Causing harm and allowing harm: a study of beliefs in obsessional problems

Behaviour Research and Therapy 38 (2000) 1141±1162

www.elsevier.com/locate/brat

Causing harm and allowing harm: a study of beliefs in obsessional problems Abigail L. Wroe, Paul M. Salkovskis* University of Oxford Department of Psychiatry, Warneford Hospital, Oxford OX3 7JX, UK

Abstract This study investigates two factors hypothesised as relevant to obsessional problems because of the way in which they in¯uence decisions whether or not to act to prevent harm. These are (i) the way in which intrusive thoughts increase the internal awareness of harm, and confront the person with the possibility of taking action to prevent such harm and (ii) the extent to which there is some obvious external factor which increases awareness of the possibility of preventing harm. Obsessional patients, anxious and non-clinical controls completed a scale which systematically measured these factors across a wide range of situations. Results across all situations evaluated con®rmed previous ®ndings that both obsessionals and nonobsessionals were more likely to report acting to prevent harm when awareness of it is prompted by an intrusion than when it is not. It was also found that participants in all groups acted more `obsessionally' when a scenario is described in ways which suggest that harm may be by `commission' than when it is described in terms of an `omission'. When scenarios about which each individual is most disturbed were analysed, anxious and non-clinical controls continued to di€erentially rate omission and commission situations; as predicted, this di€erential was not present for obsessional patients. It is concluded that obsessionals are more sensitive to omission than are nonobsessionals when considering scenarios about which they are concerned, and that this sensitivity is one factor in¯uencing the decision whether to act to prevent harm. 7 2000 Elsevier Science Ltd. All rights reserved.

* Corresponding author. Tel.: +44-1865-226-473; fax: +44-1865-226-234. E-mail address: [email protected] (P.M. Salkovskis). 0005-7967/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 5 - 7 9 6 7 ( 9 9 ) 0 0 1 4 5 - X

1142

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

1. Introduction According to the cognitive theory of Obsessive Compulsive Disorder (OCD) (Salkovskis, 1999), obsessional thoughts are normal intrusions which are misinterpreted by obsessional patients as a sign that they risk being responsible for impending harm. Such interpretations are said to be made because of general assumptions that the person has learned concerning responsibility, intrusions and harm (Salkovskis, 1998). It has been demonstrated that responsibility assumptions and responsibility appraisals may play an important role in the maintenance of OCD by leading the person to decide to act to prevent possible harm in ambiguous situations (Salkovskis et al., in press). Examples of assumptions found to be characteristic of OCD patients include the following: ``If I don't act when I can foresee danger, then I am to blame for any consequences if it happens''. ``I must always think through the consequences of even the smallest actions''. ``I feel responsible for things that go wrong'' Assumptions of this type and other distortions of thinking, often described as ``thinking errors'' (Beck, 1976), are thought to in¯uence reactions and judgements through the appraisals made of speci®c stimuli or situations. For example, an individual who holds the assumption, ``If I don't act when I can foresee danger, then I am to blame for any consequences if it happens'', and who experiences an intrusive thought about potential harm may think, ``Because I have thought of bad things that could go wrong, I must act to prevent them'', and is then likely to feel compelled to act to prevent the potential harm. It follows, then, that the way in which the individual thinks about a particular situation which involves potential harm e.g. as indicating responsibility for that harm, in¯uences the decision to act to prevent harm. In this way, responsibility-based interpretations of intrusions form the bridge between the occurrence of the intrusion and the compulsion to act. In turn, generally held assumptions drive the speci®c responsibility interpretations. Salkovskis (1996b) has suggested that the assumption that ``failing to prevent (or failing to try to prevent) harm to self or others is the same as having caused the harm in the ®rst place'' may be an important aspect of obsessional problems. This relates to ®ndings by Spranca, Minsk and Baron (1991) that non-clinical participants judge responsibility for negative consequences to be diminished when an omission is involved (i.e. not acting or the failure to act), as opposed to when some speci®c action (or commission) is involved in bringing about the negative consequence. This ®nding related to non-clinical participants even when allowances are made for the element of intention, that is the extent to which the person wishes the `negative' outcome to occur. Our clinical experience suggests that obsessional patients are as sensitive to omissions as to commissions. In Spranca et al.'s study, non-clinical participants were asked to rate the morality and causality in hypothetical scenarios. An example of such a scenario involved a tennis player having a pre-match dinner with his opponent who is allergic to cayenne pepper. The ®rst player, who knows about the allergy, discovers that there is pepper in one of the sauces but not the other. There are six situations to rate: 1. Before his opponent orders, the ®rst player recommends the sauce with the pepper and

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

1143

subsequently wins the match. 2. In the knowledge that his opponent has ordered the sauce with pepper, the ®rst player says nothing and subsequently wins the match. 3. After the opponent has ordered the sauce without the pepper, the ®rst player recommends the sauce with the pepper and subsequently wins the match. 4. Like situation (1) but the opponent wins the match. 5. Like situation (2) but the opponent wins the match. 6. Like situation (3) but the opponent wins the match. (Paraphrased from a questionnaire kindly provided by Jon Baron.) Spranca et al. (1991) showed that the participants considered harmful commissions to be morally worse than omissions that resulted in harm. Additional experiments indicated that participants' judgements about the immorality of omission and commission are dependent on several factors that ordinarily distinguish omission and commission: physical movement in commission, the presence of salient alternative causes in omission, and the fact that the consequences of omissions would occur if the actor were absent or ignorant of the e€ects of not acting. It was concluded from these studies that most people appear to regard themselves as more responsible for what they actively do than what they fail to do, and that this omission bias occurs due to: perceived di€erences in causality; di€ering degrees of responsibility; and the `bald fact that one situation is commission and one is omission'. Whereas most people regard themselves as less responsible for failing to prevent harm than for causing harm in the ®rst place, the assumptions and appraisals found to be speci®c to OCD patients suggest that they believe the contrary. The scale used by Salkovskis et al. (in press) included a considerable number of items concerning beliefs about omission. In an unpublished sub-analysis of these data, it was possible to con®rm that beliefs of responsibility for omissions were higher in OCD patients than in anxious or non-clinical controls (Wroe, 1997). It could be, then, that these responsibility assumptions and responsibility appraisals increase concern about failures to act (or omissions) transforming the situation to one involving commission, and that this is a cause of obsessional behaviour. The mechanism of such an e€ect needs clari®cation, particularly in terms of how it links to the cognitive theory of OCD. An important factor in judgements concerning responsibility for omissions is the perception of `agency'; if one perceives oneself as an agent, then one has chosen to bring something about and is therefore responsible (Salkovskis, 1996b). This is re¯ected by the entry for responsibility in the Oxford Companion to the Mind, which states that `premeditation' (that is, being able to foresee possible harmful outcomes) ``usually makes an objectionable act seem more culpable. If the actor foresaw a real possibility of his causing harm Ð for example by his way of driving Ð his act or omission will be called `reckless' and blamed accordingly''. It is highlighted also that the degree to which the individual is regarded as blameworthy depends often on ``the actor's state of mind at the time of the act . . . If the act seems to have been quite accidental Ð if for instance he knocks over a child whom he did not see in his path Ð he is not blamed, unless we think that he should have been aware of this as a real possibility''. (The Oxford Companion to the Mind, Gregory & Zangwill, 1987, p. 681.) As well as the in¯uence of responsibility assumptions and appraisals which are characteristic

1144

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

of obsessional patients, a further problem is that obsessionals frequently ``foresee a wide range of possible negative outcomes'' (Salkovskis, 1996a; Salkovskis & Kirk, 1997; Salkovskis, Richards & Forrester, 1995). Almost by de®nition, intrusive thoughts about possible harm make the person aware of possible harm, for example passing on contamination, having left the door unlocked, having left the gas on or having hurt somebody accidentally and this, in turn, provides awareness of the option to act to prevent this harm. Some patients even regard it as their duty to try to foresee negative outcomes. There is some evidence consistent with the hypothesis that occurrence of intrusive thoughts play a major role in the type of reactions shown by obsessional patients because they prompt ideas of harm. It is also clear that obsessional patients more frequently experience intrusions about the particular negative consequences that form the focus of their obsessions than nonobsessionals do about their concerns. The results of previous research (Wroe, Salkovskis & Richards, in press) are consistent with the idea that OCD relevant situations (e.g. leaving the house, seeing broken glass on the street) are associated with more intrusions only in obsessional patients, but were more likely to be acted on by both obsessional and nonobsessional participants. Semiideographic analyses were conducted in which 2 situations were identi®ed for each participant: one which they found most problematic; and one which they found least problematic. These analyses showed that controls react more like obsessional patients on the situation which troubles them most. It is therefore proposed that the reason for the apparent absence of omission bias in obsessionals is that it is masked by characteristic obsessional thinking: ®rstly, the occurrence of intrusive thoughts about harm; and secondly the responsibility assumptions which transform the situation such that the person must decide whether or not to act to prevent harm. Deciding not to act despite being aware of possible harm becomes an active decision not to act to prevent the harmful outcome, making the person a causal agent in relation to the potential harm. Thus the occurrence of the intrusive thought about possible harm, together with the perception of responsibility, transforms the situation where harm can only occur by omission into a situation where the person has `actively' decided to allow the possibility of harm to take place. If this is the case, it seems also that it must be possible to mimic omission bias in nonclinicals. Returning to the de®nition of responsibility from the Oxford Companion to the Mind, there appear to be two main factors which in¯uence it: ®rstly, `premeditation' which involves the awareness of possible harm, and secondly the perception of `agency' which relates to responsibility. Premeditation (or being able to foresee possible harmful outcomes) can be in¯uenced by the occurrence of an intrusive thought about possible harm. If nonobsessionals are provided with an intrusive thought about harm which might arise in a given situation, the e€ect is to draw attention to (and make foreseeable) possible harmful consequences. In such circumstances, nonobsessionals would, according to the hypothesis advanced here, react similarly to obsessional patients. In a recent study, Forrester, Wilson & Salkovskis (submitted for publication) found that this was indeed so. Both obsessional patients and normal controls reacted more strongly to hypothetical situations involving an unexpected intrusion concerning harm than to the same situations in which the intrusion was neutral. Increases were found in ratings of discomfort, the urge to take preventative action, experience of intrusions and perceived responsibility for harm. Only on this last measure was there an interaction between

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

1145

group (obsessional patients/nonobsessionals) and the occurrence of intrusions concerning harm, with obsessional patients showing a smaller di€erence between negative and neutral intrusions than nonobsessionals. The study described in the current paper investigates the way in which intrusive thoughts increase the internal awareness of harm, and confront the person with the possibility of taking action to prevent such harm and the extent to which there is some obvious external factor which increases awareness of the possibility of preventing harm. It is likely that both of these factors may increase the likelihood of action by increasing the perception of `agency' (that is, the extent to which an individual perceived him/her self as an active agent in the situation). It is speci®cally hypothesised from the cognitive theory that obsessionals' increased sense of responsibility is an internal factor transforming the situation such that they perceive themselves as the agent. It may be possible to in¯uence nonobsessionals' perception of agency by providing some external factor which transforms the situations such that the individual is explicitly made aware of his/her possible in¯uence over the potential harm. It is predicted that this external factor will also a€ect obsessional patients across the full range of situations, but in the speci®c situations about which they are most concerned, an internal factor (the occurrence of intrusions and the way in which this occurrence is interpreted) is likely to mask this e€ect. That is, the occurrence of an intrusion results in the transformation of the situation into one where the patient perceives themselves to be an active agent and therefore responsible for preventative action. The addition of an external commission element will have little further impact for those instances. The questionnaire used in the present study was designed to investigate the in¯uence of describing an ambiguous situation with the graded inclusion of (i) the occurrence of an intrusive thought about possible harm, and (ii) a description such that not acting to prevent harm involves a commission compared to an omission. Each questionnaire item was designed to have three sections to re¯ect these elements, and participants were then asked to rate their reactions in each of a number of situations. These situations were also rated in terms of the degree to which each person tended to ®nd them disturbing, in order to identify personally relevant and less relevant situations for a semi-ideographic analysis. There are three main possible outcomes: there may be a generalised omission±commission type factor, such that obsessionals are more sensitive to omission in all situations; the e€ect may be highly speci®c to those behaviours about which individuals are obsessional; or there may be no e€ect at all. 2. Method 2.1. Participants Participants consisted of 42 obsessionals (19 male, 23 female), 53 non-clinical controls (25 male, 28 female) and 25 anxious controls, that is individuals with an anxiety disorder other than OCD and hypochondriasis (3 male, 22 female). Non-clinical controls included students, oce sta€ and domestic sta€ at Wolfson College, Oxford University; and teaching, oce and domestic sta€ at a comprehensive school in SouthEast London, and other members of the community. The clinical patients were about to partake in treatment trials carried out by

1146

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

Department of Psychiatry, Warneford Hospital, Oxford. The obsessional participants consisted of 20 ruminators, 2 washers, 9 checkers, 11 `other OCD'. The anxious controls had anxiety disorders other than OCD and hypochondriasis since the latter is thought to involve similar thought patterns to OCD. All of them reached DSM IV criteria for their particular anxiety disorder. 2.2. Measures All participants in this study completed two questionnaires on obsessionality: Maudsley Obsessive Compulsive Inventory (MOCI) (Hodgson & Rachman, 1978); and Obsessive Compulsive Inventory (Foa, Kozak, Salkovskis, Coles & Amir, 1998). They were also asked to complete the Beck Depression Inventory (BDI) (Beck, Ward, Mendelsohn, Mock & Erbaugh, 1961) as a measure of depression, the Beck Anxiety Inventory (BAI) (Beck, Epstein, Brown & Steer, 1988) as a measure of clinical anxiety symptoms and the Spielberger State±Trait Anxiety Inventory (STAI-I; STAI-2) (Spielberger, 1983) as a general measure of anxiety. One way analyses of variance showed signi®cant di€erences in scores of obsessionality (MOCI and OCI), anxiety (STAI state and trait, and BAI) and depression (BDI) between groups. Multiple comparisons showed that, as expected, the obsessional group scored more highly on the measures of obsessionality than did the control participants, and the anxious controls and obsessionals had equivalent scores on the measures of depression (BDI) and anxiety (STAI, BAI) (see Table 1). These results indicate that both comparison groups di€ered on obsessionality, but only the non-clinical group di€ered on anxiety and depression indicating that any di€erences between obsessionals and anxious controls are likely to be related to obsessionality and not to anxiety, depression or the fact that the individuals are patients. 2.3. Procedure Participants were also asked to complete a questionnaire which addressed the omission bias. The ®rst page of this questionnaire asked participants to rate on a scale of 0 (not at all disturbed) to 10 (extremely disturbed) their usual reaction to several situations. The situations relate to the scenarios described in the main section of the questionnaire. The purpose was to identify those situations about which each individual was most and least disturbed. (All questionnaires used can be obtained from the authors.) Participants were then asked to consider all eight situations used in the questionnaires in turn; two of these involved checking, two contamination, two causing harm, and two obsessional irrelevant situations as control items. Each scenario had three levels of omission±commission presented sequentially and all participants completed ratings of six variables in each of three levels of omission±commission. The ®rst level of omission±commission was the ``omission'' level; a scenario was presented in which there may be possible harm, although the possibility of harm was not mentioned. For example, You and several other people are preparing food for an oce party. Peter and you have arrived a little early so begin the preparation. While preparing a di€erent dish, you see out



p < 0.00001a

Sample

Obsessionals …n ˆ 42; 19 male, 23 female) (mean (S.D.))

Anxious controls …n ˆ 25; 3 Group comparisons Non-clinical controls …w 2 ˆ 9:82, p < 0.01) …n ˆ 53; 25 male, 28 female) male, 22 female) (mean (S.D.)) (mean (S.D.))

Age OCI MOCI BDI BAI STAI-state STAI-trait

32.17a (10.64) 66.70b (32.13) 14.25b (6.10) 17.00a (10.25) 19.15a (9.82) 50.23a (12.06) 56.10a (10.58)

36.06a (14.07) 16.22a (12.28) 5.26a (3.82) 6.12b (6.58) 7.13b (5.64) 34. 19b (8.43) 38.87b (9.41)

a

38.91a 21.80a 6.70a 19.56a 24.04a 54.10a 55.67a

(13.48) (19.00) (3.95) (9.68) (10.47) (11.71) (11.88)

values on the same line which share a superscript are not signi®cantly di€erent from each other.

F‰2,116Š ˆ 2:18, p ˆ 0:12 F‰2,115Š ˆ 60:51 F‰2,102Š ˆ 40:23 F‰2,116Š ˆ 27:80 F‰2,102Š ˆ 42:93 F‰2,102Š ˆ 34:71 F‰2,100Š ˆ 32:24

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

Table 1 Mean and standard deviations of the scores of obsessional patients and control groups for measures of psychopathology and age.

1147

1148

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

of the corner of your eye that Peter has dropped some crisps on the ¯oor where he is standing. He picks up the crisps and puts them back. The second level involved presentation of the same scenario and also the participant was provided with a thought of a negative consequence, presented as an intrusive thought. In this scenario individuals were made aware of the possible harm but there was no external factor in¯uencing the decision whether or not to act to prevent that harm. This level is known as the ``omission-plus-thought'' level. For example, You and several other people are preparing food for an oce party. Peter and you have arrived a little early so begin the preparation. While preparing a di€erent dish, you see out of the corner of your eye that Peter has dropped some crisps on the ¯oor where he is standing. He picks up the crisps and puts them back. Suddenly the thought pops into your head, ``What if the crisps are now contaminated and someone gets ill?'' In the ®nal level of omission±commission, the `commission' level, the same scenario was presented with the intrusive thought, and also a statement was presented which meant that the participants had actively decided whether or not to act. In this situation, the individual was made aware of the possible harm, and also external factors are introduced, so that not acting would involve an active decision to not act to prevent harm. In this case, not acting is seen as a commission. For example, You and several other people are preparing food for an oce party. Peter and you have arrived a little early so begin the preparation. While preparing a di€erent dish, you see out of the corner of your eye that Peter has dropped some crisps on the ¯oor where he is standing. Suddenly the thought pops into your head, ``What if the crisps are now contaminated and someone gets ill?`` Peter looks unsure about what to do and asks you if you think that the ¯oor is too dirty and whether the crisps should be discarded and that the guests should go without. After each of the scenarios had been described, the same six questions were asked how likely they would be to act to prevent harm, for example How likely is it that you would (act to prevent harm e.g.) suggest that the crisps should be thrown away? (Scale 0 [I would de®nitely NOT suggest that the crisps should be thrown away]±10 [I would de®nitely suggest that the crisps should be thrown away.) Participants were then asked: Consider the following scenario and rate how you would feel using the scales shown: You carry on as if nothing has happened. How immoral would you feel? (scale 0 [not at all immoral]±10 [totally immoral]). How worried would you feel? (scale 0 [not at all worried]±10 [extremely worried]). How responsible would you feel for any harm that may occur? (scale 0 [not at all responsible]±10 [totally responsible]). How much would you feel that you were the cause of any harm that may occur? (scale 0 [I

Main e€ect: group Level of omission± commission Interaction: group by level of omission±commission



p < 0.0002 (a is adjusted due to multiple signi®cance tests;

Likelihood of acting

Immoral

Worry

Responsible

Cause

Blame

F‰2,117Š ˆ 1:98p ˆ 0:14 F‰2,234Š ˆ 232:13

F‰2,117Š ˆ 0:48p ˆ 0:62 F‰2,234Š ˆ 126:78

F‰2,117Š ˆ 2:02p ˆ 0:14 F‰2,234Š ˆ 180:06

F‰2,117Š ˆ 1:27p ˆ 0:28 F‰2,234Š ˆ 134:88

F‰2,117Š ˆ 0:86p ˆ 0:42 F‰2,234Š ˆ 153:80

F‰2,117Š ˆ 0:91p ˆ 0:41 F‰2,234Š ˆ 155:76

F‰4,234Š ˆ 0:63p ˆ 0:64

F‰4,234Š ˆ 1:38p ˆ 0:24

F‰4,234Š ˆ 1:92p ˆ 0:11

F‰4,234Š ˆ 0:86p ˆ 0:49

F‰4,234Š ˆ 1:71p ˆ 0:15

F‰4,234Š ˆ 1:65p ˆ 0:16

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

Table 2 Results of omnibus analyses of variance for each of the variables. a ˆ 0:001=6 ˆ 0:0002)

1149

1150

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

Fig. 1. Graphs of mean scores on each of the variables.

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

1151

would not at all feel that I was cause]±10 [I would feel that I was totally the cause]). How much would you feel that you were to blame for any harm that may occur? (scale 0 [not at all to blame]±10 [completely to blame])

3. Treatment of data The analyses are divided into two sections. The ®rst section involves analyses of all of the scenarios in order to investigate any general bias in the obsessionals relative to the controls. Since each participant had completed ratings for eight scenarios in total, these were collapsed into a single variable for each level of omission±commission by calculating the average of the scores of each scenario for each variable in question. Analyses were then carried out on these mean scores. The second set of analyses focussed on reactions to two situations: one which disturbed participants most (which for the obsessional patients was always the situations in which they tended to experience obsessional behaviour); and the other which disturbed them least. These analyses were semi-ideographic as the scenarios were not designed for each individual, but were chosen from a range of scenarios. The particular scenarios chosen for each individual depended on the scores of how disturbed individuals were about various situations; these ratings were done before the questionnaire itself (see above). If there were two situations with equivalent scores for highest or lowest rated situation, the scenario was chosen according to the level of worry rated in the `omission' level (that is the ®rst level in the questionnaire). If the rating of worry was the same, one scenario was picked at random. However, if the individual was equally concerned about all of the situations described, no situations were chosen and the participant was excluded from these analyses.

4. Results 4.1. Omnibus analyses Analyses of variance were carried out to investigate the e€ects of group (obsessionals, nonclinical controls, anxious controls) and level of omission±commission (omission, omission-plusthought, commission). Such analyses were carried out on each variable, that is the likelihood of acting to prevent harm (subsequently known as `likelihood of acting'), the extent to which individual would feel immoral if he/she did not act to prevent harm (`immoral'), the extent to which individual would worry if he/she did not act to prevent harm (`worry'), the extent to which individual would feel responsible if he/she did not act to prevent harm (`responsible'), the extent to which individual would feel that they were the cause of any harm that did occur (`cause') and the extent to which individual would feel that he/she was to blame for any harm that did occur (`blame'). The results of these analyses are shown in Table 2 and in graphical form in Fig. 1.

1152

Main e€ect: group Level of omission-commission Level of disturbance Two way e€ect: group by disturbance Group by level of omission-commission Disturbance by level of omission-commission Three way e€ect: group by omission-commission by disturbance



p < 0.00017,



p < 0.0017, p < 0.05 (a is adjusted due to multiple

Likelihood of acting

Immoral

Worry

Responsible

Cause

Blame

F‰2,90Š ˆ 4:62 F‰2,180Š ˆ 96:30

F‰2,90Š ˆ 2:65p ˆ 0:08 F‰2,180Š ˆ 64:96

F‰2,90Š ˆ 6:11 F‰2,180Š ˆ 92:66

F‰2,90Š ˆ 3:96p ˆ 0:02 F‰2,180Š ˆ 54:29

F‰2,90Š ˆ 2:25p ˆ 0:11 F‰2,180Š ˆ 58:41

F‰2,90Š ˆ 1:64p ˆ 0:20 F‰2,180Š ˆ 56:38

F‰1,90Š ˆ 87:00 F‰2,90Š ˆ 2:45

F‰1,90Š ˆ 9:41 F‰2,90Š ˆ 1:33p ˆ 0:27

F‰1,90Š ˆ 59:53 F‰2,90Š ˆ 2:35p ˆ 0:10

F‰1,90Š ˆ 53:33 F‰2,90Š ˆ 1:56p ˆ 0:22

F‰1,90Š ˆ 40:36 F‰2,90Š ˆ 0:99p ˆ 0:38

F‰1,90Š ˆ 42:21 F‰2,90Š ˆ 0:76p ˆ 0:47

F‰4,180Š ˆ 0:45p ˆ 0:77

F‰4,180Š ˆ 0:92p ˆ 0:46

F‰4,180Š ˆ 0:50p ˆ 0:73

F‰4,180Š ˆ 0:28p ˆ 0:89

F‰4,180Š ˆ 0:81p ˆ 0:52

F‰4,180Š ˆ 0:42p ˆ 0:80

F‰2,180Š ˆ 2:46p ˆ 0:09

F‰2,180Š ˆ 2:60p ˆ 0:08

F‰2,180Š ˆ 2:25p ˆ 0:11 F‰4,180Š ˆ 4:75



F‰2,180Š ˆ 1:13p ˆ 0:33 

F‰4,180Š ˆ 2:44

F‰2,180Š ˆ 0:11p ˆ 0:89 

F‰4,180Š ˆ 5:50



F‰2,180Š ˆ 5:55



F‰4,180Š ˆ 4:15



F‰4,180Š ˆ 3:38

F‰4,180Š ˆ 2:67 

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

Table 3 Results of omnibus analyses of variance for each of the variables. signi®cance tests; a ˆ 0:001=6 ˆ 0:00017; a ˆ 0:01=6 ˆ 0:0017)

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

1153

The results show, as predicted, that there is a strong main e€ect of the level of omission± commission. The more that individuals engage in a situation, the more likely they are to act to prevent harm and to feel immoral, worried, responsible, as if they were the cause, and as if they were to blame if the harm was to arise. It seems then that if individuals engage in a situation, they act more obsessionally in that situation. There were no signi®cant interactions of level of omission±commission and group when all the results are analysed together. This indicates that there is no generalised di€erential e€ect between obsessionals and nonobsessionals, when situations are described such that not acting involves commission compared to when failure to act involves omission. However, as suggested previously, it is possible that such an e€ect might only be present at a higher level of speci®city of the scenarios in question. The subsequent analyses investigates this possibility by comparing obsessionals and nonobsessionals in scenarios about which individuals are most and least disturbed. 4.2. Semi-ideographic analyses: scenarios about which individual are most and least concerned Out of the eight scenarios completed by each individual, two were picked; one relating to the situation about which they were most disturbed and the other about which they were least disturbed. A total of 102 participants had data regarding the scenarios about which they were most and least concerned (36 obsessionals, 41 non-clinical controls and 25 anxious controls). Some participants had been eliminated at this point as there was either no situation about which they were most disturbed or no situation about which they were least disturbed. Chisquared analyses showed no di€erences in the proportions of the scenarios about which individuals stated they were most disturbed …w 2 ‰6Š ˆ 3:67, p ˆ 0:72† or least disturbed …w 2 ‰6Š ˆ 6:10, p ˆ 0:41). Analyses of variance were then carried out to investigate any interactions between group (obsessionals, non-clinical controls and anxious controls), level of disturbance (most and least) and levels of omission±commission (omission, omission-plus-thought, commission). The same analyses were carried out for each of the variables in question. The results of the analyses can be seen in Table 3. Results show a main e€ect of level of omission±commission and of level of disturbance in each variable. Variables `likelihood of acting', `worry' and `responsible' gave signi®cant three way interactions at the p < 0.0017 level (adjusted due to multiple comparisons) and all variables gave a signi®cant 3 way interaction at p < 0.05 level. This indicates that there is some di€erential e€ect of omission vs. commission, and level of disturbance on the variables for the di€erent groups. In order to investigate this interaction further, a new variable was calculated: `sensitivity to omission bias', that is the di€erence between the scores on each of the variables between the commission level and the omission-plus-thought level. If there is no omission bias i.e. if omission is seen the same as commission, then this di€erence will be small i.e. individuals will score similarly when the scenario is presented as an omission and as a commission. Conversely, if the omission bias is strong, then this di€erence will be great and positive, that is individuals would score higher on the variable when in the commission level than in the omission-plusthought level. This di€erence was calculated for each individual for all the variables of the

1154

Main e€ect: group Level of disturbance Two way e€ect: group by disturbance

Likelihood of acting

Immoral

Worry

Responsible

Cause

Blame

F‰2,90Š ˆ 0:51p ˆ 0:60 F‰1,90Š ˆ 3:20p ˆ 0:08 F‰2,90Š ˆ 5:75

F‰2,90Š ˆ 1:10p ˆ 0:34 F‰1,90Š ˆ 0:06p ˆ 0:81 F‰2,90Š ˆ 3:49()

F‰2,90Š ˆ 0:67p ˆ 0:51 F‰1,90Š ˆ 0:16p ˆ 0:69 F‰2,90Š ˆ 11:45

F‰2,90Š ˆ 0:60p ˆ 0:55 F‰1,90Š ˆ 7:66p ˆ 0:007 F‰2,90Š ˆ 8:58

F‰2,90Š ˆ 0:25p ˆ 0:78 F‰1,90Š ˆ 2:14p ˆ 0:15 F‰2,90Š ˆ 5:17

F‰2,90Š ˆ 0:55p ˆ 0:58 F‰1,90Š ˆ 1:81p ˆ 0:18 F‰2,90Š ˆ 3:82()

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

Table 4 Results of analyses of variance (group by level of disturbance) of `sensitivity to omission bias'. p < 0.00017, p < 0.0017, p < 0.0083, ()p < 0.05 (a is adjusted due to multiple signi®cance tests; a ˆ 0:001=6 ˆ 0:00017; a ˆ 0:01=6 ˆ 0:0017; a ˆ 0:05=6 ˆ 0:0083

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

Fig. 2. Graphs of mean scores of `sensitivity to omission bias'.

1155

1156

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

scenarios about which they were most and least disturbed. Analyses of variance were then conducted for each variable in order to investigate any e€ect of group (obsessionals, nonclinical controls and anxious controls), of level of disturbance (most or least) and any interaction. The results are shown in Table 4 and represented in graphical form in Fig. 2. Results show that there are no main e€ects of group or of level of disturbance, but that there is an interaction in the variables `act', `worry' and `responsible' at p < 0.0083 level (where a is adjusted due to multiple signi®cance tests). The variables: `likelihood of acting', `immoral', `worry', `responsible', and `cause' showed signi®cant interactions at the p < 0.05 level, and the interaction seen for the variable `blame' is nearly signi®cant. Further analyses on each of the variables were conducted to investigate this e€ect. Firstly, paired t-tests were conducted in order to test the di€erence between the sensitivity to omission bias between the scenarios about which individuals are most and least disturbed. The results show that the obsessionals show a signi®cant di€erence between sensitivity to omission bias in the scenarios about which they are most and least concerned in respect of the variables: likelihood of acting, worry, responsible, cause and blame. For the obsessionals, sensitivity to omission bias is greater in the scenario about which individuals are least concerned than in the scenario about which they are most disturbed. In the control groups, although the e€ect was not signi®cant, the mean score of sensitivity to omission bias is always greater for the scenario about which individuals are most concerned, except the score of likelihood of acting and blame in the non-clinical group, and responsible in anxious control group. The results are shown in Table 5. One way analyses of variance were then conducted in order to compare sensitivity to omission bias within the three groups. It was found that there are signi®cant di€erences for the variable `worry' for both the scores on the situations about which individuals are most disturbed and the situations about which individuals are least disturbed (see Table 6). The ®nal analysis to be conducted in the investigation of the interactions between the sensitivity to omission bias and level of disturbance involved a comparison of the di€erence between the omission bias in the scenario about which individuals are most disturbed and the scenario about which individuals are least disturbed as a covariate. Analyses of covariance using sensitivity to omission bias of the scenario about which individuals were least disturbed could not be carried out as there were signi®cant di€erences between each group for these Table 5 Results of paired t-tests for each of the groups between sensitivity to omission bias for the scenarios about which individuals are most and least concerned. p < 0.00017, p < 0.0017, p < 0.0083 (a is adjusted due to multiple signi®cance tests; a ˆ 0:001=6 ˆ 0:00017; a ˆ 0:01=6 ˆ 0:0017; a ˆ 0:05=6 ˆ 0:0083†

Likelihood of acting Immoral Worry Responsible Cause Blame

Obsessionals

Non-clinical controls

Anxious controls

t‰30Š ˆ ÿ3:46 t‰30Š ˆ ÿ1:59, p ˆ 0:12 t‰30Š ˆ ÿ4:75 t‰30Š ˆ 4:94 t‰30Š ˆ ÿ3:81 t‰30Š ˆ ÿ3:09

t‰38Š ˆ ÿ0:58, p ˆ 0:57 t‰38Š ˆ 1:72, p ˆ 0:09 t‰38Š ˆ 1:80, p ˆ 0:08 t‰38Š ˆ 0:23, p ˆ 0:82 t‰38Š ˆ 0:13, p ˆ 0:90 t‰38Š ˆ ÿ0:05, p ˆ 0:96

t‰25Š ˆ 1:31, p ˆ 0:20 t‰25Š ˆ 1:12, p ˆ 0:28 t‰25Š ˆ 1:53, p ˆ 0:14 t‰25Š ˆ ÿ0:09, p ˆ 0:93 t‰25Š ˆ 0:84, p ˆ 0:41 t‰25Š ˆ 0:68, p ˆ 0:50

Obsessionals (mean (S.D.)) Situations about which individuals are most concerned: likelihood of acting Immoral Worry Responsible Cause Blame Situations about which individuals are least concerned: likelihood of acting Immoral Worry Responsible Cause Blame a

0.39a (2.21)

Non-clinical controls (mean (S.D.))

Anxious controls (mean (S.D.))

1.34a (2.05)

1.52a (2.22)

F‰2,101Š ˆ 2:69, p ˆ 0:07

1.34a (2.33) 1.46b (2.06) 1.07a (2.21) 0.85a (2.11) 1.17a (2.43)

0.80a (2.06) 1.12ab (2.03) 0.64a (2.63) 0.88a (2.26) 0.84a (2.37)

F‰2,101Š ˆ 0:70, p ˆ 0:50 F‰2,101Š ˆ 5:32 F‰2,101Š ˆ 3:17, p ˆ 0:05 F‰2,101Š ˆ 2:15, p ˆ 0:12 F‰2,101Š ˆ 1:48, p ˆ 0:23

2.64a (2.83)

1.77ab (2.47)

0.76a (2.47)

F‰2,108Š ˆ 3:43, p ˆ 0:04

1.47a 1.61a 1.72a 1.67a 1.67a

0.40a (1.95) 0.75ab (1.85) 1.06a (2.22) 0.81a (2.56) 1.02a (2.69)

ÿ0.28a (1.59) ÿ0.10b (1.66) 0.68a (1.57) 0.40a (2.34) 0.40a (2.66)

F‰2,108Š ˆ 2:93, p ˆ 0:06 F‰2,108Š ˆ 4:05 F‰2,108Š ˆ 1:99, p ˆ 0:14 F‰2,108Š ˆ 2:24, p ˆ 0:11 F‰2,108Š ˆ 1:80, p ˆ 0:17

0.75a 0.08a ÿ0.11a 0.03a 0.25a

(2.27) (1.57) (1.35) (1.44) (2.25)

(2.44) (1.69) (2.24) (2.34) (2.39)

values on the same line which share a superscript are not signi®cantly di€erent from each other.

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

Table 6 Means and standard deviations of sensitivity to omission bias in situations about which individuals are most disturbed. p < 0.008 (a is adjusted due to multiple signi®cance tests; a ˆ 0:05=6 ˆ 0:0083)a

1157

1158

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

scores (see (Everitt, 1996)). As an alternative the di€erences between sensitivity to omission bias for the most and least scenarios were calculated and one way analyses of variance were conducted on this variable. In other words, the most±least di€erence between each group was investigated between groups. For any variable showing a signi®cant interaction multiple comparisons between groups were carried out. Such analyses were conducted for each variable and are represented in Table 7. The results show that the di€erence between sensitivity to omission bias for the scenario about which individuals are most concerned and the scenario about which individuals are least concerned, is greater for the obsessionals than the controls. This e€ect is signi®cant at p < 0.0083 level (where a is adjusted for multiple comparisons; a ˆ 0:05=6 ˆ 0:0083† for variables `act', `worry' and `responsible'. However, in all variables there is a trend that this di€erence is greater for obsessionals than nonobsessionals, suggesting that obsessionals are less sensitive to omission bias for the scenario about which they are most concerned. In fact, where the control groups generally demonstrate an increase in sensitivity to omission bias for the scenario about which they are most concerned, the obsessionals actually show a decrease. This can most easily be explained by using an example variable: likelihood of acting has been chosen, although the overall pattern is seen for all variables. Nonobsessionals are more likely to act when the situation involves a commission than an omission Ð this is true for both the situations about which they were most disturbed and the situation about which they are least disturbed but the di€erence is greater when an individual is more disturbed about a scenario. In contrast, although the obsessionals are more likely to act when the situation about which they are least disturbed involves a commission than when a situation involves an omission, this e€ect is not found in situations about which obsessionals are most disturbed. Instead, regarding situations about which obsessionals are most disturbed, when confronted with a decision whether or not to act to prevent harm, there is no di€erence in behaviour scores between the scenarios presented with or without the external stimulus transforming the situation such that not acting to prevent harm is an active decision for the harm to occur.

Table 7 Means and standard deviations of Di€erences in sensitivity to omission bias between scenario about which individuals are most concerned and scenarios about which individual are least concerned. p < 0.00017, p < 0.0017, p < 0.0083; ()p < 0.05. (a is adjusted due to multiple signi®cance tests; a ˆ 0:001=6 ˆ 0:00017; a ˆ 0:01=6 ˆ 0:0017; a ˆ 0:05=6 ˆ 0:008†a

Likelihood of acting Immoral Worry Responsible Cause Blame a

Obsessionals (mean (S.D.))

Non-clinical controls (mean (S.D.))

Anxious controls (mean (S.D.))

ÿ2.43a (3.85) ÿ1.10a (3.79) ÿ1.93b (2.23) ÿ2.30b (2.55) ÿ2.07b (2.97) ÿ1.80a (3.19)

ÿ0.34ab (3.64) 0.82b (2.93) 1.00a (2.05) 0.11a (2.80) 0.08a (3.72) ÿ0.03ab (3.34)

0.76b (2.91) 0.52b (2.33) 0.76a (2.49) ÿ0.04a (2.15) 0.48a (2.85) 0.44b (3.23)

F‰2,92Š ˆ 5:75 F‰2,92Š ˆ 3:49() F‰2,92Š ˆ 11:45 F‰2,92Š ˆ 4:58 F‰2,92Š ˆ 5:17() F‰2,92Š ˆ 3:82()

values on the same line which share a superscript are not signi®cantly di€erent from each other.

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

1159

Obsessionals and nonobsessionals appear to act in similar ways in all scenarios, except for the one about which they are most disturbed. In these scenarios, obsessionals show a di€erent pattern of behaviour.

5. Discussion The results of this study support the hypothesis that, as individuals perceive themselves as more `engaged' in a potentially harmful situation, all participants rate their emotional and behavioural reactions as being intensi®ed. That is, they behave and feel in ways characteristic of obsessional problems (as measured by the variables: likelihood of acting to prevent harm, feeling immoral, being worried, feeling responsible, feeling that one might be the cause of and to blame for harm). Both (a) the presence of an intrusive thought, and (b) external circumstances which result in the requirement of an active decision (such that not acting is a commission instead of an omission), increase obsessive-like reaction. This is true both for obsessionals and nonobsessionals when the full range of situations included in the questionnaire were included in the analysis. However, when the focus was tightened to be speci®c to situations about which individuals rated themselves as most likely to be disturbed (which for obsessional patients involves situations in which obsessional behaviour is usually evoked), obsessional patients showed a di€erent pattern of reactions compared to both nonclinical and anxious controls. In this analysis, obsessionals react similarly to omission and commission situations for their most highly disturbing situation. For the least disturbing situation, the di€erence between omission and commission is present in the same way for both obsessional and control participants. These ®ndings are consistent with the hypothesis that obsessional patients are relatively sensitive to omission, but only for situations which activate their obsessional problem. Situations relevant to their obsessional problem with no external stimulus (i.e. only involving the possibility of harm arising from an omission) elicit the same strong reaction from them as the same situation in which external factors indicate that harm by commission may be possible. It appears then that obsessionals behave the same way as nonobsessionals in all situations except for those about which they are most disturbed. This sensitivity to omission is shown most strongly in the variables `act', `worry' and `responsible.' In fact, the remaining variables do not show signi®cant interactions when alpha is adjusted for multiple comparisons. However, it may still be productive to consider these e€ects; some statisticians have argued that adjustments for multiple comparisons can lead to clinically relevant ®ndings being discarded (Rothman, 1990). Whether or not this adjustment is made, it could be concluded that the di€erence between omission and commission in situations about which individuals are most disturbed is greater for nonobsessionals than for obsessionals, and so is worth considering the results further, both in research and therapy. A limitation of the present study is the fact that individuals were asked to rate how they would feel in certain situations; there is question as to how much this replicates how individuals actually do feel in such situations. It would be useful to extend such a study investigating the way in which individuals act and how they feel when actually confronted with situations. As a possible design, using the example of broken glass, a scenario could be set up with three conditions: one in which participants see some broken glass; another in which they see the

1160

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

broken glass and the experimenter probes an intrusive thought such as ``someone might hurt themselves on that glass''; and a third in which the participant sees some broken glass and the intrusive thought is probed, with the experimenter making a statement which transforms the situation into a commission, such as ``do you think you had better remove that glass so no one hurts themselves on it?'' A study design such as this, in which the scenarios are designed particularly for each participant's particular concerns, would not only overcome the problem of the scenarios not being idiosyncratic, but also the unreality of hypothetical scenarios which involves the uncertainty of whether individuals really would act in the way that they say they would. However, such a study may be dicult to conduct because of ethical constraints. Together with previous ®ndings by the group (Salkovskis et al., in press), these results demonstrate that factors in¯uencing obsessional behaviour include (i) more frequent intrusions about possible harm; and (ii) responsibility appraisals and assumptions triggered by such thoughts, which relate to the perception of whether not acting to prevent harm simply involves failure to prevent harm (an omission) or involves an active decision to not prevent harm (a commission). In this study, for nonobsessionals these two factors were external. It is suggested that for obsessionals, when considering the situation about which they are most disturbed, these factors are internal thus resulting in obsessionals frequently deciding to act to prevent harm. The internal factor transforming the situation is thought to result from obsessionals' increased sense of responsibility. This results in the obsessional perceiving him/herself as the agent so that not acting to prevent harm is equivalent to an active decision for the harm to occur. It appears then, that the occurrence of intrusions about possible harm, and responsibility assumptions and appraisals (factors characteristic of OCD) lead to a decision whether or not to prevent harm, in which not to act is an active decision not to prevent foreseeable potential harm. Omission bias does occur in obsessionals, but it is masked in the situations about which they are most disturbed by (i) the more frequent occurrence of intrusions, and (ii) responsibility assumptions leading to interpretations meaning that the individual perceives him/ herself as responsible for any potential harm, thus this internal factor transforms the situation so that not acting involves a commission and not an omission. These ®ndings can be related to work on decision making in OCD, and in this way in the therapy of OCD. It has been suggested (Beech & Liddell, 1974) that obsessionals have an inability to make decisions. Reed (1968) proposed that `obsessionals' have a particular inability to categorize, terminate, organize and structure their experiences, which they themselves refer to as a 1ack of decision', and that this is the cause of obsessional doubting. Reed (1969) described the patients as quite capable of `making decisions' through deductive reasoning, but harassed by a lack of conviction about the conclusions which they had drawn. The ®ndings described in this paper are consistent with Reed's (1985) description that obsessional overstructure a problem, and with the general notion from the cognitive-behavioural theory of OCD (Salkovskis, 1985, 1989) that obsessionals `try too hard'. The authors suggest, then, that obsessionals do not have an inferior decision making process, but that the reason for obsessional behaviour is a di€erence compared to nonobsessionals in the factors in¯uencing the decision. This relates to trying too hard to be sure no harm occurs, to have no bad thoughts, to do the right thing etc. The present study leads to three conclusions, all of which are consistent with the proposed

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

1161

additions to the cognitive theory of OCD. Firstly, the presence of an intrusive thought increases the likelihood of `obsessional behaviour'. Secondly, when the full range of situations are considered obsessionals and controls react more strongly when the situation is modi®ed by an external factor, so that not acting to prevent potential harm would in e€ect be an active decision not to seek to prevent harm compared to when such situations are perceived as omissions. Thirdly, when the analysis is con®ned to the individuals most disturbing situation, nonobsessionals continue to react more strongly when an external factor is involved; this di€erential was not present for obsessional patients.

Acknowledgements P.M.S. is a Wellcome Trust Senior Research Fellow. A.L.W. carried out this work whilst she was holding a Wellcome Trust Prize Studentship. The authors are grateful to Jonathon Baron for help with theoretical and measurement issues.

References Beck, A. T. (1976). Cognitive therapy and the emotional disorders. New York: International Universities Press. Beck, A. T., Epstein, N., Brown, G., & Steer, R. A. (1988). An inventory for measuring clinical anxiety: psychometric properties. Journal of Consulting and Clinical Psychology, 56, 893±897. Beck, A. T., Ward, C. H., Mendelsohn, M., Mock, J., & Erbaugh, J. (1961). An inventory for measuring depression. Archives of General Psychiatry, 18, 561±571. Beech, H. R., & Liddell, A. (1974). Decision-making, mood states and ritualistic behaviour among obsessional patients. In H. R. Beech, Obsessional states. London, UK: Methuen and Co. Ltd. Everitt, B. S. (1996). Making sense of statistics in psychology: a second level course. Oxford: Oxford University Press. Foa, E. B., Kozak, M. J., Salkovskis, P. M., Coles, M. E., & Amir, N. (1998). The validation of a new obsessive± compulsive disorder scale: The Obsessive±Compulsive Inventory. Psychological Assessment, 10(3), 206±214. Forrester, E., Wilson, C., & Salkovskis, P. M. (submitted for publication). The occurrence of intrusive thoughts transforms meaning in ambiguous situations: an experimental study. Gregory, R. L., & Zangwill, O. L. (1987). The Oxford companion to the mind. Oxford: Oxford University Press. Hodgson, R. J., & Rachman, S. J. (1978). Obsessional-compulsive complaints. Behaviour Research and Therapy, 15, 385±389. Reed, G. F. (1968). Some formal qualities of obsessional thinking. Psychiatric Clinics, 1, 382±392. Reed, G. F. (1969). `Under-inclusion': a characteristic of obsessionals personality disorder. I. British Journal of Psychiatry (115), 781±785. Reed, G. F. (1985). Obsessional experience and compulsive behaviour: a cognitive-structural approach. London: Academic Press. Rothman, K. J. (1990). No Adjustments are needed for Multiple Comparisons. Epidemiology, 1(1), 43±46. Salkovskis, P. M. (1985). Obsessional-compulsive problems: a cognitive-behavioural analysis. Behaviour Research and Therapy, 23, 571±583. Salkovskis, P. M. (1989). Cognitive-behavioural factors and the persistence of intrusive thoughts in obsessional problems. Behaviour Research and Therapy, 27, 677±682. Salkovskis, P. M. (1996a). The cognitive approach to anxiety: threat beliefs, safety seeking behaviour, and the special case of health anxiety and obsessions. In P. M. Salkovskis, Frontiers of Cognitive Therapy (pp. 48±74). New York: Guilford.

1162

A.L. Wroe, P.M. Salkovskis / Behaviour Research and Therapy 38 (2000) 1141±1162

Salkovskis, P. M. (1996b). Cognitive-behavioural approaches to the understanding of obsessional problems. In R. Rapee, Current controversies in the anxiety disorders. New York: Guilford. Salkovskis, P. M. (1999). Understanding and treating obsessive±compulsive disorder. Behaviour Research and Therapy, 37, s29±s52. Salkovskis, P. M., & Kirk, J. (1997). Obsessive±compulsive disorder. In D. M. Clark, & C. G. Fairburn, The science and practice of cognitive-behaviour therapy. Oxford: Oxford University Press. Salkovskis, P. M., Forrester, E., & Richards, H. C. (1998). The cognitive behavioural approach to understanding obsessional thinking. British Journal of Psychiatry, 173, 53±63. Salkovskis, P. M., Richards, C., & Forrester, E. (1995). The relationship between obsessional problems and intrusive thoughts. Behavioural and Cognitive Psychotherapy, 23, 281±299. Salkovskis, P. M., Wroe, A., Gledhill, A., Morrison, N., Forrester, E., Richards, H. C., Reynolds, M., Thorpe, S. (in press). Responsibility attitudes and interpretations are characteristic of obsessive±compulsive disorder. Behaviour Research and Therapy. Spielberger, C. D. (1983). Manual for the State Trait Inventory STAI (form Y). Palo Alto, CA: Consulting Psychologists Press. Spranca, M., Minsk, E., & Baron, J. (1991). Omission and commission in judgment and choice. Journal of Experimental Social Psychology, 27(1), 76±105. Wroe, A. L. (1997). Decision making, anxiety and behaviour in obsessive±compulsive disorder and health concern. Oxford: Oxford University Press. Wroe, A. L., Salkovskis, P. M., & Richards, H. C. (in press). ``Now I know it could happen, I have to prevent it'': A clinical study of the speci®city of intrusive thoughts and the decision to prevent harm. Behavioural and Cognitive Psychotherapy.