Behavior Therapy 37 (2006) 314 318

www.elsevier.com/locate/bt

Towards an Experimental Cognitive Science of CBT

Andrew Mathews, University of California, Davis

The increasing focus on cognitive processes as therapeutic

targets has not yet been matched by a cognitive science base
sufficiently developed to guide clinical practice. It is argued
that the papers in this special issue represent evidence of
progress towards this desirable goal. Collectively, they
illustrate research techniques aimed at specifying the nature
of cognitive operations likely to increase the risk of
emotional disorders, and the introduction of experimental
methods for their modification. Emergent themes include
the suggestion that negative thought content, such as that
experienced in rumination, is an unintended but maladaptive product of underlying biases in selective processing.
Despite often operating outside awareness, this biased
processing can be changed, for example by strengthening
incompatible alternatives. Beyond providing evidence for
the causal role of selective cognitive processes, this approach
offers a potentially powerful method for investigating and
developing new therapeutic tools.

I N A P A P E R P U B L I S H E D in this journal, subtitled

Cognitive Behavior Therapy in Search of Theory,
Foa and Kozak (1997) argued that progress in
cognitive-behavior therapy (CBT) has been slow
due to the lack of new theoretical input. Behavior
Therapy grew originally from a knowledge base
grounded in learning and conditioning research.
More recently, however, conditioning theories have
become increasingly integrated with experimental
cognitive psychology. At the same time, CBT itself
has increasingly targeted cognitive as well as
Address correspondence to Andrew Mathews, Department of
Psychology, One Shields Avenue, Davis, CA 95616, USA; e-mail:
andrew.mathews@sbcglobal.net.
0005-7894/06/03140318$1.00/0
2006 Association for Behavioral and Cognitive Therapies. Published by
Elsevier Ltd. All rights reserved.

behavioral and emotional symptoms. Consequently, clinicians are now concerned with the identification and change of cognitive processes likely to
be relevant to etiology and recovery. But despite
having embraced the terminology of cognitive
psychology, clinicians have not been able to draw
on experimental cognitive psychology to develop a
scientific basis that would be useful in developing
new treatment methods. The papers in this special
issue all represent attempts towards the development of just such a cognitive science base. They are
not typical of papers usually appearing in this
journal, and some readers may not see them as
appropriate contributions to Behavior Therapy as
it is usually understood.
In contrast, I believe these papers illustrate the
next necessary step in the evolution of CBT. In this
next step, relevant cognitive operations contributing towards the maintenance of psychological
problems, just as much as behaviors, need to be
specified with an increasing degree of precision. It is
important to emphasize that the term cognitive
operations is used here to refer to processes that
can be assessed by objective means, rather than only
by subjective reports of cognitive content. When
clinicians describe cognitions, they are often
referring to what clients say about the thoughts
and images of which they have become aware.
Although such reports are, of course, a critically
important source of hypotheses about cognitive
factors in etiology, they do not in themselves reveal
much about the processes that led up to that content
appearing in consciousness. In fact, the underlying
processes are typically not available to conscious
introspection at all, but must be inferred indirectly
from behavioral observations such as response
latencies and errors or from psychophysiological
and neural-imaging methods. Examples include the
selective processes whereby one among many
competing events succeeds in capturing our attention, only one of several possible meanings of

cognitive science of cbt

ambiguous information is perceived, or which of
many possible representations is selected for
encoding in (or retrieval from) memory. We are
not usually aware of controlling any of these
selective processes and indeed are often not even
aware that any selection has occurred at all.

Specifying the Nature of Emotional Processing

Underlying the research effort represented in this
special issue is the belief that understanding
selective processing of emotional events can inform
us about important factors contributing to psychological disorders. Cognitive processes involved in
biased encoding of emotional events or meanings
are often thought to play a part in the etiology of
disorders and, if so, could be profitably modified in
therapy. So far, however, most evidence linking
selective processing with emotional vulnerability
has been correlational in nature. People with
emotional disorders, relative to healthy groups,
show a tendency to encode negative events or
meanings, and these negative biases are reduced or
reversed with recovery (Mathews & MacLeod,
2005). Clearly it is important, first, to specify the
nature of such biases as precisely as possible if we
want to understand and then to study the effects of
changing them.
Consider the consistent finding that emotionally
disturbed individuals are relatively more likely than
are others to endorse threatening interpretations of
ambiguous events. Is this apparent bias an enduring traitlike tendency or does it depend on current
state? Does it reflect a real tendency to make
threatening interpretations (that is, greater sensitivity in signal detection terms) or a response bias,
reflecting greater willingness to endorse any negative item, whether based on a prior interpretation
or not? And is it a truly negative bias or the loss of a
positive bias? In a study of interpretation in sleep
disturbance, Ree and Harvey (2006) report results
suggesting that current state is associated with a
general response bias favoring the endorsement of
threatening interpretations. Looking at these data
from another direction, however, suggests that it
could be the normal groups who are biased, albeit
in a positive direction. If so, it may well reflect the
fact that healthy individuals are characterized by a
protective positive bias which is lost in negative
emotional states (see also Hirsch et al., 2006). This
possibility could have important implications when
thinking about the appropriate targets for modification in CBT. Should we try to reduce negative
biases, strengthen positive biases, or both?
Cognitive and neuro-imaging evidence has
allowed us to specify other cognitive operations
with greater precision. For example, attention can

315

be decomposed into distinct processes, including

initial engagement of attention at a particular
location, and later disengagement from that location, when attention is directed elsewhere. Despite
the evidence that threatening cues are selectively
attended by those prone to anxiety, it has not been
clear whether this is due to greater initial engagement with threat cues, slowed disengagement from
them, or both. A number of recent studies have
suggested that difficulty in disengaging attention
from threat cues is particularly characteristic of
anxious individuals (Fox, Russo, Bowles & Dutton,
2001; Yiend & Mathews, 2001; although see also
Mathews, Fox, Yiend, & Calder, 2003). Nonetheless, fear and anxiety must also serve an adaptive
function in directing attention towards locations
that may contain dangers (as when spider-fearful
individuals check dark corners for cobwebs) or
possible escape routes. These findings can readily be
understood within a theoretical framework that
views emotions as having evolved to serve specific
adaptive functions. In the case of fear and anxiety,
the adaptive function is to detect and avoid danger
(Oatley & Johnson-Laird, 1987).
It is less immediately obvious what adaptive
function might be served by selective processing in
depression. Oakley and Johnson-Laird (1987)
suggested that depression can be an adaptive state
when major current goals and plans have failed,
and thus should be abandoned. In this view,
depression serves the function of discouraging the
pursuit of goals that are now inappropriate. This
notion was tested in a study by van den Elzen and
MacLeod (2006) in which they propose that
depressed individuals should be quicker than others
to abandon a task at which they are failing. Less
obviously, this might also have the result of helping
depressives to change to an alternative course of
action. Results supported this hypothesis, with no
differences between groups in their initial learning
of a rule, but with depressives being better able to
abandon the original rule when it was changed and
learn an alternative. Could it beas van den Elzen
and MacLeod arguethat therapists should encourage this adaptive tendency to abandon failing
plans? Or would it be better to focus on helping
depressed individuals find alternative and more
adaptive goals? Experiments of this type can open
up interesting therapeutic possibilities, although
whether they will prove of real clinical value
remains to be tested.

The Nature and Consequences of Rumination

In the study by van den Elzen and MacLeod (2006),
there was no evidence of greater differences between
groups due to explicit (intentional) versus implicit

316

mathews

(incidental) rule learning, and consequently it was

argued that interference due to rumination (repetitive
self-related ideation) is unlikely to explain the rapid
abandonment of old learning in depression. There is
now little doubt, however, that rumination is an
important risk factor in prolonging episodes of
depression (Lyubomirsky & Nolen-Hoeksema,
1995). Recent research has begun to shed light on
the critical cognitive operations underlying rumination and how these can be changed.
Four papers in this special issue address this and
related topics. Rumination as commonly assessed is
not a unitary process but rather includes some
components that are more toxic than others. Along
with others, Joorman, Dkane, and Gotlib (2006)
argue that it is content described as brooding
(e.g., self-evaluative analysis), as opposed to reflection (which can include problem-solving
attempts), that has deleterious consequences.
Brooding was indeed found to be more associated
with depression than was reflective content, and
more importantly in the present contextbrooding
was also more associated with attention to faces
with sad (or happy) rather than neutral expressions
and with memory for negative self-related words.
Thus, a self-report measure of repetitive selfevaluative thought content was related to objective
measures of attention to and memory for emotional
information. The interesting question then arises:
Are these relationships causal, and if so, in which
direction? For example, could ruminative content be
maintained by biases in underlying selective processes that make it difficult to disengage from
negative self-related thoughts? Understanding directions of possible causation in this respect could have
important implications for choosing the most
appropriate targets for therapeutic modification.
As noted above, rumination (and specifically its
so-called brooding component) is characterized
by an analytic focus on the self. Interestingly,
however, this only leads to adverse emotional
consequences in those already disposed to depression. Why should this be so? Hertel and El-Messidi
(2006) suggest that the self-focused content in
rumination has greater adverse effects in depressive-prone individuals because they have developed
the habit of thinking about negative aspects of
themselves. Self-focus thus guides them along
already familiar mental paths. In the first experiment described here, Hertel and El-Messidi confirm
that a brief period of self-focused thought can have
more negative emotional effects for those already
prone to depression. Even when no differential
mood changes were seen (in a second experiment),
self-focus had more persisting effects on the
interpretation of new ambiguous information in

depression-prone individuals. Rumination thus has

two adverse consequences specific to vulnerable
individuals: it reactivates existing negative thought
patterns and makes personal interpretations of new
ambiguous events more likely. One way to prevent
these consequences could be to strengthen incompatible (i.e., less self-related) habits of thought.
In addition to self-focus, however, the more
damaging (brooding) component of rumination
includes a self-judgmental aspect (e.g., Why do I
have problems that others don't have?). In the
present issue, Gortner, Rude, and Pennebaker
(2006) go on to speculate that one way to reduce
brooding and encourage less judgmental thinking
(e.g., reflection) might be to use expressive writing.
Previously depressed students were assigned to 3 to
4 sessions of expressive writing about personal
events that they found most emotionally disturbing
or writing about nonemotional everyday events. Six
months later, those in the expressive writing group
reported significantly fewer depressive symptoms.
Extent of improvement was associated with reductions in reported brooding, but (contrary to
expectation) not with increased reflection. Furthermore, those who initially reported suppressing their
emotions also improved more following expressive
writing. This fits our common sense idea that
people who overly suppress their feelings may profit
from expressing them more but does not explain
why this should be so. It seems likely that expressive
writing encourages processing emotional events in a
nondestructive way, perhaps by interfering with an
otherwise habitual judgmental self-focus and allowing a more objective perspective to be taken. But the
exact mechanism by which vulnerability to depression was reduced remains unclear and how this
happens can only be discovered in future research.
For now, however, the finding again points to the
notion that cognitive habits having negative emotional consequences can be weakened by practicing
incompatible ways of processing the same material.
A final paper on rumination in this issue (Moberly
& Watkins, 2006) also distinguishes between types
of mental content, labeled in this case abstractevaluative versus concrete-process focused
(Watkins & Teasdale, 2004). The former (evaluative) content seems to map fairly well onto
brooding, sharing with it a focus on self-evaluative
implications and having similar adverse effects in
depression. The latter (concrete) content is less
easily identified with reflection, as it is said to
include moment-to-moment perception of feelings
(e.g., What am I feeling right now?). However, like
reflection, it can be self-focused while lacking in
evaluative content and does not lead to negative
emotional consequences. Indeed, it seems possible

cognitive science of cbt

that this type of thinking about feelings is related to
that induced by expressive writing and could offer
an explanation for its beneficial effects (if, for
example, a focus on the experience of feelings
discourages self-evaluative judgments about them).
In support of this notion, experimentally inducing
this type of thought content following a failure
experience prevented the adverse emotional effects
otherwise seen in ruminative and depressive-prone
individuals (Watkins & Teasdale, 2004).
In the present study, Moberly and Watkins (2006)
adopted the general approach developed by
MacLeod and his colleagues (Mathews &
MacLeod, 2002), in which processing style is
experimentally manipulated rather than varying
naturally between individuals or selected groups.
Critically, such an approach allows for more
definitive causal conclusions about the effect of a
specified processing style (or mode) on emotion and
emotional vulnerability. In the present study participants were allocated to think about a series of both
positive and negative personal events in different
ways: either concretely (e.g., imagine the details of
what is happening) or in an evaluative manner (e.g.,
think about the implications), with each condition
reinforced by related questions. After a contrived
failure experience, reductions in positive affect were
greater for those in the evaluative group, albeit only
for participants reporting high levels of rumination.
Although this study only involved students, it does
suggest that training to focus on the present
experience of feelings, rather than on their implications, could decrease vulnerability to depression (cf.
Teasdale et al., 2000).

Experimental Modification of Processing Style

The effect of training cognitive processing style is
also the central topic of the next two papers to be
discussed (Mackintosh, Mathews, Yiend, Ridgeway, & Cook, 2006; Holmes, Mathews, Dalgleish,
& Mackintosh, 2006) and is also highly relevant to
the last one (Hirsch et al., 2006).1 The issues
addressed by Mackintosh et al. include the persistence of such experimental training and the extent to
which it can influence later emotional vulnerability.
These issues are critically relevant to the question of
whether induced processing style persists as do
naturally occurring biases and whether processing

1
It will not have escaped the alert reader's attention that the
present writer is a co-author of the last three papers discussed. I
hope that any positive processing bias on my part is not too
apparent in this discussion, but in any event it should not detract
from the credit due to the main authors, who have contributed far
more than I have.

317

biases are a cause rather than just a consequence of

emotional vulnerability.
In the study by Mackintosh et al. (2006),
volunteers practiced accessing either the positive
or the negative meaning of ambiguous descriptions
and then returned the next day for testing. Despite
the delay, and for half of the participants a change
of environmental setting, they continued to interpret new descriptions in a manner congruent with
their prior experience in training. More strikingly,
those previously trained to access the more negative
meaning of ambiguous descriptions reported more
anxiety in response to a stressful video than did the
comparison group. These data raise the possibility
that similar methods could be used to induce more
positive processing biases in vulnerable individuals
and thus protect them from the adverse consequences of stressful life events.
In previous work using these experimental induction methods, Holmes and Mathews (2005) found
that changes in anxiety during training to induce a
negative bias depended critically on the use of
imagery rather than on the verbal meaning of texts.
Although it has long been supposed that images have
more powerful emotional effects than do verbally
mediated representations, surprisingly, supporting
evidence for this claim has been lacking. In the
present paper, Holmes et al. (2006) go on to show
that the superior effect of imagery applies equally to
the effectiveness of training on positive emotionality.
In addition to imagery having greater effects of
positive feelings during training, it also led to more
positive interpretations of ambiguous event descriptions encountered later. Imagery is, of course, widely
used in CBT for anxiety disorders, but the present
results support the possibility of wider applications,
including positive retraining in mood disorders.
The paper by Hirsch et al. (2006) builds on
evidence based on methods similar to those just
described and also makes a wider point: If
psychological disorders are associated with not
just one but several characteristic types of cognitive
content and process, how might they influence each
other? And could such mutual influences serve to
amplify their total effect, thereby creating particularly strong effects on emotional vulnerability?
Existing cognitive models make such assumptions
implicitly (for example, the cognitive model of social
phobia; Clark & Wells, 1995), but this hypothesis is
made more explicitly in the present paper. Specifically, it is proposed that biased interpretations can
influence the content of mental imagery, which has
especially powerful effects on emotion (as noted
above). In reciprocal fashion, imagery content can
powerfully influence the interpretations made of
emotionally ambiguous events. As Hirsch et al.

318

mathews

point out, mutual influences of this type are

probably quite common and perhaps can be
harnessed in therapy, for example, by targeting
one cognitive process in order to change another.

Conclusions
Despite the disparate content of these papers,
several important unifying themes emerge from
them. In particular, they are all concerned with
identifying underlying cognitive processes that are
thought to maintain emotional disorders. This
approach does not ignore reportable cognitive
content, such as repetitive emotional thoughts and
images, but this content is viewed as being
influenced by other cognitive processes that are
not usually accessible for conscious report. The
causal role of these processes, and of the content
they influence, can then be investigated by manipulating them in experimental studies. In this way we
can move beyond merely describing the cognitive
operations that characterize certain disorders towards finding the optimal ways of changing them.
Not only does this allow investigation of their
causal role in emotional vulnerability, but it should
eventually lead to new methods of treatment.
I began by anticipating that some readers of this
special issue may be uncomfortable with the
emphasis on underlying cognitive processes within
experimental studies of volunteers rather than on
observable behaviors and clinical symptom change.
However, as I hope is now clear, the experimental
approach used here often does employ behavioral
measures (e.g., decision latencies or response accuracy), and the methods of modification employed
are in many ways surprisingly close to the learning
origins of Behavior Therapy. Control over destructive cognitive habits can be achieved, not only by
discussion and instructions, but via training that
involves repeated practice serving to strengthen
incompatible positive alternatives. Results so far
may seem fairly far removed from actual clinical
application, but I believe they show that we have
made measurable progress towards providing CBT
with the cognitive science base it needs.
References
Clark, D. M., & Wells, A. (1995). A cognitive model of social
phobia. In R. G. Heimberg, M. Liebowitz, D. Hope, & F.
Schneier (Eds.), Social phobia: Diagnosis, assessment, and
treatment (pp. 6993). New York: The Guilford Press.
Foa, E. B., & Kozak, M. J. (1997). Beyond the efficacy ceiling:
Cognitive behavior therapy in search of theory. Behavior
Therapy, 28, 601611.
Fox, E., Russo, R., Bowles, R., & Dutton, K. (2001). Do
threatening stimuli draw or hold visual attention in

subclinical anxiety? Journal of Experimental Psychology:

General, 130, 681700.
Gortner, E. M., Rude, S. S., & Pennebaker, J. W. (2006).
Benefits of expressive writing in lowering rumination and
depressive symptoms. Behavior Therapy, 37, 292303.
Hertel, P. T., & El-Messidi, L. (2006). Am I blue? Depressed
mood and the consequences of self-focus for the interpretation and recall of ambiguous words. Behavior Therapy,
37, 259268.
Hirsch, C., Mathews, A., & Clark, D. (2006). Imagery and
interpretations in social phobia: Support for the combined
cognitive biases hypothesis. Behavior Therapy, 37,
223236.
Holmes, E. A., & Mathews, A. (2005). Mental imagery and
emotion: A special relationship? Emotion, 5, 489497.
Holmes, E. A., Mathews, A., Dalgleish, T., & Mackintosh, B.
(2006). Positive interpretation training: Effects of mental
imagery versus verbal training on positive mood. Behavior
Therapy, 37, 237247.
Joormann, J., Dkane, M., & Gotlib, I. H. (2006). Adaptive and
maladaptive components of rumination? Diagnostic specificity and relation to depressive biases. Behavior Therapy,
37, 269280.
Lyubomirsky, S., & Nolen-Hoeksema, S. (1995). Effects of selffocused rumination on negative thinking and interpersonal
problem solving. Journal of Personality and Social Psychology, 69, 176190.
Mackintosh, B., Mathews, A., Yiend, J., Ridgeway, V., &
Cook, E. (2006). Induced biases in emotional interpretation influence stress vulnerability and endure despite
changes in context. Behavior Therapy, 37, 209222.
Mathews, A., Fox, E., Yiend, J., & Calder, A. (2003). The face
of fear: Effects of eye gaze and emotion on visual attention.
Visual Cognition, 10, 823836.
Mathews, A., & MacLeod, C. (2002). Induced processing
biases have causal effects on anxiety. Cognition and
Emotion, 16, 310315.
Mathews, A., & MacLeod, C. (2005). Cognitive vulnerability
to emotional disorders. Annual Review of Clinical Psychology, 1, 167195.
Moberly, N., & Watkins, E. (2006). Processing mode influences
the relationship between trait rumination and emotional
vulnerability. Behavior Therapy, 37, 281291.
Oatley, K., & Johnson-Laird, P. N. (1987). Towards a cognitive
theory of emotions. Cognition and Emotion, 1, 2950.
Ree, M., & Harvey, A. (2006). Interpretive biases in chronic
insomnia: An investigation using a priming paradigm.
Behavior Therapy, 37, 237247.
Teasdale, J., Segal, Z., Williams, J. M., Ridgeway, V., Soulsby, J.,
& Lau, M. (2000). Prevention of relapse/recurrence in major
depression by mindfulness-based cognitive therapy. Journal
of Consulting and Clinical Psychology, 68, 615623.
van den Elzen, J., & MacLeod, C. (2006). Facilitated cognitive
disengagement in depression. Behavior Therapy, 37, 304313.
Watkins, E., & Teasdale, J. D. (2004). Adaptive and maladaptive self-focus in depression. Journal of Affective Disorders,
82, 18.
Yiend, J., & Mathews, A. (2001). Anxiety and attention to
threatening pictures. Quarterly Journal of Experimental
Psychology, 54, 665681.

R E C E I V E D : February 12, 2006

A C C E P T E D : February 20, 2006
Available online 30 May 2006