, Manstead A, Semin G, editors. Emo-

tions and Beliefs: How Feelings Influence Thoughts. Cambridge,

UK: Cambridge University Press, 2000, pp.212–240.

20. Field T. Infants of depressed mothers. Infant Behav Dev

1995;18:1–13.

21. Kring AM, Neale JM. Do schizophrenic patients show a disjunc-

tive relationship among expressive, experiential, and psycho-

physiological components of emotion? J Abnorm Psychol

1996;105:249–257.

22. Berenbaum H, Oltmanns TF. Emotional experience and expres-

sion in schizophrenia and depression. J Abnorm Psychol

1992;101:37–44.

23. VanSwearingen JM, Cohn JF, Bajaj-Luthra A. Specific impair-

ment of smiling increases the severity of depressive symptoms

in patients with facial neuromuscular disorders. Aesthetic Plast

Surg 1999;23:416–423.

24. Zucker N, Losh M, Bulik CM, LaBar KS, Piven J, Pelphrey KA. An-

orexia nervosa and autism spectrum disorders: Guided investi-

gation of social cognitive endophenotypes. Psychol Bull

2007;133:976–1006.

25. Russell TA, Schmidt U, Doherty L, Young V, Tchanturia K.

Aspects of social cognition in anorexia nervosa: Affective and

cognitive theory of mind. Psychiatry Res 2009;168:181–185.

26. Oldershaw A.Emotional theory of mind in anorexia nervosa:

State or trait? PhD Thesis, Institute of Psychiatry King’s College

London, 2009.

27. Geller J, Cockell SJ, Goldner EM. Inhibited expression of nega-

tive emotions and interpersonal orientation in anorexia nerv-

osa. Int J Eat Disord 2000;28:8–19.

28. Iannou K, Fox J. Perception of threat from emotions and its

role in poor emotional expression within eating pathology. Clin

Psychol Psychother 2009;16:336–347.

29. Lawson R, Emanuelli F, Sines J, Waller G. Emotional awareness

and core beliefs among women with eating disorders. Eur Eat

Disord Rev 2008;16:155–159.

30. Jansch C, Harmer C, Cooper MJ. Emotional processing in women

with anorexia nervosa and in healthy volunteers. Eat Behav

2009;10:184–191.

31. Wildes J, Ringham R, Marcus M. Emotion avoidance in patients

with anorexia nervosa: Initial test of a functional model. Int J

Eat Disord 2010;43:398–404.

32. Forbush K, Watson D. Emotional inhibition and personality

traits: A comparison of women with anorexia, bulimia and

normal controls. Ann Clin Psychiatry 2006;18:115–121.

33. Meyer C, Leung N, Barry L, Feo D. Emotion and eating psycho-

pathology: Links with attitudes toward emotional expression

among young women. Int J Eat Disord 2010;43:187–189.

DAVIES ET AL.

538 International Journal of Eating Disorders 44:6 531–539 2011

34. McClean CP, Miller NA, Hope DA. Mediating social anxiety and

disordered eating: The role of expressive suppression. Eat Dis-

ord 2007;15:41–54.

35. Soussignan R, Jiang T, Rigaud D, Royet FP, Schaal B. Sublimi-

nal fear priming potentiates negative facial reactions to food

pictures in women with anorexia nervosa. Psychol Med

2010;40:503–514.

36. Joos A, Cabrillac E, Hartmann A, Wirsching M, Zeeck A. Emo-

tional perception in eating disorders. Int J Eat Disord

2009;42:318–325.

37. American Psychiatric Association. Diagnostic and Statistical

Manual of Mental Disorders,4th ed, Text Revision. Washington,

D.C: APA, 2000.

38. First M, Spitzer R, Gibbon M, Williams J. Structured Clinical

Interview for DSM- IV Axis I Disorders, Research Version. New

York, NY: Biometrics Research, New York State Psychiatric Insti-

tute, 1997.

39. Fairburn CG, Beglin SJ. Assessment of eating disorders: Interview

or self-report questionnaire? Int J Eat Disord 1994;16:363–370.

40. Fairburn CG, Cooper Z, O’Connor M. Eating disorder examina-

tion (Edition 16.0D). In: Fairburn CG, editor. Cognitive Behav-

ior Therapy and Eating Disorders. New York: Guilford Press,

2008, pp. 265–308.

41. Mond JM, Hay PJ, Rodgers B, Owen C, Beumont PJ. Validity of

the Eating Disorder Examination Questionnaire (EDE-Q) in

screening for eating disorders in community samples. Behav

Res Ther 2004;42:551–567.

42. Lovibond SH, Lovibond PF. Manual for the Depression Anxiety

Stress Scales, 2nd ed. Sydney: Psychology Foundation, 1995.

43. Watson D, Clark LA, Tellegen A. Development and validation of

brief measures of positive and negative affect: The PANAS

scales. J Pers Soc Psychol 1988;54:1063–1070.

44. Orsillo S, Batten SV, Plumb JC, Luterek JA, Roessner BM. An ex-

perimental study of emotional responding in women with post-

traumatic stress disorder related to interpersonal violence. J

Trauma Stress 2004;17:241–248.

45. Ekman P, Friesen WV. Facial Action Coding System: A Technique

for the Measurement of Facial Movement. Palo Alto, CA: Con-

sulting Psychologists Press, 1978.

46. Landis JR, Koch GG. The measurement of observer agreement

for categorical data. Biometrics 1977;33:159–174.

47. Sloan DM, Strauss ME, Quirk SW, Sajatovic M. Subjective and ex-

pressive emotional responses in depression. J Affect Disord

1997;46:135–141.

48. Wolf M, Sedway J, Bulik CM, Kordy H. Linguistic analyses of nat-

ural written language: Unobtrusive assessment of cognitive

style in eating disorders. Int J Eat Disord 2007;40:711–717.

49. Corstorphine E, Mountford V, Tomlinson S, Waller G, Meyer C.

Distress tolerance in the eating disorders. Eat Behav

2007;8:91–97.

50. Goldin PH, Manber T, Hakimi S, Canli T, Gross JJ. Neural bases

of social anxiety disorder: Emotional reactivity and cognitive

regulation during social and physical threat. Arch Gen Psychia-

try 2009;66:170-180.

51. Strack F, Martin LL, Stepper S. Inhibiting and facilitating condi-

tions of the human smile: A non-obtrusive test of the facial

feedback hypothesis. J Pers Soc Psychol 1988;54:768–777.

52. Harris CR. Cardiovascular responses of embarrassment and

effects of emotional suppression in a social setting. J Pers Soc

Psychol 2001;81:886–897.

53. Graham S, Huang J, Clark MS, Helgeson VS. The positives of neg-

ative emotions: Willingness to express negative emotions pro-

motes relationships. Pers Soc Psychol Bull 2000;34:394.

54. Fredrickson BL. What good are positive emotions? Rev Gen Psy-

chol 1998;2:300–319.

55. Schmidt U, Treasure J. Anorexia nervosa: Valued and visible. A

cognitive-interpersonal maintenance model and its implications

for research and practice. Br J Clin Psychol 2006;45:343–366.

EMOTIONAL EXPRESSION AND EXPERIENCE IN ANOREXIA NERVOSA

International Journal of Eating Disorders 44:6 531–539 2011 539

Recommended