Responsibility and obsessive–compulsive disorder: An experimental test

ARTICLE IN PRESS

0005-7967/$ - se

doi:10.1016/j.br

�CorrespondE-mail addr

Behaviour Research and Therapy 45 (2007) 425–435

www.elsevier.com/locate/brat

Responsibility and obsessive–compulsive disorder:An experimental test

Arnoud Arntz�, Marisol Voncken, Ank C.A. Goosen

Department of Medical, Clinical and Experimental Psychology, Maastricht University, P.O. Box 616,

NL-6200 MD Maastricht, The Netherlands

Received 27 January 2006; received in revised form 27 March 2006; accepted 28 March 2006

Abstract

To test the causal status of responsibility in obsessive–compulsive disorder (OCD), an experiment was executed in which

responsibility was experimentally manipulated. OCD patients, non-OCD anxiety controls, and non-patients executed a

classification task in either a high or a low responsibility (LoRes) condition. Subjective ratings related to danger and

responsibility indicated that the manipulation was successful. Subjective OCD-like experiences and checking behaviors

were higher in OCD patients in the high responsibility (HiRes) condition than in all other groups. Although the checking

subscale of the Padua Inventory correlated with subjective ratings in the OCD patients in the HiRes condition, it was not

associated with checking behaviors. The results confirm the hypothesis that responsibility plays a causal role in OCD.

r 2006 Elsevier Ltd. All rights reserved.

Keywords: Obsessive–compulsive disorder; Cognitive models; Cognitive therapy; Responsibility; Guilt

Introduction

Recent cognitive models of obsessive–compulsive disorder (OCD) assign a causal role to responsibility andguilt in the development and maintenance of this disorder. These views state that people with OCD have aninflated sense of responsibility and are afraid to make mistakes for which they can be blamed. It ishypothesized that this makes them vulnerable to develop OCD and maintains the disorder (Rachman, 1993,2002; Salkovskis, 1985; Salkovskis & Forrester, 2002; Mancini & Gangemi, 2004; van Oppen & Arntz, 1994).More specifically, elevated responsibility drives people with OCD, and those who are at risk to develop OCD,to repeatedly check to avert to be responsible for harms to other people or themselves (e.g., Rachman, 2002).If elevated responsibility is indeed a causal factor in OCD, it follows that people with OCD, when placed in asituation in which they have a personal responsibility for averting a threat, will, more than other people,engage in various types of behaviors that serve to reduce the risk that they will make a fatal mistake oroverlook something essential. As they strive for a complete reduction of the risks, they will repeat thisbehavior and compulsive rituals may develop. Uncertainty about whether or not a specific behavior is actually

e front matter r 2006 Elsevier Ltd. All rights reserved.

at.2006.03.016

ing author.

ess: [email protected] (A. Arntz).

ARTICLE IN PRESSA. Arntz et al. / Behaviour Research and Therapy 45 (2007) 425–435426

executed in a specific case increases with the repetition of the behavior (van den Hout & Kindt, 2003), furtherfuelling the repetition of the compulsive behavior.

The idea that OCD is related to inflated responsibility and fear of guilt has early roots in psychodynamicthinking, in which OCD has been associated with problems with a too stringent superego (Freud, 1926/2001).There is indeed empirical evidence for the idea that responsibility plays a role in OCD. OCD patients and non-patients with OCD-symptoms tend to score higher on measures of responsibility and guilt (Freeston,Ladouceur, Gagnon, & Thibodeau, 1993; Freeston, Ladouceur, Thibodeau, & Gagnon, 1992; Frost, Steketee,Cohn, & Griess, 1994; Menzies, Harris, Cumming, & Einstein, 2000; Rheaume, Ladouceur, Freeston, &Letarte, 1994; Rheaume, Freeston, Dugas, Letarte, & Ladouceur, 1995; Salkovskis et al., 2000; Shafran,Watkins, & Charman, 1996; Steketee, Frost, & Cohen, 1998; Wilson & Chambless, 1999; Ladouceur et al.,1995; Rachman, Thordarson, Shafran & Woody, 1995). This association seems to be specific for OCD,especially for checkers, and not to be characteristic of anxiety disorders in general (Foa, Amir, Bogert, Molnar& Przeworski, 2001; Foa, Sacks, Tolin, Prezworski, & Amir, 2002; Salkovskis et al., 2000). When reassuredthat the experimenter takes all the responsibility, OCD patients report a reduced urge to execute their rituals(Lopatka & Rachman, 1995; Shafran, 1997). Induction of responsibility in non-patients leads to an increase inOCD-like behavior compared to control conditions (Bouchard, Rheaume, & Ladouceur, 1999; Ladouceur etal., 1995; Ladouceur, Rheaume, & Aublet, 1997; Mancini, D’Olimpio, & Cieri, 2004). As research so far iseither correlational, addresses reduction of already developed OCD-related urges, or has investigated effects ofheightened responsibility in non-clinical subjects, it does not prove that high personal responsibility is aspecific and pivotal factor in the development and expansion of compulsive behavior in OCD patients.

The present study aimed to test the hypothesis that people with OCD are vulnerable to situations in whichthey have a high personal responsibility to avert danger, in the sense that they will experience more subjectiveresponsibility and will engage more in OCD-like behaviors than other people. To test this hypothesis, anexperiment was designed in which participants were either placed in a situation with high (HiRes) or in asituation with low responsibility (LoRes). The situation was entirely new for every participant, i.e. theexperimental situation did not resemble situations the participants were familiar with. Thus, possible OCD-like phenomena triggered in the experiment were new and not merely repetitions of earlier OCD symptomsrelated to similar situations. Both subjective and objective OCD-indices were measured. To test the disorder-specificity of the hypothesized effect, both an anxiety disorder control group (Anx) and a non-patient controlgroup (NonPt) were employed. We hypothesized that OCD patients would develop the most OCD-likephenomena in the HiRes condition. We further hypothesized that the LoRes manipulation would not induceOCD-like phenomena in OCD patients. To test the hypothesis that responsibility only plays a role in patientswith checking rituals, different OCD dimensions were related to OCD-like phenomena elicited by theexperimental induction.

Methods

Subjects

Twenty-seven patients with OCD as first diagnosis (14 male, mean age 34.8 years), 37 with another anxietydisorder as first diagnosis and no (subthreshold) OCD as secondary diagnosis (18 men, mean age 35.7 years),and 28 non-patients, meeting no axis-1 diagnosis (14 men, mean age 35.0 year) participated on a voluntarybasis. Patients were recruited just before or during their start of treatment at the Community Mental HealthCenter of Maastricht and at the Vincent van Gogh Institute in Venray, The Netherlands. Diagnoses wereestablished with the SCID-I for DSM-IV, Dutch version (First, Spitzer, Gibbon, & Williams, 1997; vanGroenestijn, Akkerhuis, Kupka, Schneider, & Nolen, 1999). Potential participants with a medical/pharmacological profession/education were excluded (e.g., medical doctors, nurses, pharmacologists). Furtherexclusion criteria were color-blindness, severe visual problems (not correctable by visual aids), psychosis,IQo80, illiteracy, insufficient understanding of Dutch. All participants gave informed consent. TheMaastricht University ethical committee approved the study.

In the Anx control group, 21 patients had a panic disorder as first diagnosis, 7 a social phobia, 5 ageneralized anxiety disorder, 2 a specific phobia, 1 a PTSD and 1 agoraphobia without panic. Differences in

ARTICLE IN PRESSA. Arntz et al. / Behaviour Research and Therapy 45 (2007) 425–435 427

gender, age, and educational level between groups, conditions and group by condition were all NS. The OCDgroup differed from the other groups on all Padua subscales in the expected direction, t’s43.80, p’so0.001 forall OCD—non-patient and OCD-Anx control contrasts, with one exception: on the impulses scale the OCD-Anx controls’ difference was more modest, but in the expected direction, t ¼ 1:99, p ¼ 0:05. Within the OCDgroup, the differences between the experimental conditions on the Padua subscales were not significant.

Design

A 2� 3 factorial design was used, with two experimental conditions (HiRes vs. LoRes) and three groups(OCD patients, Anx patients and NonPt controls). Participants were randomly assigned to condition.

Procedure

Participants were individually tested. They were seated on an office chair at a rectangular table. After givingconsent, the participant filled out the Padua Inventory. Next, the experimenter gave the following instruction:

In a moment I will start the video recorder and leave the room. After I have left the room, you have to startthe cassette recorder by pressing the ‘start’ button. On the cassette it is explained what the aim of theinvestigation is and how the classification task should be executed. At a certain moment you will hear onthe cassette that you can stop the recorder. Then you push the ‘stop’ button. After this, you can start withthe classification task. For the sake of completeness, I will explain the task, although the cassette will alsogive the instructions. You will be asked to sort these pills (points to the large pot with pills) according tocolor in the small pots. You take only one pill each time and you do not look into the large pot. If youdoubt you can check the small pots and change the pills in these pots. Please remain seated during the taskand do not stand or walk around. Just leave the pots on their place. After putting all the pills from the largepot into the small pots you may set aside the pots and fill out this questionnaire. Do not throw the pills backinto the large pot, but let them stay in the small pots. On the first page of the questionnaire it is explainedhow it should be filled out (shows the questionnaire and pencil and puts them aside). When you have filledout all the questions you can call me. I will be in the room next to this room.

After this explanation, the experimenter started the video camera and left the room. The participant startedthe recorder and listened to either an instruction meant to induce low levels of responsibility instruction, or aninstruction meant to induce high levels of responsibility (see Materials for the instructions). After the task wasfinished, the subject filled out a series of questions. At the end participants were debriefed.

Materials

Experimental task

This was a modification of the task developed by Ladouceur et al. (1995). A large pot with 200 capsules of11 color combinations was placed 20 cm before the participant on the table. Behind this pot 15 smaller potswere placed in a straight line. A video camera placed before the participant at the other side of the tablevideotaped the participants’ hands.

Pilots revealed that the original HiRes instruction failed to induce heightened feelings of responsibility.After modifying the instruction (by adding a passage in which is told that earlier attempts to test the colorsystem failed because participants did not take the task seriously, and the present participants are urged to dotheir very best to make no mistakes so that the color system could be approved by the authorities), pilotsrevealed that the induction was successful. The instructions were as follows.

Low responsibility instruction

Similar to Ladouceur et al.’s (1995) LoRes instruction, this instruction explained to the participants that theinvestigators were specialized in research of the perception of colors and were developing a color-system formedicines. Before the proper test, participants were asked to practise the classification task. It was stressed

ARTICLE IN PRESSA. Arntz et al. / Behaviour Research and Therapy 45 (2007) 425–435428

that the real test would only begin after practicing. Next, an instruction similar as the experimenter gave wasgiven:

In front of you there is a large pot with pills and 15 small pots. The aim is to classify the pills from the largepot into the small pots according to color. You are allowed to take only one pill each time, without lookinginto the large pot. Every pill should be put into one of the smaller pots. You may only put one type of pill ina small pot. We ask you to order the pills as quickly and as accurately as you can. If you doubt you cancheck the small pots and change the pills in these pots. It is important that you complete the whole task.Immediately after finishing the task you can fill out the questionnaire that the experimenter has put on thetable. After finishing the questionnaire you can call the experimenter who is waiting for you. To summarize:take only one pill a time and do not look into the large pot. The aim of the task is to order the pills in thelarge pot into the small pots. If you think that you made a mistake you can move the pills to another pot.You can start the task immediately after stopping the recorder. This can be done by pushing the stopbutton. You can push the stop button now and start with the task.

High responsibility instruction

Similar to Ladouceur et al.’s (1995) HiRes instruction, this instruction explained to the participants that theinvestigators were specialized in research of the perception of colors and were developing a color-system formedicines to be used in underdeveloped areas in India where many people are illiterate. It was told thatbecause many people cannot read instructions, there is a high incidence of mistakes with medication use.Various examples of the (sometimes fatal) mistakes with medication use for children and adults were given.Next it was explained that a developmental organization has asked the researchers to develop a cultural-specific color system that would lead to less mistakes with medication use in this area. It was explained that inthis culture various colors had symbolic meanings connected to the different diseases common in these areas.Because people with various diseases, including intestinal, heart and mental diseases have to use medication,the system has to be tested in different groups of people, including the group to which the participantbelonged. It was explained that the classification task was meant to test the color system. Next it was explainedthat, in a previous study, the results were disappointing and the medication admittance committee had rejectedthe system as participants in the test made too many mistakes. It was explained that the researchers had foundout that the participants of the previous study did not feel responsible enough and had therefore done a poorjob in classifying the pills. It was stressed that both the developmental organization and the researchersthought that the color system was a good solution and therefore begged the current participants to realize thatthe admittance of the color system depended on their performance. If the current participants would not feelresponsible enough, they would make mistakes, and the color system would not be approved, which wouldlead to continuation of the current situation in these poor areas in India, with people making too manymistakes with their medication use. Next, the same instruction was given as was quoted under the LoResinstruction.

Subjective responses

Visual analog scales (VASs) of 100mm were used to measure the following subjective experiences: (1) theseverity of the consequences of a possible failure of the study as experienced during the task; (2) the chance ofnegative consequences if the study would flop as experienced during the task; (3) the personal influence on apossible failure of the study as experienced during the task; (4) the degree to which a failure of the study wouldbe dependent on the participant as experienced during the task; (5) the experienced responsibility for a failureof the study as experienced during the task; (6) the discomfort experienced during the task; (7) the number ofthoughts about making errors during the task; (8) the number of distracting thoughts during the task; (9) thesubjective need at this moment to check whether the task had been done correctly; (10) the amount of timeneeded to recheck the task if allowed; (11) the number of errors the participant thought to have made and (12)the degree that the participant was bothered by thoughts about having made mistakes at this moment.

ARTICLE IN PRESSA. Arntz et al. / Behaviour Research and Therapy 45 (2007) 425–435 429

Behavioral checklist

Two independent judges assessed the following behaviors from the videotapes on a checklist: (1) inspectionof the pill by having it in the hand for more than 2 s; (2) movements of the hand for more than 2 s between twoor more pots; (3) looking at a small pot for more than 2 s; (4) taking up a small pot to look in it; (5) emptying asmall pot in the hand; (6) looking at more than half of the small pots or moving the hand over more than halfof the small pots before putting the pill into a pot; (7) changing one or more pills from one to another smallpot; (8) final check by inspecting more than half of the pots after the last pill was classified and (9) total timeneeded to complete the task. The averaged ratings were used for further analysis. The number of faultyclassified capsules was also counted. Inter-rater agreement of these variables was very high: medianICC ¼ 0.976, range 0.915–0.999.

Padua

The Dutch version of the Padua Inventory was used to assess strength of different aspects of OCD (vanOppen, Hoekstra & Emmelkamp, 1995; Sanavio, 1988). The Padua assesses impulses, washing, checking,rumination, and precision. Validity and reliability of the Dutch version are good to excellent (van Oppen,1992).

Statistical analysis

Subjective ratings on the 12 VASs were first subjected to a principal component analysis followed by obliquerotation of the components with Eigenvalue larger than 1 to reduce the number of dependent variables.Composite scores were constituted by averaging the standardized scores of the VASs belonging to one factor.The OCD behaviors (behaviors 1–8) were summed. Subjective ratings, OCD behaviors, number of errors andtime to complete the task were analyzed. All variables were checked on distribution before analyzing. Skeweddistributions were log-transformed, or dichotomized if normality could not be approached. Normallydistributed variables were analyzed by 2� 3 ANOVAs. The hypothesis was tested in two steps. First, theHiRes OCD deviation contrast was tested as an omnibus test of the hypothesis that the OCD group woulddeviate from the grand mean. To control for the specificity of the effect, in case of the variables relevant for thehypothesis (OCD-like experiences and behavior), the other deviation contrasts were also inspected. Next, theHiRes OCD group was compared with each other group by means of planned simple contrasts. As the generalgroup� condition interaction in a 2� 3 design might fail to detect the hypothesized interaction effect (onlyHiRes OCD group standing out), planned contrasts were applied instead of relying on the non-specific generalinteraction. In case of outliers (defined by Boxplots), ANOVAs were done both with and without outliers. Ifoutliers influenced conclusions from parametric analyses, a robust ANOVA on the medians was applied usingthe McKean and Schrader (1984) technique (Wilcox, 2005) to check whether results would hold in thecomplete sample. Dichotomous variables were analyzed by logistic regression.

Results

Subjective ratings

Principal component analysis

Principal component analysis on the VAS scores revealed three clearly interpretable factors: (1) danger, onwhich the severity and chance VASs loaded; (2) responsibility, on which the influence, dependency andresponsibility VASs loaded; and (3) discomfort and subjective OCD-experiences, on which the other VASsloaded, with the exception of the VAS on distracting thoughts which loaded moderately on the two last factorsand was not further analyzed.

Danger ratings

There was a significant condition effect, caused by HiRes participants rating higher on the danger VASsthan the LoRes participants (Fig. 1, upper left panel), F ð1; 86Þ ¼ 62:31, po0:001. The group effect and thecondition� group interaction were not significant, F ð2; 86Þ ¼ 1:08, p ¼ 0:34 and F ð2; 86Þ ¼ 0:703, p ¼ 0:50.

ARTICLE IN PRESS

Composite Danger Scores

-1

-0.5

0

0.5

-0.5

0.5

1

1.5

-1

-0.5

0

0.5

1

1.5

OCD Anx NonPt OCD Anx NonPt

OCD Anx NonPtOCD Anx NonPt

HiRes

LoResHiRes

LoRes

HiRes

LoRes

HiRes

LoRes

Composite Responsibility Ratings

Composite OCD-like Experiences (means)

-1

0

1

-0.5

0.5

-1

0

1

Composite OCD-like Experiences(medians)

Fig. 1. Locations (means, medians) and standard errors of the composite scores of the subjective ratings by diagnostic group and

condition. For danger and responsibility ratings, the high responsibility condition was significantly different from the low responsibility

condition. For OCD-like experiences, only the HiRes OCD group differed significantly from the other groups. HiRes ¼ high

responsibility; LoRes ¼ low responsibility; OCD ¼ OCD patients; Anx ¼ anxiety patient control group and NonPt ¼ nonpatient control

group.

A. Arntz et al. / Behaviour Research and Therapy 45 (2007) 425–435430

Thus, the manipulation succeeded in inducing a higher sense of danger in the HiRes condition than in theLoRes condition in all three groups. The HiRes OCD deviation contrast was significant, p ¼ 0:015. Simplecontrasts revealed that the HiRes OCD group differed significantly from all three LoRes groups (p’so0.001),but not from both HiRes control groups (p’s40.11). Thus, the manipulation induced a comparable level ofthreat in both HiRes control groups as in the HiRes OCD group.

Responsibility ratings

As can be seen in Fig. 1, upper right panel, HiRes participants rated the responsibility VASs higher thanLoRes participants, F ð1; 86Þ ¼ 10:43, p ¼ 0:002. Group effect and group� condition interaction were notsignificant; F ð2; 86Þ ¼ 0:810, p ¼ 0:45; F ð2; 86Þ ¼ 0:400, p ¼ 0:67. The manipulation was successful in inducinghigher levels of perceived responsibility in all three HiRes groups than in the LoRes groups. The HiRes OCDdeviation contrast was significant, p ¼ 0:026. Simple contrasts revealed that the HiRes OCD group differedsignificantly from all three LoRes groups (p’so0.05), but not from both HiRes control groups (p’s40.18).Thus, the manipulation induced a comparable level of perceived responsibility in both HiRes control groupsas in the HiRes OCD group.

OCD-like experiences

There were two outliers. ANOVA without these outliers revealed a significant group effect,F ð2; 84Þ ¼ 12:020, po0:001. Main effect of condition was not significant, F ð1; 84Þ ¼ 1:887, p ¼ 0:17, as was

ARTICLE IN PRESSA. Arntz et al. / Behaviour Research and Therapy 45 (2007) 425–435 431

the group� condition interaction, F ð1; 84Þ ¼ 2:432, p ¼ :094. However, the HiRes OCD deviation contrastwas significant, p o0.001, and it was the only positively deviating contrast. Simple effects revealed that theHiRes OCD group reported higher levels of subjective OCD-like experiences than all other groups, p’so0.02.Using the complete data set, a 2� 3 ANOVA on the medians using the method in Wilcox (2005) wherestandard errors of the usual sample median are estimated with the McKean and Schrader (1984) techniqueyielded significant effects of group, po0:001, condition, po0:01, and group� condition interaction, p ¼ 0:05.The HiRes OCD deviation contrast was significant, po0:001, and the only positively deviating contrast.Direct comparisons of the HiRes OCD group with the other groups using the same method revealed po0:01differences. In conclusion, the HiRes OCD group showed the highest levels of subjective OCD experiences,even higher than the LoRes OCD group. Fig. 1 shows the means (left lower panel) and medians (right lowerpanel) by group and condition.

Checking behaviors

There were three outliers on the summed behaviors, one from the HiRes Anx group and two from theLoRes NonPt group. Data were first analyzed without these outliers. A 2� 3 ANOVA yielded a significanteffect of group, F ð2; 83Þ ¼ 6:065, p ¼ 0:003, a NS effect of condition, F ð1; 83Þ ¼ 2:498, p ¼ 0:118, and asignificant group� condition interaction, F ð2; 83Þ ¼ 4:173, p ¼ 0:019. The HiRes OCD deviation contrast wassignificant, po0:001, and the only positively deviating contrast. Note that the LoRes Anx group’s deviationcontrast was not significant, p ¼ 0:15. Simple contrasts indicated that the HiRes OCD group differedsignificantly from all other groups, p’so0.02, with the exception of the contrast with the LoRes Anx group,p ¼ 0:20 (despite a mean difference of 11.5 in the expected direction). Repeating the analysis with the outliersyielded an interaction in the expected direction which failed to reach significance, F ð2; 86Þ ¼ 2:536, p ¼ 0:085.Nevertheless, deviation and simple contrasts yielded similar results as the analysis without outliers. Morespecifically, the only positively deviation contrast was the HiRes OCD group, p ¼ 0:016. The deviationcontrast of the LoRes Anx group was NS, p ¼ 0:47, but the HiRes OCD–LoRes Anx contrast remained NS(as the outliers were not from these cells). Fig. 2 depicts the means.

Time to finish the task

Time to finish the task was first log-transformed to get approximately normal distributions. Fig. 2 shows themeans. A 2� 3 ANOVA revealed a significant group effect, F ð2; 83Þ ¼ 4:156, p ¼ 0:019, a NS condition effect,F ð2; 83Þ ¼ 0:152, p ¼ 0:70, and a NS interaction, F ð2; 83Þ ¼ 1:278, p ¼ 0:28. The HiRes OCD deviationcontrast was NS, p ¼ :056. Simple contrasts revealed that the HiRes OCD group did not differ significantly

Summed Checking Behaviors

0

10

20

30

40

50

60

70

80

OCD Anx NonPt OCD Anx NonPt

HiRes

LoResHiRes

LoRes

Time to Complete Task

2.5

2.6

2.7

2.8

2.9

2.6

2.7

2.8

2.9

log

(se

c)

Fig. 2. Means and standard errors of the summed checking behaviors and time to complete the task by diagnostic group and condition.

For checking behaviors, only the HiRes OCD group differed significantly from the other groups (with the exception of the HiRes

OCD–LoRes Anx contrast; but note that the LoRes Anx deviation contrast was NS). Time to complete the task was only influenced by

patient vs. nonpatient status. HiRes ¼ high responsibility; LoRes ¼ low responsibility; OCD ¼ OCD patients; Anx ¼ anxiety patient

control group and NonPt ¼ nonpatient control group.

ARTICLE IN PRESSA. Arntz et al. / Behaviour Research and Therapy 45 (2007) 425–435432

from the other three patient groups, p’s40.19, but needed more time to complete the task than both NonPtgroups, po0:023. Post-hoc simple contrasts revealed that both patient groups differed significantly from theNonPt group, p’so0.02. In sum, time needed to complete the task was higher in all patient groups comparedto the NonPt groups, and the tendency in the HiRes OCD group to need more time was not significantlydifferent from the other patient groups.

Number of errors

The distribution of number of errors was extremely skewed with the majority of each cell making zeroerrors. This variable was therefore dichotomized (zero vs. error(s)). A logistic regression showed no overalleffect, w2ð5Þ ¼ 6:421, p ¼ 0:27, main effects and interaction were NS, and none of the cells deviatedsignificantly.

Type of OCD

To investigate whether the development of new OCD-like phenomena under conditions of HiRes wasrestricted to a specific OCD subtype (notably, checking), the Padua subscales were related to the dependentvariables in the HiRes OCD group. Pearson moment correlations of the Padua checking subscale with danger,responsibility and OCD-like experiences were all significant (r’s40.77; p’so0.005), but there were noassociations with OCD-like behavior (r ¼ 0:08, p ¼ 0:79) and time to finish the task (log transformed,r ¼ 0:22, p ¼ 0:48). Because of outliers, robust correlations between Padua checking and OCD-like behaviorswere also calculated (Wilcox, 2005), but these were also not significant. Padua rumination and precisionsubscales also tended to correlate with experienced responsibility (r ¼ 0:54, p ¼ 0:055; r ¼ 0:57, p ¼ 0:042),and Precision tended to correlate with OCD-like experiences (r ¼ 0:51, p ¼ 0:076). All other correlations werenon-significant.

Discussion

To the best of the present authors’ knowledge, this study is the first test of the causal hypothesis that highpersonal responsibility for possibly negative outcomes induces new obsessive–compulsive phenomena both onthe subjective and on the behavioral level in OCD patients. The results confirmed the hypothesis: only theHiRes OCD group reported high OCD-like subjective experiences. Moreover, and most important, only theHiRes OCD group, and not the LoRes OCD group, engaged in higher levels of new OCD-like checkingbehavior. Thus, OCD patients did not respond with OCD experiences and behaviors in a LoRes situation.Lastly, it was found that the influence of HiRes is specific for OCD, as the HiRes Anx control group did notrespond similarly to the HiRes OCD group. Only the LoRes Anx group responded on one variable in a waynot completely compatible with the hypothesis: though engaging in less OCD-like behaviors than the HiResOCD group, and not significantly deviating form the general mean, the direct comparison with the HiResOCD group failed to reach significance. This effect is difficult to explain and may be a chance finding. In sum,the results supported the hypothesis that OCD patients respond with new OCD-like responses, but only whensubjected to high personal responsibility.

All four psychopathology groups were slower than both NonPt groups, an effect apparently more related topsychopathology in general than to OCD. The number of participants making errors did not differ betweengroups and conditions, and most participants executed the task accurately. Previous studies also found that aresponsibility induction generally does not lead to slowness or reduction of errors: only one of the fourexperiments observed such an effect for errors, and only one other for slowness (Bouchard et al., 1999;Ladouceur et al., 1995, 1997; Mancini et al., 2004).

Correlational analysis did not fully support the idea that responsibility theories of OCD only hold forcheckers, as the correlation in the HiRes OCD group between the Padua checking subscale and checkingbehaviors developed when the experimentally induced responsibility was virtually zero. However, correlationsbetween the Padua checking subscale and the subjective-dependent variables were substantial, but not uniqueas there were also associations between other Padua subscales and these variables. Foa et al. (2002) found thatonly checkers showed elevated responsibility ratings on a series of scenarios. It is unclear from their reporthow checkers and non-checkers were distinguished. Remarkably, their non-checkers scored as low (m ¼ 24:70)

ARTICLE IN PRESSA. Arntz et al. / Behaviour Research and Therapy 45 (2007) 425–435 433

as non-patients (m ¼ 28:96) on trait anxiety, whereas the checkers were significantly higher on trait anxiety(m ¼ 36:63). Thus, it is unclear how representative their non-checkers group was. As noted by the authors, thescenarios may have lacked ecological validity for non-checkers. Jones and Menzies (1997) demonstrated thatonly danger expectations explained response to a behavioral avoidance test in OCD-washers, perceivedresponsibility did not add to this. Control groups were lacking however. Thus, differences between the presentstudy’s findings and those of others may be related to various factors, including the difference betweenexperimentally induced responsibility and non-experimental approaches. Further studies are needed to clarifythis issue.

Others (Bouchard et al., 1999; Ladouceur et al., 1995, 1997) induced with an apparently similarmanipulation OCD-like behaviors in non-patients. In the present study, the responsibility induction did notarouse OCD-like phenomena in non-patients. It should be noted, however, that we changed Ladouceur’sresponsibility induction in a subtle way, because their original induction was ineffective in pilot subjects.Cultural differences may play a role here: Dutch subjects may have less respect for authorities and may need astronger induction of responsibility before the induction becomes effective. High levels of experiencedpersonal responsibility for negative consequences probably elicits OCD-like phenomena in everybody,whereas medium and low levels might trigger such phenomena only in vulnerable people (Bouchard et al.,1999; Foa et al., 2002), especially when combined with the knowledge that failure is not unlikely (Mancini etal., 2004).

Although the results confirm the hypothesis that high personal responsibility induces obsessive–compulsivephenomena in a specific group of vulnerable people, there are several restrictions to the present findings. First,as a consequence of the experimental manipulation, we only induced checking, hesitations and slowness asOCD-like behaviors, but not other forms of typical OCD-like behaviors as washing and other forms ofcontamination prevention. It would be most interesting to test the hypothesis that high personal responsibilitycan also induce these types of behaviors in vulnerable people. Second, the possibility cannot be ruled out thatthe display of new OCD-like behaviors by the HiRes OCD group is a consequence of these participantsalready having OCD. Although our results support the hypothesized role of responsibility in the maintenanceand expansion of OCD in people already having OCD, it is necessary to demonstrate that responsibility alsoplays this pivotal role in causing OCD in vulnerable people not yet having OCD. Bouchard et al. (1999)demonstrated that people with perfectionistic traits report more influence and responsibility when placed in anexperimental situation with HiRes than people lower on this trait. But, in their study perfectionism did notlead to more OCD-like behavior in the HiRes condition. We expect that especially people with an excessivefear of making mistakes that may have severe consequences and for which they feel that they may bepersonally blamed, i.e., those with excessive fear of guilt, are at risk.

Rachman (2002) recently formulated a highly specified cognitive theory of compulsive checking. Morespecifically, Rachman argued that the product of perceived responsibility� perceived probability of

harm� perceived seriousness of harm determines intensity and duration of checking. What do our findingstell about his analysis? The first relevant finding was that the factor analysis of the subjective experiencesyielded evidence for responsibility and danger (with both severity and chance loading on this factor) wereindependent dimensions, which in turn were relatively independent from subjective OCD-related experiences.This fits nicely with Rachman’s analysis that both perceived responsibility and perceived danger(danger ¼ probability� seriousness of harm) play an essential role in checking behavior. But, wedemonstrated that, at least with our experimental induction, only OCD patients in the HiRes conditionengaged in elevated checking. Thus, it seems that Rachman’s analysis needs a qualification, in the sense that itholds especially for specific (presumably OCD-vulnerable) people. A second statement in Rachman’s paper isthat elevated responsibility leads to increased danger perceptions. In our study both were indeed related, butas the experimental induction addressed both responsibility and danger perceptions, no conclusion can bedrawn about the direction of the association. A third statement in Rachman’s paper relevant in the presentcontext is that prolonged checking leads to further increases in experienced responsibility. From our study it isclear that checking and responsibility were related, in particular in OCD patients, but we did not address thetime relationship between these variables. Further studies should repeatedly assess perceived responsibilityduring experimentally induced checking to shed further light on this issue (or even tempt participants to checkfor a short vs. prolonged time, and then assess perceived responsibility).

ARTICLE IN PRESSA. Arntz et al. / Behaviour Research and Therapy 45 (2007) 425–435434

The clinical implication of the present findings might be that it is indicated to address guilt- andresponsibility-related issues in treatment of OCD. If these issues are indeed so essential in OCD, it isconceivable that for a profound and long-lasting change they should be addressed. If not necessary forimmediate treatment effects, addressing these issues might help to prevent relapse. Note that specific cognitivemethods to treat biased guilt and responsibility interpretations already have been developed (van Oppen &Arntz, 1994; Salkovskis, 1999).

Acknowledgments

This study was conducted with the help of the Maastricht Community Mental Health Center and theVincent van Gogh Institute at Venray. Acknowledgments are due to Jacques Oomen and the researchassistants for patient recruitment and data collection. Robert Ladouceur is thanked for his help with settingup the responsibility induction.

References

Bouchard, C., Rheaume, J., & Ladouceur, R. (1999). Responsibility and perfectionism in OCD: An experimental study. Behaviour

Research and Therapy, 37, 239–248.

First, M. B., Spitzer, R. L., Gibbon, M., & Williams, J. B. W. (1997). Structured clinical interview for DSM-IV axis I disorders (SCID-I).

New York: Biometric Research Department.

Foa, E. B., Amir, N., Bogert, K. V. A., Molnar, C., & Przeworski, A. (2001). Inflated perception of responsibility for harm in obsessive

compulsive disorder. Journal of Anxiety Disorders, 15, 259–275.

Foa, E. B., Sacks, M. B., Tolin, D. F., Prezworski, A., & Amir, N. (2002). Inflated perception of responsibility for harm in OCD patients

with and without checking compulsions: A replication and extension. Journal of Anxiety Disorders, 16, 443–453.

Freeston, M. H., Ladouceur, R., Gagnon, F., & Thibodeau, N. (1993). Beliefs about obsessional thoughts. Journal of Psychopathology and

Behavioral Assessment, 15, 1–21.

Freeston, M. H., Ladouceur, R., Thibodeau, N., & Gagnon, F. (1992). Cognitive intrusions in a non-clinical population: II. Associations

with depressive, anxious and compulsive symptoms. Behaviour Research and Therapy, 30, 263–271.

Freud, S. (19262001). Hemmung, symptom und angst (inhibitions, symptoms and anxiety). London: Vintage Random House.

Frost, R. O., Steketee, G., Cohn, L., & Griess, K. (1994). Personality traits in subclinical and nonobsessive volunteers and their parents.

Behaviour Research and Therapy, 32, 47–56.

van Groenestijn, M. A. C., Akkerhuis, G. W., Kupka, R. W., Schneider, N., & Nolen, W. A. (1999). Gestructureerd klinisch interview voor

de vaststelling van DSM-IV as-I stoornissen (SCID-I) (structured clinical interview for DSM-IV axis I disorders (SCID-I)). Lisse:

Swets & Zeitlinger.

van den Hout, M. A., & Kindt, M. (2003). Repeated checking causes memory distrust. Behaviour Research and Therapy, 41, 301–316.

Jones, M. K., & Menzies, R. G. (1997). The cognitive mediation of obsessive–compulsive handwashing. Behaviour Research and Therapy,

35, 843–850.

Ladouceur, R., Rheaume, J., & Aublet, F. (1997). Excessive responsibility in obsessional concerns: A fine-grained experimental analysis.

Behaviour Research and Therapy, 35, 423–427.

Ladouceur, R., Rheaume, J., Freeston, M. H., Aublet, F., Jean, K., Lachance, S., et al. (1995). Experimental manipulation of

responsibility: An analog test for models of obsessive–compulsive disorder. Behaviour Research and Therapy, 33, 937–946.

Lopatka, C., & Rachman, S. (1995). Perceived responsibility and compulsive checking: An experimental analysis. Behaviour Research and

Therapy, 33, 673–684.

Mancini, F., D’Olimpio, F., & Cieri, L. (2004). Manipulation of responsibility in non-clinical subjects: Does expectation of failure

exacerbate obsessive–compulsive behaviors? Behaviour Research and Therapy, 42, 449–457.

Mancini, F., & Gangemi, A. (2004). Fear of guilt from behaving irresponsibly in obsessive–compulsive disorder. Journal of Behavior

Therapy and Experimental Psychiatry, 35, 109–120.

McKean, J. W., & Schrader, R. M. (1984). A comparison of methods for studentizing the sample median. Communications in Statistics,

Simulation and Computation, 13, 751–773.

Menzies, R. G., Harris, L. M., Cumming, S. R., & Einstein, D. A. (2000). The relationship between inflated personal responsibility and

exaggerated danger expectancies in obsessive compulsive concerns. Behaviour Research and Therapy, 38, 1029–1037.

van Oppen, P. (1992). Obsessions and compulsions: Dimensional structure, reliability, convergent and divergent validity of the Padua

Inventory. Behaviour Research and Therapy, 30, 631–637.

van Oppen, P., & Arntz, A. (1994). Cognitive therapy for obsessive compulsive disorder. Behaviour Research and Therapy, 32, 79–87.

van Oppen, P., Hoekstra, R. J., & Emmelkamp, P. M. G. (1995). The structure of obsessive–compulsive symptoms. Behaviour Research

and Therapy, 33, 15–23.

Rachman, S. (1993). Obsessions, responsibility and guilt. Behaviour Research and Therapy, 31, 149–154.

Rachman, S. (2002). A cognitive theory of checking. Behaviour Research and Therapy, 40, 625–639.

ARTICLE IN PRESSA. Arntz et al. / Behaviour Research and Therapy 45 (2007) 425–435 435

Rachman, S. J., Thordarson, D. S., Shafran, R., & Woody, S. (1995). Perceived responsibility: Structure and significance. Behaviour

Research and Therapy, 33, 779–784.

Rheaume, J., Freeston, M. H., Dugas, M. J., Letarte, H., & Ladouceur, R. (1995). Perfectionism, responsibility and obsessive–compulsive

symptoms. Behaviour Research and Therapy, 33, 785–794.

Rheaume, J., Ladouceur, R., Freeston, M. H., & Letarte, H. (1994). Inflated responsibility in obsessive compulsive disorder: Psychometric

studies of a semiidiographic measure. Journal of Psychopathology and Behavioral Assessment, 16, 265–276.

Salkovskis, P. M. (1985). Obsessional–compulsive problems: A cognitive behavioural analysis. Behaviour Research and Therapy, 23,

571–583.

Salkovskis, P. M. (1999). Understanding and treating obsessive–compulsive disorder. Behaviour Research and Therapy, 37, S29–S52.

Salkovskis, P. M., & Forrester, E. (2002). Responsibility. In R. O. Frost, & G. Steketee (Eds.), Cognitive approaches to obsessions and

compulsions: Theory, assessment and treatment (pp. 45–61). Oxford: Pergamon.

Salkovskis, P. M., Wroe, A. L., Gledhill, A., Morrison, N., Forrester, E., Richards, C., et al. (2000). Responsibility attitudes and

interpretations are characteristic of obsessive compulsive disorder. Behaviour Research and Therapy, 38, 347–372.

Sanavio, E. (1988). Obsessions and compulsions: The Padua inventory. Behaviour Research and Therapy, 26, 169–177.

Shafran, R. (1997). The manipulation of responsibility in obsessive–compulsive disorder. British Journal of Clinical Psychology, 36,

397–407.

Shafran, R., Watkins, E., & Charman, T. (1996). Guilt in obsessive–compulsive disorder. Journal of Anxiety Disorders, 10, 509–516.

Steketee, G., Frost, R. O., & Cohen, I. (1998). Beliefs in obsessive–compulsive disorder. Journal of Anxiety Disorders, 12, 525–537.

Wilcox, R. R. (2005). Introduction to robust estimation and hypothesis testing (2nd ed). San Diego: Academic Press.

Wilson, K. A., & Chambless, D. L. (1999). Inflated perceptions of responsibility and obsessive–compulsive symptoms. Behaviour Research

and Therapy, 37, 325–335.