Action monitoring and perfectionism in anorexia nervosa

Transcription

1 Brain and Cognition 3 (007) Action monitoring and perfectionism in anorexia nervosa Guido L.M. Pieters a,, Ellen R.A. de Bruijn b,c, Yvonne Maas c, Wouter Hulstijn b,c, Walter Vandereycken a,d, Joseph Peuskens a,e, Bernard G. Sabbe c a Behaviour Therapy Department, University Psychiatric Centre Catholic University Leuvan, Kortenberg B-3070, Belgium b NICI (Nijmegen Institute for Cognition and Information), Radboud University, Nijmegen, The Netherlands c CAPRI (Collaborative Antwerp Psychiatric Research Institute), University of Antwerp, Belgium d Psychology Department, Catholic University Leuven, Belgium e Faculty of Physical Education and Physiotherapy, Catholic University Leuven, Belgium Accepted 5 July 00 Available online 7 September 00 Abstract To study action monitoring in anorexia nervosa, behavioral and EEG measures were obtained in underweight anorexia nervosa patients (n D 17) and matched healthy controls (n D 19) while performing a speeded choice-reaction task. Our main measures of interest were questionnaire outcomes, reaction times, error rates, and the error-related negativity ERP component. Questionnaire and behavioral results indicated increased perfectionism in patients with anorexia nervosa. In line with their perfectionism and controlled response style patients made signiwcantly less errors than controls. However, when controlling for this diverence in error rates, the EEG results demonstrated a reduced error-related negativity in the patient group. These seemingly contradictory outcomes of improved performance and reduced error monitoring are discussed in relation with indications of anterior cingulate cortex hypoactivity in anorexia nervosa patients. 00 Elsevier Inc. All rights reserved. Keywords: Anorexia nervosa; Perfectionism; Action monitoring; Error-related negativity; Anterior cingulate 1. Introduction Anorexia nervosa (AN) is characterized in DSM-IV as demonstrating severe weight loss, fear of gaining weight, preoccupation with body appearance, and amenorrhea. Perfectionism is considered to be a key characteristic in AN and to play a role in the etiology and maintenance of the disorder (see e.g. Franco-Paredes, Mancilla-Diaz, Vazquez- Arevalo, Lopez-Aguilar, & Alvarez-Rayon, 005; Shafran, Cooper, & Fairburn, 00; Shafran & Mansell, 001). For example, AN patients score higher on perfectionism, even after long-term weight recovery (Bastiani, Rao, Weltzin, & Kaye, 1995; Casper, 1990; Srinivasagam et al., 1995) and they employ a slow but accurate response style compared * Corresponding author. Fax: address: Guido.Pieters@med.kuleuven.ac.be (G.L.M. Pieters). to healthy controls (Holliday, Tchanturia, Landau, Collier, & Treasure, 005; Kaye, Bastiani, & Moss, 1995). Perfectionist subjects engage in hypervigilant monitoring of outcomes and selectively attend to failure (Shafran et al., 00). Bulik and co-workers (003) showed that elevated concern over mistakes is speciwcally associated with AN. In general, optimal performance requires minimizing the number of errors by continuously monitoring one s actions and their outcomes. Action monitoring can be investigated in more detail by examining the so-called error-related negativity or ERN (Falkenstein, Hohnsbein, Hoormann, & Blanke, 1991; Gehring, Goss, Coles, Meyer, & Donchin, 1993). The ERN is a negative response-locked event-related potential (ERP) component elicited immediately following an error. Originally, the ERN has been interpreted as the outcome of a generic error-detection mechanism, where the error signal is used as input for a remedial action system, enabling performance adjustments for correction or prevention of future 078/$ - see front matter 00 Elsevier Inc. All rights reserved. doi:.1/j.bandc

2 G.L.M. Pieters et al. / Brain and Cognition 3 (007) errors (see e.g. Coles, ScheVers, & Holroyd, 001). Alternatively, the ERN has been described as the rexection of response conxict that arises when two incompatible response tendencies are simultaneously activated (Botvinick, Braver, Barch, Carter, & Cohen, 001; Yeung, Cohen, & Botvinick, 004). A third interpretation of the ERN attributes a more central role to avective or motivational processes in action monitoring, arguing that the ERN (additionally) rexects an avective evaluation of the error (De Bruijn, Hulstijn, Verkes, Ruigt, & Sabbe, 005; Gehring & Willoughby, 00; Luu, Tucker, Derryberry, Reed, & Poulsen, 003; Pailing, Segalowitz, Dywan, & Davies, 00; Yeung, 004). Several source localization studies on ERP data (see e.g. Dehaene, Posner, & Tucker, 1994; Van Veen & Carter, 00; Holroyd, Dien, & Coles, 1998) and functional Magnetic Resonance Imaging (fmri) studies (see e.g. Carter et al., 1998; Kiehl, Liddle, & HopWnger, 000; Ridderinkhof, Ullsperger, Crone, & Nieuwenhuis, 004; Ullsperger & von Cramon, 001) have indicated the anterior cingulate cortex (ACC) as the likely generator of the ERN. All theories seem to agree that the ERN rexects the outcome of an action-monitoring process in which the signal is used to adjust behavior to improve performance and prevent future errors. In speeded choice-reaction tasks, errors are usually the result of premature responding. A possible performance adjustment is then to slow down on the trial following an error, a phenomenon known as posterror slowing (Rabbitt, 19). Alterations in action-monitoring are associated with diverences in the response style people employ. Enhanced ERN amplitudes were for instance found in normal subjects with a controlled response style (Pailing et al., 00), and in non-clinical subjects with obsessive compulsive characteristics (Hajcak & Simons, 00). In line with this, patients suvering from obsessive compulsive disorder (OCD) also showed enhanced ERN amplitudes (Gehring, Himle, & Nisenson, 000; Johannes et al., 001). These increased ERN amplitudes coincide with positron emission tomography (PET) and fmri studies demonstrating increased ACC activity in OCD patients (Adler et al., 000; Ursu, Stenger, Shear, Jones, & Carter, 003). Importantly, reviews on the phenomenological and neurobiological links between OCD and AN, and data from family and genetic studies seem to conwrm that perfectionism is a, possibly genetically transmitted, common vulnerability factor for AN and OCD (see e.g. Anderluh, Tchanturia, Rabe-Hesketh, & Treasure, 003; Bulik & Tozzi, 004; Halmi et al., 000; Monteleone, Brambilla, Bortolotti, & Maj, 000). However, neuroimaging studies in the resting condition in AN, rather suggest hypoperfusion in the ACC (Delvenne et al., 1995; Naruo et al., 001; Takano et al., 001) for a review, see (Frank, Bailer, Henry, Wagner, & Kaye, 004), which persists after weight gain (Kojima et al., 005). The aim of the current study is to investigate action monitoring in AN patients and in matched controls in more detail. In addition to ERP and behavioral measurements of a speeded choice-reaction task, we obtained results from questionnaires to assess perfectionism. Based on previous research (Bastiani et al., 1995; Casper, 1990; Srinivasagam et al., 1995), we expected questionnaire outcomes and behavioral measures to demonstrate increased perfectionism and a more controlled response style in the AN patients. With respect to the ERN analyses two possible outcomes were predicted. The Wrst prediction was based on the earlier Wndings of enhanced ERN amplitudes in patients with OCD. When the controlled response style of patients with AN coincides with increased action monitoring as issued by the ACC, larger ERN amplitudes were also expected for this patient group. On the other hand, when perfectionism is not associated with increased action monitoring as issued by the ACC, we expected smaller ERN amplitudes compared to controls. Such reduction in ERN amplitude would be in line with studies demonstrating ACC hypoactivity in AN patients.. Materials and methods.1. Participants Thirty-six participants (17 AN patients and 19 controls, matched for sex, age, and educational level) took part in the study. Patients were hospitalized in a specialized treatment center for eating disorders during a 1-month period. Inclusion criteria for the patients were: a diagnosis of ANrestricting subtype according to DSM-IV criteria, a body mass index (BMI) below 17.5 kg/m, and absence of psychotic or substance-related disorder. All eligible patients were approached, and none of them refused to participate. Controls were age-matched healthy volunteers showing no indications of an eating disorder during a short interview and on two self-reporting measures (EDI and EDES: see next paragraph) and who had a BMI above 19 kg/m. During the interview, controls were also screened for a history of severe head injury, substance abuse and psychiatric referral, which were used as exclusion criteria. All participants gave written informed consent before participation and the study was approved by the ethical committee of the University Center in Kortenberg... Clinical assessment Patients and controls were measured and weighed on the day of testing. Seven patients were taking psychotropic medication: Wve received selective serotonin reuptake inhibitors (SSRIs), four received a short-acting benzodiazepine as a sleep inductor. The patients did not take this medication within 4 h before testing. All participants completed the following self-report questionnaires: the Symptom Checklist (SCL-90; Derogatis, 1977) and the Beck depression inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 191) were used to assess a wide array of psychiatric symptoms; The eating disorder inventory (EDI; Garner, Olmsted, & Polivy, 1983) and the eating disorder evaluation scale (EDES; Vandereycken, 1993)

3 44 G.L.M. Pieters et al. / Brain and Cognition 3 (007) 4 50 assessed psychological and clinical aspects directly related to eating disorders; the multidimensional perfectionism scale (MPS; Frost, Marten, Lahert, & Rosenblate, 1990) provided an overall assessment of perfectionism as well as core dimensions of perfectionism. This last scale has been proven to discriminate successfully between subjects with and without AN (Srinivasagam et al., 1995). AN patients scored signiwcantly higher than normal controls on the MPS in a - to 4-month follow-up study (Sutandar-Pinnock, Blake, Carter, Olmsted, & Kaplan, 003). Limited research on the psychometricqualities of (3 subscales) of the Dutch translation of the MPS indicates that this translation has similar psychometric qualities as the original version (Soenens, Vansteenkiste, Luyten, Duriez, & Goossens, 005)..3. Design and procedure Participants performed the so-called Xankers task (Eriksen & Eriksen, 1974) in which they had to respond with either their left or right index Wnger to the central letter (H or S) in a congruent (HHHHH and SSSSS) or incongruent (HHSHH and SSHSS) letter string. Equal emphasis was placed on speed and accuracy in the written and verbal task instructions. Because previous ERN studies had demonstrated that accuracy could avect ERN amplitude (see e.g. Gehring et al., 1993), we Wrst calculated individual reaction-time (RT) deadlines. This procedure, which had already been employed successfully in previous studies (see e.g. De Bruijn, Hulstijn, Verkes, Ruigt, & Sabbe, 004), ensures similar performance levels of 75 80% accuracy. The personal maximum RT sets the time-frame within which a subject should respond to avoid visual feedback indicating that the response was late. To this end, the subjects Wrst performed a practice block of 0 trials. During the practice block, the initial RT deadline was set to a relatively liberal limit of 800 ms. After completion of the 0 practice trials, the participant s average RTs and standard deviations (SDs) of the correct responses were computed. Subsequently, the RT deadline for each individual participant was determined by adding 0.5 SD to this mean RT. The experimental phase consisted of six blocks of 0 trials (i.e. 50 congruent and 50 incongruent) with a selfpaced pause halfway through each block. After each block, participants received feedback on performance. Verbal encouragements were given to keep performance accuracy around 75 80%. Each trial began with the presentation of a Wxation point for 0 ms, followed 300 ms later by the stimulus also lasting 0 ms. During the following 900 ms the screen remained blank, after which a visual feedback stimulus appeared for 00 ms. The next trial was presented after an inter-trial interval of 0 ms. The feedback was a yellow, a blue, or a red rectangle indicating whether the response had been correct, incorrect, or too late, respectively. The total duration of the experimental phase was around 40 min including breaks..4. EEG recording The electroencephalogram (EEG) was recorded from 7 tin electrodes mounted in an electrode cap (Electrocap International) referenced to the average of both mastoids. Electrodes were placed at midline (FPz/AFz/Fz/FCz/Cz/Pz/ Oz) and lateral (FP1 /F7 8/F3 4/FC5 /T3 4/C3 4/ CP5 /T5 /P3 4/O1 ) locations. The electro-oculogram (EOG) was recorded from electrodes placed above and below the right eye, and from electrodes placed lateral to both eyes. All electrode impedances were kept below 5 kω. The EEG and EOG signals were ampliwed using a bandpass Wlter between.0 and 30 Hz and digitized at 00 Hz..5. Analyses EOG artifact correction was carried out using the procedure proposed by Gratton, Coles, and Donchin (1983). ERPs for correct and incorrect responses were averaged separately ov-line time-locked to response onset, starting 00 ms before and ending 500 ms after response onset relative to a 0 ms pre-response baseline. ERN amplitude was determined on the averages per individual participant for incorrect responses by subtracting the most positive peak in the time window starting 80 ms before and ending 80 ms after response onset from the most negative peak in the 0 00 ms time window after response onset at electrode Cz, where ERN amplitude was maximal. The P3 component was dewned as the most positive peak in the ms time window after stimulus onset at electrodes Fz, FCz, Cz, and Pz on correct trials only. The Pe component was dewned as the mean amplitude of the diverence wave (incorrect minus correct response) in the ms time window after response onset at electrodes Fz, FCz, Cz, and Pz. Individual averages for RTs and amplitudes were entered in a repeated measures ANOVA with group (controls vs. AN) as between-subject factor. Possible withinsubject factors were congruency (congruent vs. incongruent) and correctness (correct vs. incorrect). Since error rates divered between groups (see Section 3) and might potentially inxuence RTs and ERN amplitudes, Pearson correlation coeycients were used to determine any relationship between error rates and these measures. When a relation was found, the percentage of errors was introduced as a covariate in ANCOVAs with ERN amplitudes and RTs as dependent variables, to control for this possible confounder in the group comparisons. Similarly, correlation coeycients between depression (BDI) scores and ERN amplitudes were calculated to determine possible evects of depression on the ERN. ANOVAs were conducted with the proportions of the diverent types of responses (i.e. correct, error, too late) as dependent variables and group as factor. Post-error slowing was quantiwed as the diverence in RT between correct responses following incorrect responses and correct responses following correct responses. Please note that late

4 G.L.M. Pieters et al. / Brain and Cognition 3 (007) Table 1 Clinical variables (standard deviation) of restricting AN patients and controls AN (n D 17) Controls (n D 19) Age (years) 0. (3.8) 0.5 (3.) Duration of illness (years) 3.7 (3.4) Weight (kg) 40.4 (.5) 0.1 (5.4) b Height (m) 1.7 (0.1) 1.7 (0.1) BMI (kg/m ) 14.5 (1.) 1.9 (.0) b SCL-90 total score 4.7(90.) 1.9 (.8) b SCL-90 anxiety 7.0 (1.1) 14.4 (3.) b SCL-90 agoraphobia 14.8 (7.4) 8.1 (.0) b SCL-90 depression 5. (19.3).5 (.0) b SCL-90 somatization 30.1 (13.4) 17.4 (4.) b SCL-90 inadequacy.8 (.) 13. (3.) b SCL-90 distrust/interpersonal 51.0 (18.) 5.4 (.) b sensitivity SCL-90 hostility 13. (5.8) 8.7 (4.1) a SCL-90 sleep problems 9.3 (4.4) 4.9 (.8) b SCL-90 additional 1.8 (8.) 1.3 (3.0) b BDI total score 8. (13.1) 4.0 (3.) b EDI total score.7 (40.1) 17.0 (1.4) b EDI drive for thinness.0 (8.) 1.9 (.7) b EDI bulimia 0.4 (1.1) 0.4 (0.8) EDI body dissatisfaction 13. (7.7) 7.5 (.) a EDI inevectiveness 14. (8.3) 1.4 (.7) b EDI perfectionism 5.9 (5.8) 1. (1.7) b EDI interpersonal distrust 7. (4.7) 1.1 (.) b EDI introceptive awareness 8.7 (7.4) 0. (1.3) b EDI maturity fears 7.1 (.5).9 (.4) a EDES total score 39. (1.5) 7. (7.5) b EDES preoccupation 5.0 (4.7) 0.3 (4.4) b EDES bulimia 15.9 (4.3) 1.4 (1.4) EDES sexuality 5.4 (4.3) 15.5 (.5) b EDES adjustment 1.8 (5.1) 0.0 (3.3) b MPS overall perfectionism 80.0 (31.7) 51.3 (13.) b MPS concern over mistakes 8.3 (1.9) 14. (4.4) b MPS personal standards.8 (8.5) 1.0 (4.8) b MPS parent expectations 8. (4.5) 8.4 (3.7) MPS parental criticism 7.7 (4.) 5. (.) a MPS doubting of actions 13. (5.) 7.5 (.7) b MPS organization 3.7 (5.5) 19.9 (5.) a a p p 0.01 (t-test between groups). but correct responses were also included in the post-error slowing analyses, as slowing down after an error often results in a late response due to use of the strict RT deadline. Separate t-tests were conducted for the two groups to analyze this diverence. The mean scores on the clinical variables and rating scales were subjected to an independent samples t-test or a Mann Whitney U-test, as appropriate. Pearson correlations between ERN amplitudes and error percentages with MPS perfectionism scores were calculated for both groups. 3. Results 3.1. Clinical variables and questionnaires Table 1 presents mean age, BMI, and scores on clinical rating scales for both groups. AN patients scored signiwcantly higher on MPS overall perfectionism and on all MPS subscales except parent expectations. 3.. Behavioral results: reaction times, proportions of error responses, and performance adjustments Reaction times are summarized in Table. There was no signiwcant group diverence for reaction times [F(1, 34) D 1.48, p D.3] and also the mean individual reaction-time deadlines did not diver between the two groups (control group: 478 ms, AN group: 490 ms; F(1,34) < 1, p D.4). Incorrect responses were faster (38 ms) than correct responses [3 ms; F(1, 34) D 8.48, p <.001] and congruent stimuli (338 ms) were responded to faster than incongruent ones [35 ms; F(1,34) D 9.8, p <.001]. Interestingly, the three-way interaction between group, congruency, and correctness was signiwcant (F (1, 34) D 4., p D.047). Post hoc t-tests revealed that a congruency evect for incorrect responses was present for the control group [19 ms; t D 3., p D.005], but not for the AN group [ 7ms; t D.5, p D.5]. SigniWcant negative correlations were found between error rates and RTs (control group: r D.85, p D.001; AN group: r D.45, p D.088). Therefore, error rate was introduced as a covariate in an ANCOVA with RT as dependent variable. Again, no signiwcant diverence in RT was found between groups: F(1,33) < 1, p D.78. The analyses of the proportions of incorrect and too late trials showed only a signiwcant group diverence for the amount of incorrect responses. More errors were made by Table Behavioral analyses AN (n D 17) Controls (n D 19) Congruent Incongruent Congruent Incongruent Correct RT (ms) 31 (3) 391 (3) 34 (33) 371 (43) Proportion (%) (3.48) 7.41 (7.19) (7.9) 9.1 (8.1) Incorrect RT (ms) 337 (0) 330 (4) 31 (50) 331 (45) Proportion (%).90 (.1) 9.9 (5.01).19 (5.31) (.8) Too late Proportion (%).0 (.91) (.57).41 (.74) (3.94) Overall RTs and proportions for the diverent types of responses to congruent and incongruent stimuli for the AN and the control group. Standard deviations are given in parentheses.

5 4 G.L.M. Pieters et al. / Brain and Cognition 3 (007) 4 50 the control group (11.00%, range: 4 31%) compared to the AN group [.9%, range: 1 1%; F(1, 34) D 7.95, p D.008]. As depicted in Fig. 1, the control group displayed signiwcant post-error slowing [0 ms; t D 3.07, p D.007], while this evect was only marginally signiwcant for the AN group [11 ms; t D.00, p D.03]. Finally, analyses on the proportion of correct and incorrect responses immediately after an error demonstrated that AN patients followed an error signiwcantly less often by another error (4.73%) than controls did [11.7%; F(1,34) D 8., p D.00]. The proportions of correct responses following an error [AN: 7.4%, controls: 74.4%; F(1, 34) D.40, p D.53] or of too late responses following an error [AN: 18.8%, controls:14.49%; F(1,34) D 3.5, p D.080] did not diver between the two groups ERP results Fig. shows the response-locked ERPs for correct and incorrect responses for the control and the AN group. For both groups an ERN is present after incorrect responses. Analyses showed that there was no diverence in ERN amplitude between the control ( 11.9 μv) and the AN group ( 8.3 μv; F(1, 34) D.11, p D.15). However, as the use of the RT deadline did not provide similar error rates for the two groups, we additionally controlled for this diverence in accuracy. Indeed, as can be seen in Fig. 3, ERN amplitudes correlated signiwcantly with error rates in both groups: r D.43 (p D.04) for AN patients and r D.444 (p D.08) for controls. Therefore, the error rate was introduced as a covariate in the comparison of ERN amplitudes between groups, to correct for the signiwcantly lower percentage of errors in the AN group. The ANCOVA for ERN amplitude, with error rate as a covariate, yielded a signiwcant group evect [F(1, 33) D 7.793; p D.009], with a smaller ERN amplitude in the AN patients (.77 μv) than in the controls ( μv). As BDI scores divered signiwcantly between groups, Pearson correlation coeycients between BDI scores and ERN amplitudes were calculated. No signiwcant correlations emerged: r D.44 (p D.07) for AN patients and r D.17 (p D.47) for controls. Importantly, an additional analysis demonstrated that the ERN evect was not caused by benzodiazepine use, as ERN amplitudes remained signiwcantly smaller in AN patients than controls when patients using benzodiazepines (n D 4) were excluded from the analyses [ERN amplitude (non-benzodiazepine) AN D 7.45 μv; F(1,9) D 4.370, p D.045]. Analyses of the stimulus-locked P3 component and the response-locked error positivity (Pe) did not reveal any signiwcant main evects for group (P3: F(1, 34) D 1.87, p D.179; Pe: F < 1), nor any signiwcant interactions with group (both Fs<1) Correlation analyses The error rate did not correlate signiwcantly with MPSperfectionism neither in the control group (r D.9, p D.38), nor in the AN group (r D.13, p D.0). Correlations between ERN amplitudes and MPS-overall perfectionism scores were signiwcant for controls (r D.430; p D.033), but not for AN patients (r D.9; p D.389). 4. Discussion To our knowledge, the present study is the Wrst to investigate the relation between perfectionism and (electrophysiological measures of) action monitoring in patients with AN. In line with previous studies, questionnaire data showed that restricting AN patients score higher than 500 post correct post error 450 Reaction time (ms) AN group control group Fig. 1. Performance adjustments following erroneous responses. Reaction times for correct responses following correct responses (post-correct) and following erroneous responses (post-error) for the control and the AN group. Error bars represent standard deviations.

6 G.L.M. Pieters et al. / Brain and Cognition 3 (007) FCz Cz ERN Pz control group FCz [µv] Cz Pz ERN AN group [ms] response onset = correct response = incorrect response Fig.. Grand average response-locked waveforms for correct and incorrect responses for the control and the AN group at electrodes FCz, Cz and Pz. Responses are given at t D 0ms. Fig. 3. Scatterplots of the relationship between ERN amplitude and error rate in the AN and control group. Please note that since the ERN is a negative peak, larger ERNs correspond to more negative values. controls on most dimensions of perfectionism. Patients committed less errors than controls, while no diverences in RT emerged between the groups. Although questionnaire and behavioral outcomes provide evidence for an increased perfectionist response style in AN patients, these results do not coincide with enhanced ERN amplitudes. On the contrary, when ERN amplitudes are corrected for error rates by introducing this variable as a covariate in the comparison of ERN amplitudes between groups, the AN patients show signiwcantly smaller ERN amplitudes than normal controls. In the correlation analyses, the expected correlation between MPS perfectionism and ERN amplitude emerged in the control group, but not in the AN group. As predicted, AN patients scored considerably higher on self-report measures of perfectionism, including the MPS subscale concerns over mistakes, which seems particularly relevant to action monitoring with three items speciwcally referring to the evect of mistakes on self and others. An increased perfectionist response style in the AN patients is also supported by the Wnding that patients committed signiwcantly less errors than controls. While in previous studies the use of the RT deadline was successful in maintaining error rates comparable over groups or conditions (e.g. De Bruijn et al., 00; De Bruijn et al., 004), it seems more diycult to elicit errors in AN patients than in controls. Hence, the lower error rate likely rexects a more controlled

7 48 G.L.M. Pieters et al. / Brain and Cognition 3 (007) 4 50 response style. Second, post hoc tests revealed that the signiwcant three-way interaction between correctness, congruency and group was caused by the absence of a congruency evect for incorrect responses in the AN group. Fast incorrect responses to congruent stimuli speciwcally result from impulsive responding. Therefore, the absence of these fast incorrect responses in AN patients is indicative of their perfectionist response style. The Wnding that the control group showed post-error slowing, while this performance adjustment is only marginally signiwcant in the AN group, could be interpreted as a rexection of diminished action monitoring in AN patients. However, the proportion analyses of responses following errors contradicts such interpretation. Compared to controls, AN patients gave less erroneous responses following an incorrect response and also showed a tendency to respond more often too late immediately following an error. No signiwcant correlation between error rates and MPS perfectionism scores emerged in neither of the groups. This outcome rexects both the complexity of the (multidimensional) perfectionism concept (Dunkley, Blankstein, Masheb, & Grilo, 00; Hewitt, Flett, Besser, Sherry, & McGee, 003) and the diverence between self-report measures, indicating how participants view themselves, and objective measures, rexecting behavior (Butler & Montgomery, 005). Error rates signiwcantly correlated with ERN amplitudes in both groups, higher ERN amplitudes being associated with lower error rates, a Wnding that has been reported by others as well (see e.g. Santesso, Segalowitz, & Schmidt, 005). A possible interpretation of this relationship is that subjects with a low error rate are more focused on optimal performance. Other analyses did not reveal a signiwcant correlation between ERN amplitudes and BDI scores as a measure of depression in either group. Therefore, it seems unlikely that in the current experiment ERN amplitudes were avected by group diverences in severity of depression. The expected correlation between MPS perfectionism and ERN amplitude emerged in the control group, but not in the AN group. In the AN patients, their perfectionist trait, resulting in decreased error rates, could possibly be associated with a condition that counteracts the increased ERN amplitudes. This condition could for instance be related to the underweight state of the patients in our study. On a neural level, the Wnding of attenuated ERN amplitudes in AN patients, suggests that, during error detection, the ACC is less activated in patients with AN compared to controls. Although this interpretation may seem in conxict with increased perfectionism and a more controlled response style at Wrst, it is in line with Wndings of neuroimaging studies in the resting condition in AN, suggesting hypoperfusion in the ACC (Delvenne et al., 1995; Kojima et al., 005; Naruo et al., 001; Takano et al., 001). Similarly, Ferro et al. (005) conclude from a single photon emission computed tomography (SPECT) study in AN patients during the execution of a Stroop interference task that their results are suggestive of a blunted cingulate function in AN patients in relation to executive function. Hence, the current study suggests that hypoactivity of the ACC, as currently supported by reduced ERN amplitudes, does not necessarily lead to diminished task performance. A possible explanation for this intact performance may be related to recruitment of other brain areas in order to, for instance, increase cognitive control. This explanation of hyperactivity of alternative brain regions, is supported by at least one study in AN, which showed hypoperfusion of the ACC in combination with hyperperfusion in the amygdala hippocampus complex (Takano et al., 001). With respect to healthy subjects, Colcombe et al. (004) recently demonstrated in an fmri study using a similar Xankers task that elderly healthy subjects with higher Wtness showed an increased functioning of the frontal and parietal network. Importantly, this increased recruitment in frontal and parietal regions, associated with attentional selection and response conxict resolution, was associated with a reduced activity in the ACC. An advantage of activating the frontal and parietal network may be that errors are better prevented, so that there is less need for fast adaptive actions issued by the ACC. With respect to the ERP outcomes, the fact that part of the AN group was medicated should be mentioned as a limitation of the present study. The benzodiazepines (halflife values under 1 h) and SSRIs (half-life values under 8 h) were omitted the day before testing, meaning that no sleep inductor was taken within at least 3 h of testing. However, we acknowledge that this medication-free period may not rule out possible inxuences of medication entirely. Previous research has shown that psychopharmacological compounds may avect ERP correlates of action monitoring. While administration of a benzodiazepine to healthy volunteers caused attenuation of ERN amplitudes along with a slowing of responses, an SSRI neither avected ERN amplitudes nor behavioral or other ERP measures (De Bruijn et al., 004; De Bruijn, Sabbe, Hulstijn, Ruigt, & Verkes, 00). Therefore, we additionally analyzed the subset of AN patients who were not taking a benzodiazepine. This analysis demonstrated that the ERN amplitudes were also reduced for the benzodiazepine-free subset of patients. As a result, it seems unlikely that the current ERP results were avected by psychotropic drugs. Another limitation of the present study might be the absence of data that enable us to provide the exact amount of error corrections, as we only recorded the Wrst response during the experiment. While the question whether error correction avects the ERN is still a matter of debate (see e.g. Fiehler, Ullsperger, & Von Cramon, 005; Rodriguez- Fornells, Kurzbuch, & Münte, 00), we believe that the inxuence of error-correction related processes was limited or non-existent in the current study. Correcting was not an option in the present experiment. Participants were instructed that a corrective response would serve no purpose as it would always be disregarded. Moreover, this was always conwrmed by the trial-to-trial feedback indicating

8 G.L.M. Pieters et al. / Brain and Cognition 3 (007) whether the given response was correct, incorrect, or too late. As a result, participants almost never corrected their erroneous responses, thus severely limiting any possible evects on the current measures of interest. Further research seems warranted to elucidate the possible role of altered action monitoring in the development of AN. Direct comparison of patients with diverent types of eating disorders thought to vary in perfectionism and impulsivity, could shed further light on the role of actionmonitoring processes in the psychopathology of eating disorders. Neuroimaging studies in AN patients during the speeded-choice reaction task used in the present study could conwrm the hypothesized ACC hypofunctioning with increased functioning in prefrontal and parietal cortical regions of the attentional network of the brain. Acknowledgments G.P. and E. de B. have equally contributed to this paper. The authors thank L. Gielen and CEPOS, Psychiatric Centre St-Norbertus, DuVel (Dr. P. Van Coillie and Dr. H. Hauman) for their help in data collection and technical assistance. References Adler, C. M., McDonough-Ryan, P., Sax, K. W., Holland, S. K., Arndt, S., & Strakowski, S. M. (000). FMRI of neuronal activation with symptom provocation in unmedicated patients with obsessive compulsive disorder. Journal of Psychiatric Research, 34, Anderluh, M. B., Tchanturia, K., Rabe-Hesketh, S., & Treasure, J. (003). Childhood obsessive compulsive personality traits in adult women with eating disorders: dewning a broader eating disorder phenotype. American Journal of Psychiatry,, Bastiani, A. M., Rao, R., Weltzin, T., & Kaye, W. H. (1995). Perfectionism in anorexia nervosa. International Journal of Eating Disorders, 17, Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (191). An inventory for measuring depression. Archives of General Psychiatry, 4, Botvinick, M. M., Braver, T. S., Barch, D. M., Carter, C. S., & Cohen, J. D. (001). ConXict monitoring and cognitive control. Psychological Review, 8, 4 5. Bulik, C. M., & Tozzi, F. (004). Genetics in eating disorders: state of the science. CNS Spectrums, 9, Bulik, C. M., Tozzi, F., Anderson, C., Mazzeo, S. E., Aggen, S., & Sullivan, P. F. (003). The relation between eating disorders and components of perfectionism. American Journal of Psychiatry,, Butler, G. K. L., & Montgomery, A. M. J. (005). Subjective self-control and behavioural impulsivity coexist in anorexia nervosa. Eating Behaviors,, 1 7. Carter, C. S., Braver, T. S., Barch, D. M., Botvinick, M. M., Noll, D., & Cohen, J. D. (1998). Anterior cingulate cortex, error detection, and the online monitoring of performance. Science, 80, Casper, R. C. (1990). Personality features of women with good outcome from restricting anorexia nervosa. Psychosomatic Medicine, 5, Colcombe, S. J., Kramer, A. F., Erickson, K. I., Scalf, P., McAuley, E., Cohen, N. J., et al. (004). Cardiovascular Wtness, cortical plasticity, and aging. Proceedings of the National Academy of Sciences of the United States of America, 1, Coles, M. G. H., ScheVers, M. K., & Holroyd, C. B. (001). Why is there an ERN/Ne on correct trials? Response representations, stimulus-related components, and the theory of error-processing. Biological Psychology, 5, Dehaene, S., Posner, M. I., & Tucker, D. M. (1994). Localization of a neural system for error-detection and compensation. Psychological Science, 5, DeBruijn, E. R. A., Grootens, K. P., Verkes, R. J., Buchholz, V., Hummelen, J. W., & Hulstijn, W. (00). Neural correlates of impulsive responding in borderline personality disorder: ERP evidence for reduced action monitoring. Journal of Psychiatric Research, 40, De Bruijn, E. R. A., Hulstijn, W., Verkes, R. J., Ruigt, G. S. F., & Sabbe, B. G. C. (004). Drug-induced stimulation and suppression of action monitoring in healthy volunteers. Psychopharmacology (Berl), 177, 151. De Bruijn, E. R. A., Hulstijn, W., Verkes, R. J., Ruigt, G. S. F., & Sabbe, B. G. C. (005). Altered response evaluation: Monitoring of late responses after administration of D-amphetamine. Journal of Psychophysiology, 19, De Bruijn, E. R. A., Sabbe, B. G. C., Hulstijn, W., Ruigt, G. S. F., & Verkes, R. J. (00). EVects of antipsychotic and antidepressant drugs on action monitoring in healthy volunteers. Brain Research, 15, Delvenne, V., Lotstra, F., Goldman, S., Biver, F., De Maertelaer, V., Appelboom-Fondu, J., et al. (1995). Brain hypometabolism of glucose in anorexia nervosa: a PET scan study. Biological Psychiatry, 37, Derogatis, L. R. (1977). SCL-90-R manual. Baltimore: Johns Hopkins University Press. Dunkley, D. M., Blankstein, K. R., Masheb, R. M., & Grilo, C. M. (00). Personal standards and evaluative concerns dimensions of clinical perfectionism: a reply to Shafran et al. (00, 003) and Hewitt et al. (003). Behaviour Research and Therapy, 44, Eriksen, B. A., & Eriksen, C. W. (1974). EVects of noise letters upon identi- Wcation of a target letter in a nonsearch task. Perception & Psychophysics, 1, Falkenstein, M., Hohnsbein, J., Hoormann, J., & Blanke, L. (1991). EVects of crossmodal divided attention on late ERP components. II. Error processing in choice reaction tasks. Electroencephalography and Clinical Neurophysiology, 78, Ferro, A. M., Brugnolo, A., De Leo, C., Dessi, B., Girtler, N., Morbelli, S., et al. (005). Stroop interference task and single-photon emission tomography in anorexia: a preliminary report. International Journal of Eating Disorders, 38, Fiehler, K., Ullsperger, M., & Von Cramon, D. Y. (005). Electrophysiological correlates of error correction. Psychophysiology, 4, 7 8. Franco-Paredes, K., Mancilla-Diaz, J. M., Vazquez-Arevalo, R., Lopez- Aguilar, X., & Alvarez-Rayon, G. (005). Perfectionism and eating disorders: a review of the literature. European Eating Disorders Review, 13, Frank, G. K., Bailer, U. F., Henry, S., Wagner, A., & Kaye, W. H. (004). Neuroimaging studies in eating disorders. CNS Spectrums, 9, Frost, R., Marten, P., Lahert, C., & Rosenblate, R. (1990). The dimensions of perfectionism. Cognitive Therapy and Research, 14, Garner, D., Olmsted, M., & Polivy, J. (1983). Development and validation of a multidimensional eating disorder inventory for anorexia nervosa and bulimia. International Journal of Eating Disorders,, Gehring, W. J., Goss, B., Coles, M. G. H., Meyer, D. E., & Donchin, E. (1993). A neural system for error-detection and compensation. Psychological Science, 4, Gehring, W. J., Himle, J., & Nisenson, L. G. (000). Action-monitoring dysfunction in obsessive compulsive disorder. Psychological Science, 11, 1. Gehring, W. J., & Willoughby, A. R. (00). The medial frontal cortex and the rapid processing of monetary gains and losses. Science, 95, Gratton, G., Coles, M. G., & Donchin, E. (1983). A new method for ov-line removal of ocular artifact. Electroencephalography and Clinical Neurophysiology, 55,

9 50 G.L.M. Pieters et al. / Brain and Cognition 3 (007) 4 50 Hajcak, G., & Simons, R. F. (00). Error-related brain activity in obsessive compulsive undergraduates. Psychiatry Research, 1, 3 7. Halmi, K. A., Sunday, S. R., Strober, M., Kaplan, A., Woodside, D. B., Fichter, M., et al. (000). Perfectionism in anorexia nervosa: variation by clinical subtype, obsessionality, and pathological eating behavior. American Journal of Psychiatry, 157, Hewitt, P. L., Flett, G. L., Besser, A., Sherry, S. B., & McGee, B. (003). Perfectionism is multidimensional: a reply to Shafran, Cooper and Fairburn. Behaviour Research and Therapy, 41, Holliday, J., Tchanturia, K., Landau, S., Collier, D., & Treasure, J. (005). Is impaired set-shifting an endophenotype of anorexia nervosa? American Journal of Psychiatry, 1, Holroyd, C. B., Dien, J., & Coles, M. G. (1998). Error-related scalp potentials elicited by hand and foot movements: Evidence for an outputindependent error-processing system in humans. Neuroscience Letters, 4, 5 8. Johannes, S., Wieringa, B. M., Nager, W., Rada, D., Dengler, R., Emrich, H. M., et al. (001). Discrepant target detection and action monitoring in obsessive compulsive disorder. Psychiatry Research, 8, 1 1. Kaye, W. H., Bastiani, A. M., & Moss, H. (1995). Cognitive style of patients with anorexia nervosa and bulimia nervosa. International Journal of Eating Disorders, 18, Kiehl, K. A., Liddle, P. F., & HopWnger, J. B. (000). Error processing and the rostral anterior cingulate: an event-related fmri study. Psychophysiology, 37, 1 3. Kojima, S., Nagai, N., Nakabeppu, Y., Muranaga, T., Deguchi, D., Nakajo, M., et al. (005). Comparison of regional cerebral blood Xow in patients with anorexia nervosa before and after weight gain. Psychiatry Research, 140, Luu, P., Tucker, D. M., Derryberry, D., Reed, M., & Poulsen, C. (003). Electrophysiological responses to errors and feedback in the process of action regulation. Psychological Science, 14, Monteleone, P., Brambilla, F., Bortolotti, F., & Maj, M. (000). Serotonergic dysfunction across the eating disorders: relationship to eating behaviour, purging behaviour, nutritional status and general psychopathology. Psychological Medicine, 30, Naruo, T., Nakabeppu, Y., Deguchi, D., Nagai, N., Tsutsui, J., Nakajo, M., et al. (001). Decreases in blood perfusion of the anterior cingulate gyri in anorexia nervosa Restricters assessed by SPECT image analysis. BMC Psychiatry, 1,. Pailing, P. E., Segalowitz, S. J., Dywan, J., & Davies, P. L. (00). Error negativity and response control. Psychophysiology, 39, Rabbitt, P. M. (19). Error correction time without external error signals. Nature, 1, 438. Ridderinkhof, K. R., Ullsperger, M., Crone, E. A., & Nieuwenhuis, S. (004). The role of the medial frontal cortex in cognitive control. Science, 30, Rodriguez-Fornells, A., Kurzbuch, A. R., & Münte, T. F. (00). Time course of error detection and correction in humans: neurophysiological evidence. Journal of Neuroscience,, Santesso, D. L., Segalowitz, S. J., & Schmidt, L. A. (005). ERP correlates of error monitoring in -year olds are related to socialization. Biological Psychology, 70, Shafran, R., Cooper, Z., & Fairburn, C. G. (00). Clinical perfectionism: a cognitive-behavioural analysis. Behaviour Research and Therapy, 40, Shafran, R., & Mansell, W. (001). Perfectionism and psychopathology: a review of research and treatment. Clinical Psychology Review, 1, Soenens, B., Vansteenkiste, M., Luyten, P., Duriez, B., & Goossens, L. (005). Maladaptive perfectionistic self-representations: the mediational link between psychological control and adjustment. Personality and individual diverences, 38, Srinivasagam, N. M., Kaye, W. H., Plotnicov, K. H., Greeno, C., Weltzin, T. E., & Rao, R. (1995). Persistent perfectionism, symmetry, and exactness after long-term recovery from anorexia nervosa. American Journal of Psychiatry, 15, Sutandar-Pinnock, K., Blake, W. D., Carter, J. C., Olmsted, M. P., & Kaplan, A. S. (003). Perfectionism in anorexia nervosa: a 4-month follow-up study. International Journal of Eating Disorders, 33, 5 9. Takano, A., Shiga, T., Kitagawa, N., Koyama, T., Katoh, C., Tsukamoto, E., et al. (001). Abnormal neuronal network in anorexia nervosa studied with I-13-IMP SPECT. Psychiatry Research, 7, Ullsperger, M., & von Cramon, D. Y. (001). Subprocesses of performance monitoring: a dissociation of error processing and response competition revealed by event-related fmri and ERPs. Neuroimage, 14, Ursu, S., Stenger, V. A., Shear, M. K., Jones, M. R., & Carter, C. S. (003). Overactive action monitoring in obsessive compulsive disorder: evidence from functional magnetic resonance imaging. Psychological Science, 14, Van Veen, V., & Carter, C. S. (00). The timing of action-monitoring processes in the anterior cingulate cortex. Journal of Cognitive Neuroscience, 14, Vandereycken, W. (1993). The eating disorder evaluation scale (EDES). Eating Disorders: The Journal of Treatment and Prevention, 1, Yeung, N. (004). Relating cognitive and avective theories of the errorrelated negativity. In M. Ullsperger & M. Falkenstein (Eds.), Errors, conxicts, and the brain. Current opinions on performance monitoring (pp. 3 70). Leipzig: MPI of Cognitive Neuroscience. Yeung, N., Cohen, J. D., & Botvinick, M. M. (004). The neural basis of error detection: conxict monitoring and the error-related negativity. Psychological Review, 111,