Annu. Rev. Psychol. 1996.

47:33–57
Copyright © 1996 by Annual Reviews Inc. All rights reserved

THEORETICAL FOUNDATIONS OF
COGNITIVE-BEHAVIOR THERAPY
FOR ANXIETY AND DEPRESSION
Access provided by Washington State University on 12/09/14. For personal use only.
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

Chris R. Brewin

Department of Psychology, Royal Holloway, University of London, Egham, Surrey

TW20 0EX, United Kingdom

KEY WORDS: appraisal, emotion, schema, conditioning, unconscious

ABSTRACT
Cognitive-behavior therapy (CBT) involves a highly diverse set of terms and
procedures. In this review, the origins of CBT are briefly considered, and an
integrative theoretical framework is proposed that (a) distinguishes therapy
interventions targeted at circumscribed disorders from those targeted at gener-
alized disorders and (b) distinguishes interventions aimed at modifying con-
scious beliefs and representations from those aimed at modifying unconscious
representations in memory. Interventions aimed at altering consciously accessi-
ble beliefs are related to their theoretical bases in appraisal theories of emotion
and cognitive theories of emotion and motivation. Interventions aimed at modi-
fying unconscious representations are related to their theoretical bases in learn-
ing theory and findings from experimental cognitive psychology. In the review,
different formulations of CBT for anxiety disorders and depression are analyzed
in terms of this framework, and theoretical issues relating to self-representations
in memory and to emotional processing are considered.

CONTENTS
INTRODUCTION..................................................................................................................... 34
HISTORICAL OVERVIEW ..................................................................................................... 34
A THEORETICAL FRAMEWORK......................................................................................... 37
PROCEDURES TARGETING CONSCIOUS BELIEFS AND REPRESENTATIONS ........ 39
Circumscribed Disorders .................................................................................................... 39
Generalized Disorders ........................................................................................................ 40

0066-4308/96/0201-0033$08.00 33
 34 BREWIN

PROCEDURES TARGETING UNCONSCIOUS REPRESENTATIONS ............................. 42

Circumscribed Disorders .................................................................................................... 42
Generalized Disorders ........................................................................................................ 43
CONCEPTUALIZING CBT FOR DEPRESSION................................................................... 45
CONCEPTUALIZING CBT FOR ANXIETY ......................................................................... 47
THEORETICAL ISSUES ......................................................................................................... 49
Mental Representations of Knowledge About the Self......................................................... 49
Emotional Processing.......................................................................................................... 51
SUMMARY .............................................................................................................................. 53

INTRODUCTION
Access provided by Washington State University on 12/09/14. For personal use only.
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

Cognitive-behavior therapy (CBT) for anxiety and depressive disorders is well

established as a promising and frequently effective treatment (e.g. Chambless
& Gillis 1993, Dobson 1989, Hollon et al 1993). CBT is a generic term
referring to therapies that incorporate both behavioral interventions (direct
attempts to reduce dysfunctional emotions and behavior by altering behavior)
and cognitive interventions (attempts to reduce dysfunctional emotions and
behavior by altering individual appraisals and thinking patterns). Both types of
intervention are based on the assumption that prior learning is currently having
maladaptive consequences, and that the purpose of therapy is to reduce distress
or unwanted behavior by undoing this learning or by providing new, more
adaptive learning experiences.
CBT practitioners believe that symptomatic change follows cognitive
change, this cognitive change being brought about by a variety of possible
interventions, including the practice of new behaviors, analysis of faulty think-
ing patterns, and the teaching of more adaptive self-talk. Although incidental
cognitive change can be brought about by a variety of interventions, including
pharmacological ones (Brewin 1985, Hollon et al 1987), evidence is starting to
mount that in behavior therapy (Bandura 1977) and in CBT itself improvement
is linked to a corresponding change in cognitions (Blackburn et al 1987,
Chambless & Gracely 1988, Clark et al 1994, DeRubeis et al 1990, Firth-Coz-
ens & Brewin 1988, Mattick & Peters 1988; see Chambless & Gillis 1993 for a
review). Subsequent maintenance of gains is also often related to cognitive
measures (Ba§o™lu et al 1994, Clark et al 1994, Rush et al 1986, Simons et al
1986). Despite this growing consensus, the actual mechanisms underlying
CBT are still poorly understood, and there is considerable disagreement about
exactly what has been learned and how change occurs.

HISTORICAL OVERVIEW
One of the most enduring debates about the mechanism of psychological
treatment for psychiatric conditions such as anxiety disorders concerns the role
in learning of verbal mediation. Traditionally, behaviorists argued that such
 COGNITIVE-BEHAVIOR THERAPY 35

disorders arose from a learned association between a feared stimulus and an

avoidance response and that such conditioned fear was unaffected by a per-
son’s conscious beliefs. Numerous observations, particularly of patients with
phobias or obsessive-compulsive disorders (OCD), confirmed the irrationality
of many fears and the ineffectiveness of treatments relying on simple persua-
sion. Anxious patients’ conscious beliefs and wishes indeed frequently ap-
peared largely irrelevant in the face of overwhelming feelings of fear and the
compulsion to avoid them.
These observations led most behaviorists to reject the suggestions put for-
Access provided by Washington State University on 12/09/14. For personal use only.

ward by cognitive therapists in the 1970s that conscious thoughts could them-
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

selves have an impact on feelings and behavior. But the cognitive therapists
were only voicing a generally felt dissatisfaction with the notion that people
were entirely at the mercy of their conditioning history and that individual
differences in the interpretation of a feared situation, in perceived control, or in
the ability to formulate plans and goals had no impact on the outcome of
treatment. Despite many empirical demonstrations that conscious beliefs did
sometimes influence feelings and behavior, and despite the development of
influential cognitive theories of emotion and motivation (e.g. Abramson et al
1978; Bandura 1977, 1986; Lazarus et al 1970; Weiner 1985), the lack of
theoretical overlap between the two approaches (i.e. the absence of an explicit
role for conditioning in cognitive therapies and the absence of a role for verbal
mediation in behavior therapies) led to a prolonged period of mutual denuncia-
tion and largely fruitless argument between the two groups of practitioners.
Many of these differences were gradually resolved at a pragmatic level. For
example, cognitive therapists were unwilling to abandon highly effective be-
havioral techniques, and often recommended an integration of both cognitive
and behavioral approaches. For instance, Bandura (1977) emphasized that
changing behavior was an extremely powerful way to change maladaptive
beliefs. Many behavior therapists equally were reassured by the retention of
some behavioral interventions and the proposal of highly structured cognitive
methods that were subject to empirical test, and they came to appreciate that
the broader range of cognitive techniques facilitated the treatment of different
kinds of disorder, such as depression, panic disorder, and generalized anxiety
disorder (GAD).
At a theoretical level, however, integration between cognitive and behav-
ioral approaches has been less successful. Two main reasons can be identified.
First, behavior-therapy techniques have been most effective with disorders
involving circumscribed active- or passive-avoidance problems, such as those
found in phobias, posttraumatic stress disorder (PTSD), OCD, or pathological
grief reactions. Cognitive therapy techniques, on the other hand, have been
successfully applied not only to circumscribed problems but also to more
generalized problems such as depression, GAD, and, more recently, personal-
 36 BREWIN

ity disorder, where the focus is on behavior patterns that occur in a variety of
situations. The different clinical foci of the two kinds of therapy have tended to
produce different sorts of theory.
The second reason integration is less successful at a theoretical level is that
the development of cognitive therapies was not closely tied to a single recog-
nizable strand of basic research and theory in psychology. As is often the case
in clinical practice, the development of effective therapies has preceded theo-
retical understanding. Different therapies use distinctive terms such as “be-
lief,” “assumption,” “attribution,” “expectation,” or “schema” and describe
Access provided by Washington State University on 12/09/14. For personal use only.

what they believe to be the effective ingredient in extremely varied terms

Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

(Brewin 1989). For example, Beck et al (1979) described one of the main
goals of cognitive therapy for depression as the changing of maladaptive
underlying assumptions or schemas. Whereas underlying assumptions were
described as conditional beliefs, exemplified in a statement such as “If I am
not successful, I can never be happy,” schemas were described as stable
cognitive structures that contain rules for screening and classifying informa-
tion. Beck et al, however, often used these terms interchangeably. More re-
cently, the term schema has also been used to refer to so-called core beliefs
such as “I am bad” (Beck et al 1990, Young 1990).
In distinguishing between cognitions that are relatively accessible to con-
sciousness, such as automatic thoughts, and so-called deeper cognitions that
are less accessible, such as schemas or underlying assumptions, Beck and his
colleagues were in agreement with other important theorists such as Ellis
(1962) and Meichenbaum (1977). But Beck’s usage of the term “schema,”
although very helpful to therapists, was not clearly specified compared with its
usage by cognitive psychologists (e.g. Segal 1988). To cite only one example,
Beck implied that the content of schemas consists of propositional knowledge
that is potentially accessible to conscious introspection. He did not discuss the
possibility that such knowledge might be more implicit or procedural rather
than declarative in form, i.e. that it might be neither consciously accessible nor
able to be completely captured by a verbal description. Cognitive psycholo-
gists, however, argue that people may have no direct access to schematic
knowledge, although a person can sometimes infer what it consists of by
monitoring the products of schematic processing (Berry & Broadbent 1984,
Brewin 1989, Teasdale & Barnard 1993).
Other influential cognitive theories such as the hopelessness theory of
depression (Abramson et al 1989) or the self-efficacy theory of behavior
change (Bandura 1977) did not distinguish between accessible and underlying
cognitions and were concerned only with the modification of conscious attri-
butions or expectations. Similarly, some general cognitive therapies, such as
problem-solving therapy (Goldfried & Goldfried 1980), and more specific
therapies for panic disorder (Clark & Beck 1988) referred simply to teaching
 COGNITIVE-BEHAVIOR THERAPY 37

new skills or to identifying and modifying faulty beliefs. Thus, the different
cognitive theories and therapies, let alone cognitive and behavior therapy,
clearly share no theoretical unity, and some attempt must first be made to
address this lack. The natural place to look to fulfill this need is in contempo-
rary theories of cognition and emotion.

A THEORETICAL FRAMEWORK
Progress toward theoretical resolution has been facilitated by three develop-
Access provided by Washington State University on 12/09/14. For personal use only.

ments: the integration of animal learning theory within more general ap-
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

proaches to information processing (e.g. Dickinson 1987), the general accep-

tance of the pervasive role of unconscious processing in everyday cognitive
operations (e.g. Power & Brewin 1991), and the development of theories that
recognize the ability both of conscious appraisal and of unconscious learned
associations to generate emotions and behavior (e.g. Ekman 1986, Leventhal
1984).
Cognitive and social psychologists (e.g. Berry & Broadbent 1984, Epstein
1994, Hasher & Zacks 1979, Nisbett & Wilson 1977, Posner & Snyder 1975,
Shiffrin & Schneider 1977) have proposed the existence of two cognitive
systems with different functions and properties, one that is automatic and
outside of awareness and involves large-scale parallel information processing,
and one that is more effortful and involves conscious experience. Data from
numerous areas of psychology support the idea that most cognitive processing
has the potential to include a large amount of information and takes place
extremely rapidly and completely outside of awareness. Although we are
unaware of this kind of processing, we can become aware of its products, for
example, in the forms of thoughts and images. This kind of processing is
heavily influenced by previous learning, and new stimuli tend to elicit routi-
nized responses in a relatively inflexible way.
In contrast, conscious processing is slow and deliberate and operates on a
tiny fraction of the information available. At the same time, it is extremely
adaptable and responsive to new information, which allows for great flexibility
in behavior. It is reasonable to suppose that the output of both types of processing
is represented in memory. Extensive experimental and neuropsychological
evidence suggests the existence of separate nondeclarative (implicit) and dec-
larative (explicit) memory systems, supporting the idea that much of the infor-
mation stored in memory may not be consciously accessible (see Squire et al
1993 for a review).
Contemporary theories of emotion (Berkowitz 1990, Leventhal 1984) and
conditioning (Davey 1992) adopt the view that emotional responses may be
influenced both by unconscious learned associations derived from intense or
repeated aversive events and by relevant conscious knowledge. Applying a
 38 BREWIN

similar approach to change processes in psychotherapy, Brewin (1989) pro-

posed that persons’ knowledge gained through the extensive unconscious par-
allel processing of their responses to aversive situations, so-called situationally
accessible knowledge, is stored separately from their knowledge gained
through the more limited conscious experience of such situations, so-called
verbally accessible knowledge. Whereas verbally accessible knowledge can in
principle be deliberately interrogated and retrieved, situationally accessible
knowledge can only be retrieved automatically when environmental input
matches features of the stored memories. Thus, in the presence of reminders of
Access provided by Washington State University on 12/09/14. For personal use only.

aversive events, a person might become aware of the automatic activation of

Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

emotions, thoughts, images, and behavioral impulses. While the underlying

representation remains inaccessible, its products become available to con-
sciousness and permit a person to make inferences about the material stored in
memory.
In this dual representation theory, both kinds of knowledge—verbally ac-
cessible and situationally accessible—can give rise to maladaptive emotions
and behavior. The difference between the two kinds is that when, for example,
emotions are based on verbally accessible knowledge, people know why they
are sad or afraid, whereas when they are based on situationally accessible
knowledge, people must infer indirectly why they feel a particular way. In
either case the recognition of feeling or behaving badly is followed by deliber-
ate, strategic attempts to understand what is happening and to put things right.
This involves a variety of well-described processes such as attribution of the
cause of the problem, generation of coping options and evaluation of their
likely success, and instigation of compensatory behavior (see Brewin 1989 for
a fuller description).
The concept of nondeclarative memory or situationally accessible knowl-
edge can readily include both the learned associations of conditioning theory
and other terms for deeper representations in memory such as emotional
memories (Lang 1979) and schemas (Beck et al 1979). Although in their
original writing Lang and Beck were not explicit about whether these memo-
ries were verbally accessible, careful reading suggests that both authors in-
tended them to be more than just records of conscious experience. Rather, their
discussion suggests that the information these memories contain tends to be
quite inclusive (sometimes including a record of physiological responses, for
instance) and is sometimes subject to considerable processing, for example, in
the aggregation of information from similar events. If the information stored
were the product of unconscious parallel processing, it would not be possible
to retrieve it into consciousness. Instead, a person could only know the con-
tents of the memory by inferring them through their awareness, for example,
that certain emotions or images were triggered in certain situations. More
recently Lang (1993) described emotional memories as possessing both a
 COGNITIVE-BEHAVIOR THERAPY 39

conscious, language-based meaning component and a more primitive network-

based representation of stimuli and responses that is connected to brain centers
for afferent and efferent processing.
This framework involving dual representations in memory lends itself read-
ily to the organization of different therapeutic procedures. In the remainder of
this review these procedures are divided into those that primarily attempt to
alter conscious verbally accessible knowledge and those that attempt to alter
deeper or situationally accessible knowledge. Within each group, procedures
are divided into those that address circumscribed as opposed to generalized
problems. This way of classifying procedures is illustrated in Table 1, which
Access provided by Washington State University on 12/09/14. For personal use only.
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

also provides examples of therapies primarily based on these different ap-

proaches. In the next section, this framework is related to controversies about
the mode of action of CBT for depression and anxiety disorders. Finally, the
review concludes with discussion of two current theoretical issues in CBT, the
nature of mental representations of the self and the relation of CBT to emo-
tional processing in general.

PROCEDURES TARGETING CONSCIOUS BELIEFS AND

REPRESENTATIONS

Circumscribed Disorders
The contention that certain psychological and psychosomatic conditions arise
from specific incorrect beliefs has been persuasively argued in recent years.
These beliefs are postulated to be consciously held and may be thought of as
misconceptions requiring an explicit psychoeducational approach. For exam-
ple, a person anticipating surgery may have faulty beliefs about the nature or
painfulness of the procedures and may be calmed when given the appropriate
information (e.g. Johnson et al 1978). Anxiety-related conditions such as
stammering, insomnia, and impotence may be exacerbated when a person
attributes them wrongly to personal inadequacy or regards them incorrectly as
indicative of serious pathology (Storms & McCaul 1976). Teasdale (1985)
proposed that depressed mood may be exacerbated by depression about de-

Table 1 Classification of therapies by primary target and scope of problem

Primary therapy target

Conscious representations Unconscious representations
Circumscribed problems CTa for panic Exposure therapy
Psychoeducation Response prevention
Generalized problems Problem-solving therapy CT for depression and GAD
Self-instruction training Schema-focused therapy

a
Cognitive therapy
 40 BREWIN

pression, that is, the belief that normal depressive reactions are pathological
and reveal the weakness or inadequacy of the sufferer. Education about the
nature of depression, he suggested, can bring about speedy improvement when
coupled with simple symptom-management techniques.
Similarly, individuals high in anxiety sensitivity fear that physical symp-
toms of anxiety have harmful or even catastrophic consequences (Reiss &
McNally 1985). Conscious beliefs about the significance of certain bodily
sensations are the primary target in cognitive therapy for panic (Clark & Beck
1988). According to the cognitive theory of panic (e.g. Clark 1988), panic
Access provided by Washington State University on 12/09/14. For personal use only.

arises from specific catastrophic misinterpretations of sensations such as

Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

tachycardia or dizziness. Therapy therefore involves a variety of techniques

aimed at eliciting and challenging these misinterpretations. For example, panic
may be elicited within a therapy session by having the patient read a series of
words connected with the feared outcome, graphically demonstrating to the
patient the role played by thoughts in producing panic.
These clinical approaches have their roots in appraisal theories of emotion
such as those of Lazarus et al (1970) and Weiner (1985). These theorists
emphasized the role played by conscious assessments of the meaning and
cause of internal and external stimuli and of the coping resources available to
deal with any prospective threat.
It should be emphasized, however, that verbally accessible knowledge of
this kind may also be subject to automatic processing, outside of awareness,
particularly when it is primed by a previous stimulus. It would indeed be
strange if conscious knowledge of danger were not processed as rapidly as
other kinds of knowledge. Thus, as noted by Clark (1988), catastrophic
thoughts may sometimes come to mind so rapidly that patients are barely, if at
all, aware of the interpretive process. This phenomenon has been used by
Clark to explain panic attacks triggered during sleep. Automatic processing
does not differentiate between verbally and situationally accessible knowl-
edge. It is the fact that the thought or belief can on other occasions be directly
consciously retrieved that indicates verbally accessible knowledge.

Generalized Disorders
Other disorders such as depression or GAD cannot be so readily traced to
specific faulty beliefs. Rather, there are often complex sets of negative beliefs
about the self or the external world that are activated in a variety of situations.
One approach to treating these generalized disorders is to add to patients’
conscious knowledge about their condition by teaching them a set of proce-
dures with which to counter negative mood states. In the section “Conceptual-
izing CBT for Depression” below, we discuss Barber & DeRubeis’s (1989)
view that this model can best account for the efficacy of CBT for depression.
 COGNITIVE-BEHAVIOR THERAPY 41

Early forms of CBT for anxiety adopted this approach of equipping patients
with generally applicable skills. For example, Meichenbaum (1977) empha-
sized how behavior is controlled at least partly by a person’s internal dialogue
or inner speech. In his stress-inoculation training, he drew patients’ attention
to their tendency to make defeatist or anxiety-provoking self-statements when
faced with difficult situations. Patients were trained to imagine themselves in
feared situations and to practice making positive, adaptive self-statements that
emphasized personal control and reduced the anticipated aversiveness of the
consequences. These responses were repeated until they became part of the
Access provided by Washington State University on 12/09/14. For personal use only.

patient’s knowledge base of how to cope with anxiety and replaced their
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

previous, less-adaptive repertoire of knowledge and skills.

Another successful approach involving the teaching of generic skills is
problem-solving training (D’Zurilla & Goldfried 1971, Goldfried & Goldfried
1980). Patients were first trained to recognize the existence of a problem,
assume a solution existed, and inhibit the temptation to act impulsively. They
were then taught to reduce the problem to a concrete and manageable form that
was susceptible to a solution. Following this, patients generated as many
solutions as possible, weighed the advantages and disadvantages of each one,
and selected the most promising option. After implementation, the efficacy of
the solution was carefully monitored, with a view to trying another solution if
the outcome was not satisfactory.
These skills-based approaches in CBT are supported by an enormous body
of research into cognitive theories of motivation and self-regulation. The idea
of individuals as active planners, problem-solvers, and self-motivators is ex-
tremely influential within the area of social cognition (e.g. Bandura 1986,
Cantor & Kihlstrom 1987, Carver & Scheier 1981, Karoly & Kanfer 1982,
Lazarus 1991, Markus & Wurf 1987, Mischel 1973, Weiner 1985). Among the
many processes involved in self-regulation, goal-setting, self-monitoring, the
activation and use of standards, self-evaluation, and self-reinforcement are
paid considerable attention.
As discussed in more detail elsewhere (Brewin 1988), the awareness of
unwanted emotions and behaviors is thought to generate a number of con-
scious subroutines. These are designed to label or classify the experience,
locate the relevant causal agents, assess severity and vulnerability, and gener-
ate and assess the likely efficacy and cost/benefit ratio of a number of coping
options. Having selected a course of action, such as avoidance, distraction, or
confrontation, and having decided how much effort or persistence to put into
it, individuals can then monitor the outcome against available standards and
can administer appropriate self-reinforcement. Many forms of CBT incorpo-
rate various kinds of skills training designed to inform and support patients’
coping behavior by reattributing problems to potentially controllable causes,
 42 BREWIN

increasing expectations of success, and adding to their conscious knowledge of

effective ways to reduce unwanted emotions.

PROCEDURES TARGETING UNCONSCIOUS

REPRESENTATIONS

Circumscribed Disorders
Specific phobias are among the most common circumscribed disorders. On the
Access provided by Washington State University on 12/09/14. For personal use only.

whole, the treatment of phobias has changed little and continues to include an
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

emphasis on helping individuals to undermine habits of cognitive or behav-

ioral avoidance and to achieve habituation to a feared stimulus. On a theoreti-
cal level, however, inadequacies in the original behaviorist account of the
treatment of phobias led to the adoption of more sophisticated cognitive ap-
proaches emphasizing the importance of factors such as attention and memory
(Dickinson 1987) and conscious revaluation of the unconditioned stimulus
(Davey 1992).
Lang (1979) proposed that the representations underlying phobias, which
he termed fear memories, contain three kinds of information: details of the
location and physical characteristics of the feared situation (stimulus ele-
ments); details of the verbal, physiological, and behavioral responses that
occurred in the situation (response elements); and an interpretation of the
stimulus and response elements and of their significance for the individual
(meaning elements). Drawing on Lang’s work, Foa & Kozak (1986) proposed
that therapy for phobias works by changing the information in the fear mem-
ory: First the memory must be activated, and then new experiences arranged
by the therapist (such as within- and between-session habituation to the feared
situation) are incorporated into the memory.
The idea that memories, once created, can be altered is a controversial one
within cognitive psychology. A popular alternative view is that memory con-
sists of a series of unalterable records with more or less overlapping features.
In other words, memory can be added to but not permanently changed (e.g.
Morton et al 1985). Consistent with this view, Brewin’s (1989) dual repre-
sentation theory contained the proposal that in therapy new memories are laid
down as situationally accessible knowledge. These memories are generally
arranged to be as similar as possible to the original fear memory in all respects
other than outcome. Patients are encouraged to expose themselves to as good
an example as possible of what they fear and to experience the same sensa-
tions. In therapy, however, this experience culminates in mastery and the
habituation of fear rather than in avoidance. Patients exposed to their feared
situations are subsequently more likely to access this new memory—a mem-
ory containing fewer fearful response elements and a more benign mean-
 COGNITIVE-BEHAVIOR THERAPY 43

ing—than they are to access the original memory containing more fearful
response elements and a more threatening meaning.
This formulation, while distinctly similar to that of Foa & Kozak (1986),
has some advantages over it. First, it is more parsimonious because it does not
postulate that fear memories must be activated for therapy to be effective. A
new memory of a therapy experience in which the patient did not become
anxious is not such a good match to the original fear memory and thus is less
likely to be activated when the originally feared stimulus is encountered in the
future. Second, this formulation helps to explain why someone may recover in
therapy only to have the fear return. If memories are not changed but are
Access provided by Washington State University on 12/09/14. For personal use only.
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

simply overwritten by new memories, relapse may occur when the patient
encounters a new situation more similar to the original fear memory than to the
new memory created in therapy. After all, the creation within therapy of new
situations that match perfectly the original learning experiences is rare. In fact,
the incidence of relapse and the return of negative feelings, even after appar-
ently successful therapy, are substantial (Rachman 1980), and thus a success-
ful theory must be able to explain both partial and total therapeutic successes.

Generalized Disorders
In conditions such as depression and GAD, the underlying representations are
thought to contain more abstract information than the specific memories un-
derlying phobias. These representations are thought to consist of summaries of
numerous aversive experiences produced by complex unconscious computa-
tions that abstract common meanings from repeated experiences. One obvious
source of repeated aversive events is poor parenting. Anxious and depressed
patients, in comparison to controls, are more likely to report having parents
who were less warm and affectionate; more punishing, rejecting, and, abusive;
less encouraging of autonomy; or more controlling and overprotective (e.g.
Brewin et al 1993).
Therapists are obviously going to find it much more difficult to create new
therapeutic experiences that share many features with such memories. To
begin with, there are many more relevant memories, and these may be very
disparate. The patients are much older than they were when the memories were
first laid down, and the therapist may be not at all like parental or authority
figures from the past. For this reason, treatment of these disorders generally
takes longer and is less successful than the treatment of simple phobias.
Brewin (1989) therefore suggested that, rather than create new memories,
cognitive therapy for depression attempts instead to limit the ease with which
these memories are activated by the current environment. One feature of
depression and GAD is that negative mood changes are elicited by a wide
range of stimuli. Patients respond to many relatively harmless situations as
though these situations contained enormous potential for various psychologi-
 44 BREWIN

cal and physical threats. The therapist infers the content of these repre-
sentations by systematically gathering data about the situations that elicit
anxiety or depression in a person and about the person’s reactions to these
situations. Data may be obtained from the patient’s own account, from the
accounts of family and friends, and by the therapist’s own observations of the
patient’s behavior in the therapy situation.
By drawing attention to the patient’s apparent assumptions and challenging
these with the use of logic and behavioral experimentation, the therapist can
help the patient develop new rules for discriminating situations that are truly
Access provided by Washington State University on 12/09/14. For personal use only.

threatening from those that merely arouse the feeling of being threatened. This
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

is similar to the process of construct elaboration described by Kelly (1955).

Practice in making these discriminations then changes the content of verbally
accessible knowledge so that previously threatening situations are reclassified
and automatic activation of the unconscious representations is decreased.
This activation-based model of cognitive therapy for depression has several
corollaries. First, because the relevant situationally accessible memories are
not overwritten and remain available to be reactivated, relapse is more likely
here than in the case of specific phobias. Second, the more varied the negative
experiences contributing to the formation of the memories (for example, the
more the patient experienced rejection at the hands of multiple caregivers), the
fewer distinctive stimulus features the memories contain. This is likely to
promote overgeneralization and impede the therapist’s ability to identify and
teach the relevant discriminations. Third, the model clarifies the potential role
of cues associated with nontherapeutic events to activate and deactivate uncon-
scious representations. In principle, relevant cues associated with any event
may activate or deactivate memories. Consistent with this, negative life events
are strongly associated with the onset of depression where there is prior low
self-esteem (Brown et al 1986), and positive events are strongly associated
with recovery from depression (Brown et al 1992).
A contrasting approach was offered by Teasdale & Barnard (1993). They
suggested that depression is frequently maintained by the continued reprocess-
ing of higher-level schematic models related to the experience of depression
itself. These representations contain idiosyncratic sensory, proprioceptive, and
meaning information synthesized from past experiences of depression and are
not able to be fully captured by a verbal description. In their view, models in
which the experience of depression is linked with personal inadequacy, help-
lessness, and hopelessness have to be replaced by alternative models empha-
sizing the normality of depressive symptoms and the value of positive coping
responses. This approach, then, addresses the problem of generalized repre-
sentations by asserting that, although many experiences and elements may
contribute, an essential core concerning the nature of depression is evident for
a particular individual. Psychoeducation, coping skills, and other CBT proce-
 COGNITIVE-BEHAVIOR THERAPY 45

dures all help to create new models of depression that in time supplant the old
maladaptive models.

CONCEPTUALIZING CBT FOR DEPRESSION

The fourfold classification of CBT procedures as targeting conscious and
unconscious representations across circumscribed and generalized disorders is
something of an oversimplification, although it does appear to capture the
differing focus of many cognitive techniques. It also clarifies some of the
Access provided by Washington State University on 12/09/14. For personal use only.

conceptual debates about the theoretical basis of cognitive treatment. Whereas

Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

the debate over CBT for anxiety has mainly concerned the role of behavioral
vs cognitive components, the debate over CBT for depression has focused on
the exact role played by the cognitive interventions.
As we noted above, Beck et al (1979) identified a number of techniques
(e.g. identifying and challenging automatic thoughts, behavioral assignments,
and correcting faulty reasoning) that they believed reduced depressed mood in
the short run and in the long run altered the content of underlying assumptions
and schemas. In contrast, Teasdale (1985) proposed that faulty beliefs about
symptoms (depression about depression) lead to the exacerbation of depres-
sion and can be corrected with a relatively simple educational intervention plus
training in symptom management.
Following a review of proposals put forward by Beck et al (1979), Hollon
& Kriss (1984), Teasdale (1985), and Ingram & Hollon (1986), Hollon et al
(1988) distinguished three mechanisms that could account for the effective-
ness of CBT. These were (a) a change in the content of depressive schemata or
in the cognitive processes underlying them, (b) a deactivation of the depressive
schemata, and (c) the inculcation of new cognitive or behavioral self-manage-
ment skills, such as learning to problem-solve, generate a variety of attribu-
tions for negative outcomes, and collect evidence germane to these attribu-
tions. Following a review of the evidence, Barber & DeRubeis (1989) sug-
gested that the primary mode of action was (c), the development of
compensatory skills, but that repeated application of these skills could lead to a
change in schemata.
Brewin’s (1989) dual representation theory identified two mechanisms in
CBT for depression, both aimed at preventing the continued reactivation of
situationally accessible memories. He suggested that the therapist’s attempts
directly to alter the contents of consciousness—for example, by challenging
automatic thoughts or instructing patients to distract themselves—disrupt the
feedback loop whereby upsetting (automatic) thoughts and images constantly
pervade consciousness, reactivate situational memories and maintain de-
pressed mood. Second, as described in more detail in the preceding section, he
suggested that the therapist attempts to alter on a more permanent basis the
 46 BREWIN

contents of verbally accessible knowledge in order to prevent the reactivation

of situational memories through inappropriate stimulus classification.
Finally, Teasdale & Barnard’s (1993) interacting cognitive subsystems the-
ory represents a conceptual advance on Teasdale’s (1985) ideas. The theory
suggests that the function of therapy is to disrupt the repeated synthesis of
high-level schematic models containing generic meanings prototypical of pre-
vious depressing situations (so-called depressive interlock). Interlock involves
numerous cognitive subsystems, which include those responsible for process-
ing sensory and proprioceptive data as well as those that extract meaning at
Access provided by Washington State University on 12/09/14. For personal use only.

propositional and higher-order (implicational) levels. Disruption of interlock

Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

may be achieved in a number of ways. For example, if the models were

continually being activated by ongoing stress, problem-solving training aimed
at resolving the stress might effectively deactivate them. Physical exercise
might alter a person’s body state and create a new element capable of disrupt-
ing the particular pattern of interlock. As discussed in the previous section,
however, Teasdale & Barnard proposed that the main effect of CBT is to
disrupt interlock by replacing maladaptive implicational models of depression
with more adaptive ones.
In terms of the theoretical framework outlined above, it is clear that Teas-
dale (1985) and Barber & DeRubeis (1989) emphasized targeting conscious
representations, correcting circumscribed faulty beliefs, and teaching generally
applicable coping skills and placed less emphasis on the significance of deeper
cognitions. Beck et al (1979), on the other hand, primarily targeted the content
of underlying assumptions or beliefs that, although not necessarily immedi-
ately available to consciousness, are in principle knowable. Both Brewin
(1989) and Teasdale & Barnard (1993) explicitly identified the importance of
deactivating or changing unconscious (situationally accessible) repre-
sentations.
In spite of the different emphases contained in these theories, there is a fair
measure of agreement that therapists mainly work with consciously available
cognitions, even if the aim is to hypothesize about, deactivate, or amend
underlying cognitions. There is also agreement about the kind of procedures
that are likely to be effective and about the need for repetition and practice.
Greater agreement may not be achievable at present because depression
takes many different forms, varying for example in frequency, intensity,
chronicity, and symptom pattern. Some depressions are accompanied by nega-
tive cognitions whereas others are not (Hamilton & Abramson 1983). Some
depressions (chronic or repeated depressions rather than single short episodes)
correlate positively with reports of traumatic childhood experiences (Andrews
et al 1995). Other depressions are linked specifically to a failure to complete
the emotional processing of major life events such as bereavement (Ramsay
1977).
 COGNITIVE-BEHAVIOR THERAPY 47

These data support the idea that depression is likely to, but will not neces-
sarily, involve representations of past experience in memory. At present the
respective roles of specific event representations and generalized schemas are
not well understood. As in the case of anxiety disorders, patients may develop
idiosyncratic misinterpretations of their symptoms and adopt coping strategies
that exacerbate the problem. Generic psychological treatments for depression
such as CBT need to contain provisions for altering misconceptions, equipping
patients with more effective skills, and deactivating emotional memories. The
particular combination of approaches used depends on the individual presenta-
tion.
Access provided by Washington State University on 12/09/14. For personal use only.
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

CONCEPTUALIZING CBT FOR ANXIETY

Until recently only behavioral techniques such as desensitization, flooding, or
response prevention were widely accepted as effective in treating anxiety
disorders, but now the effectiveness for some disorders of cognitive methods,
used alone or in combination with behavioral techniques, is clear. However,
purely cognitive techniques tend not to be superior to purely behavioral meth-
ods (Rachman 1993). Just as there is disagreement among theorists about how
CBT for depression works, there is disagreement about the mechanisms under-
lying CBT for anxiety. In this section I review evidence suggesting that the
disagreement arises because the factors causing and maintaining the disorders
are complex and that CBT necessarily reflects this complexity.
Rachman (1990) noted that there are various ways of acquiring fear, either
by direct exposure to aversive or traumatic stimuli, by observing others dis-
playing fear, or hearing or reading verbal or written transmission of informa-
tion about fear and danger. In the latter case, the representations in memory
giving rise to subsequent feelings of fear likely are verbally accessible and can
presumably be altered by new verbal information, perhaps delivered in the
form of an educational intervention. Similarly, some have argued that cata-
strophic misinterpretations of symptoms based on ignorance or incorrect be-
liefs can lead to panic disorder (Clark 1988) and prolonged PTSD (Ehlers &
Steil 1995). Once again, symptoms based on verbally accessible cognitions
should be amenable to treatment by more or less structured cognitive ap-
proaches.
In other cases, anxiety reflects the situational activation of unconscious
memory structures created through direct exposure to single or repeated aver-
sive experiences. If these memories are circumscribed, as is likely to be the
case in many phobias, behavioral interventions can readily be employed.
When patients have concluded that these feared situations really are threaten-
ing, either from observations of their own behavior or from independent infor-
mation, these methods are likely to alter the contents not only of situationally
 48 BREWIN

accessible knowledge but also of verbally accessible knowledge and result in

changes in conscious cognitions as well as in behavior.
As we have seen, the nature of the disorder need have no one-to-one
correspondence with a particular type of underlying representation. Thus pho-
bias may be based on unconscious representations, independent conscious
representations, or some combination of the two. The particular pattern of
response to different treatment approaches depends on the nature of the memo-
ries. Although most panic disorder may, in a similar way, be the result of
conscious catastrophic misinterpretations, Wolpe & Rowan (1989) suggest
Access provided by Washington State University on 12/09/14. For personal use only.

that in some patients there are no misinterpretations and that panic is the result
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

of conditioning. In the terms used in this review, the disorder would be said to
arise from the activation of situationally accessible knowledge either of pre-
vious panic episodes or of related experiences such as fear of similar symp-
toms in a family member.
Patients also may demonstrate combinations of symptoms that were ac-
quired in different ways. A recent study of patients with panic plus agorapho-
bia reported that exposure therapy reduced avoidance but not panic, whereas
cognitive therapy reduced panic but not avoidance (van den Hout et al 1994).
Social phobia may involve either circumscribed or generalized fears (e.g.
Hofmann et al 1995). When discussing treatment of fears that are generalized,
as they are in GAD, I distinguish approaches based on compensatory skills
training, such as the teaching of relaxation and the modification of negative
self-statements, from those based on identifying the content of underlying
representations in memory. I suggest that although it may be difficult to alter
such representations directly, the content of verbally accessible knowledge can
be changed so that they are no longer so easily activated. Effective cognitive
therapy for GAD (e.g. Beck & Emery 1985, Butler et al 1991) typically
employs both types of procedure.
Recent analyses of OCD (Salkovskis 1985, 1989) and hypochondriasis
(Warwick & Salkovskis 1989) underscore the variety and complexity of cases
grouped under the same diagnostic category. For example, in OCD Salkovskis
identified the frequent tendency of intrusive cognitions to be accompanied by
negative automatic thoughts reflecting underlying themes of excessive respon-
sibility for harm to self and others. These thoughts trigger neutralizing behav-
iors in the form of overt or covert rituals. The neutralizing behaviors presum-
ably originate as conscious strategies for avoiding or terminating the thoughts,
but over time their occurrence may become automatic. CBT for these prob-
lems involves compensatory skills training in teaching patients not to avoid
their negative thoughts, verbal belief modification of the identified automatic
thoughts and underlying assumptions, and the overwriting of specific learned
associations with behavioral techniques such as response prevention. For ex-
ample, verbal belief modification can be helpful in reducing anxiety and thus
 COGNITIVE-BEHAVIOR THERAPY 49

preventing dropouts, facilitating cooperation with response prevention, and

maximizing the impact of exposure treatment. The differing combinations and
strengths of maladaptive beliefs and neutralizing behaviors and the different
relations between them dictate that CBT be comprehensive and flexible if it is
to be effective.

THEORETICAL ISSUES
In the final part of this review I touch on two important theoretical issues that
Access provided by Washington State University on 12/09/14. For personal use only.

follow from my analysis of CBT. I have argued that CBT procedures must be
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

related to the different kinds of representation in memory that underlie the

disorder in each patient. In the case of generalized disorders, these usually
involve the self, either alone or in relation to significant others. What is known
about the representation of the self? Can this information suggest new thera-
peutic approaches? The other issue is that negative experiences give rise to
disorders in some individuals and not in others. It has been suggested that
memories of trauma and serious life events have in the normal course of things
to be emotionally processed. What is the relation of CBT to this processing?

Mental Representation of Knowledge About the Self

The mental representations thought to underlie depression, social phobia, and
GAD are more global and more often concern the self than those thought to
underlie more circumscribed disorders such as specific phobias. Such a view is
similar to that held by social psychologists such as Markus (1977), who
proposed that schemas are structures containing conceptually or empirically
related elements that can facilitate the processing of new information about the
self.
There is considerable debate about how knowledge concerning the self is
represented in memory. For example, Kelly (1955) proposed that the self may
be described in relation to a set of idiosyncratic personal constructs, a hierar-
chically organized system of bipolar dimensions such as good/bad or
kind/cruel. A currently popular view is that people have multiple self-repre-
sentations corresponding to different social roles (e.g. Kihlstrom & Cantor
1984). Higgins et al (1985) suggested that emotions may be generated by
discrepancies among self-representations. For example, anxiety is related to
discrepancies between the “actual self” and “ought self,” and depression to
discrepancies between “actual self” and “ideal self.”
In contrast, other theorists emphasize the role played by specific memories
in self-representation. For example, exemplar-based theories hold that judg-
ments concerning one’s character traits are inseparable from specific autobio-
graphical memories: Self-judgments are based on computations carried out on
specific examples available in memory. The experiments of Klein & Loftus
 50 BREWIN

(1993) indicate that knowledge about the self in the form of global trait
judgments is independent of knowledge in the form of specific autobiographi-
cal episodes—both are present and may be accessed and used relatively inde-
pendently. However, the extent to which judgments and behavior are influ-
enced by specific, concrete past experiences rather than by more abstract,
generalized knowledge appears to be much greater than has often been
thought. Moreover, the influence of these experiences may occur without a
person’s conscious awareness (Smith 1990).
Better theoretical understanding of how self-knowledge is represented is
Access provided by Washington State University on 12/09/14. For personal use only.

essential if we are to comprehend the mechanisms responsible for the success

Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

of CBT. To take the example of depression once again, we may ask whether
depressed individuals see themselves in wholly negative terms or whether they
continue to have access to alternative, more positive self-representations that
can be exploited therapeutically. Brewin et al (1992) found that depressed
patients were readily able to distinguish among how they would currently
describe themselves, how they would usually describe themselves, and how
they would have described themselves at some salient prior time, which sug-
gests that alternative self-representations remain accessible even during the
depressed episode itself.
Another important issue concerns the role of specific autobiographical
memories. Williams (1992) observed that the depressed find it difficult con-
sciously to retrieve specific autobiographical memories, particularly of posi-
tive experiences. Williams linked this difficulty to deficits in problem-solving
and proposed that memory for specific experiences of success and failure is
important if patients are to reassess their past experience in a more adaptive
way. He also suggested that the concrete focus of many CBT interventions
enhances memory for specific experiences.
Interestingly, given this deficit in conscious retrieval, recent evidence indi-
cates that many depressed patients have high levels of spontaneous intrusive
memories of specific autobiographical events—for example, concerning early
abuse (Kuyken & Brewin 1994). Moreover, there is a significant association
between higher levels of spontaneous memories of abuse and greater difficulty
in deliberately retrieving specific autobiographical memories (Kuyken &
Brewin 1995).
A related need is to research the connection between generalized self-repre-
sentations and specific autobiographical memories. Some progress has already
been made. Strauman (1990) showed that presenting subjects with words
corresponding to their individual self-discrepancies primed the retrieval of
specific childhood memories, which suggests a possible origin of global self-
beliefs. Andrews & Brewin (1990) interviewed a sample of women with
violent partners and found that women with a history of physical or sexual
abuse in childhood were more likely to blame the violence in their current
 COGNITIVE-BEHAVIOR THERAPY 51

relationships on their own characters and not on other factors. Although far
from conclusive, these various lines of research suggest that circumscribed
memories of trauma may be more relevant to understanding depressive prob-
lems than previously suspected, and that CBT for depression may benefit from
targeting specific autobiographical memories as well as more generalized self-
representations.

Emotional Processing
The notion that specific frightening, distressing, or traumatic experiences cre-
Access provided by Washington State University on 12/09/14. For personal use only.

ate cognitive and emotional disturbances that have to be worked through or

Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

processed if they are not to interfere with ordinary life has a long history. More
recently this notion has been associated particularly with studies of bereave-
ment (Parkes 1971) and with general models of response to stress and trauma
(Horowitz 1976, 1986; Janoff-Bulman 1992).
According to these writers, major stressors or traumas (and attempts to cope
with them) can disrupt a person’s expectations and goals. Stressors or traumas
may contradict important assumptions about a person’s own nature (for exam-
ple, the self is now perceived as capable of thoughts or actions not previously
believed possible), the world (for example, the world is now seen as unsafe
and unpredictable), and other people (for example, others are now seen as
unreliable or untrustworthy). These effects have been viewed as creating a
discrepancy between a person’s prior assumptions about him- or herself and
what that person currently knows to be true. The process of working through is
necessary to facilitate the creation of new models consistent with the facts.
This process typically involves heightened arousal and an initial response of
denial followed by alternating phases of intrusion of traumatic thoughts and
memories and attempts to avoid or block them out. It finally ends with the
integration of the new information into the person’s cognitive models.
Drawing on similar observations and on the work of Lang (1979), Rachman
(1980) defined emotional processing as a return to undisrupted behavior after
an emotional disturbance has waned. Although many of the phenomena he
described were the same as those outlined above, he formulated this process in
terms of the absorption and reduction of negative emotions rather than in terms
of the creation of new and more adaptive cognitive structures. Applying the
concept to the mastery of specific events that evoked fear or sadness, he
suggested that psychological treatments such as desensitization and flooding
can facilitate this process. He also proposed that obsessions, the return of fear,
the incubation of fear, abnormal grief reactions, and nightmares all represented
a failure of emotional processing.
These ideas have been very influential, and it is useful to relate them to the
distinction between therapies targeting conscious and unconscious repre-
sentations. The term emotional processing is used by such writers as Lang
 52 BREWIN

(1979), Rachman (1980), and Foa & Kozak (1986) to refer primarily to the
reduction of negative emotions (particularly fear) by traditional behavioral
techniques such as desensitization or exposure therapy [although Rachman
(1990) expanded its use to include cognitive changes]. In this review, I have
suggested that these techniques help prevent automatic activation of uncon-
scious emotional memories of the aversive event (situationally accessible
knowledge).
This process is different from that described by theorists concerned with
more general aspects of stress and trauma (e.g. Horowitz 1986, Janoff-Bulman
Access provided by Washington State University on 12/09/14. For personal use only.

1992). These theorists appear to focus on persons’ conscious preoccupations:

Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

Why did this happen to me? What made me behave the way I did? What will
happen to me now? These preoccupations lead to a variety of emotional
reactions that may include guilt, shame, and anger and to the automatic activa-
tion of conscious memories. Recovery is the successful reworking of con-
scious memories of the event and the creation of new, more inclusive cognitive
models in verbally accessible knowledge.
In terms of our analysis, therefore, expressions such as “emotional process-
ing” and “working through” appear to be relevant to modifying both conscious
and unconscious representations. In the case of intense but circumscribed fears
that have little impact on the rest of a person’s life or views of the world,
emotional processing entails modifying or overwriting unconscious repre-
sentations that are producing conditioned fear reactions. In the case of stres-
sors such as bereavement (other than when a person witnesses a traumatic
death), emotional processing entails modifying the contents of verbally acces-
sible knowledge to incorporate the new information.
Conditions such as PTSD typically arise from an event that both creates
strong conditioned fear reactions and violates many consciously held assump-
tions. This condition is typically associated with marked cognitive and behav-
ioral avoidance of reminders of the trauma. Discussions of recovery from
PTSD tend to focus on either the first (e.g. Foa et al 1989) or the second of
these processes (Janoff-Bulman 1992). According to a recent dual repre-
sentation theory of PTSD (Brewin et al 1994), however, recovery involves
both processes.
First, a person undergoing recovery needs to reduce the negative affects
generated by the implications of the trauma through a process of consciously
reasserting perceived control and achieving an integration of the new informa-
tion with preexisting concepts and beliefs. This may involve substantial edit-
ing of autobiographical memory in order to bring perceptions of the event into
line with prior expectations. For example, the behavior of an attacker pre-
viously believed to be trustworthy may be reinterpreted, excused, or explained
away, or aspects of what was done or said may be forgotten. Alternatively,
previous expectations may be adjusted in line with the event; for example,
 COGNITIVE-BEHAVIOR THERAPY 53

behaviors, neighborhoods, or locations believed to be safe may be reclassified

as dangerous, or life goals may be abandoned in favor of less ambitious ones.
The second aspect of emotional processing in PTSD is prevention of the
continued automatic reactivation of unconscious representations about the
trauma. Following activation and the emergence of their products into con-
sciousness, these unconscious representations may be altered or added to by
the pairing of the activated information with changes in concurrent bodily
states or contents of consciousness. Changes in bodily states may consist of
states of reduced arousal and reduced negative affect. These affective and
arousal changes may be brought about by a number of means, including
Access provided by Washington State University on 12/09/14. For personal use only.
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

spontaneous or programmed habituation to the traumatic images. Similar

changes are expected to follow the conscious restoration of a sense of safety
(reduced fear), the abandonment of now unattainable goals (reduced sadness),
the absolution of others from responsibility for the trauma (reduced anger),
and other attempts to integrate the new information into preexisting expecta-
tions. As this process of conscious cognitive restructuring continues, the
trauma images can be paired with progressive representations of effective
action-outcome sequences and reduced negative affect. In the absence of the
negative affect, a reduction in attentional and memory biases, and hence in the
accessibility of the memory, may be expected.
This analysis follows Rachman (1980) in suggesting that CBT facilitates
emotional processing. I suggest, however, that emotional processing encom-
passes changes to both verbally and situationally accessible memory.

SUMMARY
The basis of behavior therapy was the assumption that actions and emotions
are under the control of learned associations represented in a consciously
inaccessible form. Hence its techniques attempt to alter situationally accessible
knowledge by changing behavior. In contrast, cognitive therapists accepted
that conscious cognitions such as beliefs, plans, and goals also influence
behavior and emotions. In addition to trying to change situationally accessible
knowledge, they also developed techniques to boost compensatory strategies
and to rectify misconceptions in verbally accessible knowledge. Although
cognitive and behavior therapies emphasize the importance of different cogni-
tive systems, both therapies are consistent with our understanding of human
cognition.
The distinction between verbally accessible knowledge and situationally
accessible knowledge also appears central to understanding the different types
of cognitive therapy and the theoretical explanations for how they work. Some
procedures are clearly designed to change conscious beliefs, others to teach
skills, and others to modify less-accessible underlying structures in memory.
 54 BREWIN

Patients with the same diagnosis differ in their past experiences, the way they
acquired their disorder, the idiosyncratic meaning they attach to their symp-
toms, and the strategies they adopt to deal with these symptoms’ effects.
The theoretical disunity underlying CBT, although regrettable, is therefore
to some extent understandable. CBT has developed pragmatically to deal with
often difficult and refractory problems. To achieve its success it has had to
deploy a large array of procedures, and to seek to modify beliefs and behavior
by whatever means were available. In the process, much has been learned
about the factors that impede successful treatment, particularly the strategies
of cognitive avoidance and neutralization that may conceal from the therapist
Access provided by Washington State University on 12/09/14. For personal use only.
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

and sometimes from patients themselves the true sources of distress. In this
process CBT was, arguably, not too badly damaged by the lack of a coherent
rationale. I suggest, however, that CBT may in fact have a firm foundation,
albeit in such quite disparate areas of psychology as cognition, emotion, and
motivation. The next task is to link therapeutic imagination and enthusiasm
with a clearer understanding of theory.

ACKNOWLEDGMENTS
I gratefully acknowledge the comments of Michael Eysenck, Andrew
MacLeod, and Bernice Andrews on an earlier draft of this article.

Any Annual Review chapter, as well as any article cited in an Annual Review chapter,
may be purchased from the Annual Reviews Preprints and Reprints service.
1-800-347-8007; 415-259-5017; email: arpr@class.org

Literature Cited
Abramson LY, Metalsky GI, Alloy LB. 1989. thought: where the action is in cognitive
Hopelessness depression: a theory-based therapy for depression. Cogn. Ther. Res.
subtype of depression. Psychol. Rev. 96: 13:441–57
358–72 Ba§o™lu M, Marks IM, Kiliç C, Brewin CR,
Abramson LY, Seligman MEP, Teasdale JD. Swinson RP. 1994. Alprazolam and expo-
1978. Learned helplessness in humans: cri- sure for panic disorder with agoraphobia.
tique and reformulation. J. Abnorm. Psy- III. Attribution of improvement to medica-
chol. 87:49–74 tion predicts relapse. Br. J. Psychiatr. 164:
Andrews B, Brewin CR. 1990. Attributions of 652–59
blame for marital violence: a study of ante- Beck AT, Emery G. 1985. Anxiety Disorders
cedents and consequences. J. Marriage and Phobias: A Cognitive Perspective.
Fam. 52:757–67 New York: Basic Books
Andrews B, Valentine ER, Valentine JD. 1995. Beck AT, Freeman A, Pretzer J, Davis DD,
Depression and eating disorders following Fleming B, et al. 1990. Cognitive Therapy
abuse in childhood in two generations of of Personality Disorders. New York: Guil-
women. Br. J. Clin. Psychol. 34: 37–52 ford
Bandura A. 1977. Self-efficacy: toward a uni- Beck AT, Rush AJ, Shaw BF, Emery G. 1979.
fying theory of behavioral change. Psychol. Cognitive Therapy of Depression. New
Rev. 84:191–215 York: Wiley
Bandura A. 1986. Social Foundations of Berkowitz L. 1990. On the formation and regu-
Thought and Action: A Social Cognitive lation of anger and aggression: a cognitive-
Theory. Englewood Cliffs, NJ: Prentice- neoassociationistic analysis. Am. Psychol.
Hall 45:494–503
Barber JP, DeRubeis RJ. 1989. On second Berry DC, Broadbent DE. 1984. On the rela-
 COGNITIVE-BEHAVIOR THERAPY 55

tionship between task performance and as- dleton H, Anastasiades P, Gelder M. 1994.
sociated verbalizable knowledge. Q. J. A comparison of cognitive therapy, applied
Exp. Psychol. 36A:209–31 relaxation, and imipramine in the treatment
Blackburn IM, Whalley LJ, Christie JE, Sher- of panic disorder. Br. J. Psychiatr. 164:
ing A, Goggo M, et al. 1987. Mood, cogni- 759–69
tion and cortisol: their temporal relation- Davey GCL. 1992. Classical conditioning and
ships during recovery from depressive ill- the acquisition of human fears and phobias:
ness. J. Affect. Disord. 13:31–43 a review and synthesis of the literature.
Brewin CR. 1985. Depression and causal attri- Adv. Behav. Res. Ther. 14:29–66
butions: What is their relation? Psychol. DeRubeis RJ, Evans MD, Hollon SD,
Bull. 98:297–309 Garvey MJ, Grove WM, Tuason VB.
Brewin CR. 1988. Cognitive Foundations of 1990. How does cognitive therapy work?
Clinical Psychology. London/Hove, Engl.: Cognitive change and symptom change in
Erlbaum cognitive therapy and pharmacotherapy for
Access provided by Washington State University on 12/09/14. For personal use only.

Brewin CR. 1989. Cognitive change processes depression. J. Consult. Clin. Psychol. 58:
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

in psychotherapy. Psychol. Rev. 96:379–94 862–69

Brewin CR, Andrews B, Gotlib IH. 1993. Psy- Dickinson A. 1987. Animal conditioning and
chopathology and early experience: a reap- learning theory. In Theoretical Founda-
praisal of retrospective reports. Psychol. tions of Behavior Therapy, e d . H J
Bull. 113:82–98 Eysenck, I Martin, pp. 57–79. New York:
Brewin CR, Dalgleish T, Joseph S. 1994. A Plenum
dual representation theory of post-trau- Dobson KS. 1989. A meta-analysis of the effi-
matic stress disorder. Submitted for publi- cacy of cognitive therapy for depression. J.
cation Consult. Clin. Psychol. 57:414–19
Brewin CR, Smith AJ, Power M, Furnham A. D’Zurilla TJ, Goldfried MR. 1971. Problem
1992. State and trait differences in depres- solving and behavior modification. J. Ab-
sive self-perceptions. Behav. Res. Ther. 30: norm. Psychol. 78:197–226
555–57 Ehlers A, Steil R. 1995. Maintenance of intru-
Brown GW, Andrews B, Harris TO, Adler Z, sive memories in posttraumatic stress dis-
Bridge L. 1986. Social support, self-esteem order: a cognitive approach. Behav. Cogn.
and depression. Psychol. Med. 16:813–31 Psychother. In press
Brown GW, Lemyre L, Bifulco A. 1992. So- Ekman P. 1986. Telling Lies. New York: Berk-
cial factors and recovery from anxiety and ley Books
depressive disorders: a test of specificity. Ellis A. 1962. Reason and Emotion in Psycho-
Br. J. Psychiatr. 161:44–54 therapy. New York: Lyle Stuart
Butler G, Fennell M, Robson P, Gelder M. Epstein S. 1994. Integration of the cognitive
1991. Comparison of behavior therapy and and the psychodynamic unconscious. Am.
cognitive behavior therapy in the treatment Psychol. 49:709–24
of generalized anxiety disorder. J. Consult. Firth-Cozens J, Brewin CR. 1988. Attribu-
Clin. Psychol. 59:167–75 tional change during psychotherapy. Br. J.
Cantor N, Kihlstrom JF. 1987. Personality and Clin. Psychol. 27:47–54
Social Intelligence. Englewood Cliffs, NJ: Foa EB, Kozak MJ. 1986. Emotional process-
Prentice-Hall ing of fear: exposure to corrective informa-
Carver CS, Scheier MF. 1981. Attention and tion. Psychol. Bull. 99:20–35
Self-Regulation: A Control Theory Ap- Foa EB, Steketee G, Rothbaum BO. 1989. Be-
proach to Human Behavior. New York: havioral/ cognitive conceptualization of
Springer-Verlag post-traumatic stress disorder. Behav. Ther.
Chambless DL, Gillis MM. 1993. Cognitive 20:155–76
therapy of anxiety disorders. J. Consult. Goldfried MR, Goldfried AP. 1980. Cognitive
Clin. Psychol. 61:248–60 change methods. In Helping People
Chambless DL, Gracely EJ. 1988. Prediction Change, ed. FH Kanfer, AP Goldstein, pp.
of outcome following in vivo exposure 97–130. New York: Pergamon. 2nd ed.
treatment of agoraphobia. In Panic and Hamilton EW, Abramson LY. 1983. Cognitive
Phobias, ed. I Hand, H-U Wittchen, 2: patterns and major depressive disorder: a
209–20. Berlin: Springer-Verlag longitudinal study in a hospital setting. J.
Clark DM. 1988. A cognitive model of panic Abnorm. Psychol. 92:173–84
attacks. In Panic: Psychological Perspec- Hasher L, Zacks RT. 1979. Automatic and ef-
tives, ed. S Rachman, JD Maser, pp. 71–89. fortful processes in memory. J. Exp. Psy-
Hillsdale, NJ: Erlbaum chol.: Gen. 108:356–89
Clark DM, Beck AT. 1988. Cognitive ap- Higgins ET, Klein R, Strauman TJ. 1985. Self-
proaches. In Handbook of Anxiety Disor- concept discrepancy theory: a psychologi-
ders, ed. CG Last, M Hersen, pp. 362–85. cal model for distinguishing among differ-
Elmsford, NY: Pergamon ent aspects of depression and anxiety. Soc.
Clark DM, Salkovskis PM, Hackmann A, Mid- Cogn. 3:51–76
 56 BREWIN

Hofmann SG, Newman MG, Ehlers A, Roth emotional imagery. Psychophysiology 16:
WT. 1995. Psychophysiological differ- 495–512
ences between subgroups of social phobia. Lang PJ. 1993. The network model of emotion:
J. Abnorm. Psychol. 104:224–31 motivational connections. In Advances in
Hollon SD, DeRubeis RJ, Evans MD. 1987. Social Cognition, ed. RS Wyer, TK Srull,
Causal mediation of change in treatment 6:109–33. Hillsdale, NJ: Erlbaum
for depression: discriminating between Lazarus RS. 1991. Emotion and Adaptation.
nonspecificity and noncausality. Psychol. New York: Oxford Univ. Press
Bull. 102:139–49 Lazarus RS, Averill JR, Opton EM. 1970. To-
Hollon SD, Evans MD, DeRubeis RJ. 1988. wards a cognitive theory of emotion. In
Preventing relapse following treatment for Feelings and Emotions, ed. M Arnold, pp.
depression: the cognitive pharmacotherapy 207–32. New York: Academic
project. In Stress and Coping across Devel- Leventhal H. 1984. A perceptual-motor theory
opment, ed. TM Field, PM McCabe, N of emotion. Adv. Exp. Soc. Psychol. 17:
Access provided by Washington State University on 12/09/14. For personal use only.

Schneiderman, pp. 227–43. Hillsdale, NJ: 117-82

Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

Erlbaum Markus H. 1977. Self-schemata and processing

Hollon SD, Kriss MR. 1984. Cognitive factors information about the self. J. Pers. Soc.
in clinical research and practice. Clin. Psy- Psychol. 35:63–78
chol. Rev. 4:35–76 Markus H, Wurf E. 1987. The dynamic self-
Hollon SD, Shelton RC, Davis DD. 1993. Cog- concept: a social psychological perspec-
nitive therapy for depression: conceptual tive. Annu. Rev. Psychol. 38:299–337
issues and clinical efficacy. J. Consult. Mattick RP, Peters L. 1988. Treatment of se-
Clin. Psychol. 61:270–75 vere social phobia: effects of guided expo-
Horowitz MJ. 1976. Stress Response Syn- sure with and without cognitive restruc-
dromes. New York: Jason Aronson turing. J. Consult. Clin. Psychol. 56:
Horowitz MJ. 1986. Stress Response Syn- 251–60
dromes. New York: Jason Aronson. 2nd Meichenbaum D. 1977. Cognitive-Behavior
ed. Modification. New York: Plenum
Ingram RE, Hollon SD. 1986. Cognitive ther- Mischel W. 1973. Toward a cognitive social
apy of depression from an information learning reconceptualization of personality.
processing perspective. In Information Psychol. Rev. 80:252–83
Processing Approaches to Clinical Psy- Morton J, Hammersley RH, Bekerian DA.
chology, ed. RE Ingram, pp. 259–81. New 1985. Headed records: a model for memory
York: Academic and its failures. Cognition 20:1–23
Janoff-Bulman R. 1992. Shattered Assump- Nisbett RE, Wilson TD. 1977. Telling more
tions: Towards a New Psychology of than we can know: verbal reports on mental
Trauma. New York: Free Press processes. Psychol. Rev. 84:231–59
Johnson JE, Rice VH, Fuller SS, Endress MP. Parkes CM. 1971. Psychosocial transitions: a
1978. Sensory information, instruction in a field for study. Soc. Sci. Med. 5:101–15
coping strategy, and recovery from sur- Posner MI, Snyder CR. 1975. Attention and
gery. Res. Nurs. Health 1:4–17 cognitive control. In Information Process-
Karoly P, Kanfer FH, eds. 1982. Self-Manage- ing and Cognition: The Loyola Symposium,
ment and Behavior Change: From Theory ed. RL Solso, pp. 55–85. Hillsdale, NJ:
to Practice. New York: Pergamon Erlbaum
Kelly GA. 1955. The Psychology of Personal Power MJ, Brewin CR. 1991. From Freud to
Constructs, Vols. 1, 2. New York: Norton cognitive science: a contemporary account
Kihlstrom JF, Cantor N. 1984. Mental repre- of the unconscious. Br. J. Clin. Psychol.
sentation of the self. In Advances in Ex- 30:289–310
perimental Social Psychology, ed. L Berk- Rachman S. 1980. Emotional processing. Be-
owitz, 17:1–47. New York: Academic hav. Res. Ther. 18:51–60
Klein SB, Loftus J. 1993. The mental repre- Rachman S. 1990. The determinants and treat-
sentation of trait and autobiographical ment of simple phobias. Adv. Behav. Res.
knowledge about the self. In Advances in Ther. 12:1–30
Social Cognition, ed. TK Srull, RS Wyer, Rachman S. 1993. A critique of cognitive ther-
5:1–49. Hillsdale, NJ: Erlbaum apy for anxiety disorders. J. Behav. Ther.
Kuyken W, Brewin CR. 1994. Intrusive Exp. Psychiatr. 24:279–88
memories of childhood abuse during de- Ramsay RW. 1977. Behavioral approaches to
pressive episodes. Behav. Res. Ther. 32: bereavement. Behav. Res. Ther. 1 5 :
525–28 131–36
Kuyken W, Brewin CR. 1995. Autobiographi- Reiss S, McNally RJ. 1985. Expectancy model
cal memory functioning in depression and of fear. In Theoretical Issues in Behavior
reports of early abuse. J. Abnorm. Psychol. Therapy, ed. S Reiss, RR Bootzin, pp.
In press 107–21. San Diego, CA: Academic
Lang PJ. 1979. A bio-informational theory of Rush AJ, Weissenburger J, Eaves G. 1986. Do
 COGNITIVE-BEHAVIOR THERAPY 57

thinking patterns predict depressive symp- tions in Attribution Research, ed. JH

toms? Cogn. Ther. Res. 10:225–38 Harvey, WJ Ickes, RF Kidd, 1:143–64.
Salkovskis PM. 1985. Obsessional-compulsive Hillsdale, NJ: Erlbaum
problems: a cognitive-behavioural analysis. Strauman TJ. 1990. Self-guides and emotion-
Behav. Res. Ther. 23:571–83 ally significant childhood memories: a
Salkovskis PM. 1989. Obsessions and compul- study of retrieval efficiency and incidental
sions. See Scott et al 1989, pp. 50–77 negative emotional content. J. Pers. Soc.
Segal ZV. 1988. Appraisal of the self-schema Psychol. 59:869–80
construct in cognitive models of depres- Teasdale JD. 1985. Psychological treatments
sion. Psychol. Bull. 103:147–62 for depression: How do they work? Behav.
Scott J, Williams JMG, Beck AT, eds. 1989. Res. Ther. 23:157–65
Cognitive Therapy in Clinical Practice: An Teasdale JD, Barnard PJ. 1993. Cognition, Af-
Illustrative Casebook. London: Routledge fect, and Change: Re-modelling Depressive
Shiffrin RM, Schneider W. 1977. Controlled Thought. Hove, Engl.: Erlbaum
Access provided by Washington State University on 12/09/14. For personal use only.

and automatic human information process- van den Hout M, Arntz A, Hoekstra R. 1994.
Annu. Rev. Psychol. 1996.47:33-57. Downloaded from www.annualreviews.org

ing. II. Perceptual learning, automatic at- Exposure reduced agoraphobia but not
tending, and a general theory. Psychol. panic, and cognitive therapy reduced panic
Rev. 84:127–90 but not agoraphobia. Behav. Res. Ther. 32:
Simons AD, Murphy GE, Levine JE, Wetzel 447–51
RD. 1986. Cognitive therapy and pharma- Warwick HMC, Salkovskis PM. 1989. Hypo-
cotherapy for depression: sustained im- chondriasis. See Scott et al 1989, pp.
provement over one year. Arch. Gen. Psy- 78–102
chiatr. 43:43–49 Weiner B. 1985. An attributional theory of
Smith ER. 1990. Content and process specific- achievement motivation and emotion. Psy-
ity in the effects of prior experiences. In chol. Rev. 92:548–73
Advances in Social Cognition, ed. TK Williams JMG. 1992. The Psychological
Srull, RS Wyer, 3:1–59. Hillsdale, NJ: Treatment of Depression. London: Rout-
Erlbaum ledge. 2nd ed.
Squire LR, Knowlton B, Musen G. 1993. The Wolpe J, Rowan V. 1989. Panic disorder: a
structure and organization of memory. product of classical conditioning. Behav.
Annu. Rev. Psychol. 44:453–95 Res. Ther. 27:583–85
Storms MD, McCaul KD. 1976. Attribution Young J. 1990. Cognitive Therapy for Person-
processes and the emotional exacerbation ality Disorders: A Schema-Focused Ap-
of dysfunctional behavior. In New Direc- proach. Sarasota, FL: Prof. Resour. Exch.