Extinction in Human Fear Conditioning
Dirk Hermans, Michelle G. Craske, Susan Mineka, and Peter F. Lovibond
Although most extinction research is conducted in animal laboratories, the study of extinction learning in human fear conditioning
has gained increasing attention over the last decade. The most important findings from human fear extinction are reviewed in this
article. Specifically, we review experimental investigations of the impact of conditioned inhibitors, conditioned exciters, context
renewal, and reinstatement on fear extinction in human samples. We discuss data from laboratory studies of the extinction of
aversively conditioned stimuli, as well as results from experimental clinical work with fearful or anxious individuals. We present
directions for future research, in particular the need for further investigation of differences between animal and human conditioning
outcomes, and research examining the role of both automatic and higher-order cognitive processes in human conditioning and
Key Words: Associative learning, fear conditioning, extinction, re-
newal, reinstatement, exposure
of the unconditioned stimulus (US) with which it was previously
paired. Following a series of CS-only presentations, the condi-
tioned response (CR) that was established during the preceding
acquisition phase gradually diminishes. For instance, during fear
conditioning, the contingent presentation of a tone CS and an
aversive shock (US) will alter the response to the former stimulus;
i.e., the CS will come to evoke fear responses when tested in
absence of the US. Subsequent nonreinforced presentations of the
tone CS will lead to a reduction in these acquired fear CRs.
Although there are many aspects of extinction that are still not
fully understood (Rescorla 2001), the past 15 years have seen
significant developments in the conceptualization of the pro-
cesses of extinction. For instance, insights into the anatomical
and biochemical basis of the extinction of conditioned fear have
been growing steadily (for an overview, see Barad, in press;
Myers and Davis 2002). Moreover, some of these insights have
directly informed possible future treatments for anxiety disorders
(Ressler et al 2004; Walker et al 2002).
Another development that has important theoretical and
clinical implications is the growing recognition that the dimin-
ished (or even vanished) responding that is observed after
extinction, whether measured behaviorally or psychophysiologi-
cally, does not reflect “unlearning” at the process level (e.g.,
Bouton 1988; Delamater 2004). Within an associative approach
to Pavlovian conditioning, this means that extinction does not
(completely) destroy the original CS-US association. Empirical
support comes from a large amount of animal work showing that
under the right conditions, extinguished conditioned responses
can be recovered (for an overview, see Bouton 1994, 2004; Falls
1998). Examples of such conditions are the mere passage of time
(spontaneous recovery) (Pavlov 1927; Rescorla 2004), context
change after extinction (renewal) (e.g., Bouton and Ricker 1994),
or the presentation of postextinction US-only trials (reinstate-
xtinction can be described as a procedure, as a result, or as
a process. As a procedure, extinction refers to the repeated
presentation of a conditioned stimulus (CS) in the absence
ment) (e.g., Rescorla and Heth 1975). Whereas these phenomena
support the view that extinction does not completely destroy the
CS-US association, more direct tests of the CS-US association in
animals actually suggest that extinction leaves the CS-US associ-
ation intact (Delamater 1996; Rescorla 1996).
These developments in extinction research have been largely
based on work with laboratory rodents. Hence, it remains an
open question as to whether these findings can be generalized to
humans. Clearly, there are large differences in brain anatomy
between rats on the one hand and humans on the other. Given
the often highly specialized functions of specific brain regions,
there are no a priori reasons to assume that identical processes
are involved in extinction in human and nonhuman animals.
Although there is now a fair amount of work concerning the
neurobiological basis of fear acquisition in humans (e.g., Phelps
et al 2001), the study of the neural mechanisms of extinction
learning in humans has received less attention (e.g., Knight et al
2004; LaBar et al 1998; Phelps et al 2004). For an overview of this
literature, we refer to Rauch et al (in press).
Another reason why animal findings might not readily trans-
late to human samples is that humans are endowed with a
cognitive system that is capable of a symbolic and propositional
analysis of conditioning experiences (see De Houwer et al 2005;
Lovibond and Shanks 2002). The extent to which higher level
propositional processes are involved in associative learning in
general, and in fear conditioning and extinction more specifi-
cally, has been a matter of debate (see Beaver et al 2005; Carter
et al 2003; Clark and Squire 1998; Lovibond 2003, 2004; Öhman
and Mineka 2001).
The translation of animal extinction research to human sam-
ples has received more attention lately due to the clinical
relevance of the topic. That is, extinction of conditioned fear can
be viewed as a laboratory analog for the study of exposure
treatment for anxiety disorders (Bouton et al 2001; Davey 1997;
Eelen et al 2001; Mineka 1985). Hence, insight into the mecha-
nisms that are responsible for fear extinction may have clinical
significance. In addition, knowledge of the conditions that
facilitate or hamper extinction learning may help to sharpen
exposure treatments in such a way that treatment outcome can
be maximized in the short run (therapy efficacy) as well as in the
long run (relapse prevention). In this article, we review recent
experimental work on extinction of human fear learning. In
general, two different approaches have been employed in this
context. The first approach mainly uses Pavlovian preparations
to study procedures and processes of acquisition and extinction
in the laboratory. These experiments typically employ a differ-
ential aversive conditioning procedure with nonclinical groups.
The second approach takes these observations one step further
by translating the extinction principles to exposure therapy
From the Katholieke Universiteit Leuven (DH), Leuven, Belgium; University
of California (MGC), Los Angeles, California; Northwestern University
Address reprint requests to Dirk Hermans, Department of Psychology, Uni-
versity of Leuven, Tiensestraat 102, 3000 Leuven, Belgium; E-mail:
BIOL PSYCHIATRY 2006;60:361–368
© 2006 Society of Biological Psychiatry
protocols in clinical groups with established fears. The combina-
tion of these two methods offers a fruitful approach to study
fundamental as well as clinical aspects of extinction of condi-
tioned fear (Vansteenwegen et al, in press).
Experimental Laboratory Studies
Protection from Extinction and Super-Extinction
Various theoretical models of Pavlovian conditioning allow
concrete predictions about the conditions that influence the
process of extinction and either facilitate or hamper its behav-
ioral outcome. Examples are the effects of trial spacing in
extinction, the impact of the associative strength of contextual
cues, the nature of the CS (e.g., fear-relevance), the presentation
of other excitatory CSs during extinction, or the introduction of
inhibitory CSs. Particularly the latter phenomenon has received
great interest within the clinical field, as it is widely believed that
such inhibitory “safety-signals” can have a deleterious effect on
the progress of exposure. For instance, the presence of the
therapist during exposure to anxiety-provoking cues can act as a
conditioned inhibitor (“when the therapist is present, nothing
bad will happen”). The presence of the therapist may lead to
rapid symptom reduction, but a return of anxiety may occur
when the client subsequently confronts the stimuli alone. In
terms of the Rescorla-Wagner model (Rescorla and Wagner
1972), adding an inhibitory stimulus during extinction blocks
extinction of the excitatory CS. Given that the inhibitory stimulus
has a negative associative strength that cancels out the positive
associative strength of the excitatory stimulus, this will lead to
little or no discrepancy between what is predicted by the sum of
both cues (close to zero) and what actually happens on extinc-
tion trials (zero). As a result, there is little change in associative
strength of the stimuli, and in particular, the excitatory CS should
retain a positive associative strength, revealed when it is tested
alone. This phenomenon is known as protection from extinction
and has been demonstrated in animal research using procedures
of aversive Pavlovian conditioning (Soltysik et al 1983) and
autoshaping (Rescorla 2003).
Employing a differential aversive conditioning procedure with
Lovibond et al (2000) (Experiment 1) tested the protection from
extinction hypothesis in a group of undergraduate students.
In the acquisition phase of this study, two CSs (stimulus C and
stimulus D) were contingently followed by the US and were
hence established as valid predictors of this aversive stimulus.
Stimulus D was subsequently extinguished by presenting it alone
without shock. Stimulus C was also presented without shock, but
during the extinction trials, it was accompanied by a stimulus K
that had previously been established as an inhibitory stimulus
(i.e., following a procedure of conditioned inhibition; A?/AK?).
Following these extinction trials, stimulus C and stimulus D were
presented alone in a test phase (see Figure 1). Stimulus A and
stimulus B were control stimuli that had been consistently paired
with shock and no shock, respectively. Compared with stimulus
A, presentation of stimulus D without shock led to a substantial
reduction in both shock expectancy (left panel of Figure 1) and
electrodermal responses during stimulus D (right panel of Figure
1). By contrast, stimulus C showed little extinction. The presence
of stimulus K had protected it from extinction.
Human studies have not yet demonstrated that the added
stimulus needs to be inhibitory to interfere with extinction. In
fact, it has been shown that an excitatory stimulus can have the
same effect. According to the Rescorla-Wagner model, adding a
conditioned exciter to the target CS during extinction should
actually enhance extinction of the target CS. Although there is
evidence for such super-extinction in animal studies (e.g., Res-
corla 2000; Thomas and Ayres 2004), two studies in human
participants have demonstrated protection from extinction by an
excitatory CS (Lovibond et al 2000, Experiment 2; Vervliet et al
2005). Hence, future research should target possible differences
between human and nonhuman conditioning procedures con-
cerning the effect of nonreinforcement in the presence of
another excitatory stimulus. For example, the ceiling effect
imposed by ethical constraints on US magnitude in humans may
disrupt the normal process of additivity that underpins the
Rescorla-Wagner model (Mitchell and Lovibond 2002). Alterna-
tively, it is possible that humans are particularly sensitive to
generalization decrement or contextual changes, such that ex-
tinction is highly specific to the precise circumstances under
which nonreinforcement has occurred.
Contextual and Stimulus Specificity of Fear Extinction
The studies described above show how differences in extinc-
tion procedures can affect responding to the extinguished CS
later at test. As shown by the phenomenon of protection from
extinction, responding that disappeared as a result of apparently
successful extinction may nevertheless reappear afterward. As
was already pointed out, context changes might also lead to
recovery of apparently extinguished responses. The phenome-
non of renewal is a good example of the context-dependency of
extinction, which is most frequently observed in ABA-renewal
designs. Bouton and Swartzentruber (1986) demonstrated that
conditioned responding is renewed when animals are tested in
the original acquisition context after extinction in a different
context. Renewal effects also are obtained when, after extinction
in a different context, transfer of extinction is tested in a third
context (ABC-renewal) and in AAB-renewal designs in which
acquisition and extinction take place in an identical context but
testing is executed in a new context (Bouton and Ricker 1994).
In a recent study, Vansteenwegen et al (2005) studied ABA-
renewal in a human differential fear conditioning preparation.
The CSs in this study were line drawings of pictorial faces.
Contexts were manipulated by turning the central lighting of the
experimental room on or off. An ABA-renewal group and an
AAA-control group were compared. The renewal group received
acquisition in one context (dark/light) and extinction in the
Figure 1. Mean online shock expectancy ratings (left panel) and mean
change in log skin conductance level during test trials from a protection
no shock. Stimuli C and D were initially paired with shock and then extin-
guished. During extinction, stimulus D was presented alone, whereas stim-
ulus C was accompanied by a safety signal K. In the test phase data shown,
from complete extinction. (Reproduced with permission from Lovibond
et al 2000).
362 BIOL PSYCHIATRY 2006;60:361–368
D. Hermans et al
opposite context (light/dark) and was tested again with the CSs
in the original acquisition context. In the control group, all
phases were executed in the same context (which could be
either light or dark, depending on the counterbalancing scheme).
Although no change in differential electrodermal responding was
observed when switching the context after acquisition (i.e., there
was generalization of acquisition to the new context), there was
a clear return of conditioned responding with the context change
after extinction. No such return was observed in the control group.
Retrospective US-expectancy ratings confirmed this data pattern.
Thus, in line with animal research, this study demonstrated that
previously extinguished fear responses can return when tested in a
context different from the extinction context. Similar results were
recently obtained in a study by Milad et al (2005).
In a subsequent study, Vervliet et al (2005) investigated
whether recovery of extinguished fear responses is observed
when stimuli rather than contexts were manipulated. In this
experiment, participants learned that one of two geometrical
figures (e.g., a parallelogram or a triangle) was followed by an
aversive US (shock). During the subsequent extinction phase,
variants of these CSs were presented (angles were slightly more
sharp/blunt) for the experimental group. For the control group,
the original acquisition CSs were employed. Testing involved the
acquisition CSs for both groups. Differential acquisition effects
generalized substantially to the extinction stimuli in the experi-
mental group. However, similar to what is observed in ABA-
renewal, participants from the experimental group hardly gen-
eralized from what was learned with the extinction stimuli to the
original acquisition stimuli during final testing. These findings
were later replicated in a follow-up study (Vervliet et al 2004).
To what extent the same mechanisms drive these contextual
effects in humans and animals needs to be subjected to further
experimental analysis (Vansteenwegen et al, in press). As already
indicated, the importance of higher-order cognitive processes in
humans also needs closer examination. For example, it may be
possible to establish or reverse safety signals in humans through
verbal instruction (e.g., Norrholm et al, unpublished data).
Reinstatement Following Fear Extinction
Another paradigm that has received attention in the context of
extinction research is reinstatement. In a reinstatement proce-
dure, a previously conditioned stimulus is first extinguished until
conditioned responding disappears. Next, a few USs are admin-
istered in the absence of the CS. When the previously extin-
guished CS is finally re-presented, conditioned responding to the
CS (partially) reappears (e.g., Rescorla and Heth 1975; Richard-
son et al 1999). It is assumed that the US-only presentations
reinstate the CS-US association that is still present but dormant
after extinction. Research has pointed to the crucial role of the
context for the emergence of reinstatement effects in animals.
Specifically, reinstatement effects only occur when USs are
presented in the original conditioning context (for an overview
and discussion, see Bouton 2004; Westbrook et al 2002). As a
matter of fact, in a human reinstatement study using a single-cue
paradigm, LaBar and Phelps (2005) demonstrated reinstatement
effects to be context-specific.
In a series of differential fear conditioning experiments,
Hermans and colleagues studied reinstatement in humans. In the
first experiment, Hermans et al (2005) used pictures of neutral
faces as CSs and an electrocutaneous stimulus as the US. During
acquisition training, one stimulus (CS?) was contingently fol-
lowed by the US, while a second stimulus (CS?) predicted the
absence of the US. Subsequently, both CSs were involved in a
lengthy extinction phase. For half of the participants (reinstate-
ment group), extinction training was followed by the presenta-
tion of four unpredicted USs (in the same context). For the
control group, no such US-only presentations were scheduled.
As can be seen from Figure 2, the groups showed equal acquisition
the CSs were tested after reinstatement, there was a selective partial
return of US-expectancy for the CS? in the reinstatement group.
This effect was absent in the control group. Analogous to the
US-expectancy ratings, selective reinstatement was observed in the
fear ratings as well. These findings were replicated in a study by
Dirikx et al (unpublished data).
However, a limitation of both of these studies is that the
reinstatement effects were restricted to self-report measures of
fear and US-expectancy. In a study by Dirikx et al (2004),
reinstatement was also demonstrated using a reaction time
measure of fear that indexed allocation of attentional resources.1
In this task, participants press a key as soon as a tone probe is
1The reaction time task that was used in this experiment was originally
developed by Dawson et al (1982) and has been successfully employed
as a measure of allocation of processing resources to the CS?/CS? in
studies of human Pavlovian conditioning (see also Hermans et al 2005;
Lipp et al 1993). There has been a growing interest in this type of
procedure. Other reaction time procedures that have been used in
human fear conditioning research include the visual dot probe reaction
time task (e.g., Beaver et al 2005) a n d t h e exogenous cuing paradigm
(Koster et al 2004). To assess the affective meaning of the conditional
stimuli, the affective priming procedure has been employed as a
nonverbal measure (e.g., Hermans et al 2002).
condition (reinstatement group, control group) and moment (postacquisi-
tion, postextinction, postreinstatement) (difference score ? CS? ? CS?).
stimulus; CS?, stimulus that was contigently followed by US during acqui-
sition; CS?, stimulus that was never followed by US during acquisition.
D. Hermans et al
BIOL PSYCHIATRY 2006;60:361–368 363
presented. It was demonstrated that during acquisition, response
latencies were significantly slowed when tones were presented
during the CS? as compared with the CS?. This difference
disappeared as a result of extinction training. After reinstatement,
however, a selective slowing for the CS? (as compared with the
CS?) was observed again in the reinstatement group.
Another intriguing finding in this series of reinstatement
studies is that the (extinction resistant) negative valence of the
CS? (or the affective discrimination between the CS? and the
CS?) was predictive of the extent to which fear returned.2
Hermans et al (2005) showed that to the extent that the CS? had
a more negative valence after extinction, there was a more
significant return of fear responses. The more negative the CS?,
the more recovery of the extinguished fear responses. This effect
could only be attributed to the negative valence of the CS? and
was unrelated to other characteristics of the CS? after extinction.
Similar findings were reported by Dirikx et al (unpublished data)
who showed that the valence of the CS? or the difference in
valence between CS? and CS? (Dirikx et al 2004) was predictive
of the amount of return of fear. Depending on the study, this
predictive power was restricted to the reinstatement group
(reinstatement effect) or concerned both groups as a whole
(general return of fear). These results clearly show that negative
stimulus valence functions as a separate source for the recovery
of extinguished fear responses. A possible explanation for these
findings is that insofar as fear is based on an orthogonal
combination of negative valence and high arousal (e.g., Lang et
al 1990), extinction might change the arousal component but not
the valence component According to that view, adding arousal
through US-only presentations would reinstate fear to the extent
that the CS is still negative in valence.
Based on this overview of experiments, we can conclude that
a number of extinction phenomena that were previously
demonstrated in animal research can be replicated in experi-
mental laboratory studies in humans (e.g., protection from
extinction, renewal, reinstatement). The fact, however, that
similar procedures lead to similar effects does not imply that
similar processes are at work. Further research is needed to
illuminate similarities and differences at the process level.
Nevertheless, these findings have clear clinical implications
about what procedures could be employed to facilitate exposure
and reduce return of fear.
Experimental Clinical Studies
The experimental investigation of extinction procedures in
human phobic samples is immediately disadvantaged by starting
from the stage of fear reduction rather than fear acquisition.
Thus, the circumstances of acquisition are often unknown and
there is usually no opportunity to test in the original context.
Nonetheless, such studies are of great importance because of
their potential relevance for enhancing treatments for anxiety
disorders. Moreover, replicable effects have been observed. This
section reviews experimental investigations of conditioned in-
hibitors, conditioned exciters, context renewal, and reinstate-
ment effects in human phobic samples.
Safety Signals and Protection from Extinction
Although clinicians are typically encouraged to wean their
clients from safety signals (such as the presence of a therapist,
pill bottles, etc.) during exposure therapy (e.g., Barlow and
Craske 1994), the role of conditioned inhibitors during exposure
therapy for phobias remains poorly studied. Investigations to
date have either failed to experimentally establish safety signals
as conditioned inhibitors and/or have confounded safety signals
with the primary conditional/phobic stimulus.
Sloan and Telch (2002) reported that claustrophobic partici-
pants who received an exposure treatment in which they were
encouraged to use safety signals reported more fear at posttest
and follow-up than those encouraged to focus on their fear
during exposure. In a subsequent study, Powers et al (2004)
found that the perception of safety (i.e., availability of safety
behaviors regardless of whether they were used) rather than use
of safety was detrimental to treatment outcome, since level of
fear reduction was unaffected by actual use of safety behaviors.
may have been attributable to distraction and purported safety
signals were not established experimentally as safety signals. More-
over, the safety signals (i.e., opening a window, unlocking a door
lock) constituted behaviors that in essence degraded the condi-
tioned stimulus (i.e., enclosed situations for claustrophobic partici-
pants). Hence, they more closely represent avoidance responses
rather than safety signals.
However, avoidance responses may share some functional
properties with Pavlovian safety signals. Consistent with this
idea, Salkovskis and his colleagues (e.g., Salkovskis 1991; Wells
et al 1995) have provided evidence that “within-situation safety
behaviors,” equivalent to avoidance responses, interfere with the
benefits of exposure therapy. Specifically, they showed that
teaching anxious clients to refrain from these behaviors leads to
greater fear reduction after an exposure session. These research-
ers argue that clients attribute the absence of harm to their safety
behaviors, rather than reducing their appraisal of the threatening
stimulus. An alternative interpretation of these results derives
from a protection from extinction perspective. Exposure to a
phobic stimulus while using safety signals (avoidance responses/
conditioned inhibitors) partially protects the phobic stimulus
from extinction. Clearly, much more direct investigation is
needed on the effects of safety signals and avoidance responses
during exposure, especially given the very direct implications for
As noted earlier, two human laboratory studies have failed to
show that concurrent conditioned exciters enhance extinction. In
the realm of clinical phobias, inclusion of multiple exciters
during exposure therapy is sometimes advocated, such as occurs
when interoceptive exposure (e.g., consuming caffeine) is incor-
porated with exposure to feared situations (e.g., traveling by
subway) for the treatment of panic disorder with agoraphobia
(Barlow and Craske 1994). However, no studies have directly
evaluated multiple exciters in a clinical population.
Contextual Effects on Return of Fear
By far, the majority of research to date has focused on context
renewal effects in human phobic samples, wherein contexts
associated with successful exposure treatment are assumed to
activate nonfearful memories and nontreatment contexts are
assumed to elicit fearful memories and return of fear. Studies to
date have evaluated effects of contextual shifts on renewal in
individuals who are highly fearful of spiders, who undergo
graduated exposure therapy with a live spider using participant
2In line with findings from nonaversive conditioning procedures (for an
overview, see De Houwer et al 2001), it has been demonstrated that
differential fear conditioning procedures can lead to an affective shift
in the valence of the CS? (which becomes a more negatively
valenced stimulus). This conditioned valence seems relatively resis-
tant to extinction (see Hermans et al 2002; Lipp and Purkis 2005).
364 BIOL PSYCHIATRY 2006;60:361–368
D. Hermans et al
modeling, conducted within one session. Outcome is assessed
immediately posttreatment, and retesting occurs 1 to 2 weeks
later. Fear typically is measured via self-report, behavioral avoid-
ance, and heart rate during a behavioral approach test to a live
In general, context renewal effects have been observed when
participants are retested 1 to 2 weeks later in a context that differs
from the treatment context (e.g., Rodriguez et al 1999; Mineka et
al 1999). The effects became stronger with more distinctly
different contexts and with various other improvements to
methodology. Specifically, Mystkowski et al (2002) compared an
outside context to the standard laboratory context. As was done
in the study by Mineka et al (1999), the contexts were addition-
ally differentiated by the sex of the experimenter and salient
visual cues (i.e., color of treatment materials and therapist lab
coats), and to overcome prior methodological limitations, self-
report fear measures were administered in the experimental
context (versus the instruction room) throughout the study. Also,
this study utilized a within-subjects paradigm, similar to designs
sometimes used in the animal conditioning literature (e.g.,
Bouton and Brooks 1993) to maximize power. Thus, 1 week
later, participants were tested for return of fear in both the
original treatment context and the different context in a counter-
balanced order (AB or BA). Self-reported return of fear was
greater when individuals confronted a previously feared stimulus
in a context different than the extinction context (see Figure 3),
and the effect sizes and power were much larger than in prior
context studies. However, the findings were restricted to self-
reported fear obtained during behavioral approach tasks, and the
B-B/A condition failed to produce renewal, for reasons that are
not immediately clear. In addition, these studies, in general, do
not find context effects in terms of how closely participants
approach the phobic stimulus. This may be due to ceiling effects,
because exposure treatments effectively eliminated behavioral
avoidance (Mystkowski and Mineka, in press).
Context-based renewal also has been demonstrated via use of
videotaped presentations. Specifically, Vansteenwegen et al (un-
published data) randomized spider-fearful individuals to condi-
tions of videotaped exposure to a spider in a specific location of
a house or videotaped exposure to locations of a house without
the spider. Measures included electrodermal responding and
disgust ratings, because disgust is elevated in persons phobic of
spiders (e.g, Matchett and Davey 1991; Tolin et al 1997). Not only
did videotaped exposure to the spider lead to decreased elec-
trodermal responding and disgust ratings compared with the
control group, but both indices subsequently increased when
tested with a videotape showing the same spider in a new
context (e.g., a location in the house that was not previously
seen) relative to the original exposure context.
Bouton and Swartzentruber (1991) suggested that mismatch
of internal states during treatment and follow-up, manipulated
via drug state, can also lead to significant return of fear. This has
direct relevance to the use of psychotropic medications in
combination with exposure therapy for anxiety disorders and
may explain the high rates of relapse when medications are
withdrawn (e.g., Marks et al 1993), particularly those medications
that have a very distinct effect on internal state (i.e., high-potency
Initial failure to induce context-based renewal via alprazolam
(Zoellner and Craske, unpublished data) was most likely attrib-
utable to dosage levels (.25 mg) that were too low to generate a
sufficiently salient context. Evidence for internal context speci-
ficity of extinction was demonstrated in a study that manipulated
drug state through caffeine versus placebo ingestion for individ-
uals fearful of spiders (Mystkowski et al 2003). Participants
received one session of exposure therapy under the influence of
a randomly chosen drug condition (i.e., placebo or caffeine, 4
mg/kg of body weight) and were reassessed 1 week later after
ingesting a drink mixture that was either the same or different
than the drink ingested during the previous treatment session. As
predicted, participants experiencing incongruent drug states
exhibited significantly greater self-reported return of fear, mea-
sured during a behavioral approach task, from posttreatment to
follow-up than those participants experiencing congruent drug
states. Interestingly, the effects were comparable whether the
shift was from caffeine to placebo or vice versa. However, the
effect sizes (calculated as the difference between mean pretreat-
ment and posttreatment scores divided by the pooled standard
deviations) (Cohen 1988) were smaller than those achieved via
the inside/outside context shifts (Mystkowski et al 2002; Myst-
kowski and Mineka, in press).
In summary, a series of experimental studies have consis-
tently demonstrated context specificity effects of exposure ther-
apy in circumscribed phobias, with larger effects obtained when
more distinctly different contexts were employed. Conceivably,
even larger effects might have been obtained if retesting had
occurred in the original fear acquisition context. This is because
in the animal literature the contextual control of return of fear is
stronger when testing for renewal of fear occurs in the context in
which the fear was originally conditioned than when tested in a
new context (Bouton and Brooks 1993).
Attenuation of Context-Based Return of Fear
An obvious implication of these studies on the effects of
contextual change on enhancing return of fear would seem to be
to conduct exposure therapy in multiple contexts so that extinc-
tion memories can be cued by multiple contexts. Unfortunately,
however, animal studies on this topic have revealed very incon-
sistent results, with only some finding that extinction in multiple
contexts has reduced fear renewal in a final novel context
relative to conducting extinction in only one context (see Bouton
et al, in press, for a review). In a recent study of human fearful
participants, Vansteenwegen et al (unpublished data) compared
multiple context exposure treatment, in which spider-fearful indi-
viduals viewed video fragments of a spider in different locations of
Figure 3. Mean self-reported fear (SUDS) for spider-fearful subjects treated
and tested in contexts B and A. Follow-ups 1 and 2 refer to follow-up tests
later that day, the first of which was in the same context as extinction
treatment, while the second was in a different context. (Reproduced with
permission from Mystkowski et al 2002). SUDS, subjective units of distress.
D. Hermans et al
BIOL PSYCHIATRY 2006;60:361–368 365
a house, to single context exposure in which participants viewed
fragments always in the same location. Although disgust ratings
evinced context renewal effects in both groups on subsequent
retesting to video fragments of a spider in a novel house location,
electrodermal responding showed a renewal effect in the single
context exposure group only.
Because it is not always feasible to conduct exposures in
original fear acquisition contexts or multiple contexts, Myst-
kowski et al (in press) sought to investigate whether a contex-
tual-based return of fear could be counteracted via “mental
rehearsal.” Mental rehearsal of extinction contexts (e.g., the
therapist, treatment information, and the physical surroundings
where treatment took place) may override veridical context
shifts. Such results might be anticipated based on findings in the
verbal memory literature. Specifically, participants who are in-
structed to recall the original learning environment just prior to
free recall of a list of words in an unfamiliar environment
experience a release from contextual dependence otherwise
observed following contextual shifts and perform identically to
participants tested in the original learning environment (Smith
Spider-phobic participants were treated and followed up in
the same context (i.e., matched context groups) or treated and
followed up in different contexts (i.e., mismatched context
groups). Half of the participants in each of these two groups
were instructed to mentally rehearse the treatment context and
the material learned in that context before entering the test
context at follow-up, whereas the other half was instead asked to
recall a neutral scenario. Self-report data replicated previous
research on contextually driven return of fear, with strong effect
sizes between groups and a high degree of statistical power.
Furthermore, participants who mentally rehearsed the treatment
context before encountering the phobic stimulus in a new
context at follow-up had less return of fear than those who did
not (see Figure 4). Indeed, their fear levels were comparable to
those obtained in participants tested at follow-up in the original
Other research has demonstrated the value of objects as
retrieval cues in relation to reactivity to alcohol cues. Specifically,
Collins and Brandon (2002) showed that the return of alcohol
cue reactivity after extinction due to a context change could be
reduced by the use of retrieval cues that were present during
extinction. However, mental rehearsal effects would seem to
have more practical value than retrieval objects, given that fear
cues might be encountered at times when retrieval objects are
not available (Mystkowski and Mineka, in press). Notably, these
human studies corroborate evidence from rodent studies show-
ing that retrieval cues during extinction trials attenuate context-
based renewal effects and more so than relatively novel cues or
cues that were present during conditioning rather than extinction
(Brooks and Bouton 1994).
The experimental investigation of reinstatement effects is
prohibited by ethical constraints with human phobic samples.
However, naturalistic observations support the notion of fear
reinstatement. For example, Wade et al (1993) found that panic
disorder and agoraphobia clients who experienced more nega-
tive life events 3 to 5 years after a 12-week behavioral treatment
were functioning less well than those who experienced fewer
negative life events. Similarly, relapsers from treatment for
obsessive-compulsive disorder reported more distress over inter-
vening life events since the end of treatment than clients who
maintained their treatment gains (Steketee 1993). Notably, these
observations suggest that reinstatement effects may occur with
aversive events that are distinctly different from the aversive
event involved in initial fear acquisition, as was found by
Rescorla and Heth (1975) (Experiment 3) in rodents.
Although most extinction research is conducted in animal
laboratories, the study of extinction learning in human fear
conditioning has received more attention during the last decade.
We have reviewed experimental investigations of the impact of
conditioned inhibitors, conditioned exciters, context renewal,
and reinstatement on fear extinction in human samples. The
results of these studies have generally confirmed predictions
derived from animal research, with one exception being the
effect of conditioned exciters during extinction training. First and
foremost, however, the findings from human samples support
the view that extinction does not entail unlearning and that the
changes, or the presentation of US-only trials after extinction
training can lead to a partial recovery of the apparently “extin-
guished” fear responses. These findings raise the immediate
question of how the procedure of extinction can be altered to
improve its behavioral outcome. The work on extinction in
multiple contexts and use of reminder cues has been important
in this respect, not only because of its theoretical relevance but
also because of the potential clinical applications.
Given that research has established comparable findings in
human and nonhuman animals, we see a number of directions
for future research. First, attention should be focused on those
areas of divergence between research with humans and
laboratory rodents. We already pointed to discrepancies con-
cerning the effects of adding extra exciters during extinction
(Lovibond et al 2000; Rescorla 2000; Vervliet et al 2005).
Another example is AAB-renewal. Although the renewal effect
is generally less pronounced in AAB designs (as compared with
ABA designs) in animal studies, human extinction studies have
failed to replicate the AAB effect (Vansteenwegen et al 2005;
Vervliet et al 2004).
Irrespective of whether these discrepancies are replicable or
not, a second important research focus will be to study corre-
Figure 4. Mean self-reported fear (SUDS) for spider-fearful subjects from
follow-up contexts. A plus sign (?) after the group name indicates treat-
ment context mental reinstatement instructions, compared with a minus
sign (?) for nontreatment context mental reinstatement instructions. (Re-
units of distress; BAT, Behavioral Approach Test.
366 BIOL PSYCHIATRY 2006;60:361–368
D. Hermans et al
spondences and divergences at the process level. Similar out-
comes do not necessarily reflect similar processes. As already
indicated, the role of higher-order, propositional processes in
human conditioning deserves further exploration. Although no
one will dispute that reasoning processes are an essential
component of how humans approach the conditioning prepara-
tions in which they participate, the extent to which these
propositions have a causal impact needs further examination. It
is our view that these studies should not assume an all-or-none
approach but should focus on the conditions under which
propositional mechanisms have a key influence compared with
the conditions under which their impact is less pronounced
(Hermans and Van Gucht, in press).
The research that we reviewed stemmed from two related
approaches: experimental laboratory research in nonclinical
groups and experimental clinical work in anxious individuals.
We believe that the combination of both approaches, together with
animal conditioning work, provides a solid basis to answer the
many research questions that we summarized in this discussion.
Preparation of this manuscript was (in part) based on
research work supported by K.U. Leuven grant GOA/2001/01
“Extinction and the return of conditioned responses” (awarded
to Drs. Eelen, Van den Bergh, and Baeyens), as well as by
National Institute of Mental Health (NIMH) RFA-MH-04-005
“Developing translational research on mechanisms of fear ex-
tinction” (awarded to Drs. Barad and MGC) and NIMH R01
MH65651 “Common and specific risk factors for emotional
disorders” (awarded to MGC and SM). In addition, preparation
of this manuscript was supported by Australian Research Coun-
cil grant A10007156 (to PFL).
Aspects of this work were presented at the conference, “Extinc-
tion: The Neural Mechanisms of Behavior Change,” February 2–6,
2005, in Ponce, Puerto Rico. The conference was sponsored by
the National Institute of Mental Health, National Institute of
Drug Abuse, Ponce School of Medicine, University of Puerto Rico
COBRE Program, Pfizer Global Pharmaceutical, and the Munic-
ipality of Ponce.
Clinical Implications. Washington, DC: American Psychological Associa-
Beaver JD, Mogg K, Bradley BP (2005): Emotional conditioning to masked
stimuli and modulation of visuospatial attention. Emotion 5:67–79.
Bouton ME (1988): Context and ambiguity in the extinction of emotional
learning: Implications for exposure therapy. Behav Res Ther 26:137–
Bouton ME (1994): Context, ambiguity, and classical conditioning. Curr Dir
Pavlovian discrimination reversal. J Exp Psychol Anim Behav Process 19:
tive on the etiology of panic disorder. Psychol Rev 108:4–32.
Bouton ME, Ricker ST (1994): Renewal of extinguished responding in a
Bouton ME, Swartzentruber D (1986): Analysis of the associative and occa-
sion-setting properties of contexts participating in a Pavlovian discrim-
Bouton ME, Swartzentruber D (1991): Sources of relapse after extinction
in Pavlovian and instrumental learning. Special issue: Applied learn-
ing theory–Research issues for the 1990s. Clin Psychol Rev
Bouton ME, Woods AM, Moody EW, Sunsay C, Garcia-Gutierrez A (in press):
Clinical Implications. Washington, DC: APA Books.
Brooks DC, Bouton ME (1994): A retrieval cue for extinction attenuates
response recovery (renewal) caused by a return to the conditioning
Carter RM, Hofstötter C, Tsuchiya N, Koch C (2003): Working memory and
of awareness. Science 280:77–81.
Cohen J (1988): Statistical Power Analysis for the Behavioral Sciences, 2nd ed.
Hillsdale, NJ: Erlbaum.
Collins BN, Brandon TH (2002): Effects of extinction context and retrieval
cues on alcohol cue reactivity among nonalcoholic drinkers. J Consult
Davey GCL (1997): A conditioning model of phobias. In: Davey GCL, editor.
Phobias: A Handbook of Theory, Research, and Treatment. Chichester,
Dawson ME, Schell AM, Beers JR, Kelly A (1982): Allocation of cognitive
De Houwer J, Thomas S, Baeyens F (2001): Associative learning of likes and
ing. Psychol Bull 127:853–869.
tive processes in human associative learning. In: Wills A, editor. New
Directions in Human Associative Learning. Mahwah, NJ: Lawrence Erl-
Delamater AR (1996): Effects of several extinction treatments upon the
Delamater AR (2004): Experimental extinction in Pavlovian conditioning:
Behavioural and neuroscience perspectives. Q J Exp Psychol B 57B:97–
ment of extinguished conditioned responses and negative stimulus
valence as a pathway to return of fear in humans. Learn Mem 11:549–
Eelen P, Hermans D, Baeyens F (2001): Learning perspectives on anxiety
Wiley and Sons, 249–264.
Falls WA (1998): Extinction: A review of theory and evidence suggesting
that memories are not erased with nonreinforcement. In: O’Donohue
W, editor. Learning and Behavior Therapy. Boston: Allyn & Bacon,
Hermans D, Crombez G, Vansteenwegen D, Baeyens F, Eelen P (2002):
Expectancy-learning and evaluative learning in human classical con-
ditioning: Differential effects of extinction. In: Shohov SP, editor.
Advances in Psychology Research, vol. 12. New York: Nova Science
and implicit cognition. In: Wiers RW, Stacy AW, editors: The Handbook of
Hermans D, Vansteenwegen D, Crombez G, Baeyens F, Eelen P (2002): Ex-
ing: Affective priming as an indirect and unobtrusive measure of condi-
tioned stimulus valence. Behav Res Ther 40:217–234.
Knight DC, Smith CN, Cheng DT, Stein EA, Helmstetter FJ (2004): Amygdala
and hippocampal activity during acquisition and extinction of human
Koster EHW, Crombez G, Van Damme S, Verschuere B, De Houwer J (2004):
Does imminent threat capture and hold attention? Emotion 4:312.
LaBar KS, Gatenby JC, Gore JC, LeDoux JE, Phelps EA (1998): Human amyg-
dala activation during conditioned fear acquisition and extinction: A
mixed-trial fMRI study. Neuron 20:937–945.
D. Hermans et al
BIOL PSYCHIATRY 2006;60:361–368 367
LaBar KS, Phelps EA (2005): Reinstatement of conditioned fear in humans is
context dependent and impaired in amnesia. Behav Neurosci 119:677–
reflex. Psychol Rev 9:377–395.
Lipp OV, Purkis HM (2005): No support for dual process accounts of human
affective learning in simple Pavlovian conditioning. Cogn Emotion 19:
Lipp OV, Siddle DAT, Dall PJ (1993): Effects of miscuing on Pavlovian condi-
tioned responding and on probe reaction time. Aust J Psychol 45:161–
Lovibond PF (2003): Causal beliefs and conditioned responses: Retrospec-
Lovibond PF (2004): Cognitive processes in extinction. Learn Mem 11:495–
Lovibond PF, Davis NR, O’Flaherty AS (2000): Protection from extinction in
human fear conditioning. Behav Res Ther 38:967–983.
Lovibond PF, Shanks DR (2002): The role of awareness in Pavlovian condi-
tioning: Empirical evidence and theoretical implications. J Exp Psychol
Marks IM, Swinson RP, Basoglu M, Kuch K, Noshirvani H, O’Sullivan G, et al
Milad MR, Orr SP, Pitman RK, Rauch SL (2005): Context modulation of mem-
ory for fear extinction in humans. Psychophysiology 42(4):456–464.
and limitations. In: Maser J, Tuma A, editors. Anxiety and the Anxiety
Disorders. Hillsdale, NJ: Lawrence Erlbaum, 199–244.
ing contexts on return of fear following exposure treatment for spider
Mitchell CJ, Lovibond PF (2002): Backward and forward blocking in human
electrodermal conditioning: Blocking requires an assumption of out-
Myers KM, Davis M (2002): Behavioral and neural analysis of extinction.
Mystkowski JL, Craske MG, Echiverri AM (2002): Treatment context and
return of fear in spider phobia. Behav Ther 33:399–416.
Mystkowski JL, Echiverri AM, Labus JS, Craske MG (in press): Mental rein-
statement of context and return of fear in spider phobia. Behav Ther.
Mystkowski, JL, Mineka S (in press): Behavior therapy for fears and phobias:
Context specificity of fear extinction. In: Baker TB, Bootzin R, Treat T,
editors. Psychological Clinical Science: Recent Advances in Theory and
Practice: Integrative Perspectives in Honor of Richard M. McFall. Hillsdale,
NJ: Lawrence Erlbaum.
Mystkowski JL, Mineka S, Vernon LL, Zinbarg RE (2003): Changes in caffeine
state enhance return of fear in spider phobia. J Consult Clin Psychol
Öhman A, Mineka S (2001): Fears, phobias and preparedness: Toward an
evolved module of fear and fear learning. Psychol Rev 108:483–522.
Pavlov IP (1927): Conditioned Reflexes: An Investigation of the Physiological
humans: Role of the amygdala and vmPFC. Neuron 43:897–905.
Phelps EA, O’Connor KJ, Gatenby JC, Gore JC, Grillon C, Davis M (2001):
Powers MB, Smits JAJ, Telch MJ (2004): Disentangling the effects of safety-
behavior utilization and safety-behavior availability during exposure-
based treatments: A placebo- controlled trial. J Consult Clin Psychol 72:
Rauch SL, Shin LM, Phelps EA (in press): Neurocircuitry models of posttrau-
matic stress disorder and extinction: human neuroimaging research
past, present and future. Biol Psychiatry.
Rescorla RA (1996): Preservation of Pavlovian associations through extinc-
Rescorla RA (2000): Extinction can be enhanced by a concurrent excitor.
J Exp Psychol Anim Behav Process 26:251–260.
Handbook of Contemporary Learning Theories. Mahwah, NJ: Lawrence
Erlbaum Associates, 119–154.
Rescorla RA (2003): Protection from extinction. Learn Behav 31:124–132.
Rescorla RA (2004): Spontaneous recovery. Learn Mem 11:501–509.
Rescorla RA, Heth CD (1975): Reinstatement of fear to an extinguished
Rescorla RA, Wagner AR (1972): A theory of Pavlovian conditioning: Varia-
tions in the effectiveness of reinforcement and nonreinforcement. In:
Black AH, Prokasy WF, editors. Classical Conditioning II: Current Research
and Theory. New York: Appleton-Century-Crofts 64–99.
et al (2004): Cognitive enhancers as adjuncts to psychotherapy: Use of
D-cycloserine in phobics to facilitate extinction of fear. Arch Gen Psychi-
Rodriguez BI, Craske MG, Mineka S, Hladek D (1999): Context-specificity of
Salkovskis PM (1991): The importance of behaviour in the maintenance of
anxiety and panic: A cognitive account. Behav Psychother 19:6–19.
Sloan T, Telch MJ (2002): The effects of safety-seeking behavior and guided
threat reappraisal on fear reduction during exposure: An experimental
Smith SM (1979): Remembering in and out of context. J Exp Psychol Hum
ing of inhibitory conditioning within a serial conditioned stimulus-con-
ditioned inhibitor compound: Maintenance of acquired behavior with-
out an unconditioned stimulus. Learn Motiv 14:1–29.
Steketee G (1993): Social support and treatment outcome of obsessive
compulsive disorder at 9-month follow-up. Behav Psychother 21:
the effects of extinction with co-present fear inhibitors or excitors: Im-
plications for theories of extinction and for treating human fears and
phobias. Learn Motiv 35:22–52.
Tolin DF, Lohr JM, Sawchuk CN, Lee TC (1997): Disgust and disgust
Vansteenwegen D, Dirikx T, Hermans D, Vervliet B, Eelen P (in press): Re-
newal and reinstatement in human fear conditioning. In: Craske MG,
Hermans D, Vansteenwegen D, editors. Fear and Learning: From Basic
Processes to Clinical Implications. Washington, DC: American Psycholog-
caused by a return to the original acquisition context. Behav Res Ther
Vervliet B, Vansteenwegen D, Baeyens F, Hermans D, Eelen P (2005): Return
of fear in a human differential conditioning paradigm caused by a stim-
ulus change after extinction. Behav Res Ther 43:357–371.
Vervliet B, Vansteenwegen D, Eelen P (2004): Generalization of extinguished skin
Wade S, Monroe S, Michelson L (1993): Chronic life stress and treatment
Wells A, Clark DM, Salkovskis PM, Ludgate J, Hackmann A, Gelder M (1995):
Social phobia: The role of in-situation safety behaviours in maintaining
anxiety and negative beliefs. Behav Ther 26:153–161.
Westbrook RF, Iordanova M, McNally G, Richardson R, Harris JA (2002):
Reinstatement of fear to an extinguished conditioned stimulus: Two
368 BIOL PSYCHIATRY 2006;60:361–368
D. Hermans et al