The Benzodiazepine Alprazolam Dissociates
Contextual Fear from Cued Fear in Humans as
Assessed by Fear-potentiated Startle
Christian Grillon, Johanna M.P. Baas, Daniel S. Pine, Shmuel Lissek, Megan Lawley, Valerie Ellis, and
Background: The startle reflex is potentiated by aversive states. It has been proposed that phasic startle potentiation to a threat cue
and sustained startle potentiation to contextual stimuli reflect distinct processes mediated by different brain structures. The present
study tested the hypothesis that alprazolam would reduce the sustained startle potentiation to contextual threats but not the startle
potentiation to a threat cue.
Methods: Sixteen healthy subjects received each of four treatments: placebo, .5 mg of alprazolam, 1 mg of alprazolam, and 50 mg
of diphenhydramine (Benadryl) in a crossover design. Participants were exposed to three conditions, including one in which
predictable aversive shocks were signaled by a cue, a second in which shocks were administered unpredictably, and a third condition
in which no shocks were anticipated. Acoustic startle were delivered regularly across conditions.
Results: Phasic startle potentiation to the threat cue in the predictable condition was not affected by alprazolam. In contrast, the
sustained increase in startle in the predictable and unpredictable conditions was reduced significantly by the high dose of alprazolam.
Conclusions: Startle responses to an explicit threat cue and to an aversive context are psychopharmacologically distinct, suggesting
that they may represent functionally dissociable aversive states.
Key Words: Alprazolam, anxiety, benzodiazepine, context, fear,
conditioned cue that signals an aversive event (e.g., a shock).
Evidence suggests that this effect is mediated by the central
nucleus of the amygdala (CeA). For example, lesions of the CeA
block fear-potentiated startle to a conditioned cue previously
paired with shock (Davis 1998). Although the CeA responds to
various types of stressors, this structure is not always critical for
fear-potentiated startle (Davis 1998) or for responses to stress in
general (Hammack et al 2004). Davis and his collaborators have
reported a series of studies showing no effect of lesions of the
CeA on startle potentiation caused by various stressors, such as
bright lights (i.e., light-enhanced startle), shock sensitization, and
corticotropin-releasing hormone injection (reviewed in Walker et
al 2003). Rather, another structure, the bed nucleus of the stria
terminalis (BNST), was found to mediate startle potentiation in
these conditions (Walker et al 2003). In an analysis of the
experimental situations that do or do not require the CeA and
BNST, Walker et al (2003) suggested that these structures were
involved in functionally different aversive states in rodents. It was
proposed that the CeA was crucial for the phasic form of
fear-potentiated startle to a predictable threat cue, whereas the
BNST was responsible for the more sustained form of startle
potentiation induced by unpredictable or unconditioned aver-
sive stimuli (Davis 1998; Gewirtz et al 1998; Walker and Davis
he startle reflex, a cross-species response to a sudden
intense stimulus, is sensitive to aversive states (Davis et al
1993). Rodents show robust startle potentiation to a
1997a). It was suggested further that the two aversive states
mediated by the CeA and the BNST in rodents were reminiscent
of fear and anxiety states in humans, respectively (Davis 1998).
According to this view, fear is a response to a clearly identifiable
danger that subsides shortly after the offset of a threat cue.
Anxiety is a more sustained form of general distress and anxious
apprehension in response to less identifiable cues (Barlow 2000;
Davis 1998; Lang et al 2000).
The distinction between a phasic and a more sustained form
of startle potentiation also has been made in humans (Cuthbert et
al 2003; Grillon et al 1991, 1997; Grillon and Davis 1997; Lang et
al 2000; Pole et al 2003). Startle is potentiated by an explicit threat
cue that signals an impending aversive event (e.g., a shock;
Grillon et al 1993; Hamm and Vaitl 1996), such as when phobic
individuals are confronted with their phobic objects (de Jong et
al 1996; Globisch et al 1999). More sustained forms of startle
potentiation can be found among individuals who are exposed
to stressful experimental settings (Bocker et al 2001; Grillon and
Ameli 1998; Pole et al 2003), when the experimental room is in
complete darkness (Grillon et al 1997), or after context condi-
tioning (Grillon and Davis 1997).
Consistent with the animal literature, phasic and sustained
forms of startle potentiation in humans appear to reflect distinct
processes. For example, individuals with posttraumatic stress
disorder (PTSD) or with panic disorder display normal startle to
explicit threat cues that signal a shock but display enhanced
startle reactivity in the experimental context in which the shocks
are administered (Grillon et al 1994, 1998b; Pole et al 2003).
Patients with PTSD also exhibit increased context conditioning
(Grillon and Morgan 1999) and increased facilitation of startle in
the dark (Grillon et al 1998a).
There currently is little information on the neurobiological
mechanisms that may differentiate phasic cued fear from more
sustained contextual anxiety in humans. Evidence for such a
neurobiological dissociation would be bolstered if one could
demonstrate that these two forms of aversive states are differen-
tially responsive to psychopharmacologic treatments. The main
objective of this study was to obtain such evidence by using the
Health, Bethesda, Maryland.
Address reprint requests to Christian Grillon, Ph.D., National Institutes of
Mental Health–Mood and Anxiety Disorder Program, 15K North Drive,
Building 15K, Room 113, MSC 2670, Bethesda, MD 20892-2670; E-mail:
Received July 12, 2005; revised October 27, 2005; accepted November 11,
BIOL PSYCHIATRY 2006;60:760–766
© 2006 Society of Biological Psychiatry
There is an emerging literature on the effect of benzodiaz-
epines on fear-potentiated startle in humans. We have reported
results of four studies showing that oxazepam and diazepam did
not affect fear-potentiated startle to a threat cue (Baas et al 2002).
In contrast, other groups reported that alprazolam (Riba et al
2001), diazepam (Bitsios et al 1999), and lorazepam (Graham et
al 2005) reduced fear-potentiated startle in threat of shock
experiments. However, it is not clear whether the reduction in
fear-potentiated startle in these latter studies was caused by an
anxiolytic effect per se or was an artifact of the sedative effect of
benzodiazepines on baseline startle. A drug-induced reduction in
baseline startle reactivity may lead to an inaccurate measurement
of fear-potentiated startle (Grillon and Baas 2002; Walker and
Davis 2002b). Important methodologic differences between stud-
ies also may explain contradictory findings. For example, in two
studies the subjects were verbally informed of the threat and safe
conditions while the shock electrodes were being attached or
removed, and these studies were conducted in near darkness
(Bitsios et al 1999; Graham et al 2005). It has been argued that
both the shock electrodes and darkness are contextual stimuli
(Grillon and Ameli 1998; Grillon et al 1997; Walker and Davis
1997b), which may have increased contextual anxiety. Hence, it
is unclear whether the response that was effectively reduced by
diazepam in the Bitsios et al (1999) design constitutes a cued fear
response or contextual anxiety caused by placement of the shock
There is evidence to suggest that benzodiazepines affect
contextual anxiety. In rodents, benzodiazepines reduce baseline
startle. This reduction is a result not only of sedation but also of
an anxiolytic effect on contextual fear (Guscott et al 2000). Baas
et al (2002) reported a similar effect in humans. In addition, Baas
et al (2002) showed that diazepam reduced the facilitation of
startle in the dark in humans. The facilitation of startle in the dark
in humans is a sustained form of startle potentiation similar to the
light-enhanced startle in rodents (Grillon and Baas 2002), which
itself is alleviated by benzodiazepines (Walker and Davis 2002a;
de Jong et al 2002).
The present study improved on past studies by addressing
two main issues. First, we compared fear-potentiated startle
elicited by predictable and unpredictable shocks to clearly
dissociate phasic cued fear from sustained anxiety (Grillon et al
2004). Second, a sedative drug (diphenhydramine) that is not
used for anxiolysis was included to control for confounding
effects of sedation on startle reactivity (Grillon and Baas 2002;
Walker and Davis 2002b). The experimental design was based on
the observation that unpredictable aversive events increase
context conditioning (Grillon and Davis 1997; Odling-Smee
1975), which can be conceived of as a form of sustained anxiety.
In this design, subjects are presented with three conditions: no
shocks, predictable or signaled shocks, and unpredictable or
nonsignaled shocks. Previous results show two types of aversive
responses: (1) a phasic startle potentiation during the threat
signal relative to the absence of threat signal in the predictable
shock condition and (2) a sustained startle potentiation during
the predictable and the unpredictable shock condition compared
with the no-shock condition in the absence of specific cues
(Grillon et al 2004). For the remainder of this article, cued fear
will refer to the phasic startle potentiation to the explicit threat
signal, and contextual anxiety will refer to the sustained startle
potentiation in the absence of cues.
The main hypothesis of the study was that alprazolam would
not affect fear-potentiated startle to the threat signal in the
predictable shock condition (cued fear) but would reduce fear-
potentiated startle in the absence of cues in the predictable and
unpredictable shock condition (contextual anxiety). Each subject
was tested in four treatments: (1) placebo, (2) a low dose (.5 mg)
of alprazolam, (3) a high dose (1 mg) of alprazolam, and (4)
diphenhydramine (Benadryl, 50 mg). The two doses of alprazo-
lam were selected on the basis of a study that reported a
reduction of fear-potentiated startle by alprazolam (Riba et al
Methods and Materials
Participants were healthy volunteers who gave written in-
formed consent that had been approved by the National Insti-
tutes of Mental Health Human Investigation Review Board.
Inclusion criteria included the following: (1) no past or current
psychiatric disorders as per Structured Clinical Interview for
DSM-IV (SCID; First et al 1995), (2) no medical condition that
interfered with the objectives of the study as established by a
physician, and (3) no use of illicit drugs or psychoactive medi-
cations as per urine screen.
Participants underwent a screening session that consisted of a
SCID, a physical exam, and a shock workup procedure to
establish a level of shock that was “highly annoying but not
painful”. The mean intensity of the shock was 4.2 ?A, with a
range of 3–5 ?A. In addition, subjects were screened for baseline
startle reactivity with nine startle stimuli (40-ms duration, 103
dB). Subjects who displayed small startle responses (a mean of
less than 5 ?V over nine startle responses) or displayed no startle
response on at least one trial were not invited to participate in the
study. Four to 10 days after screening, participants returned for
the first of four testing sessions. Sixteen subjects (five were male)
with a mean age of 23.0 years (SD ? 4.7 y) ultimately were
included in the study. Mean scores on the state and trait portions
of Spielberger’s State and Trait Anxiety Inventory (Spielberger
1983) were 26.0 (SD ? 4.0) and 28.2 (SD ? 4.8), respectively.
The treatments were placebo, .5 mg of alprazolam, 1 mg of
alprazolam, and 50 mg of diphenhydramine, tested in a double-
blind crossover design (within subjects). Treatment administra-
tion was performed according to a randomization table compris-
ing a 4 ? 4 Latin square repeated four times.
On the test day, subjects filled out a mood rating scale
(pretreatment) that evaluated subjective feelings of mental and
physical sedation. Next, they ingested a capsule containing one
of the active drugs or placebo. Subjects rested for 1 hour to allow
drug absorption, after which the procedure to apply measure-
ment- and shock-electrodes was started.
Details of the procedures are provided elsewhere (Grillon et
al 2004). During testing, subjects first were presented a habitua-
tion block consisting of nine startle stimuli delivered every 18–25
sec to reduce excessive initial startle reactivity before the threat
study (data not reported). Participants then were given explicit
instructions regarding the conditions under which they would
and would not receive an aversive event. After the instructions,
the threat experiment began. The experiment consisted of three
different conditions: no shock (N), predictable shock (P), and
unpredictable shock (U), each lasting approximately 150 sec. In
the N condition, no shocks were delivered. In the P condition,
shocks were administered predictably, that is, only in the pres-
ence of a threat cue. In the U condition, the shocks were
C. Grillon et al
BIOL PSYCHIATRY 2006;60:760–766 761
unpredictable. In each 150-sec condition, an 8-sec cue was
presented four times. The cues were different geometric colored
shapes in each condition (e.g., a blue square for N, a red circle
for P, and a green star for U). The cues signaled the possibility of
receiving an aversive stimulus only in the P condition. They had
no signal value in the N and U conditions. Instructions were
displayed on a computer monitor to inform participants of the
current condition by displaying the following information
throughout the testing procedure: “no shock” (N), “shock only
during shape” (P), or “shock at any time” (U). During each
predictable and unpredictable condition, one shock was admin-
istered. When a shock was administered, it was delivered during
the cue in the predictable condition and in the absence of the
cues in the unpredictable condition. In each N, P, and U
condition, six acoustic startle stimuli were delivered, three during
intertrial intervals (ITI; i.e., between cues) and one during three
of the four cues, 5–7 sec after cue onset. The threat experiment
consisted of two recording blocks with a 5- to 10-min rest
between blocks. Each block consisted of three N, two P, and two
U conditions in one of the following two orders: P N U N U N P
or U N P N P N U. Each participant was presented with the two
orders, with half the participants starting with the P condition.
One shock was administered in each individual P and U condi-
tion, for a total of four shocks in the four P conditions and of four
shocks in the four U conditions. The shock was delivered 7.5 sec
after cue onset in the P condition. It was administered either 7 sec
or 10 sec after cue offset in the unpredictable condition. No
startle stimuli could follow a shock by less than 10 sec.
Subjects were asked to fill out the mood rating scale another
time during the interval between the two threat blocks. In
addition, after each recording block, subjects retrospectively
rated how anxious they felt in the presence and absence of the
cue in each condition (N, P, U) on an analog scale ranging from
0 (not at all anxious) to 10 (extremely anxious).
Stimuli and Physiological Responses
Stimulation and recording were controlled by a commercial
system (Contact Precision Instruments, London, United King-
dom). The acoustic startle stimulus was a 40-ms duration,
103-dB(A) burst of white noise with a near-instantaneous rise
time, presented binaurally through headphones. The eyeblink
reflex was recorded with electrodes placed under the left eye.
Amplifier bandwidth was set to 30–500 Hz. The electric shock
was produced by a constant current stimulator and was admin-
istered on the right wrist.
Peak amplitude of the blink reflex was determined in the 20-
to 100-msec time frame after stimulus onset relative to baseline
(average baseline EMG level for the 50 ms immediately preced-
ing stimulus onset) and was averaged within each condition. The
startle data and retrospective measures of subjective anxiety
were analyzed with analyses of variance (ANOVA) with repeated
measures. Preliminary analyses indicated no gender difference
for the startle and subjective measures. Hence, gender was not
entered as a factor in the data analysis. Given our specific a priori
hypotheses, separate comparisons were conducted to examine
cued fear and contextual anxiety. Cued fear was tested by first
calculating the difference scores between startle magnitude
during the cues and startle magnitude during ITI. These differ-
ence scores then were analyzed with two-way ANOVAs, with
treatment (placebo, diphenhydramine, low alprazolam, or high
alprazolam) and condition (N, P, or U) as repeated factors.
Contextual anxiety was evaluated by using the startle magnitudes
during ITI in each of the three conditions. This involved two-way
ANOVAs, with treatment (placebo, diphenhydramine, low alpra-
zolam, or high alprazolam) and condition (N, P, or U) as
repeated factors. Given a significant main effect of condition, the
presence of a linear trend in startle magnitude over N, P, and U
was tested. It was hypothesized that the linear trend would be
affected by alprazolam but not by diphenhydramine, compared
with placebo. These analyses were repeated on standardized
scores by using within-subjects t scores ([Z scores ? 10] ? 50).
Because similar results were obtained with the raw scores and
with the t scores for within-subjects comparisons, only results of
the raw scores are presented. The same analysis was conducted
for the retrospective subjective reports of anxiety. Alpha was set
at .05 for all statistical tests. Greenhouse-Geisser corrections
(GG-?) were used for main effects and for interactions involving
factors with more than two levels.
treatment are presented in Table 1. Figure 1A displays the
magnitude of fear-potentiated startle (cue minus ITI). There was
a significant main effect of condition [F(2,30) ? 11.9, p ? .001,
GG-? ? .65], reflecting greater fear-potentiated startle during the
cue in the U condition, compared to the N and P condition.
These effects were not affected by treatment, as reflected by a
lack of treatment main effect [F(3,45) ? .03, ns] and treatment ?
condition interaction [F(6,90) ? .3, ns].
Contextual Anxiety. The ITI data that were used to evaluate
contextual anxiety are shown in Table 1. Consistent with our
previous report (Grillon et al 2004), there was a linear increase in
startle magnitude from the neutral, to the predictable, to the
unpredictable condition [main effect of condition: F(2,30) ?
14.0, p ? .0009, GG-? ? .7; linear effect of condition: F(1,15) ?
18.7, p ? .0009]. There also was a trend for a main effect of
treatment [F(3,45) ? 2.4, p ? .09, GG-? ? .8]. However, this
effect was qualified by a significant treatment ? condition
interaction [F(6,90) ? 3.4, p ? .01, GG-? ? .7] and by a treatment
? condition linear effect [F(1,15) ? 5.7, p ? .03]. These
The results in each condition and in each
Table 1. Mean (SEM) Startle Magnitude (?V) during the Cue and during ITI across Treatments and Conditions
Cue ITICue ITI CueITI
ITI, intertrial intervals.
762 BIOL PSYCHIATRY 2006;60:760–766
C. Grillon et al
interactions reflect differential effects of treatment on startle
potentiation during ITI. Follow-up tests focused on the increase
in startle magnitude from the neutral to both the predictable and
unpredictable conditions across treatments. This analysis was
implemented by calculating the difference scores for predictable
minus no-shock and unpredictable minus no-shock conditions
(Figure 2A). These difference scores were entered into a treat-
ment (placebo or high alprazolam) ? condition (P or U) ANOVA
yielding a main effect of treatment [F(1,15) ? 10.2, p ? .0009]
and no significant treatment ? condition interaction. The signif-
icant treatment main effect confirmed that the ITI startle poten-
tiation from the neutral to both the predictable and unpredictable
condition was reduced by high alprazolam.
The active control substance diphenhydramine was used to
examine whether the treatment effect could be attributed solely
to sedation. This was not the case. First, the treatment (placebo
or diphenhydramine) ? condition (P or U) ANOVA showed no
significant effect involving treatment. Second, the treatment
(high alprazolam or diphenhydramine) ? condition (P or U)
ANOVA showed a significantly lower potentiated startle in the
[F(1,15) ? 5.5, p ? .03]. The placebo, diphenhydramine, and
low-alprazolam treatment conditions did not differ significantly
from each other. Note that high alprazolam also reduced poten-
tiated startle compared with the low-alprazolam treatment con-
dition [F(1,15) ? 5.7, p ? .03].
Retrospective Ratings of Anxiety
Cued Fear. The anxiety rating scores are shown in Figure 1B.
There was a significant main effect of condition [F(2,30) ? 56.0,
p ? .0009, GG-? ? .95], reflecting greater subjective anxiety
during the cue in the P condition, compared to the N and U
conditions. There was no significant main effect of treatment
[F(3,45) ? .61], but the treatment ? condition interaction was
significant [F(6,90) ? 2.4, p ? .04]. Because the comparison of
interest was anxiety ratings in the P condition across treatments,
follow-up tests contrasted anxiety scores between treatments in
the P condition in a one-way ANOVA. The results indicated no
reduction in anxiety with any of the treatments as reflected by a
nonsignificant treatment main effect [F(3,45) ? 1.9, p ? .1].
Contextual Anxiety. Consistent with the startle data, there
was a progressive increase in anxiety from the neutral, to the
predictable, to the unpredictable condition [F(2,30) ? 56.7, p ?
.0009, GG-? ? .69; linear trend: F(1,15) ? 69.4, p ? .0009]. There
also was a main effect of treatment [F(3,45) ? 7.6, p ? .001, GG-?
? .83] that was caused by an overall reduction in anxiety with
diphenhydramine [F(1,15) ? 17.1, p ? .001], low alprazolam
[F(1,15) ? 4.8, p ? .04], and high alprazolam [F(1,15) ? 14.9, p
? .002] treatments compared to placebo (Figure 2). However,
unlike the startle results, there was no treatment ? condition
interaction, indicating no significant difference in subjective
Figure 2. Responses during ITI (context) in each treatment. (A) Contextual
potentiated startle. Difference scores between startle magnitudes in the
the no-shock condition. *The increased startle from the neutral to the pre-
high alprazolam (alpraz), compared to the other treatments. (B) Subjective
(predictable and unpredictable) and reported anxiety in the no-shock con-
dition. There was no significant difference among treatments.
Figure 1. Responses to the cues in each treatment expressed as a change
cue minus ITI startle magnitudes in the no shock, predictable, and unpre-
dictable conditions. The main effect of condition was significant. *Signifi-
magnitudes in the presence minus absence of cues in each condition. The
main effect of condition was significant. *Significantly (p ? .05) increased
startle from ITI. There was no significant difference among treatments.
C. Grillon et al
BIOL PSYCHIATRY 2006;60:760–766 763
anxiety across conditions among the three active treatments
Mental and Physical Sedation
The mental and physical sedation data from the mood rating
scale are shown in Table 2. Results were analyzed with treatment
(placebo, diphenhydramine, alprazolam low, or alprazolam
high) ? condition (baseline or posttreatment) ANOVAs. For both
mental sedation and physical sedation, there were significant
main effects of treatment [F(3,45) ? 6.5, p ? .001, GG-? ? .78
and F(3,45) ? 5.4, p ? .003, GG-? ? .92, respectively] and of
condition [F(1,15) ? 37.5, p ? .0009 and F(1,15) ? 44.0, p ?
.0009, respectively], as well as a significant treatment ? condition
interaction [F(3,45) ? 8.4, p ? .001, GG-? ? .70 and F(3,45) ?
9.9, p ? .0009, GG-? ? .79, respectively]. The interaction
reflected the fact that subjects exhibited greater increases in
physical and mental sedation from the baseline to the posttreat-
ment period with diphenhydramine, low alprazolam, and high
alprazolam compared with placebo (all p ? .05). There also was
a greater increase in mental and physical sedation with the
diphenhydramine and high-alprazolam, compared with low-
alprazolam, treatments (all p ? .05). Finally, there was no
significant difference in mental and physical sedation between
diphenhydramine and high-alprazolam treatments.
The present study sought to establish a psychopharmacolog-
ical distinction by using the benzodiazepine alprazolam between
a phasic aversive response to a threat cue (cued fear) and a more
sustained aversive response associated with the experimental
context in which shocks are anticipated (contextual anxiety). The
experimental model to elicit these two aversive states was based
on preclinical data in rodents (Davis 1998) and on empirical
work in our laboratory that used the threat of predictable
(signaled) and unpredictable (unsignaled) shocks (Grillon et al
2004). Consistent with our hypothesis, alprazolam did not affect
phasic fear-potentiated startle to the threat cue but reduced the
sustained potentiation of startle in the predictable and the
unpredictable condition. This effect cannot be attributed to an
artifactual sedative effect of alprazolam on baseline startle
(Walker and Davis 2002b) for two reasons. First, baseline startle
responses, as well as mental and physical sedation, were affected
to the same extent by high alprazolam and diphenhydramine,
indicating that diphenhydramine was an appropriate nonspecific
active control substance. Yet, contextual startle potentiation in
the predictable and unpredictable conditions was reduced sig-
nificantly in the high alprazolam compared to the diphenhydra-
mine treatment. Second, if the reduction in contextual fear-
potentiated startle was a result of sedation, it is unclear why such
an effect would not have affected fear-potentiated startle to the
threat cue. These results suggest that alprazolam exerted a
differential effect on potentiated startle to an explicit threat cue
and to contextual cues.
It could be argued that the treatment effect of the high dose of
alprazolam on ITI startle in the predictable and unpredictable
condition was artifactually caused by a floor effect. This expla-
nation is unlikely because the floor effect for startle magnitude is
an absence of eyeblink response. In the present study, the
magnitude of startle in the high-alprazolam condition was clearly
above 0 (about 20 ?V). Alternatively, it could be argued that the
lack of effect of alprazolam on fear-potentiated startle to the cue
in the predictable condition was caused by a ceiling effect. This,
again, is unlikely. First, we analyzed the startle habituation data
(not shown) and found much larger responses during startle
habituation compared response to the threat cues in the predict-
able condition. For example, the group mean magnitude of the
first habituation startle in the placebo condition was 91 ?V,
which is larger than either the mean response to the first startle
in the predictable cue (74 ?V) or than the mean startle response
to all the cues in the predictable condition (57.9 ?V).
The present findings are consistent with preclinical and
clinical studies distinguishing cued fear from contextual anxiety.
Preclinical studies show that the CeA mediates phasic responses
to explicit threat cues and that the BNST mediates sustained
responses to contextual stimuli (Walker et al 2003). Clinical
investigations have reported normal fear-potentiated startle to an
explicit threat cue but elevated contextual fear-potentiated startle
in patients with anxiety disorders (Cuthbert et al 2003; Grillon et
al 1991, 1997; Grillon and Davis 1997; Pole et al 2003).
The present results also are consistent with animal data
showing that benzodiazepines reduce sustained forms of startle
potentiation. Walker and Davis (2002a) reported that the light-
enhanced startle effect was alleviated by the benzodiazepine
chlordiazepoxide. However, the lack of effect of alprazolam on
fear-potentiated startle to an explicit threat cue apparently con-
flicts with cued fear conditioning data in rodents (Hijzen et al
1995; Melia and Davis 1991). One important difference between
the present study and preclinical studies is the reliance of animal
studies on associative-learning processes such as fear condition-
ing. The reduced fear-potentiated startle after benzodiazepine
administration in animal models that rely on conditioning cannot
be attributed unambiguously to an anxiolytic effect as opposed
to an effect on learning or memory (Walker and Davis 2002a). In
addition, it is possible that the amount of drug used in animals is
not comparable to that used in humans. Nevertheless, a recent
fear-conditioning study in humans supports our findings. Scaife
et al (2005) found that diazepam blocked the acquisition but not
the expression of fear-potentiated startle to an explicit cue.
In terms of current results for subjective ratings of anxiety the
cue and contextual manipulations induced the same basic pat-
tern of results found in the startle data. In the predictable
condition, subjective anxiety was larger in the presence than in
the absence of the threat cue. Like the startle data, this increase
in subjective anxiety was unaffected by drug treatments. As for
the context effect, both startle magnitude and the subjective
ratings of anxiety to the context increased linearly from the
Table 2. Mean (SEM) Scores of Subjectively Reported Mental and Physical Sedation
Mental SedationPhysical Sedation
764 BIOL PSYCHIATRY 2006;60:760–766
C. Grillon et al
no-shock to the predictable to the unpredictable conditions.
However, treatment effects on these subjective reports did not
follow the startle results; instead of a significant treatment ?
condition interaction, there was only a main effect of treatment.
This indicates that though the subjects were feeling less anxious
overall under all three drug treatments compared with placebo,
they did not report context-specific reductions in anxiety under
alprazolam. Apparently, startle was sensitive to subtle effects of
the treatments that were not accessible to subjective measures of
One obvious reason for this difference is that whereas startle
was used to probe contextual anxiety online, the subjective
anxiety measures were retrospective. The passage of time may
have obscured subtle differences in responding because of the
complexity of the design. The design included six different
conditions (three contextual manipulations, each with and with-
out the presence of a cue), for which subjective anxiety was
measured retrospectively. The retrospective subjective anxiety
data appeared to have been strongly influenced by sedation. This
is suggested by the fact that the pattern of retrospective ratings of
anxiety exactly paralleled the ratings of mental and physical
sedation (Table 1). Future studies may benefit from using an
online measure of subjective anxiety.
How could alprazolam reduce contextual anxiety? It is un-
likely that the present results were the result of a gross cognitive
deficit, such as memory impairment. Subjects did not have to
remember the contingency between threat and the different
conditions because the instructions were written on a monitor
screen throughout the duration of each condition. A purely
anxiolytic effect on sustained anxiety could be achieved, for
example via action at the level of structures involved in this type
of response (i.e., the BNST). Alternatively, alprazolam could
reduce anxiety by facilitating distraction from stressful cognitions
associated with the threat experiment. Distraction from stressful
cognitions may have been more likely to occur in the unpredict-
able condition when the threat was long lasting, compared to the
predictable condition, in which the threat cue signaled an
imminent danger, resulting in a reduction in contextual anxiety
but not phasic fear. Such a distraction could have been amplified
by drowsiness, but this is unlikely given that the sedative drug
diphenhydramine did not reduce the enhanced startle in the
The limitations of the present study must be considered when
interpreting the results. It is possible that alprazolam effects on
potentiated startle did not depend on the qualitative nature of the
induced emotion (e.g., fear vs. anxiety or explicit vs. contextual
cues), but on the intensity of the aversive reaction. According to
this view, benzodiazepines could reduce potentiated startle to
stimuli or situations that elicit little fear or anxiety, such as
contextual anxiety, but are ineffective when higher levels of fear
or anxiety are involved, such as during a threat cue. Such a
possibility should be tested experimentally by examining the
effect of benzodiazepines on threat cues that elicit different levels
of fear. However, the Scaife et al (2005) fear-conditioning study
may shed light on this issue. Startle potentiation to a cue is much
greater during a threat of shock experiment than during fear
conditioning (Grillon and Davis 1977; Grillon et al 1991). Yet
Scaife et al (2005) found that diazepam did not block the
expression of cued fear conditioning. Hence, a weak aversive
response is not necessarily affected by benzodiazepines. Another
limitation is that we used diphenhydramine as an active control
condition because of its sedative effect. The subjective data
suggest that diphenhydramine was mildly anxiolytic, and animal
data suggest that the histaminergic system may be implicated in
anxiety (Fish et al 2004; Privou et al 1998). Because benzodiaz-
epines have muscle relaxant properties, a muscle relaxant may
have been a more appropriate control drug.
The present results are consistent with the hypothesis of a
functional differentiation between phasic startle potentiation to a
threat cue and sustained startle potentiation to contextual threat
proposed by Davis (1998). Because of the translational nature of
startle studies, it is likely that human and animal models will
ultimately yield more insight into the neurobiology of cued fear
and contextual anxiety. Differentiating between cued fear and
contextual anxiety may help us identify the components of
aversive states and their relevance to pathologic anxiety. Assum-
ing that contextual anxiety is relevant to pathologic anxiety
(Grillon and Morgan 1999; Grillon et al 1998b), experiments in
animals and humans that manipulate contextual fear-potentiated
startle may be helpful for screening novel anxiolytics.
This research was supported by the Intramural Research
Program of the National Institutes of Mental Health.
Baas JM, Grillon C, Bocker KB, Brack AA, Morgan CA III, Kenemans JL, Ver-
baten MN (2002): Benzodiazepines have no effect on fear-potentiated
startle in humans. Psychopharmacology (Berl) 161:233–247.
the perspective of emotion theory. Am Psychol 55:1247–1263.
Bitsios P, Philpott A, Langley RW, Bradshaw CM, Szabadi E (1999): Compari-
the fear-inhibited light reflex in man. J Psychopharmacol 13:226–234.
Bocker KB, Baas JM, Kenemans JL, Verbaten MN (2001): Stimulus-preceding
negativity induced by fear: A manifestation of affective anticipation. Int
psychophysiology of anxiety disorders: Fear memory imagery. Psycho-
Davis M (1998): Are different parts of the extended amygdala involved in
fear versus anxiety? Biol Psychiatry 44:1239–1247.
Davis M, Falls WA, Campeau S, Kim M (1993): Fear-potentiated startle: A
neural and pharmacological analysis. Behav Brain Res 58:175–198.
de Jong PJ, Visser S, Merckelbach H (1996): Startle and spider phobia: Uni-
lateral probes and the prediction of treatment effects. J Psychophysiol
startle paradigm as a putative animal model of anxiety: Effects of chlor-
diazepoxide, flesonoxan and fluvoxamine. Psychopharmacology (Berl)
First MB, Spitzer RI, Williams JBW, Gibbon M (1995): Structured Clinical Inter-
view for DSM-V (SCID). Washington, DC: American Psychiatric Associa-
of escitalopram, citalopram, and R-citalopram in maternally separated
terminalis block sensitization of acoustic startle reflex produced by re-
peated stress, but not fear-potentiated startle. Prog Neuropsychophar-
Grillon C, Ameli R (1998): Effects of threat of shock, shock electrode place-
ment, and darkness on startle. Int J Psychophysiol 28:223–231.
Grillon C, Ameli R, Goddard A, Woods S, Davis M (1994): Baseline and fear-
potentiated startle in panic disorder patients. Biol Psychiatry 35:431–
Grillon C, Ameli R, Woods SW, Merikangas K, Davis M (1991): Fear-potenti-
ated startle in humans: Effects of anticipatory anxiety on the acoustic
blink reflex. Psychophysiology 28:588–595.
C. Grillon et al
BIOL PSYCHIATRY 2006;60:760–766 765
Grillon C, Ameli R, Woods SW, Merikangas K, Davis M (1993): Measuring the
time-course of anxiety using the fear-potentiated startle reflex. Psycho-
Grillon C, Baas JMP (2002): Comments on the use of the startle reflex in
psychopharmacological challenges: Impact of baseline startle on mea-
surement of fear-potentiated startle. Psychopharmacology (Berl) 164:
Grillon C, Baas JP, Lissek S, Smith K, Milstein J (2004): Anxious responses to
Grillon C, Davis M (1997): Fear-potentiated startle conditioning in humans:
Explicit and contextual cue conditioning following paired vs. unpaired
training. Psychophysiology 34:451–458.
Grillon C, Morgan CA (1999): Fear-potentiated startle conditioning to ex-
Grillon C, Morgan CA, Davis M, Southwick SM (1998a): Effect of darkness on
tal context and explicit threat cues on acoustic startle in Vietnam veter-
ans with posttraumatic stress disorder. Biol Psychiatry 44:1027–1036.
Grillon C, Pellowski M, Merikangas KR, Davis M (1997): Darkness facilitates
the acoustic startle in humans. Biol Psychiatry 42:453–460.
Guscott MR, Cook GP, Bristow LJ (2000): Contextual fear conditioning and
parison of benzodiazepine/gamma-aminobutyric acid-A receptor ago-
nists. Behav Pharmacol 11:495–504.
Hamm AO, Vaitl D (1996): Affective learning: Awareness and aversion. Psy-
Hammack SE, Richey KJ, Watkins LR, Maier SF (2004): Chemical lesion of the
bed nucleus of the stria terminalis blocks the behavioral consequences
of uncontrollable stress. Behav Neurosci 118:443–448.
Hijzen TH, Houtzager SW, Joordens RJ, Olivier B, Slangen JL (1995): Predic-
anxiolytic drugs. Psychopharmacology (Berl) 118:150–154.
Lang PJ, Davis M, Ohman A (2000): Fear and anxiety: Animal models and
human cognitive psychophysiology. J Affect Disord 61:137–159.
and on the anxiolytic effects of buspirone and diazepan. Physiol Behav
Odling-Smee FJ (1975): The role of background stimuli during Pavlovian
Pole N, Neylan TC, Best SR, Orr SP, Marmar CR (2003): Fear-potentiated
startle and posttraumatic stress symptoms in urban police officers.
Privou CC, Knoche AA, Hasenöhrl RRU, Huston JJP (1998): The H1- and
H2-histamine blockers chlorpheniramine and ranitidine applied to the
ment related processes. Neuropharmacology 37:1019–1032.
Riba J, Rodriguez-Fornells A, Urbano G, Morte A, Antonijoan R, Barbanoj MJ
tiated startle reflex in humans: A dose-response study. Psychopharma-
Scaife JJC, Langley RRW, Bradshaw CCM, Szabadi EE (2005): Diazepam sup-
presses the acquisition but not the expression of “fear-potentiation” of
the acoustic startle response in man. J Psychopharmacol 19:347–356.
CA: Consulting Psychologists Press.
Walker DL, Davis M (1997a): Double dissociation between the involvement
of the bed nucleus of the stria terminalis and the central nucleus of the
tioned fear. J Neurosci17:9375–9383.
Walker DL, Davis M (1997b): Anxiogenic effects of high illumination levels
assessed with the acoustic startle response in rats. Biol Psychiatry 42:
Walker DL, Davis M (2002a): Light-enhanced startle: Further pharmacologi-
Walker DL, Davis M (2002b): Quantifying fear potentiated startle using ab-
solute versus proportional increase scoring methods: Implications for
the neurocircuitry of fear and anxiety. Psychopharmacology (Berl) 164:
Walker DL, Toufexis DJ, Davis M (2003): Role of the bed nucleus of the stria
766 BIOL PSYCHIATRY 2006;60:760–766
C. Grillon et al