265

The British Psychological Society

British Journal of Clinical Psychology (2008), 47, 265–279 q 2008 The British Psychological Society

www.bpsjournals.co.uk

Obsessive compulsive symptoms and the simulation of future negative events Nadine Keen1, Gary P. Brown1* and Jonathan Wheatley2 1 2

Royal Holloway University of London, Surrey, UK Psychology Department, Whittington Hospital, London, UK Objectives. To explore whether Kahneman and Tversky’s (1982) simulation heuristic might help account for why the obsessions of people with obsessive compulsive disorder (OCD) are so compelling to them. It was predicted that participants would be better able to simulate a scenario relevant to a central OCD fear than they would scenarios related to other OCD and non-OCD fears and that how well participants simulated feared scenarios would be associated with higher ratings of subjective probability for that outcome and consequently greater worry. Design. Individuals with obsessive compulsive symptoms mentally simulated hypothetical scenarios so as to enable comparison of a personally relevant to less relevant scenarios. Methods. Thirty participants recruited from OCD support groups simulated four scenarios each and completed symptom and relevant construct measures. Results. Personally relevant scenarios were better simulated than less relevant scenarios. ‘Goodness of simulation’ (GOS) was related to worry about the feared outcome, but this was not mediated by raised subjective probabilities. GOS correlated with OCD symptomatology, anxiety, and depression but not with cognitive variables thought to be related to OCD phenomenology. Conclusion. The overall findings converge with recent literature (O’Connor, 2002) emphasizing the importance of imagination and imaginary narratives in fuelling OCD symptoms.

Cognitive appraisal models of obsessive compulsive disorder (OCD; Salkovskis, 1985; Wells, 1997) suggest that faulty inferences concerning intrusive thoughts and beliefs about the importance of thoughts drive OCD symptoms. Several underlying mechanisms have been implicated, including heightened responsibility, thought–action fusion (TAF), self-doubt, over-importance of thoughts, cognitive control, perfectionism, overestimation of threat, and intolerance of uncertainty (Clark, 2005; Frost & Steketee, 2002). According to these models, individuals who make these faulty appraisals or hold * Correspondence should be addressed to Dr Gary P. Brown, Psychology Department, Royal Holloway University of London, Egham, Surrey TW20 0EX, UK (e-mail: [email protected]). DOI:10.1348/014466508X282833

266 Nadine Keen et al.

maladaptive beliefs about unwanted intrusive thoughts are motivated to engage in selfperpetuating neutralizing behaviour. These appraisal-based models have been influential in informing clinical practice. Despite these advances, even when cognitive aspects of OCD are addressed, a significant proportion of people with OCD do not respond to treatments (Abramowitz, 1997). Thus, there is likely ample scope for building on the literature aimed at understanding the cognitive mechanisms underlying OCD. One potential area of development is methodological. Much of the research into appraisal models has relied heavily on the use of questionnaire methodology. Brown, MacLeod, Tata, and Goddard (2002) argue that questionnaires frequently lack ecological validity and ‘focus on phenomena only indirectly related to the type of thinking processes encountered in individuals with emotional problems’ (p. 1). They argue further that questionnaire-based approaches promote a focus on static constructs and do not capture the recurrent, cyclical, and time-consuming thinking processes that cause the most distress for individuals with disorders like OCD. It is also unlikely that questionnaires, by their nature, can tap into the phenomenal reality of individuals with OCD, which is usually idiosyncratic, and would therefore not be adequate for answering a central question about the disorder, namely why the content of obsessions frequently seems to be hypothetically possible to individuals with OCD. Most research to date has not focused on this potentially key aspect of the subjective experience of the disorder. More recently, O’Connor and colleagues (e.g. O’Connor & Robillard, 1995, 1999) have worked towards this objective in their inference-based approach, in which reasoning tasks are employed to examine the imaginal narratives of people with OCD. Within this approach, obsessions are viewed as inferences about reality, arrived at on the basis of an inductive narrative (O’Connor, 2002). Although the individual may initially perceive reality adaptively, as a result of reasoning errors that lend tenability to the obsessional inferences (O’Connor & Robillard, 1995) he or she is liable to be influenced by self-generated narratives that lead them to doubt their perception of external experience in favour of a hypothetical, internally generated version (Pelissier & O’Connor, 2002). Ultimately, these errors give rise to inferential confusion whereby a remote possibility (e.g. that one might become contaminated with the AIDS virus by sitting on a stained chair) is conflated with a completely fictional narrative (e.g. that a drug user may have previously sat in that chair and may have left behind a needle which is hidden in the seat), resulting in what they had previously only imagined now being perceived as a real possibility. Thus, ‘while the senses may have already rendered the person “certain” that there is no danger, the doubt from the internally generated narrative trumps this “certainty” and replaces it with a “maybe”’ (Clark & O’Connor, 2005, p. 156). Subsequently, the individual acts ‘as if’ this imagined supposition is potentially real and is driven to try to modify it, albeit unsuccessfully (O’Conner & Aardema, 2003). Inference- and appraisal-based models are not mutually exclusive; whereas appraisal models are mostly concerned with interpretation of an intrusion after the event, inferential confusion relates to the perceived likelihood of the relevant hypothetical events at the time of the intrusion (Clark & O’Connor, 2005). The inferential model suggests that imagination may play more of a role in OCD than is presently provided for within appraisal models and can potentially help explain a striking central feature of OCD, namely that people with OCD are typically concerned with imagined outcomes that are often bizarrely unlikely ( Jakes, 1996). Indeed, Wheatley (2000) has argued that the main cognitive models of OCD have neglected the role of imagination in the maintenance of OCD, despite the fact that it has been shown that procedures such as imaginal exposure used alongside exposure with response

OC symptoms and simulation

267

prevention produce longer-term positive outcomes in the treatment of obsessive checking (Foa, Steketee, Turner, & Fischer, 1980). From the general literature on anxiety disorders (e.g. Butler & Mathews, 1983), it seems likely that imagery and imagination would be associated with increased subjective probabilities for imagined negative events in OCD. Twenty years ago, De Silva (1986) noted that clinical accounts of OCD and diagnostic definitions had for decades uniformly highlighted the role of images in OCD, but that despite this, ‘the literature is almost totally bereft of any detailed examination of obsessional-compulsive imagery as a specific phenomenon in its own right’ (p. 334). Although the importance of imagery in clinical psychology research is increasingly being recognized (e.g. Hackmann & Holmes, 2004), it is still not a focus of research on OCD. Further grounds for a focus on imagery comes from experimental psychology, within which there is considerable evidence suggesting that imagining a future event increases the subjective likelihood that the event will occur. For example, when Gregory, Cialdini, and Carpenter (1982) asked participants to imagine themselves at a positive event (e.g. winning a contest) or a negative event (e.g. being wrongly arrested for a crime), they rated the event that they had been asked to imagine as being more likely to occur. However, it does not appear to be the case that imagination alone always increases subjective probability. It has been found that an important moderating factor is the ease with which an event can be imagined. For example, Sherman, Cialdini, Schwartzman, and Reynolds (1985) asked participants to imagine contracting a disease whose symptoms were either easy or difficult to imagine. The ease with which participants could imagine the symptoms was linked to their subjective probability appraisals. Relatedly, people who find it easier to generate reasons why an event might happen, compared to why it might not, will think that the event is more likely (Campbell & Fairey, 1985; MacLeod, 1999). The impact that the ease with which mental operations can be performed has on decision-making processes is a central focus of Tversky and Kahneman’s (1973; Kahneman & Tversky, 1982) work on heuristics. They distinguished the availability heuristic, whereby people estimate the likelihood of a particular event happening in the future by how easily they are able to bring relevant instances from long-term memory to mind, from the simulation heuristic, according to which the subjective probability of a given outcome depends upon the ease with which a mental model of the hypothetical situation can be constructed. Kahneman and Tversky (1982) argue that the latter is employed when the situation in question is uncommon or unique and therefore lacking the database of past experiences drawn upon by the availability heuristic. The concerns of individuals with OCD are almost always about imagined events that have never occurred before, at least to them, and so it is possible that the simulation heuristic would have a role in the creation of mental models of the feared outcomes anticipated in the course of obsessive intrusions. Brown and colleagues (2002) developed a methodology for operationalizing the simulation heuristic that seeks to tap into these imaginal processes. In Brown et al.’s (2002) study, women who were pregnant for the first time were asked to mentally simulate going into labour and arriving at the hospital on time. It was found that ‘goodness of simulation’ (GOS; the ease at which scenarios could be simulated) of a successful outcome was associated with higher subjective probabilities for that outcome and consequently less worry. Using this methodology, Wheatley (2000) conducted a pilot study to explore the potential role of the simulation heuristic in OCD. He found that scenarios related to an individual’s obsessions were, on average, better

268 Nadine Keen et al.

simulated, rated as more worrying, and had higher subjective probabilities than similar obsessive scenarios unrelated to the individuals’ particular concerns. The principal aim of the current study was to explore the role of the simulation heuristic in OCD using the methodology outlined by Brown et al. (2002), hence extending Wheatley’s (2000) pilot study. It was suggested that future outcomes that were particularly salient to a person’s symptom-specific concerns would be more easily simulated and therefore more readily accepted as true. Understanding the role of the simulation heuristic in OCD could help explain why the obsessions of people with OCD are so compelling and why they are frequently associated with heightened probability estimates. The central hypothesis was that GOS, as evidenced in the participants’ verbal protocols, would be higher for scenarios relevant to the participants’ personal OCD fear relative to a non-personal OCD scenario, a general worry scenario, and a worry scenario with a positive outcome. An association of GOS with subjective probability for the outcome in question would be consistent with the operation of the simulation heuristic, and high subjective probability should, in-turn, be associated with greater worry about the outcome.

Method Participants Thirty adult volunteers (17 men and 13 women) were recruited through a British national OCD charity. The mean age of the sample was 42.3 years (SD ¼ 11:92; range ¼ 21 – 71 years). Materials Development of the scenarios The procedure for eliciting simulations followed the method outlined by Brown et al. (2002), which was loosely modelled on the means-ends problem solving approach (MEPS; Platt & Spivak, 1977). In line with the standard MEPS procedure, respondents were provided with the beginning of an imaginary scenario and the end of the scenario and were asked to give a step-by-step account of what would happen in between these two points. As an example, the final version of the contamination scenario read: ‘I am going to describe to you the beginning of a future situation and the end of the situation and I want you to tell me what you imagine the middle will be. At the beginning of the situation you have just been served your main course at a restaurant and have begun eating, when you realize that your waitress has just used the toilet before serving you. Take a moment to imagine that. At the end of the situation it is the next morning and you feel distinctly unwell. Now go back to the beginning of the situation, where you have just been served your main course, and describe step-by-step exactly what will happen from that point onwards.’ This method was previously piloted in a small OCD sample by Wheatley (2000). The scenarios in Wheatley’s (2000) study reflected the following dimensions of OCD: harming, contamination, checking, hoarding, and offending others. A review of factor and cluster analysis studies that have examined the dimensions of OCD symptomatology (Baer, 1994; Calamari et al., 1999; Hantouche & Lancrenon, 1996; Leckman et al., 1997; Mataix-Cols et al., 1999, 2002; Sumerfeldt et al., 1999; Van Oppen et al., 1995) indicated that, with the exception of the last category (offending others), these were the most consistently reported dimensions of OCD. A further dimension of symmetry/ordering

OC symptoms and simulation

269

was not reflected in Wheatley’s (2000) study but was consistently reported in the above factor analysis studies. Hence, except for symmetry/ordering, Wheatley’s (2000) scenarios appeared to reflect the range of obsessive fears reported in the literature and would thus likely have personal relevance to study participants. Wheatley’s (2000) scenarios were revised through consultation with a group of people with obsessive compulsive symptoms using the Delphi method, a means for developing consensus among experts in a way that seeks to bypass the group dynamics that may influence or bias individual opinion formation (Murphy et al., 1998). The four members of the group (two men and two women) were recruited from an OCD support group and were asked to comment on whether Wheatley’s (2000) scenarios accurately reflected typical OCD fears. Responses were gathered individually, aggregated, and then fed back to the group members for further comment. This continued until a consensus (defined as at least 75% agreement) was reached. This was achieved after two rounds of feedback, and the original scenarios were revised accordingly: the ‘offending others’ scenario was removed, a second checking scenario was added, and the ending of the existing checking scenario was modified. Finally, three control scenarios were developed: a worry scenario revolving around a non-obsessive worry and a positive-worry scenario involving a situation which would be worrying but which had a positive outcome. A second worry scenario was developed in case the situation described in the first scenario had actually been experienced by the participant (thus making it something that was available in memory and would not be simulated de novo). These scenarios were not submitted to the Delphi procedure. Following each simulation task, participants were asked to make ratings on two Likert scales. The first asked, ‘On a scale from 1 to 9, with 1 ¼ not worried at all and 9 ¼ extremely worried, how worried are you about something like this happening to you?’ The second scale asked, ‘How probable do you think it is that something like this might happen to you? Please give a probability estimate from 1 to 7, with 1 ¼ definitely would not happen, and 7 ¼ definitely will happen’. The scales were reproduced on a card. Where participants had directly experienced any of the situations described, a replacement scenario was administered. For the OCD scenarios, the back-up scenario was the next highest ranking scenario on the event-ranking questionnaire (see below). The second worry scenario was back-up for the original worry scenario.

Scenario coding The coding system incorporated factors discussed in the literature as underlying ‘good’ simulations (Kahneman & Tversky, 1982) and was based on the codes developed by Brown et al. (2002). In the Brown et al. study, the scenario concerned a desired outcome, and the protocols provided were concerned with how to bring this about. In the present study, the endpoints were all undesired outcomes, making elements of the Brown et al. codes inapplicable (e.g. contingencies and back-up plans). As such, the scenarios of the first six participants in the present study were used to help revise the original coding system to make it applicable to negative outcomes. The five aspects of ‘GOS’ coded for were: (1) Logical sequencing: the extent to which successive elements of a scenario are connected logically to each other with each step following logically from the previous one; (2) Temporal ordering: the extent to which temporal order is communicated so that a sense of temporal flow is established; (3) Minimization of uncertainty: the extent to which the scenario increases or decreases a sense of uncertainty about what is being described; (4) Details and ease of imagining: the

270 Nadine Keen et al.

extent to which the scenario gives a comprehensive account of all the basic elements of the situation; and (5) Flows smoothly: a subjective judgment on the part of the rater of how well the scenario flows. Each dimension was rated on a 5-point scale, with each point anchored by detailed criteria and higher ratings indicating better GOS. Event-ranking questionnaire Participants were presented with short descriptions of 12 situations and asked to rank the scenarios in order of how upsetting they would find them and to make ratings of the degree to which they would find them upsetting. The statements included ones describing the central feared outcome of the eight scenarios to be used in the main procedure (five OCD items, two worry items, and one positive-worry item) and four distractor items. The scenarios corresponding to the highest ranking and lowest ranking OCD items would be used as the personal and non-personal OCD scenarios, respectively. The frequencies of the most upsetting event were Harming (N ¼ 11 of 30), Checking gas (N ¼ 6), Checking doors (N ¼ 5), Contamination (N ¼ 2), and Hoarding (N ¼ 6). Distractor task To avoid priming the simulation tasks, a distractor task was administered after the eventranking questionnaire: an opinion poll asking participant’s to give their views on whether UK should join the European single currency, a topic thought to be sufficiently involving that the participants would be distracted from thinking further about any items on the event-ranking questionnaire. Obsessive belief questionnaire-44 (OBQ-44; OCCWG, 2005) The OBQ-44 consists of items reflecting belief statements intended to be characteristic of obsessive thinking. Respondents are asked to indicate their level of agreement with each statement relative to what they are like most of the time on a 7-point rating scale ranging from 2 3 (disagree very much) to þ 3 (agree very much). The authors report three factor subscales (‘importance and control of thoughts’, ‘perfectionism and intolerance of certainty’, and ‘responsibility and threat estimation’), good evidence of convergent and discriminant validity, and good to excellent internal consistency and stability (Sica et al., 2004). Thought–action fusion questionnaire – revised (TAFS; Shafran, Thordarson, & Rachman, 1996) The TAFS is intended to assess the tendency to psychologically fuse thoughts and actions. The construct of thought–action fusion (TAF) has two components: the belief that thinking about an unwanted event makes it more likely to happen (likelihood TAF) and the belief that having an unacceptable thought is the moral equivalent of acting on that thought (moral TAF). Several studies have demonstrated a significant relationship between TAF and OCD (Shafran et al., 1996). The 19-item self-report measure includes 12 items that assess moral TAF and 7 items that assess likelihood TAF. Shafran et al. (1996) report internal consistency coefficients for the TAFS ranging from .85 to .96. Hospital anxiety and depression scale (HADS; Zigmond & Snaith, 1983) The 14-item HADS was used as a measure of depression (7 items) and anxiety (7 items), and was chosen because of its brevity and well-established psychometric properties (Mykletun, Stordal, & Dahl, 2001).

OC symptoms and simulation

271

Clark–Beck obsessive-compulsive inventory (CBOCI; Clark & Beck, 2002) Frequency and severity of OCD symptoms were assessed using the CBOCI. The CBOCI is briefer than alternative OCD measures but has comparable discriminant and convergent validity (Clark, Antony, Beck, Swinson, & Steer, 2005). The CBOCI has 25 items, each of which has a series of four statements scored 0–3 that reflect a progression of severity for that symptom. Respondents select the statement that best reflects their experiences. A principal factor analysis (Clark et al., 2005) of the CBOCI 25 items identified two correlated scales, one of obsessions (a ¼ :85), and one of compulsions (a ¼ :86). According to their CBOCI scores, 73% of participants fell in the mild-to-moderate category of OCD and 27% of participants fell into the severe category for OCD. This distribution is nearly identical to the breakdown by category of the Clark et al. (2005) sample of individuals diagnosed with OCD, and the profile of individual item scores was very similar. Procedure Measures were administered as follows: the event-ranking questionnaire, the distractor task, and the simulation scenarios, followed by the self report questionnaires in the order listed in the previous section. The procedure for eliciting simulations followed the method outlined by Brown and colleagues (2002). For the simulation tasks, the positiveworry scenario was administered first and the worry scenario last. The personal and non-personal OCD scenarios were administered second and third, the order of the two counterbalanced across participants.

Results Inter-rater reliability Using transcripts of participants’ verbatim responses, two independent raters assigned scores for each scenario on the five GOS dimensions. Raters were blind to all other information, and the entire set of protocols (four for each participant) was ‘scrambled.’ Half-point ratings were permitted. Overall agreement within one rating point was 84%, and the intra-class correlations (ICC) between the ratings ranged from .40 to .76 (see Table 1; ICC’s . .75 are usually judged to be acceptable and those , .75 but . .60 are marginally acceptable). In each case, the temporal order dimension had the lowest reliability and was the only dimension with agreement below acceptable levels. Following the initial ratings, each rating difference of greater than one point was discussed and ratings were consensually adjusted so that they differed by at most one point. The final rating for each dimension was the average of the ratings made by the two raters, and the five ratings were summed to create a total GOS score for each scenario. The internal consistency of the GOS dimensions was high (. .92) for each of the four scenarios. To establish whether GOS ratings were confounded by the raters unintentionally picking up on the personal relevance of the scenarios, one of the raters rated the protocols according to whether they appeared to be based on respondents’ personal or non-personal OCD fears from 1 (‘definitely non-personal’) to 4 (‘unable to tell’) to 7 (‘definitely personal’). No significant correlations were found between this rating and the GOS ratings for the personal or non-personal scenarios, suggesting the raters were not inadvertently basing their GOS ratings on a perception of the relatedness of the scenarios to the respondents’ central fears.

272 Nadine Keen et al. Table 1. Reliability of goodness of simulation dimensions Inter-rater agreement GOS subscale

Within one rating point (%)

Logical sequence Temporal order Minimization of uncertainty Details and ease of imagining Flows smoothly

84 76 88 89 84

ICC .64 .40 .74 .76 .63

Internal consistency Item-total r range .82–.96 .83–.92 .72–.92 .65–.85 .78–.94

Note. N ¼ 30. ICC, intra-class correlation for the average of the two ratings, one-way random effects model. Item-total correlations are for the goodness of simulation scores computed for each of the four scenarios: personal OCD; non-personal OCD; worry; and positive worry. Overall a’s for these dimensions were .96, .93, .96, and .92, respectively.

Tests of the main hypothesis Differences between the scenarios on worry, subjective probability, and GOS Mean differences between the scenarios on the GOS index, worry, and subjective probability ratings are shown in Table 2. Worry ratings for the worry and positive-worry scenarios fell between those of the personal and non-personal OCD scenarios. A repeated measures analysis of variance indicated a significant effect for scenario, Fð3; 87Þ ¼ 23:42, p , :001 (sphericity assumption not rejected). Bonferroni-corrected t tests ( p ¼ :05=6 ¼ :008) indicated that the personal OCD and positive-worry ratings did not differ but that they both differed significantly from the non-personal OCD and worry ratings, which did not differ. The position of the positive-worry scenario was not expected, as it was a scenario with a positive outcome and so intended to serve as a control. However, anecdotally, participants related that the situation described (having to arrive at the airport on time after the planned means of travel was not available) had triggered some of their obsessive fears (e.g. a number of participants mentioned themes, such as having to travel on dirty public transport, which they said might trigger their OCD). Because this scenario was not operating as intended as a reference point for the ratings on the other scenarios but at the same time was not central to the main hypotheses, it was not considered in any further analyses. Table 2. Comparison of the four scenarios on ratings of goodness of simulation, worry, and subjective probability GOS indexa Scenario Personal OCD Non-personal OCD Worry Positive-worrya

M 18.38b 16.70c 15.81c 16.70

Subjective probabilitya

Worry SD

M

SD

3.53 3.32 4.45 3.38

7.03b 3.67c 4.93c 6.30b

1.85 2.17 2.70 1.85

Note. N ¼ 30. a Positive-worry not included in the analyses of GOS and subjective probability. b,c Means sharing the same superscript do not significantly differ.

M 3.80b 2.97c 3.33b,c 4.23

SD 1.58 1.38 1.12 1.70

OC symptoms and simulation

273

Within-subjects analyses of variance indicated that there were significant differences between the three remaining scenarios with respect to GOS index (Fð2; 58Þ ¼ 5:02, p ¼ :01) and subjective probability (Fð2; 58Þ ¼ 3:75, p ¼ :03) (sphericity assumed). A priori contrasts indicated that the personal OCD scenario differed from the nonpersonal OCD scenario (Fð1; 29Þ ¼ 3:99, p ¼ :05) and the worry scenario (Fð1; 29Þ ¼ 9:56, p ¼ :004) on the GOS index. The personal OCD scenario also had a higher subjective probability rating than the non-personal OCD scenario, Fð1; 29Þ ¼ 5:39, p ¼ :04, but not the worry scenario. The main hypothesis was thus essentially supported – participants found it easier to simulate and had higher subjective probabilities for a scenario related to their main obsessive fear than scenarios related to another obsessive fear that was not personally relevant and a non-OCD fear. Relationship of worry, subjective probability, and GOS As shown in Table 3, the prediction that GOS should be associated with higher subjective probability for the outcome in question which, in-turn, would be associated with greater worry, which had been found by Brown et al. (2002) for a positive outcome, was not supported. The possibility was next considered that the expected effect might not be uniform for all OCD subtypes represented in the sample. In particular, Wu and Watson (2005) have presented compelling evidence that hoarding represents a distinctive subgroup from other forms of OCD. As such, the analysis was repeated without the participants (N ¼ 6) who selected hoarding as their personal scenario. A significant positive correlation was now found between GOS and the worry rating, but not the subjective probability rating, for the personal OCD scenario, (rð22Þ ¼ :43, p ¼ :017). Worry and subjective probability ratings were found to correlate for the non-personal OCD (rð22Þ ¼ :51, p ¼ :006) and the worry (rð22Þ ¼ :35, p ¼ :048) scenarios. Furthermore, there was an unexpected significant inverse relationship between the GOS index and the worry rating for the non-personal OCD scenario (r ¼ 2:41). Construct validity of GOS The three scenario-derived variables were correlated with measures of other constructs shown to play a role in OCD, namely the subscales and total scores of the OBQ, the TAFS, and the CBOCI, as well as the depression and anxiety scales of the HADS. In the sample at large, including hoarders, GOS for the personal OCD scenario correlated only with HADS anxiety, whereas the worry rating had significant correlations with most of the variables of interest, and the subjective probability scale was, again, unrelated to most of the criterion variables (see Table 3). When the hoarding group was excluded from the analysis, higher GOS for the personal scenario was now also associated with higher CBOCI obsession scores (rð22Þ ¼ :40) and HADS depression scores (rð22Þ ¼ :38). Tests for potential methodological artefacts A basic assumption of the GOS index is that it is a ‘deep’ property of the verbal protocols provided by the participants and not simply a reflection of surface properties such as the mere amount of information provided (Brown et al., 2002). To examine this, GOS for the personal OCD scenario was correlated with scenario length (the number of characters in the scenario), the number of sentences produced, and sentence length (the number of characters per sentence). GOS was found to correlate with scenario length and number of sentences (both r’s ¼ :41) in the full sample and in the subsample excluding

2 .13

.19 .37* .43* .38*

.16 .33* .39* .26

.23 .30*

.46* .14 .35*

.21 2 .10

.01 .08 2 .14 2 .05

2 .05 .09 2 .03 2 .02

.27 .46*

.22 2.02 .11

WR

.01 .24 .14

.06 .13

.10 .01 .13 .10

.07 2.04 .19 .12

SPR

.40* .01 .23

.38* .59*

.13 .27 .12 .17

.21 .27 .03 .19

.43* 2 .04

GOS

.43* .24 .38*

.12 .29

.12 .35* .33* .23

.12 .27 .37* .29

.03

WR

.07 .28 .19

.11 .27

.12 .00 .22 .14

.08 .04 .30 .22

SPR

.01 .12 .40* .15 .31

2 .04 2 .36* 2 .22

.44* .41* .30* .46*

.25 .51* .15 .37*

.51*

WR

.12 .07

2 .03 2 .28 2 .45* 2 .18

.13 2 .13 2 .26 2 .06

2 .41* 2 .14

GOS

.10 2.10 .01

.06 .17

2.27 2.08 .30 .04 .19

2.23 .11 .01 2.14

.07 2.12 2.13 2.08

.19 .01

GOS

.36* .40* .42* .43*

.30 .46* .03 .33*

SPR

2 .01 .01 .01

.37* .36*

.10 .18 .19 .15

2 .01 2 .11 .17 2 .05

.35*

WR

SPR

.33* .19 .29

.12 .29

.31 .28 .26 .34*

.28 .44* 2 .04 .29

Worry scenario

Notes. N ¼ 30 including hoarders, N ¼ 24 excluding hoarders. GOS, Goodness of Simulation; WR, Scenario worry rating; SPR, Scenario subjective probability rating; OBQ, Obsessive beliefs questionnaire; RT, Responsibility/threat estimation; PC, Perfectionism/certainty; ICT, Importance/control of thoughts; TAFS, Thought–action fusion scale; HADS, Hospital anxiety and depression scale; CBOCI, Clark–Beck obsessive compulsive inventory. *p , :05, one tailed.

Scenario ratings Worry Subjective probability OBQ RT PC ICT Total TAFS Morality Self-likelihood Other likelihood Total HADS Depression Anxiety CBOCI Obsessions Compulsions Total

GOS

Non-personal OCD scenario

Personal OCD scenario

Personal OCD scenario

Excluding hoarders

Including hoarders

Table 3. Construct validity correlations

274 Nadine Keen et al.

OC symptoms and simulation

275

hoarders (r ¼ :37 and :36, respectively). Some relationship between GOS and these surface features would be expected, as some of the components of GOS (e.g. level of detail of the scenario) would naturally presuppose longer scenarios having been produced. However, the modest magnitudes of the correlations would suggest that GOS was not simply reducible to these surface features. Moreover, the variables related to scenario length were unrelated to the CBOCI obsessions scale (suggesting that longer scenarios were not merely a reflection of greater symptomatology related to the particular fear) and the HADS scales and so could not account for the relationship between these variables and GOS for the personal scenario.

Discussion The principal aim of the current study was to explore the potential role of Kahneman and Tversky’s (1982) simulation heuristic in OCD using a novel methodology. It was thought that the simulation heuristic might represent a mechanism that contributes to the inflated sense among individuals with OCD that their obsessive fears might come true. The central hypothesis was that participants would be better able to simulate a scenario relevant to a core OCD fear than they would scenarios related to other OCD and non-OCD fears. This hypothesis was largely supported. It was also anticipated that how well participants simulated feared scenarios should be associated with a higher rating of subjective probability for that outcome and consequently greater worry. This was supported with respect to the relationship of GOS to worry about the feared outcome (in the non-hoarder group); however, the hypothesis was not fully supported in that subjective probabilities were not found to mediate the relationship between GOS and worry, as the simulation heuristic would imply. Supporting the validity of GOS as a construct, the variable was found to correlate with levels of obsessive symptomatology, anxiety, and depression. However, it was not associated with cognitive variables thought to be related to OCD phenomenology. The methodological approach that was used has been employed only once before (Brown et al., 2002), and the present study represents the first time it has been applied to an undesired, feared outcome. The discrepancies between the present study and the Brown et al. (2002) study with respect to the scenario-derived variables (GOS, subjective probability, and worry) could, among other possible reasons, be due to substantive differences between the simulation of positive and negative outcomes or the failure to adapt the methodology sufficiently in this respect. Whereas certain changes to the stimuli and scoring criteria were made that were necessitated by the change in the nature of the outcome (e.g. the dimensions included in the GOS index), there may have been other needed changes that were not anticipated. For example, the wording of the rating of subjective probability should probably have been modified to reflect the fact that the outcome of interest in the present study was hypothetical rather than nearly certain, as it was in the Brown et al. study. Reasons for the inverse relationship found between the GOS index and worry for the non-personal scenarios are more difficult to identify. This finding is conceivably an artefact of the source of participants. Support group members might be highly familiar with the obsessive fears of others that they do not share. As such, they might be able to simulate them well and thereby be in a better position to discount them and not worry about them. While the results for the current study are encouraging for a new methodology, the pattern of findings imposes limitations on what can inferred with respect to the underlying theory. The failure to find the expected results for the subjective probability

276 Nadine Keen et al.

variable and, to a lesser extent, the cognitive appraisal and belief variables means that corroborating evidence is lacking that GOS taps into imaginal processes, as implied by the basic premise of the simulation heuristic. Indeed, the same pattern of findings could be accounted for merely on the basis of greater familiarity of individuals with their own fears as compared to other fears and thus greater facility in describing them, which is, in fact, a more parsimonious explanation of the present results. However, the two explanations may prove to be complementary. Specifically, simulation is likely to play a role in primary appraisals of intrusions, which are then amplified through secondary appraisals and reinforced through ritualizing. Subsequent repetition and ‘practice’ of the same obsession results in the relevant simulation becoming more coherent and elaborated, such that a greater sense of potential threat is produced as ever better running simulations are produced in successive cycles of the obsession. Whereas the present study appears to provide evidence consistent with the operation of the simulation heuristic in OCD, the methodology is potentially applicable to a range of other clinical issues within which the simulation heuristic could potentially operate, the main defining feature being the salience of a never-before-encountered future state of affairs that can be either desired or undesired. Examples of further applications to anxiety disorders could include anticipated catastrophic outcomes in panic disorder, a range of feared outcomes in generalized anxiety disorder, and positive and negative social encounters in social phobia. It is important to note that the nature of the GOS variable is quite different from typical cognitive variables used in this area in that it is not concerned with the content of thinking per se but with the configuration and patterning of this content. GOS might also be sensitive to features of OCD imagery that cannot be easily captured by other methods, such as flow and movement (Rachman, 1976; Rachman &Hodgson, 1980, as cited by De Silva (1986)). From the standpoint of the simulation heuristic, patterns that represent a compelling depiction of the imagined future situation and which provide a satisfactory account of how the situation might be brought about are presumed to have greater heuristic value and, consequently, to strike the individual who experiences them as more true and therefore more likely. In this respect, GOS may be getting at how ongoing experience ‘seems’ to the individual, a quality which cannot necessarily be directly or fully verbalized. Within Teasdale and Barnard’s (1993) interacting cognitive subsystems framework, it might be seen as tapping into the implicational level of meaning as opposed to the more verbally accessible propositional level. Another area requiring further explanation is the potential heterogeneity of effects in different subgroups of OCD. In the present study, certain predicted effects were only found when hoarders were excluded from the sample. This is consistent with the research reviewed by Wu and Watson (2005), which suggested that hoarding is distinct from other manifestations of OCD-like symptomatology. Perhaps this distinctiveness relates in some way to the types of cognitive factors examined in the present study. Whereas some evidence for the construct validity of GOS was found with respect to its correlations with relevant symptom measures, GOS was uniformly unrelated to cognitive variables, such as those represented by the subscales of the OBQ and the TAFS. How to interpret the absence of these associations depends on the expected theoretical relationship. The putative operation of the simulation heuristic in OCD relates to the initial intrusion rather than to the subsequent appraisals of harm, responsibility, or danger flowing from the intrusion. Specifically, it is seen to contribute to how likely and realistic the intrusion scenario seems to the person experiencing it, what O’Connor and colleagues (e.g. O’Connor & Robillard, 1999) call the ‘primary inference of doubt’,

OC symptoms and simulation

277

in contrast to the secondary appraisals that are the focus of the dominant cognitive behavioural theories of OCD. It is only recently (Clark & O’Connor, 2005) that these models have been merged, and the predicted relationship between these key aspects of cognitive models of OCD still need to be spelled out in a way the permits an empirical test. Summary and conclusion The present study evaluated the potential role of the simulation heuristic in OCD. Although there were notable gaps in the results relative to what was predicted, there were sufficient positive findings to justify further development of the approach. The present approach has potential for tapping into the type of dynamic and cyclical thinking processes at the heart of disorders like OCD that questionnaire methods are inadequate for accessing.

References Abramowitz, J. S. (1997). Effectiveness of psychological and pharmacological treatments for obsessive-compulsive disorder: A quantitative review. Journal of Consulting and Clinical Psychology, 65(1), 44–52. Baer, L. (1994). Factor analysis of symptom subtypes of obsessive compulsive disorder and their relation to personality and tic disorders. Journal of Clinical Psychiatry, 55, Suppl. 3, 18–23. Brown, G. P., Macleod, A. K., Tata, P., & Goddard, L. (2002). Worry and the simulation of future outcomes. Anxiety Stress and Coping, 15(1), 1–17. Butler, G. & Mathews, A. (1983). Cognitive processes in anxiety. Advances in Behavior Research and Therapy, 5, 551–565. Calamari, J. E., Wiegartz, P. S., & Janeck, A. S. (1999). Obsessive-compulsive disorder subgroups: a symptom-based clustering approach. Behaviour Research and Therapy, 37, 113–125. Campbell, J. D., & Fairey, P. J. (1985). Effects of self-esteem, success-failure, and self-consciousness on task performance. Journal of Personality and Social Psychology, 48, 1097–1111. Clark, D. A. (Ed.). (2005). Intrusive thoughts in clinical disorders: Theory, research, and treatment. New York: The Guildford Press. Clark, D. A., Antony, M. A., Beck, A. T., Swinson, R. P., & Steer, R. A. (2005). Screening for obsessive and compulsive symptoms: Validation of the clark-beck obsessive–compulsive inventory. Psychological Assessment, 17, 132–143. Clark, D. A., & Beck, A. T. (2002). Manual for the Clark–Beck obsessive–compulsive inventory. Can Antonio, TX: Psychological Corporation. Clark, D. A., & O’Connor, K. (2005). Thinking is believing: Ego-dystonic intrusive thoughts in obsessive–compulsive disorder. In D. A. Clark (Ed.), Intrusive thoughts in clinical disorders: Theory, research, and treatment (pp. 145–174). New York: The Guildford Press. De Silva, P. (1986). Obsessional–compulsive imagery. Behaviour Research and Therapy, 24, 333–350. Foa, E. B., Steketee, G. S., Turner, R. M., & Fischer, S. C. (1980). Effects of imaginal exposure to feared disasters in obsessive–compulsive checkers. Behaviour Research and Therapy, 18, 449–455. Frost, R. O. & Steketee, G. (Eds.). (2002). Cognitive approaches to obsessions and compulsions. Oxford: Pergamon. Gregory, W. L., Cialdini, R. B., & Carpenter, K. M. (1982). Self-relevant scenarios as mediators of likelihood estimates and compliance: Does imagining make it so? Journal of Personality and Social Psychology, 43(1), 89–99. Hackmann, A., & Holmes, E. (2004). Reflecting on imagery: A clinical perspective and overview of the special issue of memory on mental imagery and memory in psychopathology. Memory, 12(4), 389–402.

278 Nadine Keen et al. Hantouche, E. G., & Lancrenon, S. (1996). Modern typology of symptoms and obsessivecompulsive syndromes: Results of a large French study of 615 patients. Encephale, 22(1), 9–21. Jakes, I. (1996). Theoretical approaches to obsessive–compulsive disorders. New York: Cambridge University Press. Kahneman, D., & Tversky, A. (1982). The simulation heuristic. In D. Kahneman, P. Slovik, & A. Tversky (Eds.), Judgement under uncertainty: Heuristics and biases (pp. 201–208). New York: Cambridge University Press. Leckman, J. F., Grice, D. E., Boardman, J., Zhang, H., Vitale, A., Bondi, C., et al. (1997). Symptoms of obsessive-compulsive disorder. American Journal of Psychiatry, 154, 911–917. MacLeod, A. K. (1999). Prospective cognition. In T. Dalgleish & M. Power (Eds.), Handbook of cognition and emotion (pp. 267–280). Chichester, UK: Wiley. Mataix-Cols, D., Rauch, S. L., Manzo, P., Jenike, M. A., & Baer, L. (1999). Use of factor-analyzed symptom dimensions to predict outcome with serotonin reuptake inhibitors and placebo in the treatment of Obsessive-Compulsive Disorder. American Journal of Psychiatry, 156, 1409–1416. Mataix-Cols, D., Marks, I. M., Greist, J. H., Kobak, K. A., & Baer, L. (2002). Obsessive-compulsive symptom dimensions as predictors of compliance with and response to behaviour therapy: Results from a controlled trial. Psychotherapy and Psychosomatics, 71, 255–262. Murphy, M. K., Black, N. A., Lamping, D. L., McKee, C. M., Sanderson, C. F. B. J., & Marteau, T. (1998). Consensus development methods and their use in clinical guideline development. Health Technology Assessment, 2, 1–88. Mykletun, A., Stordal, E., & Dahl, A. A. (2001). Hospital anxiety and depression (HAD) scale: Factor structure, item analyses and internal consistency in a large population. British Journal of Psychiatry, 179, 540–544. Obsessive Compulsive Cognitions Working Group (2005). Psychometric validation of the Obsessive Beliefs questionnaire and interpretations of intrusions inventory: Part 2, factor analyses and testing of a briefer version. Behaviour Research and Therapy, 43, 1527–1543. O’Connor, K. P. (2002). Intrusions and inferences in obsessive–compulsive disorder. Clinical Psychology and Psychotherapy, 9, 38–46. O’Conner, K., & Aaardema, F. (2003). Fusion or confusion in obsessive compulsive disorder. Psychological Reports, 93(1), 227–232. O’Connor, K., & Robillard, S. (1995). Inference processes in obsessive–compulsive disorder: Some clinical observations. Behaviour Research and Therapy, 33, 887–896. O’Connor, K., & Robillard, S. (1999). A cognitive approach to modifying primary inferences in obsessive compulsive disorder. Journal of Cognitive Psychotherapy, 13, 1–17. Pelissier, M. C., & O’Connor, K. P. (2002). Deductive and inductive reasoning in obsessive– compulsive disorder. British Journal of Clinical Psychology, 41, 5–27. Platt, J. J., & Spivak, G. (1977). Workbook for training in interpersonal problem-solving thinking. Philadelphia, PA: Hahnemann Community Mental Health Centre. Rachman, S. (1976). The modification of obsessions: A new formulation. Behaviour Research and Therapy, 14, 437–443. Rachman, S., & Hodgson, R. J. (1980). Obsessions and compulsions. Englewood Cliffs, NJ: Prentice-Hall. Salkovskis, P. M. (1985). Obsessive–compulsive problems: A cognitive behavioural analysis. Behaviour Research and Therapy, 23, 571–584. Shafran, R., Thordarson, D. S., & Rachman, S. (1996). Thought–action fusion in obsessive compulsive disorder. Journal of Anxiety Disorders, 10, 379–391. Sherman, S. J., Cialdini, R. B., Schwartzman, D. F., & Reynolds, K. D. (1985). Imagining can heighten or lower the perceived likelihood of contracting a disease: The mediating effect of ease of imagery. Personality and Social Psychology Bulletin, 11, 201–212.

OC symptoms and simulation

279

Sica, C., Coradeschi, D., Sanavio, E., Dorz, S., Nanchisis, D., & Novara, C. (2004). A study of the psychometric properties of the obsessive beliefs inventory and interpretaions of intrusions inventory on clinical Italian individuals. Journal of Anxiety Disorders, 18, 291–307. Summerfeldt, L. J., Richter, M. A., Antony, M. M., & Swinson, R. P. (1999). Symptom structure in Obsessive-Compulsive Disorder: a confirmatory factor-analytic study. Behaviour Research and Therapy, 37, 297–311. Teasdale, J. D., & Barnard, P. J. (1993). Affect, cognition, and change: Re-modelling depressive thought. Hillsdale, NJ: Erlbaum. Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency and probability. Cognitive Psychology, 5, 207–232. van Oppen, P. V., Hoekstra, R. J., & Emmelkamp, P. M. G. (1995). The structure of obsessivecompulsive symptoms. Behaviour Research and Therapy, 33, 15–23. Wells, A. (1997). Cognitive therapy of anxiety disorders: A practice manual and conceptual guide. Chichester, UK: Wiley. Wheatley, J. (2000). Phenomenological enquiry into persistent OCD. Unpublished Doctoral Thesis, University of East London. Wu, K. D., & Watson, D. (2005). Hoarding and its relation to obsessive–compulsive symptoms. Behaviour Research and Therapy, 43, 897–921. Zigmond, A. S., & Snaith, R. P. (1983). The hospital anxiety and depression scale. Acta Psychiatrica Scandinavica, 67, 361–370. Received 4 September 2006; revised version received 3 December 2007