Psychological Bulletin Copyright 1986 by the American Psychological Association, Inc.
1986, Vol. 99, No. 1, 20-35 0033-2909/86/$00.75
Emotional Processing of Fear: Exposure to Corrective Information
Edna
B.
Foa and Michael J. Kozak
Temple University
In this article we propose mechanisms that govern the processing of emotional information, particularly
those involved in fear reduction. Emotions are viewed as represented by information structures in
memory, and anxiety is thought to occur when an information structure that serves as program to
escape or avoid danger is activated. Emotional processing is defined as the modification of memory
structures that underlie emotions. It is argued that some form of exposure to feared situations is
common to many psychotherapies for anxiety, and that confrontation with feared objects or situations
is an effective treatment. Physiological activation and habituation within and across exposure sessions
are cited as indicators of emotional processing, and variables that influence activation and habituation
of fear responses are examined. These variables and the indicators are analyzed to yield an account
of what information must be integrated for emotional processing of a fear structure. The elements of
such a structure are viewed as cognitive representations of the stimulus characteristic of the fear
situation, the individual's responses in it, and aspects of its meaning for the individual. Treatment
failures are interpreted with respect to the interference of cognitive defenses, autonomic arousal, mood
state, and erroneous ideation with reformation of targeted fear structures. Applications of the concepts
advanced here to therapeutic practice and to the broader study of psychopathology are discussed.
The last two decades have brought remarkable advances in
the behavioral treatment of pathological fears and an abundance
of data on treatment outcomes. This accumulation of data, how-
ever, has not been paralleled by theoretical understanding of the
processes that relate interventions to outcome. In this article we
examine the data on treatment of fear and offer a framework for
organizing them. Within this framework we advance hypotheses
about the mechanisms of therapeutic change and consider why
treatment succeeds with some individuals and fails with others.
The search for mechanisms of fear reduction can begin with
recognition of some commonalities in how different schools of
psychotherapy view anxiety and its treatment. Regardless of their
theoretical persuasion, clinicians have long ascribed a central
role to anxiety or other unpleasant affect in the etiology and
maintenance of neurotic behavior. A basic assumption in psy-
chodynamic approaches has been that neuroses reflect attempts
to avoid disturbing experiences (Freud, 1956). In describing pa-
thology, Perls (1969) asserted that, "If some of our thoughts,
feelings are unacceptable to us, we want to disown them but
only at the cost of disowning valuable parts of ourselves ....
Your ability to cope with the world becomes less and less" (p.
1 l). Most explicit are the behaviorists who have viewed anxiety
Preparation of this manuscript was supported in part by National
Institute of Mental Health Research Grant MH31634 awarded to Edna
B.
Foa.
Critical reviews of earlier drafts of this manuscript were contributed
by Dave Barlow, Bruce Cuthbert, Uriel Foa, Dick Hallam, Peter Lang,
Isaac Marks, Richard McNally, Greg Miller, Jackie Persons, and Gall
Steketee. Their criticism and suggestions are gratefully acknowledged as
are those of the anonymous reviewers and the Editor.
Correspondence concerning this article should be addressed to Edna
B. Foa, Department of Psychiatry, Program for the Clinical Study of
Anxiety Disorders, Temple University, c/o Eastern Pennsylvania Psychi-
atric Institute, 3300 Henry Avenue, Philadelphia, Pennsylvania 19129.
disorders as continuous attempts to avoid confrontation with
fear-evoking cues (Mowrer, 1960).
Despite their theoretical differences, a common principle for
the treatment of neuroses has emerged across schools of psy-
chotherapy: the principle of exposure. Indeed, if neurotics are
avoiders who fail to recognize and/or retrieve discomfort-evoking
information about themselves or their environment, psycho-
therapy might be construed as providing a setting in which con-
frontation with such information is promoted so that changes
in affect can occur. Psychodynamically oriented therapists expose
their patients to information about unconscious conflicts, painful
memories, and unacceptable wishes through interpretation of
their behavior in therapy, of dreams, or of free associations.
Likewise, Gestalt therapists use imagery, role-enactment, dream
interpretation, and group-interaction to coax a person into "the
here-and-now," that is, to promote confrontation with infor-
mation that has been avoided. Techniques that more directly
promote confrontation with fearful events have been developed
by behavior therapists.
A wealth of evidence attests to the efficacy of exposure tech-
niques (for reviews see Foa & Kozak, 1985; Marks, 1978). They
lead to long-term improvement in about 75% of agoraphobics
(Emmelkamp & Kuipers, 1979) and obsessive-compulsives (Foa
et al., 1983); these two disorders had long been considered in-
tractable. In contrast, relaxation treatment with obsessive-com-
pulsives (Marks, Hodgson, & Rachman, 1975) and long discus-
sions of anxiety symptoms with agoraphobics (Chambless, Foa,
Groves, & Goidstein, 1980) produced little change. Long before
this experimental evidence was available, clinical observations
had led both Freud (1956, p. 399) and Fenichel (1963, p. 215)
to recognize in vivo exposure as a highly potent procedure for
treating phobias.
How does exposure help to reduce anxiety? By what mecha-
nisms might emotional change occur? The behaviorist view that
anxiety disorders are founded in abnormal associations among
20
EMOTIONAL PROCESSING
21
stimuli and responses has led to explanations of fear reduction
in terms of stimulus-response dissociation. However, limitations
of contiguity theories in explaining learning phenomena have
been widely recognized (e.g., Wagner & Rescorla, 1972), as have
also the difficulties in explaining fear acquisition and maintenance
by traditional learning theories (Eysenck, 1976; Rachman, 1976).
The argument that the "signal value" of a stimulus is an im-
portant predictor of conditioning (e.g., Grings, 1976; Kamin,
1969) has indicated the need for an "informational model" of
learning (Reiss, 1980; see also Levey & Martin, 1983). This view
seems to imply that in addition to formed associations, some
aspects of the meaning of the associated events are involved in
learning. Thus, satisfactory explanations of fear and its reduction
may require attention not only to stimulus-response associations,
but also to their meanings.
In the present article we try to explicate how exposure leads
to fear reduction. We adopt the position that fear is represented
in memory structures that serve as blueprints for fear behavior,
and therapy is a process by which these structures are modified.
We argue that two conditions are required to reduce pathological
fear: First, the fear structure must be activated, and next, infor-
mation incompatible with its pathological elements must be in-
corporated. It follows that understanding the therapeutic process
involves the identification of information that promotes fear-
activation and that modifies the fear structure. This identification,
in turn, requires theoretical unraveling of the fear structure. To-
ward this end, a model of pathological fears is discussed and a
pathway for their correction via treatment proposed)
Fear and its Modification
Structure of Fear Memory
A starting point for considering the mechanisms of exposure
therapies can be found in Lang's (1977, 1979) bioinformational
conceptualization of fear, which is couched in terms of the
"propositional representation" position (Anderson & Bower,
1974; Kieras, 1978; Pylyshyn, 1973) on the nature of cognition.
Accordingly, all knowledge can be expressed in an abstract code
representing concepts; cognitions are construed as propositional
entities and propositions are understood to be logical relations
that express concepts. This view is contrasted with some other
positions, such as that fundamental differences between visual
and verbal cues are involved in cognitions (Paivio, 1971 ), or that
word and sentence meanings are "pictures" themselves (Bugelski,
1970). Although theoretical controversy about the propositional
position flourishes (Kosslyn, 1980), and the empirical work of
Lang and his associates is not directed specifically at its resolution,
the bioinformational theory has led to fruitful investigation of
fear.
Adopting Pylyshyn's (1973) construal of a propositional net-
work as an organization of concepts related to one another by
other concepts, Lang (1977, 1979) suggested an analysis of the
fear structure into propositions. Accordingly, fear is represented
as a network in memory that includes three kinds of information:
(a) information about the feared stimulus situation; (b) infor-
mation about verbal, physiological, and overt behavioral respon-
ses; and (c) interpretive information about the meaning of the
stimulus and response elements of the structure. This information
structure is conceived of as a program for escape or avoidance
behavior.
If the fear structure is indeed a program to escape danger, we
propose that it must involve information that stimuli and/or
responses are dangerous, as well as information about physio-
logical activity preparatory for escape. Thus, a fear structure is
distinguished from other information structures not only by re-
sponse elements but also by certain meaning or information it
contains. For example, the programs for running ahead of a
baton-carrying competitor in a race and for running ahead of a
club-carrying assailant on a racetrack are likely to involve similar
stimulus and response information. That which distinguishes
the fear structure is the meaning of the stimuli and responses:
Only the fear structure involves escape from threat.
It is apparent that most people experience fear in some cir-
cumstances. What then distinguishes the structures of normal
fears from those of pathological fears? We suggest that patho-
logical structures involve excessive response elements (e.g.,
avoidance, physiological activity, etc.) and resistance to modifi-
cation. The persistence of fears may stem not only from their
marked structural coherence (as noted by Lang, 1977) but also
from impairments in mechanisms for the processing of fear-rel-
evant information (Foa & Kozak, 1985).
As a hypothetical construct, a fear structure must be investi-
gated through converging measures. In trying to account for
physiological responses measured during fear, Lang (1979) sug-
gested that fear is accompanied by physiological activity deter-
mined by the response structure that underlies it. Accordingly,
physiological responses measured during fear evocation can pro-
vide an index of the fear structure. These responses are thought
to reflect prototypes of overt behavior, that is, the much atten-
uated versions of ordinary actions reflect stored perceptual-motor
schemata for those actions. According to this view, an evoked
memory structure is influenced by the structure of the fear-
evoking material. Like memory structures, evocative material
(e.g., feared situations or their descriptions) can also be analyzed
with respect to its stimulus, response, and meaning elements.
Furthermore, the propositional structure of an evoked memory
is assumed to parallel the structure of the material that evokes
it. Physiological activity recorded during fear evocation, as well
as the self-reports of fear, are taken as measures of hypothesized
memory structures. Relations of these measures to one another
and to input variables can then be evaluated, where
input
vari-
ables are events hypothesized to evoke the structure. Thus, the
propositional structure of the evoked memory is hypothesized
to depend in part on the structure of evocative material and to
be reflected in measurable physiological efferents and self-reports.
Our position certainly does not imply that a fear structure is
entirely available to consciousness. Although certain aspects may
be identified through introspection, ample evidence (cf. Van Den
Berg & Eelen, 1985) suggests that associations among stimuli,
responses, and their meanings can exist in the absence of con-
scious knowledge about them. Just as a person may be unaware
of some response information in a fear structure (e.g., infor-
mation that underlies increased blood pressure), so also may one
The development of normal and pathological fear memory structures
will not be discussed in this article.
22 EDNA B. FOA AND MICHAEL J. KOZAK
be unaware of the meaning of those responses. This is not to say
that people are always unaware of meanings associated with
stimuli and responses, for they can indeed report beliefs and
evaluations that reflect elements in their fear structures. Because
of people's imperfect knowledge about their fear structures, non-
introspective assessment of these structures is also required. In
addition to recorded physiology, nonverbal behavior such as facial
expressions, postural adjustments, overt actions, and so on would
also be expected to reflect some elements. Any of these data can
provide a basis for hypotheses about the elements of a fear struc-
ture and the relations among them.
In the last decade, the assessment of clinical anxiety has been
influenced by Lung's (1968) proposal that anxiety involves three
systems: physiological activity, subjective report, and overt be-
havior. Accordingly, fear activation will be reflected in physio-
logical responses measurable at the periphery, in reports about
experience of fear, and/or in overt acts such as avoidance or
escape. Fear activation, however, does not always give rise to
reports of fear or to escape itself. Because we construe a fear
structure as a program to escape or avoid, it follows that activation
of fear cannot occur without preparatory changes in physiological
activity. In this article we focus on physiological indices of fear
and to a lesser extent, subjective reports; little attention is given
to overt behavioral measures. This emphasis stems mainly from
the kind of data available in studies of the process of fear re-
duction during therapy. Data on overt behavior are more likely
to appear as treatment-outcome measures. Thus, a hypothesized
structure can be validated via multisystem assessment in a variety
of ways. For example, one can provide information thought to
match the network and observe whether the responses evoked
support the construct.
Modifying an Affective Memory: Emotional Processing
We propose that regardless of the type of therapeutic inter-
vention selected, two conditions are required for the reduction
of fear. First, fear-relevant information must be made available
in a manner that will activate the fear memory. Indeed, as sug-
gested by Lang (1977), if the fear structure remains in storage
but unaccessed, it will not be available for modification. Next,
information made available must include elements that are in-
compatible with some of those that exist in the fear structure,
so that a new memory can be formed. This new information,
which is at once cognitive and affective, has to be integrated into
the evoked information structure for an emotional change to
Occur,
The hypothesized change in the fear structure can be concep-
tualized as the mechanism for what Rachman (1980) has defined
as emotional processing: a process by which emotional responses
decrease. In contrast to this definition, our view that emotional
processing involves incorporation of new information into an
existing structure allows for either increased or decreased emo-
tional responding. Indeed, exposure to information consistent
with a fear memory would be expected to strengthen the fear.
Emotional processing occurs spontaneously throughout life:
Emotional responses increase and decrease with experience. Be-
havioral treatments are designed to provide information that is
sufficiently incompatible with the fear structure to reduce fear.
Because this article is concerned primarily with processes un-
derlying fear reduction, our use of the term
emotionalprocessing
refers mostly to changes that result in fear decrement.
Indicators of Emotional Processing
To assess whether emotional processing is complete following
therapy, Rachman (1980) suggested the use of"test probes," that
is, presentations of relevant stimulus material in an attempt to
evoke an emotional reaction. For Rachman, ifa fear response is
elicited, it indicates that emotional processing has not been suc-
cessfully completed; conversely, if the probe fails to elicit fear,
emotional processing is assumed to have taken place. The test-
probe approach to the assessment of emotional processing poses
some problems. Whereas evoked fear evidences incomplete
emotional processing, the opposite is not necessarily true: Ex-
posure to fear-relevant information does not always activate an
existing fear structure (e.g., Chambless et al., 1980; Grossberg
& Wilson, 1968). Therefore, failure to evoke fear with a test
probe does not itself indicate that emotional processing has oc-
curred. Rather, it may reflect the inadequacy of the probe material
or an avoidance of the information presented. This problem with
psychometric reliability limits the validity of the test-probe ap-
proach. In addition, assessing emotional processing solely by
response to posttreatment test probes may fetter this concept in
tautological subservience to treatment outcome.
We surmount these difficulties by identifying indicators of
emotional processing (during therapy) that predict therapy out-
come and at the same time are logically independent of it. Our
solution rests on a distinction between emotional processing and
treatment outcome. Emotional processing of fear is a hypothetical
construct referring to the ongoing course of change in a fear
structure. To measure this course, behavior that directly reflects
the structure should be assessed at several points during therapy.
The choice of measures depends on how the fear structure is
conceptualized. Our view that physiological response information
is coded in the structure dictates the use of physiological measures
in addition to self-report for assessing emotional processing.
Treatment outcome is distinguished from emotional processing
in two ways. First, outcome involves an endpoint at which struc-
tural changes are assumed to have occurred, and its measures
are designed to assess the new structure. Second, whereas emo-
tional processing of fear is indicated only by measures of that
fear, treatment outcome is a broader concept. It also includes
functioning indirectly related to the fear structure that is hy-
pothesized to change as a result of fear reduction, such as, job
performance, social interactions, sleep disturbance, and general
mood state. In summary, the logical distinction between emo-
tional processing and outcome pivots on the ideas of temporal
continuity and breadth of measurement. Processing is ongoing,
requiring repeated measurement of fear; outcome is discrete,
requiring measurement at some endpoint of behavior both di-
rectly and indirectly related to the fear structure.
Data collected in various clinical studies reveal a set of re-
sponses occurring in patients who improve with exposure treat-
ment, and thus they may serve as indicators of emotional pro-
cessing. First, these patients give physiological responses and re-
ports of fear that evidence activation of fear during exposure.
Second, their reactions decrease gradually (habituate) within ex-
EMOTIONAL PROCESSING
23
posure sessions? Third, initial reactions to the feared
object
at
each exposure session decrease across sessions. Various lines of
evidence from both clinical outcome studies and laboratory ex-
periments point to the validity of these indicators.
The first indicator is activation. Lang, Melamed, and Hart
( i 970) and Borkovec and Sides (1979) found that phobic subjects
who profited most from systematic desensitization (gradual ex-
posure to feared images) showed increased heart rates during
the initial feared images; weak reactors benefited less from treat-
ment. Similarly, physiological responsiveness during flooding
(exposure to situations or images that evoke intense fear) was
found to be positively related to outcome of treatment with spe-
cific phobics and agoraphobics (Watson & Marks, 1971).
The second indicator is habituation within sessions. Decreases
in cardiac activity have generally been observed during repeated
presentations of feared material (e.g., Anderson & Borkovec,
1980; Borkovec & Sides, 1979; May, 1977). Investigating changes
in reported anxiety during imaginal flooding with obsessive-
compulsives and agoraphobics, Foa and Chambless (1978) found
that once fear was activated, self-reported anxiety decreased
within sessions in a roughly linear fashion. This pattern approx-
imated habituation patterns found during exposure to actual
feared situations for heart rate (Grayson, Foa, & Steketee, 1982;
Stern & Marks, 1973; Watson, Gaind, & Marks, 1972) and for
self-reported anxiety (Hafner & Marks, 1976; Nunes & Marks,
1975; Stern & Marks, 1973). Simple phobics who improved with
treatment showed greater heart rate habituation during imagery
of fearful material than their less successful counterparts (Bor-
kovec & Sides, 1979; Lang et al., 1970; Watson et al., 1972).
Similarly, habituation of reported anxiety during both in vivo
and imaginal exposure has been found to predict outcome for
obsessive-compulsives (Foa et al., 1983).
The third indicator is decrease in initial reactions to feared
stimuli (across-sessions habituation). This has been observed
during treatment by exposure even though increasingly difficult
situations were added in the course of treatment (Foa & Chamb-
less, 1978; Grayson et al., 1982; Hafner & Marks, 1976; Shahar
& Marks, 1980). Habituation across sessions has also been found
related to treatment outcome (Foa et al., 1983). Examples of
habituation within and across sessions are given in Figure 1.
These data provide diverse support for the proposed indicants
of emotional processing. The activation of affect, its reduction
during exposure sessions, and its decrease across sessions, appear
positively related to treatment outcome, denoting evocation and
modification of fear memories during therapy. Conversely, de-
viations from this pattern may signify that the fear structure has
been unavailable for modification or that components of a
preexisting memory have not been modified by new information.
The tactic offered here for the assessment of emotional pro-
cessing requires sufficient repeated measures to establish whether
the pattern observed during fear reduction in a given individual
corresponds to the hypothesized pattern. With too few data
points, one cannot distinguish between lack of fear evocation
and habituation when fear responses are not observed. A different
method of distinguishing between lack of evocation and fear re-
duction was employed by Levin (1982), who measured fear-rel-
evant as well as specific fear-irrelevant physiological responses
related to the imagined situation. The presence of the specific
fear-irrelevant responses was taken to indicate that the fear
structure had been accessed, whereas the disappearance of fear
responses indicated degree of habituation. Thus, the indicant of
memory accessing was logically independent of the indicant of
fear.
We have proposed that emotional processing of pathological
fear requires the activation of the fear structure and then, the
incorporation of corrective information. We have also proposed
three indicators of emotional processing: fear responses (indi-
cating that the structure has been accessed), short-term (within-
sessions) habituation, and long-term (across-sessions) habituation.
All three indicators have been related to treatment efficacy. Ac-
cordingly, conditions that enhance fear evocation and habituation
are expected to improve treatment outcome. Such conditions
are examined in the next section.
Variables That Influence Accessing and Modification
of Affective Memory
Content of the Evocative Information
A fear memory is accessed when a fearful individual is pre-
sented with fear information that matches some of the infor-
mation structure in memory (Lang, 1977). This information may
be about the feared situation, the person's responses in the sit-
uation, or their meaning. Lang (I 977) suggested that some critical
number of information units must be matched for the entire fear
memory to be activated, and that some information elements
may be especially important in evoking the fear structure. Intense
phobias, he proposed, may be characterized by strongly coherent
structures that can be evoked with minimally matching infor-
mation. For example, the sight of a coiled garden hose may elicit
intense fear in a snake phobic; a warm sensation may evoke a
panic attack in the agoraphobic who fears physiological sensations
of anxiety.
Invoking a matching explanation to account for fear activation
risks circularity in the absence of other ways to assess the struc-
ture. To obviate such circularity one must first identify the struc-
ture from self-reports, behavioral observations, and so on. Data
about responding to matched information can then be used to
validate hypotheses about the structure. Accordingly, situations
that vary in their degree of similarity to the hypothesized struc-
ture would elicit varying degrees of fear.
In studying variables that influence the accessing of fear struc-
tures, Lang and his associates trained subjects to form fearful
fantasies through practice with scripts that included descriptions
of either a stimulus context only (e.g., a green snake on a rock),
or of both a stimulus and specific physiological responses (e.g.,
your heart pounds). Subjects who had received the training in-
volving responses and who were later tested with scripts con-
taining response descriptions showed greater physiological ac-
2 Habituation has often been used to denote a short-term sensory effect,
whereas extinction usually refers to the longer term unlearning of a con-
tingency. This distinction is controversial in the light of the findings that
habituation is retained over time (e.g., Groves & Lynch, 1972). Our use
of the term
habituation
refers simply to response decrement. This is
consistent with the Thompson and Spencer 0966) usage of the term:
"given that a particular stimulus elicits a response, repeated applications
of the stimulus result in decreased response (habituation)" (p. 18).
24 EDNA B. FOA AND MICHAEL J. KOZAK
tivity and reported higher fear than subjects trained and tested
with stimulus descriptions only (Lang, Kozak, Miller, Levin, &
McLean, 1980). This finding can be interpreted to indicate that
scripts that include response references enhance accessing of a
fear structure, but only in conjunction with prior response
training.
In another experiment (Lang, Levin, Miller, & Kozak, 1983),
snake and socially anxious subjects received either response
training or stimulus training. When later presented with both
fear-relevant and fear-irrelevant image scripts, those who had
been trained to focus on responses showed increased physiological
responding, especially during fear-relevant images. These results
seem to indicate that greater matching of input information with
a preexisting fear memory enhances fear evocation. Training
subjects to focus on responses may have helped them to attend
to response elements in the scripts, thus enhancing the match
between input information and stored information, thereby call-
ing up the structure more fully. Similarly, the larger responses
to fear-relevant scripts were hypothesized to reflect a better match
of response scripts with a fear structure than that between re-
sponse scripts and a nonfear structure (Lang et al., 1983). In
summary, it appears that promoting focus on response infor-
mation (which corresponds to that in the fear structure) increases
the likelihood of fear evocation.
Individuals differ in their responsiveness to imagery training.
Miller et al. (1981) found that, prior to training, good and bad
imagers did not differ with respect to their physiological responses
during fear imagery. After training, only the good imagers showed
increased physiological responding. In other studies, with phobic
populations, good imagers showed greater physiological re-
sponding to fear-relevant scripts than did poor imagers, even
when no imagery training was administered (Levin, 1982; Levin,
Cook, & Lang, 1982). Thus, good imagers, if phobic, do not
seem to need training to respond to phobic material. These find-
ings suggest individual differences in the ability to use fear-rel-
evant information to access a fear memory. Such a preexisting
capacity may influence emotional processing during exposure.
How does the information content of exposure influence fear
modification? Little attention has been devoted to analyzing the
content of the information presented during exposure sessions.
Figure 1. Data from several experiments on habituation during exposure. A: Mean values for six agoraphobics
treated with eight 90-rain sessions of imaglnal flooding (from Foa & Chambless, 1978). B: Mean values for
24 speech phobic volunteers treated with four 50-rain sessions of imaginal flooding (from Chaplin & Levine,
1980). C: Data for a single cat phobic during one session of in vivo exposure (from Watson et al., 1972). D:
Means for 16 obsessive-compulsives during in vivo exposure to contaminants (from Grayson et al., 1982).
(These data illustrate habituation of autonomic and self-report indices of fear and suggest congruence of
habituation patterns across measures and procedures.)
