The method of factors

A dynamic hierarchy is not a predetermined list of behavioural tasks. Rather, the subject and therapist collaboratively identify factors that influence the intensity of the phobic experience – such as closeness of approach, size of the stimulus, colour, texture, verbal descriptions, use of imagery, anticipation of feared consequences, touching or not touching, eyes open or closed, induction of symptoms (e.g. hyperventilation) and so on.

A rough set of versions can be distinguished for each of these factors, e.g. degrees of severity of contamination in a case of OCD. (See the Appendix for an example of the clinical application of dynamic hierarchies with a client.)

Arousal set-point

The client should choose the target level of arousal. The therapist should suggest that the appropriate level is one as high as the subject can tolerate for an hour or so, but no higher; but the decision about what that means should be the unforced choice of the client. One important purpose of keeping it so high is to drive the pace of change. That pace is seen in how frequently the subject needs to step up their arousal to prevent it declining. As habituation occurs, one or more of the factors can be intensified in order to keep the discomfort high. Since the rate of habituation is proportional to the intensity of discomfort (Eberle et al. Reference Eberle, Rehm and McBurney1975), this should maximize the rate of improvement.

Moving up the scale

As the arousal measured by subjective units of distress (SUDs) starts to reduce, subjects should ask themselves ‘what could I now do to push it back up to the set-point?’ That question should prompt the subject to review their factors and find some way of raising their arousal, for example by commencing hyperventilation, evoking fearful memories, or by driving faster (if it is safe and legal to do so). It is in order to reduce the strength of one factor so that a different one can be introduced; for example, to step back from the cat, commence hyperventilation, and evoke feared memories; and then move closer to the cat again as habituation occurs.

Escalating the challenge in order to keep arousal high might be thought to prevent within-session habituation, and therefore between-session habituation. In fact, there is evidence that incomplete within-session habituation is preferable, in that it prevents the return of fear (Rachman et al. Reference Rachman, Robinson and Lopatka1987; Rachman & Lopatka, Reference Rachman and Lopatka1988).

It is sometimes suggested that safety behaviours should be discouraged during exposure (Craske & Barlow, Reference Craske and Barlow2007, p. 42), but this results from thinking of the exposure task and the consequent arousal as being separate. Using the dynamic hierarchies model we can treat safety behaviours as just another in the list of factors that modulate arousal.

This latter approach is well illustrated in Milosevic & Radomsky (Reference Milosevic and Radomsky2008). They randomized 62 snake-phobic volunteers to two conditions where they approached a snake either with or without gloves and goggles. The subjects with safety equipment achieved closer approach to the snake (Fig. 1a), but SUDs ratings during exposure were identical (Fig. 1b). Post-treatment questionnaires showed significant improvement (p < 0.001) but no significant difference between conditions. This is precisely the pattern we would wish to exploit in using dynamic hierarchies.

Fig. 1.

(a) Approach to the stimulus and (b) subjective arousal levels during exposure while using (—•—) or not using (- -•- -) safety behaviours. BAT, Behavioral Avoidance Test; SUDs, subjective units of distress. (From Milosevic & Radomsky, Reference Milosevic and Radomsky2008, reproduced with permission from Elsevier Ltd., UK.)

Among the benefits of a dynamic hierarchy approach is that it acknowledges the commonality between different avoidance-based conditions, i.e. that the subject is avoiding the discomfort occasioned by certain triggers. Once clients have learned how to carry out exposure, and experienced the result, they are in a position to apply the same lessons to other areas where they may experience fear, anger, guilt, shame or depression. This might be described as the generalizability of the therapeutic change.

A further benefit is that the therapist does not need to warn against the use of safety behaviours, distraction, self-reassurance and other arousal-prevention behaviours. Using the dynamic hierarchies framework allows the client to examine the function and utility of a range of behaviours in maintaining or resolving their problems. The use of such strategies is discussed with the client and these may be selectively utilized to maintain arousal levels at the agreed ‘set-point’.

A dynamic hierarchy also provides for a smoother escalation in the challenge that subjects present to themselves. Rather than having to simply introduce a new element (such as driving alone rather than accompanied) – which may produce a sharp increase in arousal, possibly beyond what subjects can persuade themselves to accept – a dynamic hierarchy allows the subject to merely ‘top up’ their arousal as it starts to habituate.

Further improvement to the dynamic hierarchy is obtained by paying attention both to the appraisals that we expect to change during exposure, and to the use of cognitive dissonance to maximize such change. We now consider these points in a little more detail.

Appraisals

Appraisal theory attempts to explain emotional experiences by referring both to the evaluative cognitions that accompany (or perhaps cause, underlie or correlate with) them, and to two types of cognitions – knowledge and appraisal (Lazarus & Smith, Reference Lazarus and Smith1988) – in which knowledge is argued to play a distal role and appraisal a proximal role in emotion. Parkinson's (Reference Parkinson1997) review of the evidence is instructive. He questions whether evaluative cognitions have a strictly causal role in emotion. Certainly, from a philosophical perspective (as well as from a three-systems perspective; Lang, Reference Lang and Shlien1968) it makes sense to see appraisals as constitutive of emotions, rather than causal. This does not mean we cannot use behavioural methods to change appraisals, just that we should see ourselves as influencing the whole system of which they are a part. The difference between ‘is’ cognitions and ‘ought’ cognitions is fundamental, and it would be a gross error to confuse them (Hare, Reference Hare and Hudson1969). However, these are deep philosophical waters.

Appraisals may be changed by dissonance-led methods; including disputation (Ellis, Reference Ellis1962), voluntary exposure, and acceptance. An example of the use of disputation to increase dissonance can be found in Festinger (Reference Festinger1957); high-school students were asked to persuade a fellow student of a political position that they themselves disagreed with. The political opinions of the listener changed slightly in the direction opposite to that being argued for, while those of the persuader changed significantly in the direction of their arguments.

If, then, the client's emotional response to a stimulus is to change, we must look to change the client's appraisal (evaluation) of the stimulus. For example, therapeutic change will include a change in attitude to cats, so that they cease to be evaluated as ‘horrible’. Such change can occur if the client engages in behaviour that is dissonant with such an appraisal, and the probability of the desired change can be increased by manipulating the associated cognitive dissonance.

We should also note that descriptive cognitions are susceptible to such cognitive effects as the availability heuristic (Williams et al. Reference Williams, Watson, MacLeod and Mathews1997, ch. 8) – the client's feel for the likelihood of an adverse outcome will be affected by how vividly it is imagined; and this in its turn will be affected by how it is appraised. We predict that reducing a person's emotional responding will cause a reduction in their estimate of the chance of the feared outcome – so changes in descriptive cognition will sometimes be parasitic on changes in appraisal.

Cognitive dissonance

It has been hypothesized that an important mechanism for cognitive change during exposure is cognitive dissonance (Saxby, Reference Saxby1973; Tryon & Misurell, Reference Tryon and Misurell2008), although it must be conceded that the hypothesis still awaits adequate testing.

From the control system perspective, what has been termed ‘cognitive dissonance’ (Festinger, Reference Festinger1957) is better seen as conflict between systems at different levels within the hierarchy – one at a higher (principle or system) level than the other (e.g. programme level). Translating this into a more familiar language, e.g. where there is a conflict between, on the one hand, underlying structures such as schemata, assumptions or value systems and, on the other, the behavioural expressions of those structures, such as self-talk or approach behaviour, the conflict will be resolved by changing either the underlying structure or the behaviour. For the purposes of this paper, however, we shall keep to the language of cognitive dissonance.

Our suggestion is that exposure is therapeutic when it produces a change in attitude (i.e. high-level system), and that such change may be brought about by the mechanism of cognitive dissonance (Festinger, Reference Festinger1957). It is worth noting here that although Festinger's explanation of his theory suggests that the dissonance in question is between two descriptive beliefs (‘I am smoking’ and ‘smoking is bad for me’ being one example he gives), the instances he offers all show discrepancy between a behaviour and a value judgement (witness the above example). This observation suggests that an appraisal (such as ‘this cat is horrible’) is changed by deliberately acting in a manner incompatible with such an opinion, i.e. by exposure.

Thus, the evidence for the effects of cognitive dissonance is primarily around appraisals. Descriptive cognitions or knowledge – statements of putative fact such as ‘everybody hates me’ or ‘I'll never pass my exams’, on which evidence can be brought to bear – are less obviously prone to dissonance-driven change, at least in its direct form. This change of evaluative attitudes (measured by semantic differentials) during exposure is illustrated in Watson et al. (Reference Watson, Gaind and Marks1972, fig. 9) and Marks (Reference Marks1981, fig. 2.2).

One implication of the cognitive dissonance hypothesis for the use of exposure is that the client's choice of level of anxiety should be subjected to as little external influence as possible (Festinger & Carlsmith, Reference Festinger and Carlsmith1959; Zimbardo, Reference Zimbardo1969).

The dynamic hierarchy method would conform to this by requiring an ongoing series of decisions-to-continue, each of them (where possible) unforced. The character of these decisions as (relatively) forced or unforced should influence the rate and degree of cognitive change. This is a matter for further research.

If this picture is correct, it suggests that the best way to produce therapeutic change can be specified by looking at how to maximize cognitive dissonance. We know that unforced choices of aversive experience create greater dissonance, and therefore greater drive to change. This means that the ability to escape the exposure at any moment is an important element, provided that the escape is not carried out. Other factors, such as being free to choose the time, duration, frequency and intensity of the exposure will also magnify dissonance (on the model of post-decision dissonance see Festinger, Reference Festinger1957). For this reason, the dynamic hierarchy model is predicted to be more powerful than the static, as it requires the client to make an unforced choice about what to do next.

Miss R – avoidance of trauma-related cues

Miss R is a young working-class black woman living in industrial South Yorkshire, who was the victim of an unprovoked attack by three white women outside a nightclub. Her throat was cut in the attack, although she has no clear memory of this aspect. When confronted with reminders, such as images of injury, she re-experiences the attack and can feel and taste blood in her throat. Her ‘hot-spot’ events for this episode include being on her hands and knees on the pavement, being loaded into the ambulance, and in a cubicle at the hospital. She can remember the feeling of her throat being stitched.

Prior to this, she had always had white boyfriends; since the attack she has had only black boyfriends. She will not go to any nightclub alone, and she avoids those clubs where there is a preponderance of white customers.

She received instruction in the practice of exposure using dynamic hierarchies. It was emphasized that she should practise frequently and for a good period of time: and that she should try to keep her arousal as high as possible throughout the practice. We discussed what factors would increase or decrease arousal. These arousal-relevant factors included those given in Table 1.

Table 1.

Arousal-relevant factors for Miss R

Clearly some of these factors interact (e.g. II and V, and V and VI). Some are more under her control (I, VI), some less so (II), and others not at all (IV). The control may involve controlling the stimulus (asking a friend to stay close), or the percept (not looking at the women). She was instructed that if someone started staring at her, for example, she might reduce her arousal back to her set-point level by moving closer to her friend, looking away, or moving to a less crowded part of the dance floor. The list of versions of each factor does not need to be in order of severity, although it should ideally represent a good range of different levels of severity. The grading is molar – between whole experiences, each of which has a mix of factors represented, rather than between individual factors. The exercise of grading them is not about producing a static hierarchy, but about teaching the client to construct their hierarchy dynamically and extempore.

Miss R tried several times to arrange a night out with her friend, but without success. She therefore started working on her hierarchy alone, beginning with bars in the daytime and progressing to nightclubs. She kept a record of her exposure sessions, including arousal levels (see Fig. 2). She used a bluetooth earpiece to record a running commentary on her mobile phone, and transcribed her notes later, but was unwilling to supply a digital copy of her recording.

Fig. 2.

Excerpt from Miss R's exposure record.

To construct a graphical representation of her progress we associate each rise in SUDs (on stepping up the hierarchy) with the time where it happened, and plot the cumulative rise against minutes into the exposure (Fig. 3a). Notice especially the retrenchment that occurs 9 minutes in: this is entirely compatible with the dynamic hierarchy model, as it was bringing her back to her target arousal level.

Fig. 3.

Within-session charts showing cumulative subjective units of distress (SUDs) scores for (a) approach and (b) habituation.

Alternatively, we can chart the cumulative habituation by associating falls in SUDs with the times they happened (Fig. 3b)