An Introduction to Cognitive Behaviour Therapy, 2 edition


Behavioural Experiments

At the 2004 conference of the European Association for Behavioural and Cognitive Therapies, there was a symposium entitled ‘Where is the B in CBT?’. This chapter will outline an answer to that question, i.e. the place of behavioural methods in current CBT. We shall focus on one specific area where behaviour change is crucial: behavioural experiments (BEs), a CBT strategy that can be used to great effect in most if not all problems. Another common behavioural technique in CBT, activity scheduling, is described in the chapter on depression (Chapter 12), because that is where it is most widely used.

What are BEs?

The following discussion of BEs draws heavily on the recent volume devoted to the use of BEs in CBT, to which all three of the present authors contributed (Bennett-Levy, Butler, Fennell, Hackmann, Mueller & Westbrook, 2004). We shall adopt Bennett-Levy et al.’s operational definition of BEs:


Behavioural experiments are planned experiential activities, based on experimentation or observation, which are undertaken by patients in or between cognitive therapy sessions. Their design is derived directly from a cognitive formulation of the problem, and their primary purpose is to obtain new information which may help to: test the validity of the patients’ existing beliefs about themselves, others, and the world; construct and/or test new, more adaptive, beliefs; contribute to the development and verification of the cognitive formulation.

(Bennett-Levy et al., 2004, p. 8)

This means that BEs are designed, like experiments in science, to generate evidence that will help us decide what hypothesis is best supported. But instead of testing scientific theories, BEs in CBT are designed to gather evidence that will help patients test the predictions that follow from their unhelpful cognitions or to test elements in a formulation. Chapters 7 and 8 have already addressed verbal methods for exploring cognitions and expanding the range of evidence that the client considers. BEs offer us a way of taking this a step further, by exploring beliefs through action and observation rather than just through verbal discussion, and by helping the client generate new evidence. BEs are therefore often used to follow up verbal discussion. Having explored a particular negative cognition and generated possible alternative views during a session, BEs may offer a useful way of testing out and consolidating these conclusions. They can help the client gather more cogent evidence as to whether the original negative cognition or the new alternative offers the best (most accurate or most helpful) view of a situation.

A client with social anxiety had the belief that he looked ‘peculiar’ (and that others would therefore disapprove of him). One piece of evidence for this belief was that he noticed when he went into the canteen at work that other people ‘stared at’ him. His response was to look down so as to avoid their gaze, to sit and eat alone and to focus closely on his plate. During CBT sessions, an alternative account was developed, namely that perhaps people tended to look at anyone who entered the canteen, because they were curious, rather than this behaviour’s being exclusive to him and due to his ‘peculiarity’; and, furthermore, that maybe his subsequent avoidance of looking at other people meant that he had no opportunity to observe whether this was true. This discussion led on to a BE designed to gather evidence about which account was most convincing. It was agreed that he would enter the canteen as usual but this time try to keep looking up and count roughly how many people looked at him. Then after he sat down, he would make an effort to continue looking around and to count how many people looked at anyone else who entered the canteen. He was able to do this and, somewhat to his surprise, found the new alternative belief amply supported. Some people in the canteen seemed to look up at anyone who entered, and there was no evidence that he attracted more curiosity than anyone else. He found this helpful in beginning to question the belief that he was ‘peculiar’.

Another client with social anxiety was worried about the consequences of blushing during social interactions. She believed that if she blushed, other people would be bound to make negative judgements about her, for example that she was silly, or abnormal. Although she had occasionally been teased about her blushing, no one had ever actually expressed a negative evaluation because of it, but she tended to dismiss this on the grounds that they were just being kind to her. This client found it helpful to do a survey experiment. A question about reactions to people blushing was carefully constructed so that both she and the therapist agreed that it was reasonably unbiased (i.e. not obviously expecting either negative or positive responses – for example, not starting off with ‘Would you think badly of such a person?’, but a more neutral ‘Would blushing have any impact on your opinion of such a person?’). Then the therapist distributed the question sheet to a number of work colleagues and friends to collect their responses – for this client, it was important that the people surveyed did not know her and were therefore less likely to be ‘kind’ in their answers. She found that most people thought blushing was quite charming and that the worst anyone thought was that someone blushing might be anxious and that they would therefore tend to feel sympathetic.

Behavioural experiments compared to behaviour therapy

BEs are derived from CBT’s behavioural legacy, and some BEs may look like traditional behavioural methods such as exposure in vivo to anxiety-provoking situations. However, it is important to remember that the aims of, and the conceptual framework surrounding, BEs are quite different from traditional behavioural therapy. In the latter, the most common conceptual model is of exposure leading tohabituation. To put it in very crude terms (with apologies to learning theorists, whose ideas are actually far more complex than this) the idea is that exposure to anxiety-provoking stimuli leads to the anxiety response gradually dying away as the person gets used to the situation. An analogy sometimes used is that if I suddenly make a loud noise, you will probably be startled, but if I repeatedly make a loud noise every 10 seconds for the next 10 minutes, then you will probably gradually stop being startled and react less.

In contrast to this model, BEs in CBT are quintessentially a cognitive strategy, explicitly aimed at generating information and/or testing out beliefs, not at promoting habituation of anxiety responses. If we consider the treatment of someone with agoraphobia and panic who fears supermarkets, both traditional behaviour therapy and CBT might suggest that it would be useful for the client to visit a supermarket, but the goals and thinking behind the strategy (and hence the exact procedure to be followed) would be quite different:


  • Behavioural exposurewould aim for the person to learn a new response to supermarkets, which would involve staying in the situation long enough (and repeating exposure often enough) for the anxiety response to die away. No particular attention would be paid to thoughts or beliefs; all that would be thought necessary is for a client to overcome his avoidance for long enough for habituation to occur. In order to assist this, exposure would usually be graduated, i.e. working up a hierarchy of increasing levels of anxiety, trying to make sure that he was not too anxious at any point (although there is also a form of exposure known as flooding, in which clients are exposed to the most fear-provoking situation from the beginning).
  • If using a behavioural experimentwithin CBT, the visit to a supermarket would follow a cognitive understanding of the person’s negative predictions about what might happen. The primary goal of the visit would be to help him test out these negative beliefs by seeing whether what he fears actually happens: does he actually collapse/die/pass out, or whatever it may be? His level of anxiety, although of course an important clinical concern, would not be the primary focus during the experiment – or at least not unless being anxious played a part in his negative beliefs (e.g. ‘If I become very anxious I will lose control and go mad’). In the latter case, it might actually be important to an adequate BE that he did become very anxious, in order to test out this belief. Therefore, although it still might be clinically necessary to tackle the situation in a graded way, neither this nor repeated exposure is essential to a BE: the heart of the matter is simply to test out thoughts and beliefs as thoroughly and convincingly as possible, and this can sometimes happen with just one experiment.

Part of the appeal of BEs in CBT is that they offer a possible way of getting around a common problem in interventions that depend primarily on verbal methods, namely responses like, ‘Well I can see intellectually that this is a more logical way of looking at it, but it still feels like my negative thought is true’. By testing out thoughts and beliefs in action, rather than just through words, BEs can help to develop a more ‘gut feel’ kind of learning. They are also useful in almost every kind of psychological problem, in contrast to exposure, which is focused on anxiety problems.

Efficacy of BEs

The evidence on whether BEs are any more or less effective than exposure is limited at present. In a recent systematic review – the first on this topic – McMillan and Lee (2010) included 14 relevant studies, covering panic, social anxiety, OCD and specific phobias. Although they are appropriately cautious in their conclusions because of various methodological problems, their summary is that ‘there was some evidence that behavioral experiments were more effective than exposure alone …’ (McMillan & Lee, 2010, last line of Abstract). More and better evidence is needed, but this is the best we have at the moment.

Figure 9.1    Types of behavioural experiment

Types of behavioural experiment

We can usefully distinguish two dimensions along which BEs may vary: hypothesis-testing versus discovery BEs; and active versus observational BEs (Bennett-Levy et al., 2004). Putting these together, we have a diagram of possible BEs (Figure 9.1).

Hypothesis-testing versus discovery

Hypothesis-testing BEs are perhaps the closest to the classical scientific experiment. In such experiments we either start from one hypothesis, or from both of two relatively clear hypotheses – often known as Theory A and Theory B. Theory A is the client’s initial belief or explanation, for example ‘People look at me because I look peculiar’. Theory B is the new, alternative belief, often based on the CBT formulation or perhaps worked out during a CBT session between client and therapist, e.g. ‘People look at anyone coming into a room, out of curiosity – there is nothing special about me’. When we can state at least one of these hypotheses reasonably clearly, then we have the necessary conditions for a BE in which the aim is to find some clear evidence bearing on the hypothesis. We may either test Theory A or Theory B alone (the question to be tested is then ‘Does this theory correctly predict what happens in this situation?’); or we may compare the two theories to see which one works best in predicting the observed outcomes – as in the canteen experiment above. The aim is to find some predicted consequence of the hypothesis that is in principle observable so that the client can tell whether his prediction comes true.

Hypothesis-testing experiments are the most common, and often the most useful, but clients sometimes have no clear hypothesis to test, perhaps because they have not yet worked out a clear statement of their negative cognitions or because they can’t yet even conceive of an alternative. In such cases it may be useful to do discovery BEs, aiming to explore in a more open-minded way ‘What would happen if I did X?’. For example, ‘What would happen if I were to talk a bit more openly about myself to other people? How would I feel? How would they react? Perhaps I can find out …’.

Active versus observational

The second distinction is between:


  • BEs in which the client is an active participant, in the sense of going out and actively doing something to generate information – often something different to his usual behaviour.
  • BEs in which the client is observing events or gathering already available evidence rather than actively doing something different.

The canteen experiment is an example of an active experiment; the blushing survey is an example of an observational experiment.

Observational experiments can include therapist modelling, where the client observes the therapist doing something, so that he can see what happens without too much ‘risk’ to himself:

For example, a client who fears collapsing in a supermarket might find it useful to observe what happens. After identifying what the client’s negative predictions are, therapist and client can go to a supermarket together, the therapist can pretend to collapse, and the client can observe what actually happens.

Many other kinds of information-gathering are possible:

A client with social anxiety was worried about not having anything important or clever to say. He found it useful to observe how other people carry on conversations, which led to his realising that most ordinary conversations are pretty mundane, not necessarily containing profound topics or deep thoughts.

It may also be useful for the client to gather information from books or the Internet.

A client with claustrophobia found some detailed information on the Internet about the risks of suffocation in confined spaces, including a calculation of how long one might survive in an airtight room!

Most classic behavioural experiments fall into the top-left quadrant of Figure 9.1 (Theory A), but there are useful examples in the other quadrants as well. See Bennett-Levy et al. (2004) for a comprehensive collection. Using one or more of these approaches, the aim is to work out something the client can do, in or between sessions, that will help him generate or gather more evidence relevant to his negative cognitions.

Planning and implementing behavioural experiments


Careful planning is a crucial preliminary to most successful BEs. There are several essential components:


  • Ensure that both you and the client clearly understand the purpose and rationale of the experiment, and always plan experiments collaboratively. BEs should not be unilaterally assigned in the last two minutes of a session, nor should they be done just because a protocol says you should be doing them! They should grow out of the session as a logical way to move things forward. Remember, it is desirable to involve the client in thinking about BEs and homework: ‘Given what we’ve been discussing in this session, what do you think might be useful to take this further between now and the next session?’
  • Particularly for hypothesis-testing experiments, spend time getting clear about what cognition(s) are to be tested and the client’s negative predictions about what might happen. This step is crucial, because a BE aimed at a poorly defined cognition will rarely be effective. For example, your client may fear approaching a particular situation without safety behaviours, and his initial prediction may well be something vague, like ‘It will be awful’. You are unlikely to be able to test this prediction, because it is not precisely defined. How can you or the client tell whether it is ‘awful’ or not? What exactly constitutes ‘awful’? Furthermore, this kind of BE may indeed turn out to be ‘awful’, at least in the sense that the client feels anxious. It will usually be much better to work out a clear prediction such as, ‘I will collapse’ or ‘People will laugh at me’: something that (a) distinguishes between the belief to be tested and possible alternatives; and (b) can be developed into reasonably clear criteria that will enable both client and therapist to determine unambiguously whether it has happened.


  • Having identified a clear cognition, have the client rate how strongly he believes it on a scale from 0 to 100% (0% Not at all, to 100% Absolutely certain it’s true) to provide a baseline against which any change can be measured.
  • Choose the best type of experiment to test the cognition, e.g. an active experiment or an observational experiment. Partly this decision will depend on what point the client has reached in shifting his thinking and how threatening the experiment appears to him. Observational BEs are often less threatening and therefore may be a useful first step before moving on to active BEs.

A survey experiment for a man with body dysmorphic disorder involved getting a number of people who did not know him to look at photographs of him and several other people to see whether he was singled out – as he feared he would be – for the ‘ugliness’ of his nose. As is common in such experiments, the wording of the questions to respondents needed careful consideration so that the questions were meaningful without leading respondents to any particular response. In this case he agreed that he did not want respondents to be focusing especially on him or initially on noses, so early questions were framed along the lines of ‘Do any of these people’s faces seem unusual in any way? If so, in what way?’; only later did the survey ask respondents to rate noses specifically.


  • In planning BEs you need to pay reasonable attention to safety and risk. This should be done in a balanced way, bearing in mind that CBT is about realistic thinking, not generalised positive or negative thinking; and that the risks of doing a particular BE need to be weighed against the risks of notdoing it (in terms of potential reduced treatment efficacy and therefore an increased risk of continuing to have significant anxiety problems). It is important not simply to feed your client’s anxiety by suggesting excessive caution, but it is equally important not to ignore significant risk. So, for example, you need to take care with experiments involving strenuous physical exercise if there is any reason to suppose your client has relevant physical problems (e.g. pulmonary or cardiac problems – if so, ask the GP or other doctor for a physical assessment); or if an experiment involves walking down the street to test a fear of being attacked then you obviously need to be reasonably confident that it is in fact safe (e.g. it is a street where most people would not hesitate to walk). Keep a sense of proportion. We need to consider significant risks but we also need to remember the message we often give to patients: that not much in life is absolutely 100% safe!
  • Design BEs so that as far as possible they are ‘no lose’, i.e. whatever happens, the client will have gained something. If the experiment ‘works’ in the sense that negative predictions are not borne out, then that is useful; but equally if some part of the negative prediction is confirmed, that can still be useful if we have learned something and now need to think about what made that happen, which in turn can lead to further productive exploration and new BEs.
  • For the same reason, try to be genuinely open-minded about the outcome of BEs. Do not approach them in a way that suggests to the client that you already know for sure what will happen. If the BE does not work out the way you predict, then the client may lose confidence in you and may also feel that he must have failed. It is much better to be genuinely curious: ‘I really don’t know for sure what will happen here, but maybe it won’t be as bad as you fear – how about finding out?’
  • Similarly, try to anticipate with the client what might be difficult or go wrong and then develop and rehearse strategies for coping with such setbacks. If you are doing a BE involving other’s reactions, what will the client do if indeed he does get a negative response? If the BE involves a client with agoraphobia going into a supermarket alone, how will he cope if he does have a panic attack? BEs are much more likely to be helpful if you have considered such problems beforehand.
  • Notwithstanding all of the above, do not ignore the potential for doing spontaneous BEs, prompted by something that has happened within a session. For example, when discussing with someone who fears cardiac problems the effects of a safety behaviour like avoiding exertion, it may be possible to suggest doing a BE right now – for example, running up and down the stairs a couple of times to see what happens. Sometimes your client may be more willing to try something on the spur of the moment than he would be if he has a week to worry about it. This obviously needs to be carefully done, and the client must be clear that he can refuse if he wants to, but it can be very productive.

The experiment itself

Experiments may be carried out by your client independently, for example as part of homework, or your client may carry them out with you – in session or outside in the real world. The latter in vivoexperiments can be very useful, both because you can support and encourage your client and because they offer invaluable opportunities for you to learn more about the problems: in vivo BEs frequently generate previously unknown thoughts and beliefs, safety behaviours and so on. If you are accompanying your client, there are several things you can be aware of in order to increase the chances of a successful outcome; if your client is trying a BE alone, then you can make him aware of these factors:


  • Encourage your client to be fully engaged in the situation rather than just ‘going through the motions’. He needs to understand that usually if a BE does not result in any anxiety (for example, because he is distracting himself or not really pushing against his limits) then the BE is less likely to be useful.
  • You and/or your client need to be continually monitoring his thoughts and emotions, both in order to be aware of any changes, whether positive or negative, and to be sure that the BE is going along the right lines. For example, it is unusual for a client to go through a BE without feeling at least some discomfort during it; if he remains completely unaffected, it would be wise to investigate whether he is subtly avoiding, or performing safety behaviours. On the other hand, if there is no positive change at all in the client’s thoughts or emotional state during a BE, it may indicate that the cognitions have not really been touched, and it may be useful to think about taking it further, or doing something different.
  • As noted above, BEs by their very nature are to some degree unpredictable, and the unexpected can and does happen. You and/or your client need to be flexible and ready to respond to unexpected events.

After the experiment

In order to make the most of a BE, it is important to take time to ‘de-brief’ and help the client reflect on what happened:


  • First, you need to go through with the client what actually happened. What were his thoughts? How did he feel? Did events go as predicted or were there significant differences to his predictions? If so, what were they? Did he still use any safety behaviours to prevent some disaster (if so, it may be important to try again with reduced or eliminated safety behaviours)?
  • Second, it is important to help the client reflect on the meaning of the BE. What does this tell him that he didn’t know before (about himself, or others, or the world in general)? How can he make sense of what happened? Does it have any implications for how he might tackle similar situations in the future? Are there any follow-up BEs that might be useful to extend or generalise his conclusions? Finally, have the client re-rate his belief in the cognitions tested so that both of you can see whether there has been any change.

This post-experiment reflection can help the client gain the maximum possible value from the experiment and may also help reduce the risk of his devaluing the results of the experiment as old habits reassert themselves.

Below is a record sheet that you may find helpful for you and your client to record the planning and carrying out of BEs (Figure 9.2).

Common problems in behavioural experiments

BEs can be an extremely powerful way of changing cognitions and emotions, but, as noted above, their complexity and unpredictability also mean that there is plenty of scope for things to go in unexpected directions. Many of these risks can be avoided by careful planning and preparation, but this section gives some further ideas about how to cope with some common problems.

Figure 9.2    Behavioural experiment record sheet

Therapist worries

It is important to recognise that therapists, as well as clients, may have worries about BEs. If these become too intense, you may communicate your doubts to your client and thus reinforce his fears. It is acceptable – maybe even desirable – for you sometimes to be pushing your own limits, for example, by doing things in public that trigger your own social anxieties. But it is also important that you approach BEs in a positive and encouraging way: ‘This may be a bit scary, but it’s not going to be a catastrophe.’

Finding a graceful retreat

Even with the best planning in the world, sometimes things go wrong: the test turns out to be harder than you or the client thought; other people react in precisely the ‘wrong’ way; or the client’s nerve fails him. It is at these times that therapeutic skill and creativity are most needed, to find a way to retreat with grace in such a way that the client does not feel he has completely ‘failed’. A good general rule is to try always to finish with a success, no matter how small. If the original aim is clearly too ambitious, try to find a smaller goal that the client can accomplish before finishing the exercise.

Experiments that ‘fail’

If negative predictions do come true, then we can still learn something useful by examining carefully what happened. Was it just an unlucky chance outcome, or did the client do something that produced that result? Is there some other aspect of cognition or behaviour whose effect we have not fully taken into account? Are there subtle forms of avoidance or other safety behaviours that are reducing the impact of the experiment? It is important to use such ‘failures’ constructively – even negative information can tell us something we can use to make therapy ultimately more effective.

Therapist–client relationship

There are different demands on the therapeutic relationship between a typical office-based therapy and BEs in CBT where, for example, you may be going to a supermarket with your client and falling over in the shop so that he can observe how others react. What professional issues does this raise? What kind of conversation is acceptable when you are outside the office and not ‘on task’? It is important to reflect on these issues and discuss them in clinical supervision so that you can arrive at a way of relating that feels reasonably comfortable to both you and the client whilst respecting essential professional and ethical boundaries (see Chapter 3).




  • Behavioural experiments (BEs) involve the active gathering or generating of evidence that will help to test our clients’ – or our own – thoughts and beliefs.
  • BEs therefore offer a way to go beyond purely verbal examination of cognitions by testing them out in the real world. As such, they may lead to more ‘gut level’ learning.
  • There are different types of BE, including hypothesis-testing vs. exploratory BEs, and active vs. observational BEs.
  • To help clients get the most out of BEs, it is usually best to plan and carry them out carefully – although there is also a useful role for occasional impromptu BEs, done on the spur of the moment in response to something that comes out of a session.
  • BEs sometimes demand an active role for therapists in doing activities with clients outside the therapy office. This can challenge therapists’ fears, as well as patients’.

Learning exercises

Review and reflection:


  • If you are new to BEs, then take a few minutes to think about your thoughts and feelings in relation to them. You might think particularly whether you have your own worries about doing BEs, as we mentioned above. If so, what could you do about that? Can you come up with any alternative views? Have you any evidence to support or contradict your worries? Remember this is one of those areas where our ownsusceptibility to negative thoughts and beliefs may be prominent, and may block us from trying out effective procedures.
  • What are your own views about how to relate to clients when you are doing BEs together outside the therapy office? Reflect on these questions and perhaps discuss them with your supervisor.

Taking it forward:


  • If you are new to BEs, or have doubts about them as discussed above, perhaps you can do BEs about BEs. Review your caseload with your supervisor; think about which of your clients’ beliefs you might be able to test through BEs; devise appropriate BEs; and then review the results with your supervisor.
  • When you try out doing BEs, do they seem to offer any of the benefits to clients that we have suggested above? Do any disasters follow? On balance does the use of BEs seem useful or not?

Further reading

Bennett-Levy, J., Butler, G., Fennell, M., Hackmann, A., Mueller, M., & Westbrook, D. (Eds). (2004). The Oxford guide to behavioural experiments in cognitive therapy. Oxford: Oxford University Press.

The essential guide to BEs, with ideas about conceptualising them, guides to using them in different disorders, and many practical examples of actual BEs used with clients.