Developing a theoretical understanding of therapytechniques: An illustrative analogue study

Daniel Freeman*, Philippa A. Garety, Philip McGuireand Elizabeth KuipersInstitute of Psychiatry, King’s College London, UK

Objectives. In psychological interventions, clients are often asked to reviewunhelpful beliefs. Surprisingly, there is no theoretical understanding of how beliefs arereviewed in therapy. Moreover, by understanding a therapeutic technique, potentialinteractions with symptom processes can be considered. An analogue study assessingthe feasibility of researching therapy techniques is described, in which links betweensymptoms, reasoning style, and an experimental version of the cognitive therapytechnique of belief evaluation are examined.

Design. Individuals without psychiatric illness (N ¼ 30) completed (i) dimensionalmeasures of depression, anxiety, and delusions, (ii) a measure of confirmatoryreasoning (Wason’s 2–4–6 task) both before and after instruction in disconfirmatoryreasoning, and (iii) a belief evaluation task.

Results. Compared with individuals with a confirmatory reasoning style, individualswith a disconfirmatory reasoning style in Wason’s task were less hasty in their datagathering, considered a greater number of hypotheses during the task, had higherintellectual functioning, and had lower levels of depressive symptoms. Conversely, theindividuals with the strongest confirmatory reasoning had higher levels of depressionand preoccupation with delusional ideation. Successful adoption of disconfirmatoryreasoning was associated with less hasty decision-making and lower levels ofpreoccupation and distress by delusional ideation. Individuals with a disconfirmatoryreasoning style reported more evidence both for and against their beliefs in the beliefevaluation task.

Conclusion. The preliminary evidence, from this small non-clinical group, indicatesthat evaluating beliefs may partially involve the use of confirmatory and disconfirmatoryreasoning processes. Disconfirmatory reasoning, associated with less hasty datagathering and consideration of alternatives, may lead to better belief evaluation. In thecontext of clinical research indicating that individuals with delusions are hasty in theirdata gathering and have difficulty considering alternatives, a potential implication of thefindings is that individuals with delusions may find belief evaluation in therapy

* Correspondence should be addressed to Dr D. Freeman, Department of Psychology, PO Box 77, Institute of Psychiatry,Denmark Hill, London SE5 8AF, UK (e-mail: D.Freeman@iop.kcl.ac.uk).

TheBritishPsychologicalSociety

241

British Journal of Clinical Psychology (2005), 44, 241–254

q 2005 The British Psychological Society

www.bpsjournals.co.uk

DOI:10.1348/014466505X29981

particularly difficult. The current study has clear limitations, but a research focus onspecific techniques of therapy does appear feasible and show promise.

Individuals’ beliefs concerning their experiences are central to many theories ofclinical problems, and therefore the exploration of such beliefs is central to most

psychological therapies. For example, negative beliefs about the self are hypothesized

to underlie depression (Beck, 1964), catastrophic beliefs to lead to panic disorder

(Clark, 1986), and delusions are viewed as beliefs formed to explain anomalous

experiences (Maher, 1974). Clinical trials indicate that cognitive interventions that

focus upon understanding and evaluating such unhelpful beliefs have efficacy

(e.g. Clark et al., 1994; Kuipers et al., 1998; Rush et al., 1977). However the theoretical

basis of the techniques used to explore beliefs has not been empirically investigated: itis not known what psychological processes are utilized when clinicians explore beliefs

with individuals.

This might reflect a wider blind spot in the study of psychological interventions.

Cognitive therapists primarily draw upon findings from two main areas of empirical

inquiry: theory of clinical problems and therapy efficacy studies. It is assumed that

findings in one area validate related work in the other. For instance, if a therapy is

shown to have efficacy, then it is assumed that the theory from which it is derived is

correct, even though this conclusion need not follow. Studies of cognitive mediators oftherapy have attempted to link the two literatures (e.g. DeRubeis et al., 1990; Teasdale

et al., 2001). Although greatly under-researched, the study of mediating variables has

the potential to inform both the understanding of presenting symptoms and the

influence of therapy. If psychological processes that change in therapy are assessed,

this can provide theoretical underpinning for modifications in the intervention and

advance the understanding of the nature of clinical disorders (Kraemer, Wilson,

Fairburn, & Agras, 2002). We suggest that there is another area of research that could

link models of clinical symptoms and efficacy studies: the understanding of themethods used in therapy (‘technique theory’). There are two sequential questions.

What psychological processes are utilized by a therapy technique? How might the

actions of the technique interact with the processes associated with the clinical

phenomena? A small number of techniques directly map onto the theory of clinical

problems (e.g. exposure), but in many cases they are assumed to affect the processes

associated with clinical problems without empirical investigation of how this might be

achieved.

Understanding techniques could provide a firmer theoretical grounding forinterventions, analogous to studies of the mechanism of action of medications. Knowing

the method of action of a technique can allow close matching with the clinical problem. It

is probable that some techniques will prove more useful with one clinical group than

another, and it would be valuable to understand and predict such interactions (see Fig. 1).

A further complexity should be noted, but is not addressed in this paper, that of individual

factors (Garfield, 1986). That is, in therapy, the complete interaction is between the

psychological processes associated with individuals, symptoms, and specific techniques.

Figure 1. A simplified model of the interaction between therapy and symptoms at a psychological level.

Daniel Freeman et al.242

An analogue study

We report an analogue pilot study that illustrates how technique theory may be

developed. The study is a first step in understanding the standard cognitive therapytechnique of belief evaluation. The aim was to begin to understand belief evaluation in

normal groups and to test the methodology before application with clinical groups. We

suggest that belief evaluation may be partly understood by drawing upon the reasoning

literature.

Reasoning and belief evaluationThere is a large psychological literature indicating that there are systematic errors and

biases in individuals’ attempts to reason. When compared with predictions from

normative frameworks, rules of logic, or methods of scientific inquiry, individuals fallshort in their reasoning. For instance, there is evidence that individuals seek evidence

that is consistent with their beliefs rather than evidence that is inconsistent with their

beliefs (see reviews by Evans, 1989; Manktelow, 1999; Nickerson, 1998). This is

generally known as the ‘confirmation bias’. It has been most commonly investigated

using Wason’s 2–4–6 Task (e.g. Wason, 1960; Klayman & Ha, 1989; Mahoney &

DeMonbreun, 1977; Rossi et al., 2001). Participants are told that the experimenter has

a rule in mind that classifies sets of three whole numbers (triplets). Participants are

told that the triplet ‘2–4–6’ conforms to the rule. They are then asked to try to discoverthe rule by generating triplets of their own. For each triplet the experimenter tells the

participant whether or not it conforms to the rule. The participants are told to

announce the rule when they are highly confident that they know it. The example

‘2–4–6’ suggests the rule ‘numbers ascending by two’, whereas the actual rule is ‘any

ascending numbers.’ Usually participants test many triples that conform to their

hypothesis (e.g. ‘6–8–10’, ‘20–22–24’). What they do not normally do is test examples

that do not conform to their hypothesis (e.g. ‘2–3–4’; ‘8–6–4’). Therefore most

participants do not receive any evidence that disconfirms their initial rule. Mahoneyand DeMonbreun (1977) report that even doctoral scientists tend to use confirmatory

reasoning. Success in the task is associated with greater use of disconfirmatory

reasoning, trying a greater number of triplets before being certain of the rule, and

the consideration of alternative hypotheses (Wason, 1960; Gorman et al., 1987;

Tweney et al., 1980; Wharton et al., 1993).

The confirmation bias is an unscientific reasoning strategy because of the problem of

induction: a hypothesis cannot be proved by an unlimited number of confirmatory

observations. Falsification is needed to overcome the problem of induction since ahypothesis can be disproved by a single disconfirmatory observation (Popper, 1962).

Nonetheless, the confirmation bias is not irrational behaviour; it has the advantage of

reducing cognitive demands and in many situations can still bring to attention instances

of disconfirmation (Evans & Over, 1996).

The presence of a confirmation bias is problematic for individuals from the

perspective of the understanding of psychiatric disorders. As noted, cognitive theories

place a large emphasis on the role of negative beliefs in causing and maintaining clinical

disorders. Clearly, if individuals tend to only consider evidence consistent with theirnegative beliefs, and not consider evidence inconsistent with their negative beliefs, then

the negative beliefs and hence the disorder are likely to persist. Such ideas are present in

cognitive theories. Beck (1976) argues that contributing to the maintenance of

depression are cognitive distortions such as arbitrary inference and suggests that

Therapy technique theory 243

‘intrinsic to this thinking is the lack of consideration of alternative explanations that are

more plausible and more probable.’ Maher and Ross (1984) propose that delusional

hypotheses are confirmed by ‘selective observation’ and that ‘delusional certainty

resembles the normal attachment of a scientist to his or her own hypothesis.’

Studies of the confirmation bias have not found that individuals are incapable of

thinking about falsifying their beliefs, only that there is a strong tendency not to try.Interestingly, an important technique in cognitive therapy for clinical disorders is to

encourage patients to evaluate their beliefs, that is, to consider both confirmatory and

disconfirmatory evidence. Cognitive therapy may partly work by getting individuals to

fully apply ‘higher level’ analytic reasoning processes (Beck, 1967). Evans and Over

(1996) distinguish between personal rationality (rationality1), reasoning that is generally

reliable and efficient for achieving one’s goals, and impersonal rationality (rationality2),

when there is a reason for what one does sanctioned by a normative theory. They

hypothesize that rationality1 relies on implicit systems and rationality2 relies on explicitsystems. Often, cognitive therapy may promote rationality2 analytic reasoning to modify

particular conclusions derived from rationality1 processes.

The central objective of the study is to examine whether the cognitive therapy

technique of belief evaluation utilizes confirmatory and disconfirmatory reasoning

processes. The technique of belief evaluation may counter the confirmation bias.

The first prediction is that individuals who use disconfirmatory reasoning during

Wason’s 2–4–6 task will be superior at belief evaluation (can generate more evidence

against their beliefs), compared with individuals who only use confirmatory reasoningin Wason’s task.

Reasoning, belief evaluation and symptomsIt is plausible that the confirmation bias may contribute to the persistence of a range of

psychiatric symptoms. Conversely, individuals who naturally adopt a disconfirmatory

reasoning style may be less susceptible to such symptoms. Therefore, the secondprediction is that individuals who use disconfimatory reasoning may have lower levels of

symptoms such as depression, anxiety, and delusions.

Furthermore, an advantage of thinking about technique theory for clinical

interventions is that potential interactions between therapy and processes associated

with specific clinical problems can be considered. This can inform clinical practice.

Based upon theoretical models of clinical problems it may be possible to predict

whether some symptoms may be more or less responsive to particular techniques

compared with other symptoms. In the pilot study we highlight a link between delusionprocesses and reasoning. As has been noted, success in Wason’s reasoning task is

associated with participants testing a greater number of triplets before reaching 100%

certainty in the correctness of their rule and the consideration of alternative hypotheses.

In essence, disconfirmatory reasoning may be related to caution in forming strong

judgments, the use of additional data gathering, and the consideration of other

explanations. There are intriguing parallels here with the theoretical literature on

delusional beliefs. Empirical studies of probabilistic reasoning indicate that individuals

with delusions seek less data before making a decision (e.g. Garety & Hemsley, 1994;Dudley, John, Young, & Over, 1997; see review by Garety & Freeman, 1999). Individuals

with delusions are said to ‘jump to conclusions’. Furthermore, individuals with

delusions may have difficulty generating and considering alternative explanations

(Freeman et al., 2004). Therefore, the third prediction is that individuals with higher

Daniel Freeman et al.244

delusional ideation may have the strongest confirmatory bias and be less likely to use

disconfirmatory reasoning. This is consistent with the emphasis in definitions of

delusions that ‘they are not amenable to reason or modifiable by experience’

(Mullen, 1979).

Obviously, if individuals who use disconfirmatory reasoning are skilled in evaluating

beliefs, then clinicians will be interested in ways of improving the belief evaluation skills

of individuals who generally use confirmatory reasoning. Individuals can be successfully

instructed to adopt a disconfirmatory reasoning style in Wason’s 2–4–6 task

(e.g. Gorman & Gorman, 1984; Gorman et al., 1987; Tweney et al., 1980). However,

the previous studies have not examined which individuals are better at adopting a

disconfirmatory strategy. A further aim of the current study will be to examine this issue

by including a training phase and repeating Wason’s task. The final prediction is that

individuals with delusional ideation may be poorer at adopting a disconfirmatory

reasoning style (the confirmatory bias is stronger and less likely to change).In summary, there are four predictions:

(1) There will be an association between confirmatory and disconfirmatory reasoning

as measured by a formal reasoning task and an experimental version of the

cognitive therapy technique of belief evaluation. Disconfirmatory reasoning,

compared with confirmatory reasoning, will be associated with the generation of

more evidence against strongly-held beliefs.

(2) Disconfirmatory reasoning will be associated with lower levels of psychological

symptoms compared with confirmatory reasoning.

(3) Individuals higher in delusional ideation may have the strongest confirmatory

reasoning bias and least use of disconfirmatory reasoning.

(4) Individuals higher in delusional ideation may be less likely to adopt

disconfirmatory reasoning after training in this reasoning style.

Method

ParticipantsThirty individuals without a history of psychiatric illness participated. Twenty-three

participants were students at King’s College, London. They were recruited by a circular

e-mail. Twelve participants were undergraduates and 11 were postgraduates. Eight

participants studied medicine, eight studied sciences, and seven studied arts subjects.

Psychology students were not permitted to take part. The other seven participants had

taken part in an unrelated research study and had been recruited from the local

employment service. An equal number of males and females participated in the study.

Testing was completed in one appointment lasting 1.5 to 2 hours.

MeasuresIndividuals completed tasks assessing intellectual and executive functioning and

psychological symptom questionnaires. They then completed Wason’s 2–4–6 task,

received training in disconfirmatory reasoning for this task, and then completed a

parallel version of the reasoning task. Finally, individuals completed the belief evaluation

procedure.

Therapy technique theory 245

Intellectual functioningParticipants completed the vocabulary and block design sub-tests of the Wechsler adult

intelligence scale – revised (Wechsler, 1981). Full scale IQs for this short-form WAIS-R

were taken from Silverstein (1982). The Brixton Test (Burgess & Shallice, 1997) was also

administered. The Brixton Test is an executive function task examining rule detection

and rule following. To succeed in this task the ability to shift rule use is needed. Such anexecutive function may be involved in the use of disconfirmatory reasoning. Scaled

scores based upon the total number of task errors were used.

Symptom measuresDepression, anxiety, and delusional ideation were assessed with the Beck Depression

Inventory (BDI; Beck et al., 1979), the Beck Anxiety Inventory (BAI; Beck et al., 1988),

and the 21-item Peters et al. Delusions Inventory (PDI; Peters et al., 1996), respectively.The PDI provides a measure of the number of thoughts of a delusional content

endorsed, and the strength of delusional conviction, preoccupation, and distress.

Wason’s 2–4–6 taskAfter an explanation of the task and a brief practice with an alternative example,

participants received the 2–4–6 Task with the rule ‘any ascending numbers’. They were

told that they would be asked to generate 12 triplets. Each time they generated a tripletthey were asked, ‘What were you testing or trying to find out?’ and, ‘How sure are you

that your rule is the correct one? (0% to 100%)’. They were then told whether the triplet

fitted the rule and asked to suggest another triplet in order to discover the rule.

Participants were not told whether their rule was correct. By asking what the

participants’ were testing, it was possible to determine whether a triplet was an attempt

at confirmatory or disconfirmatory reasoning: a triplet was classified as disconfirmatory

if the participant had chosen it to be incorrect according to the rule that they believed

was most likely to be correct. The percentage estimate obtained for the rule allowed anexamination of how quickly the participants reached certainty for their rule. Therefore,

the variable of how many triplets a person required before being 100% certain of their

rule provided a measure of data gathering. When a person was not 100% sure after the

12 triplets, then they were given a score of 13. The number of different hypotheses

tested by the participants was also noted.

After receiving an explanation of disconfirmatory reasoning using a modified form of

the instructions used by Tweney et al. (1980) participants were asked to repeat Wason’s

task. During the repeat task, the participants were reminded twice to try and usedisconfirmatory reasoning. The further ‘embedded-hypothesis’ example of the Wason

task was taken from Klayman and Ha (1989). The repeat example triplet was 10–20–30

(rule: even numbers).

Belief evaluation taskAt the beginning of testing the participants completed a belief rating form. The form

comprised three positive beliefs about the self (e.g. ‘I am a success’), three oppositenegative beliefs about the self (e.g. ‘I am a failure’), three beliefs of a paranoid nature

(e.g. ‘People often want to get at me’), three political beliefs (e.g. Not enough is being

done to look after the environment’), and three beliefs representing the opposite

political beliefs (e.g. ‘Too much concern is given to the environment’). Participants were

Daniel Freeman et al.246

asked to rate how strongly they held each belief (0% to 100%) and how important the

belief is to them (0–10). Nine beliefs were selected for the belief evaluation task:

The three beliefs about the self held with the strongest conviction, the three paranoid

beliefs, and the three strongest political beliefs. Only 2 of the 90 beliefs about the self

that were chosen were negative (i.e. positive beliefs about the self were overwhelmingly

used), which is consistent with the sample being a non-clinical group.In the belief evaluation task the following procedure was used for each of the nine

beliefs chosen from the questionnaire. The person was asked to give a percentage

estimate of how much they believed the belief (0% to 100%). They were then asked to

spend 30 seconds thinking, in silence, of pieces of evidence that support the belief.

They were then asked how many pieces of evidence that they had produced. They were

then asked to spend 30 seconds thinking of pieces of evidence against the belief. Again

they were asked how many pieces of evidence they had produced. Participants were

then asked to re-rate how strongly they held the belief (0% to 100%). Participants weretold that the experimenter would sometimes ask for a brief summary of what they had

been thinking about after the 30 seconds; the purpose of asking participants to provide

the occasional summary was to check, in a manner that would not result in participants

modifying their thoughts, that the procedure was being correctly followed. In sum the

experimenter recorded how strongly the belief was held, how many pieces of evidence

were reported supportive of the belief, how many pieces of evidence were reported that

went against the belief, and how strongly the belief was held after evaluation.

AnalysisAll analyses were conducted using SPSS for Windows (version 10.0; SPSS, 2000).

Significance test results are quoted as two-tailed probabilities. Group differences were

examined by Mann-Whitney U tests because of differences in sample sizes; therefore themedians and inter-quartile ranges (IQR) for each variable are reported. Associations

were principally examined by Pearson’s product moment correlation coefficients.

Results

There were no missing data. The mean age of the participants was 24.8 (SD ¼ 5:7,

minimum ¼ 19, maximum ¼ 41). Twenty-five participants were White-British, three

participants were Asian, and two were Black-African. IQ scores were above average

(M ¼ 115:4, SD ¼ 10:8), and performance on the Brixton test was high average

(M ¼ 7:1, SD ¼ 1:7). Levels of depression (M ¼ 4:1, SD ¼ 3:8), anxiety (M ¼ 4:4,

SD ¼ 4:1), and delusional ideation (M ¼ 5:7, SD ¼ 3:1) were low.

Wason’s 2–4–6 taskSeven participants (23%) used disconfirmation at least once (‘disconfirmers’). Four

disconfirmers were female and three were male. Only four participants obtained the

correct rule, and all of these individuals were disconfirmers.

The disconfirmers were compared with the individuals who did not use anydisconfirmatory reasoning (‘confirmers’; see Table 1). The disconfirmers had

significantly higher IQ scores and lower depression scores.

On the 2–4–6 task the disconfirmers, compared with the confirmers, tried a

greater number of triplets before being 100% certain that their rule was correct

Therapy technique theory 247

(U ¼ 37:5, Z ¼ 2:15, p ¼ :032), and considered a greater number of hypotheses

(U ¼ 13:5, Z ¼ 4:05, p , :001; see Table 2). All the confirmers were 100% sure of

their rule before the seventh triplet, while three disconfirmers were not 100% sure,

even after 12 triplets.

Seven individuals were 100% sure of their rule at the first triplet. These individuals

can be considered as having the strongest confirmatory reasoning bias and as ‘jumping

to conclusions’. None of these individuals were disconfirmers. There was some

evidence on depression and delusion variables that these individuals experienced a

greater level of symptoms in comparison with individuals who were less hasty in the

task (see Table 3).

The belief evaluation taskThere was no significant difference in the degree of conviction with which the

confirmers and disconfirmers held the evaluated beliefs (U ¼ 60:0, Z ¼ 1:01, p ¼ :314).

Table 1. Comparison of the confirmers and disconfirmers on demographic and symptom data

Confirmers(N ¼ 23) Median

(IQR)

Disconfirmers(N ¼ 7) Median

(IQR) U Z p

Age 23.0 (5.0) 22.0 (7.0) 79.5 0.05 .961IQ 114.0 (13.0) 123.0 (12.0) 25.5 2.70 .007Verbal IQ scaled score 11.0 (4.0) 13.0 (3.0) 27.0 2.65 .008Performance IQ scaled score 13.0 (4.0) 14.0 (2.0) 48.5 1.59 .112Brixton test 7.0 (2.0) 7.0 (4.0) 74.0 0.33 .774BDI 4.0 (5.0) 1.0 (3.0) 39.0 2.05 .040BAI 3.0 (4.0) 4.0 (3.0) 64.5 0.79 .429PDI – number endorsed 6.0 (3.0) 4.0 (3.0) 48.5 1.58 .113PDI – conviction 17.0 (13.0) 12.0 (8.0) 64.0 0.81 .418PDI – Preoccupation 11.0 (12.0) 7.0 (3.0) 49.0 1.55 .121PDI – distress 9.0 (11.0) 5.0 (14.0) 62.0 0.91 .363

Table 2. Data on the two completions of Wason’s task: 2–4–6 (no instruction) and 10–20–30

(after disconfirmation instruction)

Task

Total sample(N ¼ 30) Median

(IQR)

Confirmatory(N ¼ 23) Median

(IQR)

Disconfirmatory(N ¼ 7) Median

(IQR)

No. confirmatory triplets 2–4–6 12.0 (0.3) 12.0 (0.0) 9.0 (3.0)10–20–30 7.5 (3.5) 8.0 (4.0) 6.0 (5.0)

No. of disconfirmatory triplets 2–4–6 0.0 (0.3) 0.0 (0.0) 3.0 (3.0)10–20–30 4.5 (3.5) 4.0 (4.0) 6.0 (5.0)

No. of triplets until decision 2–4–6 2.5 (3.3) 2.0 (3.0) 5.0 (11.0)10–20–30 12.5 (7.5) 8.0 (9.0) 13.0 (0.0)

No. of hypotheses tested 2–4–6 1.0 (1.3) 1.0 (0.0) 4.0 (3.0)10–20–30 4.0 (3.3) 4.0 (4.0) 4.0 (2.0)

Daniel Freeman et al.248

However, compared with the confirmers, the disconfirmers produced more evidence

supportive of their beliefs (U ¼ 34:5, Z ¼ 2:26, p ¼ :024) and more evidence against

their beliefs (U ¼ 34:0, Z ¼ 2:29, p ¼ :022; see Table 4). Since this concerned the main

study hypothesis, a regression analysis was used to examine whether the relationship

between disconfirmatory reasoning and performance on the belief evaluation task was

explained by IQ or depression. IQ, depression, and whether the person was a confirmer

or disconfirmer were entered into a regression analysis with the number of items of

evidence against the beliefs as the dependent variable. The model was significant,

Fð3; 26Þ ¼ 4:7, p ¼ :009, adjusted R2 ¼ :28. Only the variable of whether the

participant was a confirmer or disconfirmer remained significant in the regression

analysis, t ¼ 3:0, p ¼ :006. The dependent variable in the regression analysis was not

normally distributed, but a repeat analysis with a transformed and normally distributed

dependent variable did not alter the pattern of results.

In the total sample the number of items of evidence against the beliefs was also

significantly correlated with the number of triplets needed before the person was 100%

certain (r ¼ :40, p ¼ :028) and the number of hypotheses tested (r ¼ :40, p ¼ :028) in

Wason’s 2–4–6 task. The number of items of evidence supportive of the belief was

significantly correlated with the number of triplets needed before the person was 100%

certain in Wason’s 2–4–6 task (r ¼ :40, p ¼ :027), but not with the number of

hypotheses tested (r ¼ :29, p ¼ :118).

Table 3. Symptom scores for individuals who were 100% certain after one triplet in Wason’s task

100% certain after one triplet(N ¼ 7) Median

(IQR)

Other participants(N ¼ 23) Median

(IQR) U Z p

Depression 6.0 (9.0) 2.0 (5.0) 31.5 2.42 .015Anxiety 4.0 (14.0) 3.0 (4.0) 57.5 1.14 .256PDI – Number endorsed 6.0 (8.0) 5.0 (4.0) 52.5 1.39 .166PDI – Conviction 17.0 (26.0) 16.0 (13.0) 61.5 0.93 .351PDI – Preoccupation 17.0 (18.0) 7.0 (7.0) 40.0 1.99 .046PDI – Distress 13.0 (21.0) 8.0 (11.0) 44.5 1.77 .077

Table 4. Number of items of evidence elicited in the belief evaluation task

Confirmers(N ¼ 23) Median

(IQR)

Disconfirmers(N ¼ 7) Median

(IQR)

Total items supportive 25.0 (10.0) 36.0 (18.0)Supportive – beliefs about self 12.0 (7.0) 12.0 (7.0)Supportive – paranoia beliefs 4.0 (3.0) 7.0 (6.0)Supportive – political beliefs 9.0 (7.0) 15.0 (7.0)

Total items against 13.0 (7.0) 25.0 (21.0)Against – beliefs about self 5.0 (3.0) 9.0 (10.0)Against – paranoia beliefs 3.0 (1.0) 7.0 (5.0)Against – political beliefs 4.0 (2.0) 9.0 (11.0)

Therapy technique theory 249

The majority of participants, whether they were confirmers or disconfirmers, did not

change their level of belief conviction during the task, preventing any statistical analysis

(across the nine beliefs, 71% of the group on average did not change their degree of

belief conviction, while 16% reduced and 13% increased their belief conviction;

a summary of these data can be obtained from the authors on request).

Instruction in disconfirmationInstruction increased the number of disconfirmatory triplets that were tried in Wason’s

task (see Table 2). In order to examine the predictors of the learning of disconfirmatory

reasoning, the number of disconfirming attempts after instruction by the original

confirmers (N ¼ 23) was correlated with age, IQ, the Brixton test, the symptom

measures, and, the number of triplets required before being 100% certain on the firstadministration of the task. Three correlations were significant: PDI-distress (r ¼ 2:45,

p ¼ :030), the number of triplets before being 100% certain on the first administration of

the task (r ¼ :42, p ¼ :045), and PDI-preoccupation (r ¼ 2:42, p ¼ :047). The less

distress or preoccupied the person was by delusional ideation, and the greater the

number of triplets required to be 100% sure of the rule in the first administration of the

task, then the greater the use was made of disconfirmatory reasoning. There were also

two correlations with the number of disconfimatory triplets that tended towards

significance: verbal IQ (r ¼ :37, p ¼ :082) and anxiety (r ¼ 2:36, p ¼ :097).After instruction, nine people obtained the correct rule (four original disconfirmers,

five original confirmers). Compared with the first administration of the task, a greater

number of triplets were needed before participants were 100% sure of the rule: half of

the participants were not 100% sure of the rule after all 12 triplets.

Discussion

In this paper we have suggested that it may be beneficial to develop the theoretical

understanding of therapy techniques. Understanding the psychological processes

manipulated by therapy strategies could lead to improvements in the techniques used

and thereby enhance the efficacy of interventions. Insights into the interaction between

therapy and symptoms can also be obtained. A novel exploratory study was reported, in

which reasoning styles, the therapy technique of belief evaluation, and symptoms of

psychological disorders were linked. The group studied did not have a history of mental

illness and therefore a key caveat is that the applicability of both the findings and themethodology to clinical groups remains to be determined.

There was support for each of the study predictions, though slight in some instances.

The main prediction was that reasoning style would be related to performance on the

belief evaluation task, which was a simplified version of the cognitive therapy

technique. It was found that natural disconfirmers in Wason’s task produced more

evidence for and more evidence against their beliefs in comparison with confirmers.

Moreover, individuals in Wason’s task who tended to be less hasty in their decision

making and tried several hypotheses produced more evidence in the belief evaluationtask. The results are consistent with the idea that the technique of belief evaluation can

partly be understood with reference to the reasoning literature: belief evaluation may

involve skills of disconfirmatory reasoning, data gathering, and the availability of

alternative hypotheses.

Daniel Freeman et al.250

Consistent with previous studies using Wason’s 2–4–6 task, it was found that most

people did not spontaneously use disconfirmatory reasoning. Just under a quarter of the

present sample used disconfirmatory reasoning. These individuals were of higher

intelligence. Moreover, there was some evidence to support the second prediction that

disconfirmatory reasoning would be associated with lower levels of psychological

symptoms: the disconfirmers had lower levels of depression in comparison with theother participants. Being able to use disconfirmatory reasoning may protect against the

symptoms of depression, although clearly this finding, like the others, should be treated

with caution until replication in a larger and more symptomatic sample.

It was also predicted that delusional ideation would be particularly associated with

use of confirmatory reasoning and less use of disconfirmatory reasoning. Individuals

who had what may perhaps be considered the strongest confirmatory reasoning styles

(being 100% certain of the rule after one triplet) had significantly higher levels of

preoccupation with delusional ideation. It is encouraging that further replication was

made of the finding that people can learn to use disconfirmatory reasoning in Wason’stask. However, higher levels of preoccupation and distress with delusional ideas, and

hasty decision making, were associated with poorer learning of disconfirmatory

reasoning. Therefore, a possible implication of the results is that individuals with

delusions may have a strong confirmatory bias and be poor at adopting disconfirmatory

reasoning, which would also follow from clinical research indicating that individuals

with delusions are hasty in data-gathering and have difficulties considering alternatives

(Freeman et al., 2004; Garety & Freeman, 1999). Potentially, though clearly more

research is needed, this is an instance of an interaction between the psychological

processes associated with a therapy technique and the psychological processesassociated with a symptom.

We have noted the difficulty of generalizing from analogue studies to clinical

populations. There are two further obvious weaknesses. The sample size is small,

particularly for the comparisons between confirmers and disconfirmers. There is also

multiple hypothesis testing, which is liable to lead to Type I errors (detecting an effect

when none exists). This is an inevitable consequence of attempting to link multiple

areas of research and the exploratory stage of the research. The results are broadly in

line with the initial predictions, and are intriguing, providing evidence of promise in this

topic of investigation. We think the results provide pointers for future studies in clinicaland non-clinical groups.

Future studies with symptomatic non-clinical individuals and clinical groups may

find stronger associations between reasoning styles and symptoms. In clinical groups it

would be possible to examine negative beliefs. Factors identified for further

investigation of belief evaluation include: diagnosis, reasoning style, data gathering,

and the provision of alternatives. It would also be helpful to develop a version of the task

that is better able to assess the strength of the belief confirmation bias since there was a

ceiling effect in the current study. The degree of strength of the belief confirmation bias

could then be examined across symptoms. Further, Evans and Over’s (1996) suggest thatconfirmatory and disconfirmatory reasoning strategies may involve different

psychological processes, and consequently, it would be intriguing to investigate

whether separate brain regions can be identified by functional neuroimaging

techniques.

Yet clearly, there are likely to be other factors than analytic reasoning involved in the

belief evaluation task. It is of note that there was no relation between reasoning style

and belief change, although this may partly be a result of the small participant numbers

Therapy technique theory 251

and a rather shortened belief evaluation methodology. How individuals use

disconfirmatory evidence would be a valuable area of future research. There were

some notable reactions to disconfirmatory information in the second presentation of

Wason’s task. Some participants were startled to receive disconfirmatory information

and did not fully hear it, asking the experimenter to repeat the feedback – this did not

happen with confirmatory evidence. Some participants would not remember thedisconfirmatory information and soon reverted back to the rule that they had found to

be incorrect. These types of observations are common in clinical work with patients.

Also, a good therapeutic relationship is often considered a sine qua non for belief

evaluation – the influence of trust and anxiety on reasoning would be an interesting area

of further investigation and link with therapy process research (Orlinsky & Howard,

1986; Roth & Fonagy, 1996).

The overall aim in this area of research is to take the evidence-based and

theoretically-driven approach of cognitive therapy a step further. The results of thecurrent study underline the potential importance of reasoning processes in therapy.

Beliefs are inherently a judgment and therefore it is unsurprising that reasoning

processes should be implicated in their persistence; nonetheless, there is often a

tendency in therapy to concentrate upon content rather than process. With regard to

interventions in psychosis, the study provides further indication that psychological

therapies may have greater efficacy when there is a focus on data gathering, developing

plausible and compelling non-delusional explanations for experience (‘making sense of

psychosis’), and reviewing the evidence for the new explanations than if delusionalbeliefs are simply evaluated or directly challenged (Fowler et al., 1995; Freeman &

Garety, 2004). Although, where possible, reasoning processes should be a topic of

discussion with clients, by concentrating on making sense of psychosis, developing an

alternative psychological account of how a delusion developed, confirmatory reasoning

can perhaps support, rather than hinder, progress in therapy. These sorts of differences

in therapeutic approach are likely to become an important issue in the evaluation of

therapy. The work we have described is clearly at an early stage of development.

Nevertheless, in our opinion it has a potential, when suitably developed, to informpsychological theory and intervention.

Acknowledgements

This work was supported by a programme grant from the Wellcome Trust (No. 062452).

Ruth Williams and Ed Watkins provided helpful comments on an earlier draft of the paper.

References

Beck, A. T. (1964). Thinking and depression 2: Theory and therapy. Archives of General

Psychiatry, 9, 324–333.

Beck, A. T. (1967). Depression: Clinical, experimental, and theoretical aspects. New York: Harper

and Row.

Beck, A. T. (1976). Cognitive therapy and the emotional disorders. New York: International

University Press.

Beck, A. T., Epstein, N., Brown, G., & Steer, R. (1988). An inventory for measuring clinical anxiety:

Psychometric properties. Journal of Consulting and Clinical Psychology, 56, 893–897.

Beck, A. T., Rush, A. J., Shaw, B. F. & Emery, G. (1979). Cognitive therapy of depression. New York:

The Guildford press.

Daniel Freeman et al.252

Burgess, P. W., & Shallice, T. (1997). The hayling and brixton tests. Suffolk, UK: Thames valley test

company.

Clark, D. M. (1986). A cognitive model of panic. Behaviour Research and Therapy, 24, 461–470.

Clark, D. M., Salkovskis, P. M., Hackmann, A., Middleton, H., Anastasiades, P., & Gelder, M. G.

(1994). A comparison of cognitive therapy, applied relaxation and imipramine in the treatment

of panic disorder. British Journal of Psychiatry, 164, 759–769.

DeRubeis, R. J., Evans, M. D., Hollon, S. D., Garvey, M. J., Grove, W. M., & Tuason, V. B. (1990). How

does cognitive therapy work? Cognitive change and symptom change in cognitive therapy

and pharmacotherapy for depression. Journal of Consulting and Clinical Psychology, 58,

862–869.

Dudley, R. E. J., John, C. H., Young, A. W., & Over, D. E. (1997). Normal and abnormal reasoning in

people with delusions. British Journal of Clinical Psychology, 36, 243–258.

Evans, J. St. B. T. (1989). Bias in human reasoning: Causes and consequences. Hove: LEA.

Evans, J. St. B. T., & Over, D. E. (1996). Rationality and reasoning. Hove: Psychology Press.

Fowler, D., Garety, P. A., & Kuipers, L. (1995). Cognitive behaviour therapy for psychosis: Theory

and practice. Chichester: Wiley.

Freeman, D., & Garety, P. A. (2004). Paranoia: The psychology of persecutory delusions. Hove:

Psychology Press.

Freeman, D., Garety, P. A., Fowler, D., Kuipers, E., Bebbington, P., & Dunn, G. (2004). Why do

people with delusions fail to choose more realistic explanations for their experiences?

An empirical investigation. Journal of Consulting and Clinical Psychology, 72, 671–680.

Garety, P. A., & Freeman, D. (1999). Cognitive approaches to delusions: A critical review of

theories and evidence. British Journal of Clinical Psychology, 38, 113–154.

Garety, P. A., & Hemsley, D. R. (1994). Delusions: Investigations into the psychology of delusional

reasoning. Oxford: Oxford University Press.

Garfield, S. L. (1986). Research on client variables in psychotherapy. In A. E. Bergin & S. L. Garfield

(Eds.), Handbook of psychotherapy and behavior change (4th ed.). New York: Wiley.

Gorman, M. E., & Gorman, M. E. (1984). A comparison of disconfirmatory, confirmatory and

control strategies on Wason’s 2-4-6 task. Quarterly Journal of Experimental Psychology, 36A,

629–648.

Gorman, M. E., Stafford, A., & Gorman, M. E. (1987). Disconfirmation and dual hypotheses on a

more difficult version of Wason’s 2-4-6 task. Quarterly Journal of Experimental Psychology,

39A, 1–28.

Klayman, J., & Ha, Y-W. (1989). Hypothesis testing in rule discovery: Strategy, structure, and

content. Journal of Experimental Psychology: Learning, Memory and Cognition, 15,

596–604.

Kraemer, H. C., Wilson, G. T., Fairburn, C. G., & Agras, W. S. (2001). Mediators and moderators of

treatment effects in randomised clinical trials. Archives of General Psychiatry, 59, 877–883.

Kuipers, E., Fowler, D., Garety, P. A., Chisholm, D., Freeman, D., Dunn, G., Bebbington, P. E., &

Hadley, C. (1998). The London-East Anglia randomised controlled trial of cognitive behaviour

therapy for Psychosis III: Follow-up and economic evaluation at 18 months. British Journal of

Psychiatry, 173, 61–68.

Maher, B. A. (1974). Delusional thinking and perceptual disorder. Journal of Individual

Psychology, 30, 98–113.

Maher, B. A., & Ross, J. S. (1984). Delusions. In H. E. Adams & P. B. Sutker (Eds.), Comprehensive

handbook of psychopathology (pp. 383–409). New York: Plenum Press.

Mahoney, M. J., & DeMonbreun, B. G. (1977). Psychology of the scientist: An analysis of problem-

solving bias. Cognitive Therapy and Research, 1, 229–238.

Manktelow, K. (1999). Reasoning and thinking. Hove: Psychology Press.

Mullen, P. (1979). Phenomenology of disordered mental function. In P. Hill, R. Murrary &

G. Thorley (Eds.), Essentials of post-graduate psychiatry (pp. 25–54). London: Academic.

Therapy technique theory 253

Orlinsky, D. E., & Howard, K. I. (1986). Process and outcome in psychotherapy. In A. E. Bergin &

S. L. Garfield (Eds.), Handbook of psychotherapy and behavior change (3rd ed.). New York:

Wiley.

Nickerson, R. S. (1998). Confirmation bias: A ubiquitous phenomenon in many guises. Review of

General Psychology, 2, 175–220.

Peters, E. R., Day, S., & Garety, P. A. (1996). The Peters et al. delusions inventory (PDI): New norms

for the 21-item version. Schizophrenia Research, 18, 118.

Popper, K. R. (1962). Conjectures and refutations. London: Hutchinson.

Rossi, S., Caverni, J. P., & Girotto, V. (2001). Hypothesis testing in a rule discovery problem: When

a focused procedure is effective. Quarterly Journal of Experimental Psychology, 54A,

263–267.

Roth, R., & Fonagy, P. (1996). What works for whom?: A critical review of psychotherapy

research. New York: The Guilford Press.

Rush, A. J., Beck, A. T., Kovacs, M., & Hollon, J. D. (1977). Comparative efficacy of cognitive

therapy and imipramine in the treatment of depressed patients. Cognitive Therapy and

Research, 1, 17–37.

Silverstein, A. B. (1982). Two- and four-subtest short forms of the Wechsler adult intelligence

scale – revised. Journal of Consulting and Clinical Psychology, 50, 415–418.

SPSS (2000). SPSS Base 10.0 user’s guide. Chicago, IL: SPSS.

Teasdale, J. D., Scott, J., Moore, R. G., Hayhurst, H., & Paykel, E. S. (2001). How does cognitive

therapy prevent relapse in residual depression? Evidence from a controlled trial. Journal of

Consulting and Clinical Psychology, 69, 347–357.

Tweney, R. D., Doherty, M. E., Worner, W. J., Pliske, D. B., & Mynatt, C. R. (1980). Strategies of rule

discovery in an inference task. Quarterly Journal Of Experimental Psychology, 32, 109–123.

Wason, P. C. (1960). On the failure to eliminate hypotheses in a conceptual task. Quarterly

Journal of Experimental Psychology, 12, 129–140.

Wechsler, D. (1981). WAIS-R Manual: Weschsler Adult Intelligence Scale-Revised. New York:

Psychological Corporation.

Wharton, C. M., Cheng, P. W., & Wickens, T. D. (1993). Hypothesis-testing strategies: Why two

goals are better than one. Quarterly Journal of Experimental Psychology, 46A, 743–758.

Received 22 May 2003; Revised version received 12 February 2004

Daniel Freeman et al.254