Emotion and two kinds of meaning: Cognitive therapy and applied cognitive science

Emotion and two kinds of meaning: Cognitive therapy and applied cognitive science

0005-7%7/93 $6.00 + 0.00 Copyright 0 1993 Fkrgamon Press Ltd Behav. Res. Thher.Vol. 31. No. 4. pp. 339-354. 1993 Printed in Great Britain. All rights...

2MB Sizes 1 Downloads 26 Views

0005-7%7/93 $6.00 + 0.00 Copyright 0 1993 Fkrgamon Press Ltd

Behav. Res. Thher.Vol. 31. No. 4. pp. 339-354. 1993 Printed in Great Britain. All rights reserved



Research Council


Psychology (Received



15 Chaucer

f 7 September

Road, Cambridge

CB2 2EF.



Summary-The clinical cognitive approach assumes that emotional reactions are mediated through the meanings given to events. Cognitive therapy aims to change emotion by changing meanings. It focuses on specific level meanings, evaluating the truth value of particular beliefs. Bower’s science-driven associative network theory of cognition and emotion is also primarily concerned with specific meanings. This focus on meaning at a specific level causes problems, e.g. the contrasts between ‘intellectual’ and ‘emotional’ belief, between ‘cold’ and ‘hot’ cognition, and between explicit and intuitive knowledge. These problems are resolved in the Interacting Cognitive Subsystems (ICS) approach. ICS distinguishes between a specitic and a more holistic. intuitive, level of meaning. In contrast to alternative approaches, ICS suggests that holistic level meanings are of primary importance in emotion production. Representations at this level consist of schematic mental models, encoding high-order inter-relationships and prototypical patterns extracted from life experience. The ICS approach to meaning is described and its implications for understanding and treating emotional disorders discussed, together with relevant empirical findings. ICS suggests a therapeutic focus on holistic rather than q&tic meanings, a role for ‘non-evidential’ interventions, such as guided imagery, and a rational basis for certain experiential therapies.


Alan Ross, in a recent book review, characterised the current state of behaviour therapy as Growth wifhouf Progress: “One thing is certain: The field has grown, but it has not progressed. The growth has taken the form of moving into additional areas of application. Progress would have meant applying additional psychological principles in old and new areas, supplementing those of conditioning and learning with which the field began . . . what is being applied is whatever promises to work. Conceptual source and empirical support seem irrelevant. This is no longer behavior therapy, at least not as its founders conceived it” (Ross, 1991, p. 743). Ross singles out “the meteoric development of cognitive behavior therapy” as the factor “chiefly responsible for the present state of affairs . . . creative and well-controlled laboratory research on cognitive processes is being published by diligent investigators, but the therapists who seek to reduce their clients’ disorders by trying to change their concepts and attributions employ neither the researchers’ methods nor their principles. What they do employ is some of the terminology of cognitive psychology, schema and Structure for example. Such terms are usually called on to serve as post hoc rationa~isations when adhoc clinical procedures seemed to have worked” (Ross, 1991, p. 743). This paper considers the area singled out by Ross as particularly problematic, the relationship between cognitive-behaviour therapy and the relevant science base. We shall concentrate primarily on cognitive therapy of depression (Beck, Rush, Shaw & Emery, 1979). This approach has undoubtedly made considerable clinical achievements (e.g. see HolIon, Shelton & Loosen, 1991). The contributions of Beck and others who have developed cognitive approaches to understanding and treating depression are rightly regarded as one of the most significant breakthroughs this century. However, there are also voices that suggest that all is not well in this field and that progress may be reaching some kind of plateau. The Beckian cognitive model and treatment of depression depend on ‘everyday’ or ‘lay’ concepts of ‘cognition’ and meaning. Some of the many problems now recognised with Beck’s cognitive model and therapy can be related to this usage. In this paper I shall describe alternative conceptual frameworks, rooted in experimental cognitive psychology and cognitive science, that overcome 339



these problems. Initially, I shall focus on applications of Bower’s (198 1) associative network theory of mood and memory to understanding depression. Application of this science-driven conceptual framework allows us to overcome a number of the difficulties of the Beckian clinical cognitive model. However, over time, it has become apparent that Bower’s theory, itself, has formidable difficulties at both the empirical and theoretical levels. The advantage of working with the relatively precise concepts and theories developed from laboratory investigations, compared to more ‘everyday* clinically-derived concepts, is that their precision allows us to know more quickly when theories are wrong and, most importantly, how they are wrong. Consequently, by a continuing interaction between theory and experiment, we can progressively refine and advance our understanding of emotion-related information-processing. The method of advance consists, literally, of learning from our mistakes. More fuzzy ‘everyday’ concepts of cognition, that are readily communicated to clients, therapists and researchers alike, may be invaluable at getting therapy movements launched on their ‘meteoric’ trajectories. However, ‘fuzzy’ theories lack the self-correcting potential inherent in more precisely specified scientific theories. For that reason, they may be less able to point the further way forward when, as Ross (1991) suggests, the initial exhilarating rate of progress slows and we sense that we may be running into the sand. This paper will illustrate the self-correcting potential of scientific theories by showing how the deficiencies identified in Bower’s approach can be used to shape our ideas of what will be required in other theories if they are to avoid those same problems. I shall describe an alternative science-driven framework that overcomes the difficulties of the associative network theory of mood and memory. The Interacting Cognitive Subsystems (ICS) framework (Barnard & Teasdale, 1991; Teasdale & Barnard, 1993) recognises qualitatively distinct types of information (cognition), with different functional relationships to emotion. In particular, this framework distinguishes two levels of meaning, a specific, and a more generic, holistic, level. In contrast both to Beck’s clinically derived cognitive model (Beck ef al., 1979), and to Bower’s experimentally derived associative network theory of mood and memory (Bower, 1981) the ICS framework proposes that only the more generic level of meaning is directly linked to emotion. As well as describing the ICS approach, I shall describe an investigation contrasting the explanations for depressive thinking offered by the Bower and ICS accounts. The findings of this study support the view that affect is linked to an holistic, rather than specific level of representation. The ICS approach overcomes the difficulties of the clinical cognitive model. ICS also has very important implications for how we conduct therapy now and how we can develop new approaches to treatment in the future. I shall conclude by drawing out some of these implications. BECK’S






Beck’s cognitive model is a ‘clinical’ theory. The main purpose of the theory is to guide the clinician in his understanding and treatment of patients, rather than to provide a detailed exposition articulated in precise theoretical terms. Reflecting this aim, the cognitive model tends to use ‘everyday’ concepts of ‘cognition’ and ‘meaning’. As is well known, Beck’s cognitive model suggests that the emotional impact of an event is mediated through the meaning placed on the event, this meaning being available in consciously accessible thoughts and images. In this approach, cognitions are defined as “any ideation with verbal or pictorial content” (Beck et al., 1979, pp. 12-14). Similarly, “Cognitions are stream-ofconsciousness or automatic thoughts that tend to be in an individual’s awareness . . .” (Beck, Epstein & Harrison, 1983, p. 2). This use of the term ‘cognitions’ to refer solely to consciously experienced thoughts and images clearly diverges from the much wider use of the term in cognitive psychology. There, it is assumed that the majority of cognitive processing is not experienced as consciously accessible thoughts or images. Beck’s cognitive model suggests that “distorted cog&ions are produced when a stressful event (e.g. divorce, loss of a job) activates an individual’s unrealistic schemata. . . Schemata are . . . stable, general underlying beliefs and assumptions . . . Examples of schemata are ‘If I am not loved by others, I am not a worthwhile person’ and ‘I must achieve great things or I will be a failure in life’. . . the person’s underlying assumptions constitute a vulnerability to events” (Beck et al., 1983, p. 2).

Two kinds of meaning and cognitive therapy


Again, the equation of ‘schema’ with beliefs and assumptions is quite different from the usage of this term by cognitive psychologists. In fact, the development of cognitive therapy for depression has proceeded largely in isolation from basic cognitive science. The cognitive model was developed initially from Beck’s astute clinical observations and investigations of depressed patients. Cognitive therapy for depression was born from an inspired amalgamation of the cognitive model with the best features of behaviour therapy as it had already developed more generally: a clearly articulated treatment rationale, a high degree of structure, explicit procedures that can be concretely described in a treatment manual, goal-oriented, symptom-/problem-focused, homework-based, subject to systematic empirical evaluation, etc. This recipe has clearly worked very well up to now, delivering a psychological treatment that is as effective as tricyclic antidepressants in the treatment of outpatient major depression, and that offers, even more excitingly, the hope of better prophylaxis of future episodes than drug therapy (Hollon et al., 1991; Williams, 1992). One might reasonably respond: ‘OK, so there may not be much relationship to basic cognitive science, so what? Who needs it? Can’t we get on very well without?‘. Unfortunately, it seems that we cannot; increasingly, problems have been encountered by the Beckian cognitive approach. For example: (I) We now know that cognitive therapy does not seem uniquely more effective than other psychological treatments for depression that do not target negative thinking (e.g. Williams, 1992). Embarrassingly, when such treatments alleviate depression they also reduce most measures of negative thinking to a similar extent to comparably effective cognitive therapy (e.g. Rehm, Kaslow 8c Rabin, 1987; Imber, Pilkonis, Sotsky, Elkin, Watkins, Collins, Shea, Leber & Glass, 1990). These findings suggest that cognitive therapy may achieve its effects, not by changing negative thinking, but by some other mechanism, shared with noncognitive treatments. The changes in negative thinking may be a consequence of the reduction in depression rather than antecedent to it. (2) Improvements in depression as a result of treatment by anti-depressant drugs are associated with changes in measures of depressive thinking similar to those shown with comparably effective cognitive therapy (e.g. Simons, Garfield & Murphy, 1984). Again, this raises the possibility that negative thinking is a consequence of depression, not an antecedent. (3) The cognitive model suggests that vulnerability to depression depends on individuals possessing underlying dysfunctional assumptions and attitudes. There has been a conspicuous failure to demonstrate the predicted presence of these attitudes in vulnerable individuals once their depression has remitted; the evidence suggests such attitudes are often mood-state dependent, rather than enduring characteristics of vulnerable individuals (e.g. see Teasdale, 1988). (4) It is a common clinical observation that patients can experience emotional reactions without being able to identify either any related negative automatic thoughts or any negative thoughts commensurate with the intensity of the reaction. And yet, such thoughts and images are postulated as the mediating link in emotion production by the cognitive model. (5) Within cognitive therapy, ‘rational’ argument or ‘corrective’ information is frequently ineffective in changing emotional response, even when the client acknowledges ‘intellectually’ the logical power of the evidence. Relatedly, Beck’s cognitive model recognises only one level of meaning, and for that reason has considerable difficulties with the distinction between ‘intellectual’ and ‘emotional’ belief, or, more generally, between ‘cold’ and ‘hot’ cognition. So, when a depressed person says something like ‘I know I’m not worthless but I don’t believe it emotionally’, the Beck approach suggests that this simply reflects quantitative variations in a single level of meaning: “The therapist can tell the patient that a person cannot believe anything ‘emotionally’ . . . when the patient says he believes or does not believe something emotionally, he is talking about degree of belief’ (Beck er al., 1979, p. 302, original italics). Many clinicians have found this analysis unconvincing, regarding ‘emotional’ belief as qualitatively distinct from ‘intellectual’ belief, and functionally more important. (6) As cognitive therapy has been applied to more long-standing clinical problems, it has become clear that the conventional cognitive therapy for depression program, focusing on negative automatic thoughts, is frequently ineffective. In response, treatment procedures have been imported wholesale on an ad hoc basis from other therapy traditions based on quite different underlying rationales, notably Gestalt therapy (e.g. Edwards, 1990; Young, 1990). Any effectiveness of such procedures seems more appropriately attributed to the rationale of the therapy from which they were borrowed than to the cognitive model.



The accumulation of problems such as these has led some to suggest a ‘crisis of confidence’ in the clinical cognitive approach to understanding and treating depression. It seems that all is not well, and that there is, indeed, a case for asking whether, after all, the clinical cognitive approach can afford to ignore the relevant science base and the applied science model much longer. The Mulla Nasrudin tradition (Shah,1972, p.116) provides us with a cautiona~ tale that may be relevant. It is entitled “All I needed was time”: “The Mulla bought a donkey. Someone told him that he would have to give it a certain amount of food every day. This he considered to be too much. He would experiment, he decided, to get it used to less food. Each day, therefore, he reduced its rations. Eventually, when the donkey was reduced to almost no food at all, it fell over and died. ‘Pity’, said the Mulla. ‘If I had had a IittIe more time before it died I could have got it accustomed to living on nothing at all.“’ There is a real possibility that, just as the Mulla’s donkey appeared to survive on a diminishing input of food until the day it died of starvation, so the clinical cognitive approach may appear to thrive without bothering too much about its isolation from basic cognitive science until the point is reached where, deprived of new input, progress grinds to a halt. So, let us now turn to some more basic science-driven approaches to the relation between cognition and emotion to see what sustenance they can provide. BOWER’S ASSOCIATIVE



Bower’s associative network theory of mood and memory (Bower, 1981) has been the most influential of the experimentally-derived cognitive psychological approaches to understanding cognition-emotion relationships. The essence of the theory is that, when we are in a given mood now, we are more likely to remember events previously associated with that mood in the past. The concepts and constructs previously used to interpret events in that mood will also be reactivated and become more accessible, and so will be more likely to be used to interpret current events. So, for example, in a depressed mood n01v a11 the concepts and constructs previously activated in interpreting the events that made us depressed in the past will be reactivated. Consequently, current experience will be more likely to be interpreted negatively. There are two important points to make about this associative network theory. First, it enables US to overcome many of the difficulties that the Beckian cognitive model encountered. Second, we now know that this network model is inadequate in important respects. The network model, and related experimental work, suggest that depressed mood biases cognitive processing in a negative direction in all of us. We can see the negative thinking of clinical depression as an extreme form of a normal effect of mood on information processing, rather than the result of ‘matching’ environmental events activating related dysfunctional assumptions in vulnerable individuals. The network theory suggests that negative thinking is a consequence of depressed mood. It follows that the theory has no difficufty with findings showing that negative thinking reduces as depression improves, even when the improvement results from pharmacotherapy, or from non-cognitive psychological treatments. Similarly, negative thinking in depressions with a primary biological aetiology can be explained. However, and most importantly, the application of network theory to depression suggests that negative thinking can be an antecedent to depression, as well as a consequence: negative cognitive appraisals can produce depressed affect. The network theory suggests a reciprocal relationship between negative thinking and depressed mood: once in depressed mood we are likely to think negatively because of the effects of mood on the accessibility and activation of negative memories and interpretative constructs, but that same negative thinking is likely to maintain depression. Thus, self-perpetuating vicious cycles of the form shown in Fig. I can become established. In this way, application of the network model provides an explanation for the maintenance of depression (Teasdale, 1983; Ingram, 1984). It also offers an alternative view of cognitive vulnerability, which overcomes the difficulties of the Beckian cognitive model.

Two kinds of meaning

and cognitive






Fig. 1. A vicious cycle based on the reciprocal relationship between and depressed mood [from Teasdale





negative (1983)].



of experience


Application of the network model to depression suggests that vulnerability to severe or persistent depression can be thought of in terms of the likelihood that the vicious cycle shown in Fig. 1 will become established and maintained. Many factors will affect the probability that this will occur. The network theory suggests as one important factor the type of memories and concepts that become re-activated by depressed mood. Although these will tend to be negative in all of us, it may be only certain types of negative cognition that produce sufficient further depression to keep the cycle revolving. The tendency to access this type of cognition in depressed mood may be an important aspect of cognitive vulnerability to depression. Thus: “Once a person is initially depressed, an important factor that determines whether that depression remains mild and transient or becomes more severe and persistent is the type of cognitions that become accessible in that state . . . vulnerability to intense and persistent depressions may be determined by individual differences in the cognitions that become accessible once the person is depressed” (Teasdale, 1983). This differential activation hypothesis of cognitive vulnerability (Teasdale, 1988) suggests that the negative findings from most previous studies of cognitive vulnerability are not unexpected. On this view, the crucial factor related to vulnerability is the pattern of information processing occurring Once a person is at least mildly depressed. The differential activation hypothesis does not necessarily predict differences between recovered depressed patients and controls on measures taken in normal mood state. This view suggests that we will be more likely to find differences if we take our measures in mild depressed mood. There is encouraging preliminary support for this prediction. Teasdale and Dent (1987) compared recovered depressed and never depressed Ss in normal mood and in experimentally induced depressed mood. In the absence of depressed mood, scores on a measure related to global negative self-view were identical for the recovered depressed and never depressed Ss (Teasdale, 1988, p. 259). In induced mild depressed mood, recovered depressed Ss differed significantly in the predicted direction from controls on the measure of negative self-view (Teasdale & Dent, 1987). Miranda and Persons (1988) and Miranda, Persons and Byers (1990) have reported comparable findings from two studies that examined the relationship between naturally occurring mild depressed mood and scores on a measure of dysfunctional assumptions, the Dysfunctional Attitude Scale (DAS; Weissman & Beck, 1978). All Ss scored in the non-depressed range of the Beck Depression Inventory at the time of study, indicating that the previously depressed Ss were fully recovered from their episode of depression. The pattern of results was similar in both studies. Recovered and never depressed Ss did not differ in DAS score at low levels of depressed mood, replicating the general negative findings of previous studies that had compared recovered depressed and never depressed Ss in normal mood. However, with increasing depressed mood, the DAS scores of the recovered depressed group increased whereas those of the non-depressed group did not. Consequently. mean DAS scores of the two groups differed substantially for Ss tested in mildly depressed mood, just as the differential activation hypothesis predicts. Williams (1988) obtained further supportive evidence from a prospective study of Oxford undergraduates. Scores on the measure of global negative self-view used by Teasdale and Dent (1987) were obtained in both normal and experimentally induced mild depressed mood, at Time

Jam D. %ASDAL&


I. Subjects were then followed over the course of 1 yr, completing the Beck Depression Inventory every few months. Measures of negative self-view taken in normal mood at Time I failed to predict depression at any of these assessments. By contrast, measures of the extent to which, at Time I, negative self-view increased as Ss moved from normal to induced depressed mood significantly predicted subsequent depression on three out of the four assessment occasions. Again, these findings provide promising initial support for the idea that cognitive vulnerability to depression may not be related so much to enduring differences in dysfunctional attitudes or assumptions as to differences between individuals in the patterns of thinking that become activated once they are in a mildly depressed mood. Recourse to experimentally-derived observations and theory allows us to retain an essentially cognitive approach to certain aspects of VuInerabiIity, and so to sustain the viability of a cognitive perspective on depression. Experimentally-derived theories can give new insights into clinical conditions. Of equal importance, the precision of more ‘scientific’ theories aliows us to identify their areas of weakness and error, and in this way to know what will be required in the theories that will supersede them. It is now clear that the Bower network theory has major inadequacies; key empirical predictions have not been fulfilled, and major difficulties are apparent when one examines the theory carefulty (Teasdale, 1993; Teasdale & Barnard, 1993). Many problems can be attributed to the fact that semantic networks represent all knowledge in a single uniform format, that of specific concepts and the relationships between them. It follows that they cannot easily represent knowledge at levels of meaning more abstract than the word or sentence, although the need for higher order levels of meaning is widely recognized in mainstream cognitive psychology. The single representational format also creates enormous difficulties for Bower’s model when it comes to handling the distinction between ‘hot’ and ‘cold’ emotional processing. It is clear that we can think about emotional events or concepts without experiencing emotion. Equally, we can remember emotional events ‘cold’ (‘that I was upset’) or ‘hot’ (with re-experience of the upset). The distinction between ‘hot’ and ‘cold’ processing of emotion-related material is clearly centrally important in cognitive therapy, and has similarities to the problem of ‘intellectual’ vs ‘emotional’ belief that we touched on earlier. In the Bower model, both ‘hot’ and ‘cold’ processing depend on the same underlying representations. Consequently, use of emotion-related concepts or remembering emotional events would be expected to intensify emotion. Cleariy this does not inevitably occur. The clarity with which Bower’s associative network theory is specified allows us to find out that it is wrong in important respects. The power of scientific theories is that, through identifying the deficiencies of existing accounts relatively precisely, we are in a good position to describe what will have to be incorporated in a new and better theory. In the present case, we know that mood effects will probably not be satisfactorily explained by simple models, with only one form of cognitive representation, in which mood directly primes representations of concepts and events. Let us now consider an alternative framework that avoids these difficulties. THE INTERACTING




The Interacting Cognitive Subsystems (ICS) framework was first described by Barnard (1985). It is a comprehensive framework, shaped by the findings of cognitive psychology and cognitive science, within which, in principle, accounts of all aspects of information processing can be developed. It has only recently been extended to the interaction of cognition and emotion {Barnard & Teasdale, 1991; Teasdale & Barnard, 1993). ICS proposes nine qualitatively distinct types of information, or mental codes, each representing a different aspect of experience. Each type of information is processed by its own speciaiised processing subsystem. There are separate memory stores for each type of information, making nine distinct memory systems in all. Information processing involves the transfer of information between subsystems and its transformation from one mental code to another. ICS proposes mental codes related to two levels of meaning, a specific and a more generic level. The incorporation of two levels of meaning into the framework was not initially motivated by any concern to understand cognition-emotion relationships. Rather, this distinction was simply a

Two kinds of meaning

and cognitive



necessity if one wanted to give an adequate explanation of many aspects of information processing in general. Patterns of Propositional code represent specific meanings in terms of discrete concepts and the relationships between them, for example, the specific meaning behind the speech form ‘Roger has brown hair’. Meaning at this level can be grasped relatively easily as there is a fairly direct relationship between language and concepts at this level. The meanings represented in semantic networks, such as Bower’s theory of mood and memory, are at this level, consisting of specific constructs and the relationships between them. Patterns of Implicational code represent a more generic, holistic, level of meaning. Meaning at this level is difficult to convey adequately because it does not map directly onto language. This generic level of representation encodes recurring very high order regularities across all other information codes. ICS proposes that only this generic level of meaning is directly linked to emotion. ICS suggest that, subjectively, synthesis of generic meanings is marked by experience of particular holistic ‘senses’ or ‘feelings’ with implicit meaning content: ‘something wrong’, ‘confidence’, ‘on the right track’, ‘hopelessness’. Three important features of Implicational level meaning have to be grasped. First, representation is at a very high level of abstraction. So, hearing the Nasrudin tale about the donkey, the generic Implicational meaning that we extract is beyond the specific referents of donkeys, Mullas, and food, and at a more thematic level. At this thematic level, we can recognise potential parallels between the fate of Nasrudin’s donkey and the current state of cognitive approaches to depression, although, of course, these two domains have very little overlap in terms of specific content. Second, sensory features, such as tone of voice or proprioceptive feedback from facial expression or bodily arousal, make a direct contribution to Implicational meanings, together with patterns of specific meanings. So, the higher order meanings we derive from the specific meanings conveyed in what someone is telling us may be directly influenced by whether their voice tone is tense and strained, rather than warm and relaxed, and whether, at a bodily level, we are calm and alert rather than tired and uncomfortable. Third, the implicit knowledge encoded in coherent patterns of Implicational code can be thought of as representing schematic models of experience. Mental models, in general, represent the inter-relationships between semantic elements (Johnson-Laird, 1983). These schematic models represent inter-relationships between generic features of experience, capturing very high level recurring regularities in the world, the body and ‘the mind’. These models are about what goes with what at a high level of abstraction. At this level, communalities can exist between Mullas and donkeys, on the one hand, and cognitive therapy and science, on the other. The knowledge in schematic models is implicit, rather than explicit. Presented with the text: “John knocked the glass off the table. Mary went to the kitchen to fetch a broom”, the [‘brokeness’] schematic model will be synthesised. This encodes all the high order recurring regularities that have previously been extracted from experiences involving breakage of fragile objects. The implicit knowledge inherent in this schematic model allows us to immediately infer that the glass was broken, without recourse to any explicit rules such as ‘if something is knocked, and someone goes for a cleaning implement, then the thing must be broken’. Unlike lower level meanings, high level Implicational meanings cannot be communicated by single sentences. Traditionally, such meanings have been communicated by poems, parables and stories, such as those in the Mulla Nasrudin tradition. Some ‘feel’ for representations at the Implicational level can be gained by considering the analogy between a sentence and a poem. A sentence conveys one or more specific meanings by appropriate arrangements of letters or phonemes in the appropriate sequence. A poem conveys ‘holistic’ meanings, that cannot be conveyed by single sentences, by arranging sentences in appropriate sequences, together, very importantly, with appropriate direct sensory contributions from the sounds of the words, the rhythms and metres of the whole, and from the visual imagery elicited. The imagery also facilitates simultaneous ‘parallel’ communication of clusters of specific meanings more readily than the normal ‘serial’ presentation of linear sequences of specific meanings in the sentences of prose.

JOHN D. TEASDALE Table I. Poetry as Implicational Meaning. The origmal poem in the upper part of the table and the alternative version in the lower part have the same sequence of proposiuonal meanings. However, only the original version conveys a coherent Implicational ‘sense’. [From Tea&k and Barnard (1993)] ‘0 what can ail thee, knight-at-arms. Alone and palely toitcring? Tbc sedge has wither’d from the lake. And no birds sing.’

‘What is the matter. armed old-fashioned soldier. Standing by yourself and doing nothing with a pallid expression? The reed-like plants have decomposed by the lake. And there are not any birds singing.’

The total meaning conveyed by a poem is qualitatively different from the sum of the separate specific meanings, just as the meaning of a sentence is qualitatively different from that of its component letters or words. This is illustrated in Table 1, where an extract of a poem is presented. The holistic meaning created by the poetry is marked by a ‘sense’ of melancholy and abandonment. Table 1 also includes a ‘prose’ version which retains the same sequence of specific level meanings, but the ways in which they are expressed lack the coherence, evocative sound qualities and imagery of the poem. The total effect of the prose version is quite different! Table 1 can also be used to illustrate the ‘implicit’ knowledge inherent in the schematic models created by the poem. If one reads the poetry aloud, gets the ‘sense’ that it conveys, and then answers the question: ‘Would he be fun to meet at a party?‘, one can answer the question very directly and immediately by consulting the implicit knowledge of the schematic model constructed. By contrast, if one reads the ‘prose’ version and then answers the question, most likely one will have to do this piecemeal, considering each of the propositions in turn and arriving at a judgement more slowly and ‘rationally’. ICS



The ICS approach assumes that emotional reactions originate as innate prepared biological responses to certain arrays of sensory stimuli. In a given family and culture, these prepared sensory stimuli will regularly occur concurrently with constellations of other, unprepared, sensory stimuli, and with specific level meanings that the child derives as it develops cognitively. Features of information codes, both sensory and semantic, that regularly co-occur in situations eliciting a given emotion will be extracted and represented in patterns of Implicational code. These will capture the ‘prototypical’ features of emotion-eliciting situations in related schematic models, Such emotionrelated Implicational patterns ‘inherit’ the ability to elicit emotion. In the adult, the core of emotion production is the synthesis of patterns of Implicational code extracted as prototypical of previous experiences of a given emotion. When such patterns are processed, emotion is produced. Within ICS, the ability directly to elicit emotion is restricted to emotion-related Implicational codes; on this view, Propositional representations of emotion-related information cannot, alone, elicit emotion. Thus, processing of such specific Propositional representations would be associated with ‘cold’ consideration of emotion-related material. ‘Hot’ processing of emotion-related material depends on the synthesis of appropriate generic level meanings in Implicational code. Relatedly, ‘intellectual’ belief or ‘knowing with the head’ is equated with agreement or disagreement with specific Propositional meanings, whereas ‘emotional’ or ‘intuitive’ belief, ‘knowing with the heart’, is related to the state of holistic Implicational representations. Having briefly outlined aspects of the ICS approach [see Barnard and Teasdale (1991) and Teasdale and Barnard (1993) for a fuller presentation], let us now consider its implications for understanding and treating depression. ICS





Much depressive thinking is mood-dependent, most measures of negative thinking returning to normal levels as depressed mood remits. By explicitly recognising the distinction between specific

Two kinds

of meaning

and cognitive



and higher level meanings, the ICS approach forces us to confront the question: At what level of representation do these mood-related cognitive biases primarily arise? As we have already noted, Bower’s associative network model suggests that depressed patients think more negatively because their depressed mood selectively increases the activation or accessiblity of specific negative interpretative constructs, representations of negative specific meanings, and memories of negative events previously associated with depressed mood. By contrast, the ICS account of depressive thinking suggests that mood biases arise, not primarily at the level of specrjic constructs, but at the generic level of schematic models of experience. According to ICS, the maintenance of depression depends on the regeneration of depression-related Implicational schematic models. These models output biased lower level meanings in the form of negative evaluations, attributions, instructions to access negative material in memory, etc. The bias in these specific meanings, reflecting the schematic models from which they were derived, leads to negative depressive thinking. ICS and the associative network model offer contrasting accounts of the level of representation at which mood-related cognitive biases primarily arise. Let us focus on the accounts offered by these two approaches to explain the reduction in the negative thinking shown by depressed patients as their mood improves with recovery. The associative network model suggests that thinking becomes less negative as a result of changes in the accessibilty or activation of representations at the level of specific constructs or clusters of constructs. The ICS analysis, on the other hand, suggests that thinking becomes less negative as depressed mood remits as a result of shifts in the prevailing schematic mental models dominating information processing. The focus of such schematic models is the inter-relationship between constellations of constructs (and sensory elements), rather than simply the level of activation of individual constructs. On the ICS account, change in mood is accompanied by a shift in schematic models, the implicit knowledge of what leads to what. Often changes at the schematic model level will be reflected in congruous changes at the level of specific constructs. However, because schematic models and specific constructs are at different levels of abstraction, it is possible to arrange situations in which effects at these two levels operate in opposite directions. By exploiting such a situation, Teasdale, Taylor, Cooper, Hayhurst and Paykel (1993) were able to pit predictions from Bower’s associative network model and from ICS against each other. The aim of this study was to contrast the predictions emerging from the two accounts with respect to the increased endorsement by depressed patients of dysfuctional attitudes that are likely to perpetuate depression. As patients recover, endorsement of these attitudes decreases so that, when they have returned to normal mood, previously depressed Ss endorse dysfunctional attitudes no more than normal controls (Teasdale, 1988). Let us consider how the two rival accounts explain the fact that a depressed person will endorse a dysfunctional attitude such as ‘If a person I love does not love me it means I am unlovable’ when they are depressed, but not when they are recovered. The Bower network model suggests that endorsement when depressed is primarily a reflection of a general increase in activation and accessibility of all the negative constructs, such as ‘unlovable self, that have previously been associated with the depressed state. As mood recovers, activation to this construct from the depressed state will reduce and so endorsement of the dysfunctional attitude is less likely. By contrast, the ICS account suggests that the shift in endorsement of the dysfunctional attitude reflects a shift in the prevailing high order mental models of self and world dominating information processing as one moves from one mood state to another. The focus of such schematic models is the inter-relationship between constellations of constructs, rather than simply the level of activation of individual constructs. On this account, change in mood is accompanied by a shift in schematic models, the implicit knowledge of what leads to what. The implications and consequences predicted from a given state of affairs by the model prevailing in one mood state might be quite different from the consequences predicted from the same state of affairs by the model prevailing in another mood state. According to this account, the depressed state in these patients is associated with dysfunctional models, whereas the non-depressed state is associated with functional models. Changes in endorsement of the item ‘If a person I love does not love me it means I am unlovable’ from the depressed to the non-depressed state reflect the shift from a dysfunctional to a functional



2. Changes

recovery change

from from

in sentence

depressed positive


to normal





If I could

is a recipe always

would For


and advice Always people


to pun others’

m= to look















to me for guidance



the approval

is the road




be right


reduces. completions

f 199311

33 interests

by an individual

As depression

to negative










of other


model, with associated changes in the consequences and implications anticipated from a given state of affairs. Dysfunctional attitudes are typically measured with Weissman and Beck’s (1978) Dysfunctional Attitude Scale (DAS). In Teasdale et al.‘s (1993) study, selected DAS items were changed to a sentence completion format, the S having to complete each sentence stem with the first word that came to mind. The Bower model clearly predicts that, other things being equal, Ss in depressed mood will make negative completions to these stems, but more positive completions once they have returned to normal mood. By carefully rewriting DAS items a situation was created in which the dysfunctional schematic models suggested by the ICS analysis would lead to completion of sentence stems by @rive constructs. Conversely, functional models would be expressed by statements which required completion by negarive constructs. For example, dysfunctional completions of the stem ‘Always ’ would involve positive constructs to put others’ interests before your own is a recipe for such as ‘happiness’, whereas more functional completions would involve negative constructs such as ‘disaster’. in this way, a situation was created in which the ICS approach, focusing on the high order inter-relationships between patterns of constructs, makes the counterintuitive prediction that depressed patients should make more positive completions than the non-depressed. With recovery, there should be a decrease in the number of positiue completions. By contrast, the associative network approach, focusing on the activation of specific constructs, predicts that depressed patients will be more likely to make negative completions, and that this tendency will decrease with recovery. These predictions were contrasted in a study in which patients with major depression were tested initially while depressed, and, again, 3 months later, when the majority were considerably less depressed. Consistent with the predictions from the ICS account, depressed patients initially made signi~cantly more positive completions than non-depressed controls. Patients were divided into those whose mood had improved over the 3 months follow-up, and those whose mood had stayed the same or deteriorated. As predicted by the ICS account, patients who were less depressed showed a decrease in positive completions, whereas those who were unimproved showed an increuse in positive completions, the changes in the two groups being significantly different. Table 2 gives an example of the changes in one patient that illustrate this point particularly well. It does, indeed, seem that as mood shifts, so do our schematic models of self and world-in different moods we inhabit, mentally, different ‘alternative realities’. Results of this study support the proposal that affect-related biases in information processing are better understood in terms of effects at the generic schematic level rather than at the fevel of the activation of specific constructs. Encouraged by this support, let us now turn to consider the implications of the ICS approach for psychologica1 treatment. IMPLICATIONS






Treatmenr targeis ICS and the clinical cognitive model suggest quite different relationships between negative automatic thoughts and depression (Fig, 2). The clinical cognitive model suggests that negative thoughts and images, or associated meanings, are the immediate antecedents to depression.

Two kinds of meaning and cognitive




Schematic models - generic meanings


anitlve Subsvm Fig. 2. The role of thoughts and images in emotion production: contrasting positions from Beck’s cognitive model (upper panel) and Interacting Cognitive Subsystems (ICS) (lower panel) [from Teasdale and Barnard (1993)].

ICS, by contrast, suggests that the immediate antecedent to depressed emotion is the processing of the higher level meanings encoded in depression-related Implicational schematic models. From this perspective, verbalisable negative automatic thoughts, negative visual images, and negative specific meanings are not immediate antecedents to depression and have no direct effect on thematically related depression. Depression-related schematic models may produce, ‘downline’, specific meanings, and negative verbalisable thoughts or images. This may give the impression that such thoughts and images are antecedent to depression. However, according to ICS, this apparent causal relationship is illusory. Reflecting the differences in their views on the immediate antecedents to depression, ICS and the clinical cognitive model offer contrasting ‘prescriptions’ for the central goal of therapy. The clinical cognitive model suggests that negative thoughts or images are the antecedents to depression and so should be the target of treatment. Cognitive therapy, of course, aims to do just this, with many procedures designed to invalidate, discredit or disprove negative automatic thoughts. By contrast, the ICS analysis suggests that the immediate antecedents to depression are depressogenic higher level Implicational meanings, or schematic models, extracted as prototypical of previous depressing situations. From this analysis, the central goal of therapy should be to replace Implicational code patterns related to depressive schematic models with alternative patterns related to more adaptive higher level meanings or schematic models. On the ICS view, there may be a place for modifying thoughts and images but this would only be if, in doing so, one achieves change at the level of higher order meanings. This could occur in a number of ways. First, it may be that the procedures that successfully modify negative thoughts also modify the ‘parent’ schematic models from which they were derived, the thoughts and images acting as a useful ‘marker’ of the state of the target of ultimate interest. Second, it could be that the thoughts generated from one depressogenic schematic model become important components contributing to the subsequent synthesis of the same, or a similar depressogenic schematic model, and in that way act to maintain depression (see Fig. 2). Thirdly, it may be that the very action of attempting to deal with negative thoughts, in common with other active coping procedures, leads to the synthesis of schematic models related to ‘taking control’. These would replace the schematic models related to themes of helplessness and hopelessness that would otherwise maintain depression. Cognitive therapy, in common with a range of other ‘non-cognitive’ psychological treatments for depression, includes features such as the provision of a credible treatment rationale,



high treatment structure. training in coping responses, homework assignments, feedback on progress, etc. At least some of the effectiveness of such treatments can be attributed to the effects of these ‘non-specific’ factors creating ‘mastery/problem-focused’ schematic models in place of ‘helpless/hopeless’ models (see Teasdale & Barnard, 1993, Chap. 16. for a fuller discussion of these issues). Meaning

and therapeutic


We shall return to considering how one might change patterns of Implicational code related to depressive schematic models in a moment, but first we must consider meaning within the clinical cognitive model. This model suggests that not only negative automatic thoughts but the meanings attached to events mediate emotional response. The ICS framework immediately prompts us to ask-‘which level of meaning, specific, generic, or both?’ First, we should note that the clinical cognitive model explicitly rejects any distinction between kinds of meaning. So, in discussing the question of ‘emotional’ vs ‘intellectual’ belief the standard text on the cognitive therapy of depression advises: “Patients often confuse the terms ‘thinking’ and ‘feeling’ . . _ the therapist can tell the patient that a person cannot believe anything ‘emotionally’ . . . when the patient says he believes or does not believe something emotionally he is talking about degree of belief” (Beck et al., 1979, p. 302, original italics). The clinical cognitive model suggests that we are dealing simply with quantitative variations of a single level of meaning, rather than the qualitative distinction between two kinds of meaning that the ICS analysis proposes. Further, it is clear that the level of meaning that the clinical model has in mind is specific-propositional statements that have a truth value that can be determined. The goal of therapy is then to invalidate the truth value of specific depressive meanings. This aim is reflected in cognitive therapy’s use of procedures such as the collection of evidence for and against beliefs, hypothesis testing, analogies with the investigative procedures of scientists or courts of law, Socratic dialogue, syllogistic reasoning, etc. This focus on specific meanings can lead to a ‘search and destroy’ approach in which the therapeutic task narrows down to that of identifying and discrediting negative automatic thoughts. As we noted earlier, the contrast between ‘intellectual’ and ‘emotional’ belief, ‘knowing with the head’ vs ‘knowing with the heart’. poses recurring problems for clinicians. The solution to this problem offered by the clinical cognitive model does not seem wholly satifactory. The distinction between these two kinds of knowing is not a difticulty for ICS. ICS suggests that these forms of meaning and knowing correspond to qwlitaticely distinct types of representation, respectively, the Propositional and Implicational levels. In contrast to the clinical cognitive model, the ICS analysis attaches primacy to Implicational level meaning as the antecedent to emotion production. This focus on generic, holistic, meaning as the primary target of therapy within the ICS analysis suggests that the therapist should be sensitive to the total information context, even when attempting to shift higher order meanings by changing elements related to specific meanings. Changing



Implicational schematic models, or higher order meanings, represent inter-relationships between generic features of experience. Such models are composed of informational elements, or variables, each element corresponding to a high-order feature, or dimension, extracted from experience. These elements can take different values, corresponding to the current state of the underlying dimensions into which experience is ‘parsed’. We can think of the high order meanings of Implicational schematic models in terms of patterns of values across variables. Figure 3 uses the convention of representing a schematic model by such a pattern of values (arbitarily assigned 0 and I in the figure), the pattern of values across variables defining the nature of the schematic model. The figure illustrates how, when the pattern is appropriate, an emotional response is produced, in the present case depression. It also reminds us that the total pattern receives contributions not only from elements related to patterns of lower order specific meanings, but also, most importantly, from elements derived directly from sensory elements such as those related to voice tone and proprioceptive feedback from facial expression, posture, or bodily arousal. It is the total pattern that determines high level meaning and emotional response. Just as making small changes to the sequence of letters making up a sentence can radically alter the specific meaning it represents, so changing just a small portion of a total pattern of


kinds of meaning and cognitive



EMOTION (if right pattern) Coherent pattern of Implicational code carrying Implicational meaning




10101110 /-T-

Elements derived from atterns of SF&C MEANINGS -focal content, wider semantic context

Elements derived from atterns of BOD!’ STATE sensory inputfacial expression, posture, arousal

11001100 -T-T

Elements derived from patterns of ACOUSTIC sensory inputvoice tone,volume

Elements derived from patterns of VISUAL sensory inputambient illumination, rate of visual flow

Fig. 3. A diagrammatic representation of an emotion-related Implicational schematic model, showing contributions from inputs derived both from patterns of specific meanings and from patterns of sensory information. The schematic model is represented as a pattern of values (0 or I) across a range of variables, each variable corresponding to a higher order dimension of experience.

code may be sufficient to alter radically the high level meaning represented, and so to change the emotional response. It follows that the ICS analysis suggests that it may well be therapeutically useful to make changes to one or more specific meanings if they alter the total pattern of Implicational code sufficiently to create a new higher level meaning. For example, for someone who has synthesised the high level meaning [‘self-as-a-total-failure’] following failure on an examination, helping them discover that 95% of other candidates also failed may create changes in the related parts of the total Implicational pattern that, although limited, are sufficient to create a radically different higher order meaning. Equally, and in complete contrast to the cognitive model, the ICS analysis suggests that high level meanings can be altered by modifying purely sensory elements. The higher order meaning that I create following a failure experience may be quite different if I combine the specific failure-related meanings with the elements related to the sensory feedback from a smiling facial expression rather than a frown, or from high bodily arousal and an erect posture rather than from a sluggish body-state and bowed, stooped posture. The ICS analysis suggests a very useful place for purely physical interventions, such as training in maintaining a half-smiling expression, or vigorous physical exercise, in the cognifiue therapy of depression. There may even be a place for that traditional behaviour therapy procedure, muscle relaxation. Often, it may be necessary to alter more than single elements of depression-related Implicational code patterns in order to transform depressogenic patterns into those related to more adaptive high order meanings. Interventions may be needed that create alternative, coherent, ‘packages’ of semantically-and sensorily-derived elements. These interventions may have zero value as evidence as far as evaluating the truth value of specific propositional meanings and, for that reason, they cannot be rationally included in orthodox cognitive therapy. Guided imagery provides an example of such an intervention. This procedure has recently been imported into the repertoire of cognitive therapy from Gestalt therapy (e.g. Edwards, 1990) and is frequently used, for example, in the treatment of victims of early sexual abuse: in imagery clients relive scenes of childhood abuse but introduce into the images the elements of control and power that they now have as adults but lacked as children, at the time of the abuse. Such imagery clearly has no worth as evidence for refuting specific propositional meanings. However, from the ICS perspective, it can be seen as potentially a very powerful way of introducing new elements into a pattern of Implicational code to create new coherent higher order meanings that allow clients to free themselves from the domination of dysfunctional childhood schematic models. In this, and similar ways, the ICS analysis, as well as sensitising us to a wider information context, suggests a use and need for the incorporation of procedures, such as guided imagery, that have no rational value within the clinical Implicational



cognitive model. Interventions primarily targeted on sensory elements, such as physical exercise, manipulations of facial expression or music, would clearly also fall in this category.

The therapeutic goal suggested by the clinical cognitive model is the identification and systematic invalidation of specific meanings. By contrast, ICS suggests as a therapeutic goal the substitution of an adaptive alternative schematic model for the one currently maintaining depression; at this level, questions of truth and falsehood are not so obviously relevant as at the lower level. The difference in approach can be illustrated by an example from the standard text on cognitive therapy for depression (Beck et al., 1979, p. 153). A depressed student expressed the beiief that she would not get into college. Acting on the assumption of the clinical cognitive mode1 that this belief maintained her depression, the therapist explored the basis for her belief: Therapist: Patient: Therapist: Patient: Therapist: Patient: Therapist: Patient: Therapist: Patient: Therapist: Patient: Therapist: Patient: Therapist: Patient:

Why do you think you won’t be able to get into the university of your choice? Because my grades were really not so hot. Well, what was your grade average? Well, pretty good up until the last semester in high school. What was your grade average in general? As and Bs. Well, how many of each? Well, I guess, almost all of my grades were As, but I got terrible grades my last semester. What were your grades then? I got two As and two Bs. Since your grade average would seem to me to come out to almost all As, why do you think you won’t be able to get into the university? Because of competition being so tough. Have you found out what the average grades are for admissions to the college? Well, somebody told me that a B-f- average would suffice. Isn’t your average better than that? I guess so.

The therapist’s aim, guided by the clinical cognitive model, was clearly to invalidate the client’s negative belief, and so to reduce the depression assumed to be produced by this belief. The transcript suggests that the interchange produced ‘intellectual assent’, but one is left wondering how much emotional shift was actually achieved. By contrast, the ICS view suggests that the student’s depression was maintained by a schematic model l‘myself as incompetent’]. The thought ‘I won’t be able to get into college’ would be seen as a product of this schematic model. The thought would not be seen as making any direct contribution to the production of depression, although it might do so indirectly by contributing to the regeneration of the ‘parent’ schematic model. From this view, attacking the reasoning leading to the thought by looking at evidence inconsistent with the specific belief would be seen as therapeutically relevant and effective only if, in doing so, the synthesis of more adaptive schematic models were facilitated. Indeed, the ICS view suggests that, depending on wider aspects of the client-therapist relationship, the therapist’s attempts to demonstrate the illogicality of the client’s thinking could actually reinforce the [‘myself as incompetent’] schematic model and so increase, rather than reduce, depression. The therapeutic strategy suggested by ICS is to foster alternative non-depressogenic schematic models consistent with the available evidence. In the present example, this alternative might be of the form f’myself as competent but with doubts in my competence’]. Interestingly, the most dramatically successful applications of cognitive therapy, to panic and hypochondriasis (Clark & Salkovskis, 1991) actually adopt this strategy of fostering, from the beginning of therapy, the creation of alternative functional schematic models rather than focusing primarily on invalidating specific negative beliefs (D. M. Clark, Personal Communication, July, 1991).


Two kinds of meaning and cognitive therapy

Skilled and sensitive clinicians are ‘shaped up’ by the realities of clinical experience to adopt, often implicitly, ‘working’ theories that may not coincide exactly with explicit statements of the clinical cognitive model. The ICS analysis provides an explicit formulation of ‘implicit’ theories that are probably shared by many experienced cognitive therapists. The advantage of an explicit articulation of underlying theoretical views is that it may decrease the variance between therapists in the way cognitive therapy is implemented. Most importantly, it may also provide a firm foundation for further improvements in treatment. By explicitly identifying the creation and maintenance of alternative schematic models as A= central therapeutic process in the psychological treatment of depression, the ICS analysis points to the need to design treatment interventions with the deliberate aim of creating such models. A rational role for experiential


By placing central importance on an holistic level of implicit meaning, marked subjectively by ‘senses’ or ‘feelings’, the ICS analysis immediately sensitises us to the possibilities of incorporating procedures and interventions from more experientially oriented therapies that explicitly focus on change at this level. It comes as no surprise from the ICS perspective that major recent importations into modified cognitive therapy have come from Gestalt therapy, which, like ICS, emphasises the importance of holistic levels of representation, the role of body state, and of the wider semantic context. However, ICS does more than legitimise this ad hoc borrowing from other therapeutic approaches. It provides a detailed, explicit, comprehensive information-processing framework, operating in accordance with strict principles, within which what is treated in an intuitive, often anti-intellectual, manner in experiential therapies can be conceptualised as simply another, albeit very complex, form of information processing. In this way, ICS opens the possibility of formulating detailed, explicit, information-processing accounts of the central therapeutic processes of these, often otherwise slightly mysterious, approaches. When we begin to apply the investigative strategies developed within cognitive psychology to exploring and evaluating these accounts, as is already possible to some extent, there is a real prospect that we will be able to identify and measure the state of key schematic models. We can then assess the effects of a range of tightly described and controlled interventions on these models. This strategy offers the prospect of identifying the centrally effective components of procedures that produce change at the level of higher order meaning, and then applying this knowledge to the design of new treatments that achieve these effects with maximum effectiveness and efficiency. In the case of behaviour therapy of phobia, this strategy of identifying centrally effective therapeutic processes and then incorporating them into highly efficient formats of treament delivery produced enormous improvements in the cost-effectiveness of treatment (Mathews, Gelder & Johnston, 1983). Table 3 summarises the therapeutic implications of the ICS analysis that we have discussed. CONCLUSIONS The clinical cognitive model, relying on ‘everyday’ conceptions of cognition and meaning, has delivered therapies of impressive and exciting effectiveness. It is unlikely that cognitive therapy

Table I.



(a) if the ‘parent

3. Therapeutic


(b) if the thoughts/images (c) as part 2.




of a total




of the Interactinn





and images may be useful: model

is also alTaxed;


package creating





than should







to the regcncrntion





bc to create alternative

of affect-related





text. bc the primary




than ‘search

of therapy. and destroy’

negative automatic

thoughts. 4.

Thcrapcutic meanings

interventions related

cifics’-structure. 5.


is a ra~mnal



(e.g. music) 6.








etc.) and of sensory


basis for using manipulations

and ‘set shifts’

more entxricntial


to the focal problem.

attention thcraoics.

interventions (e.g. modifying

(e.g. ‘taking IO ‘feelings’



is sensitive

of the wider aspeas-voice

that have no evidential facial expression.




to the rofol parrern



facial expression.

for invalidating



of information,

(thcrapist-clicn~ posture.



not just

relationship. bodily

propositional expression).




meanings-guided purely



control‘). with






and therapeutically.

and so otTen

a link





could have achieved its ‘meteoric rise’ in any other way. However, it is clear that ‘more of the same’ will not maintain this exhilarating rate of progress. I have illustrated how, by going back to relevant basic science, we can sharpen our concepts, resolve current difficulties of the clinical cognitive model, and develop self-correcting theories. In this way, we can return to the task of developing psychological treatments equipped with a better unde~tanding that offers the prospect of innovations in therapy, linked to theories and investigative techniques developed in cognitive psychology and cognitive science. The successful outcome of this enterprise depends on the kind of productive interchange between basic science and clinical inspiration that characterised the initial development of behaviour therapy. if this relationship can be re-established, I am confident that we can look forward to a future for cognitive-behavioural therapy that includes both growth and progress. Acknowledgement-This paper is based on the author’s Keynote Address to the 25th Annual Convention of the Association for the Advancement of Behavior Therapy, New York City, November, 1991.

REFERENCES Barnard, P. (1985). Interacting cognitive subsystems: a psycholinguistic approach to short-term memory. In Ellis, A. (Ed,), Progress in the psychology of language, Vol. 2 (pp. 197-258). London: Lawrence Erlbaum Associates. Barnard, P. J. & Teasdale. J. D. (1991). Interacting cognitive subsystems: a systemic approach to cognitive-affective interaction and change. Cognition and Emorion, 5, I-39. Beck, A. T.. Rush, A. J.. Shaw, B. F. & Emery, G. (1979). Cognifirje therapy of depression. New York: Guiiford Press. Beck, A. T., Epstein, N. & Harrison, R. (1983). Cognitions, attitudes and personality dimensions in depression, British Journal of Cognitive Psychotherapy. I, I-16. Bower, G. H. (1981). Mood and memory. American Psychologist, 36. 129-148. Clark, D. M. & Salkovskis, P, M. (1991). Cognitive therapy with panic ond hypochondriasis. Oxford: Pergamon Press. Edwards, D. J. A. (1990). Cognitive therapy and the restructuring of early memories through guided imagery. Journal of Cognitive Psychotherapy: An International Quarterly, 4, 33-50. Ho!!on, S. D., Shelton, R. C. & Loosen, P. T. (1991). Cognitive therapy and pha~acothcrapy for depression. Journal of Consulting and Clinical Psychology, 5. 88-99. Imber, S. D.. Pilkonis. P. A., Sotsky, S. M., Elkin. I., Watkins. J. T.. Collins, J. F., Shea, T. M., Leber, W. R. & Glass, D. R. (1990). Mode-specific effects among three treatments for depression. Journalof Consuhing and Clinical Psychology. 58, 352-3.59. Ingram, R. E. (1984). Toward an information-processing analysis of depression. Cognitive Therapy and Research, 8, 443-478. Johnson-Laird, P. N. (1983). Menraf models. Cambridge: Cambridge University Press. Mathews, A. M., Celder, M. G. &Johnston, D. W. (1981). Agoraphobia-Nature andtreutment. New York: Guilford Press. Miranda, J. & Persons, J. 8. (1988). Dysfunctional attitudes are mood-state dependent. Journal of Abnormal Psychology, 97, 76-79. Miranda, J., Persons, J. B. & Byers, C. N. (1990). Endorsement of dysfunctional beliefs depends on current mood state. Journal of Abnormal Psychology, 99, 237-24 1. Rehm, L. P., Kaslow, N. J. & Rabin, A. S. (1987). Cognitive and behavioral targets in a self-control therapy program for depression. Journal of Consulting and Clinical Psychology, 5, W-67. Ross, A. (1991). Growth without progress. Contemporary Psychology, 36, 743-744. Shah. I. (1972). The exploits of the Incomparable Mulla Nasrudin. New York: E. P. Dutton. Simons. A. D., Garfield, S. L. & Murphy, G. E. (1984). The process of change in cognitive therapy and pharmacotherapy for depression: Changes in mood and cognition. Archives of General Psychiutry, 41, 45-51. Teasdale, J. D. (1983). Negative thinking in depression: cause, effect or reciprocal relationship? A&unees in Beh~i~~ur Research and Therapy, S, 3-25. Teasdale. J. D. (1988). Cognitive vulnerability to persistent depression. Cognition and Emotion. 2. 247-274. Teasdale, J. D. (1993). Selective effects of emotion on information processing. In Baddeley, A. D. and Weiskrantz, L. (Eds), Attention: SeJection, aN*areness and controt. A tribute to Donald Broadbent. Oxford: Oxford University Press. Teasdale, J. D. Br Barnard, P. J. (1993). Affecr, cognition, and change. Hove: Lawrence Erlbdum Associates. Teasdale, J. D. & Dent, J. (1987). Cognitive vulnerability to depression: an investigation of two hypotheses. British Journul of Clinical Psychology, 26. !!3-126. Teasdale, J. D., Taylor, M. J., Cooper. 2.. Hayhurst, H. & Payke!, E. (1993). Depressive thinking: shifts in construct accessibility or in schematic mental models? To be submitted. Weissman. A. & Beck, A. T. (1978). The Dysfunctional Attitudes Scale. Paper presented at the Annual Meeting of the Association for the Advancemenr of Behavior Therapy, Chicago. Williams. J. M. G. (1992). The psychological treatment depression(2nd Edn). London: Routledge. Williams, R. M. (1988). Individual differences in the effects of mood on cognition. Unpubiished D.Phi! Thesis, University of Oxford. Young, J. (1990). Cognitive therapy for personality disorder. Sarasota, FL: Professional Resource Exchange. Inc.