Hippocampus and Contextual Fear Conditioning:
Recent Controversies and Advances
Stephan G. Anagnostaras, Greg D. Gale, and
Michael S. Fanselow
Department of Psychology and Brain Research Institute,
University of California, Los Angeles, California
in recently, but not remotely, acquired contextual fear without impairing
memory of discrete training stimuli, i.e., DH lesions produce an antero-
grade and time-limited retrograde amnesia specific to contextual mem-
ory. These data are consistent with the standard model which posits
temporary involvement of the hippocampus in recent memory mainte-
nance. However, three recent controversies apparently weaken the case
for a selective mnemonic role for the hippocampus in contextual fear.
First, although retrograde amnesia (from posttraining lesions) is severe,
anterograde amnesia (from pretraining lesions) may be mild or nonexist-
ent. Second, a performance, rather than mnemonic, account of contextual
freezing deficits in hippocampal-lesioned animals has been offered. Third,
damage to the entire hippocampus, including the ventral hippocampus,
can produce a dramatic and temporally stable disruption of context and
tone fear. These data are reviewed and explanations are offered as to why
they do not necessarily challenge the standard model of hippocampal
memory function in contextual fear. Finally, a more complete description
of the hippocampus’ proposed role in contextual fear is offered, along
with new data supporting this view. In summary, the data support a
specific mnemonic role for the DH in the acquisition and consolidation of
contextual representations. Hippocampus 2001;11:8–17.
© 2001 Wiley-Liss, Inc.
Dorsal hippocampal (DH) lesions produce a severe deficit
consolidation; amnesia; configural; spatial; amygdala;
It is widely recognized that after damage to the human hippocampal
formation, an amnesic syndrome ensues. In amnesics, an anterograde am-
nesia (AA; an inability to form new memories) is accompanied by a retro-
grade amnesia (RA; a loss of memory acquired before the trauma). These
memory failures are specific to declarative, as opposed to nondeclarative,
memory. The extent of RA can be quite variable, but is typically temporally
graded, i.e., memories acquired just prior to the lesion (recent memory) are
more severely impacted than those acquired several years before (remote
memory; Scoville and Milner, 1957; Squire and Alvarez, 1995; Knowlton
and Fanselow, 1998; but see also Nadel and Moscovitch, 1997, 1998). In
humans, temporally graded RA is inferred through the
use of retrospective memory tests, such as those examin-
shows (e.g., Rimpel-Clower et al., 1996; Reed and
Squire, 1998; Squire, 1992).
Many studies in animals have examined the effect of
hippocampal lesions made prior to training (e.g., Olton
et al., 1979; Morris, 1983; Phillips and LeDoux, 1992;
Kim et al., 1993), but relatively few have examined tem-
porally graded RA after damage to the hippocampal for-
mation (Zola-Morgan and Squire, 1990; Winocur,
1990; Kim and Fanselow, 1992; Kim et al., 1995;
Bolhuis et al., 1994; Cho et al., 1993; Cho and Kesner,
et al., 1999; for a summary, see Milner et al., 1998).
Aspects of hippocampal lesion-induced amnesia can
be found in Pavlovian fear conditioning, in which a tone
conditional stimulus (CS) is paired with a shock uncon-
ditional stimulus (US) several times in a novel context.
Rats trained in this manner develop a fear of both the
tone and training context, which we measure as freezing,
an adaptive species-specific defense reaction (Bolles,
1970; Fanselow, 1980). Contextual fear acquisition may
1980; Sutherland and Rudy, 1989), and many lines of
evidence support hippocampal involvement in contex-
tual fear conditioning (see, for example, Maren et al.,
1998; Phillips and LeDoux, 1992).
In our laboratory, we examined temporally graded RA
of contextual fear in three studies. Kim and Fanselow
days before the animals received an electrolytic lesion of
the dorsal hippocampus (DH) or sham surgery. In this
study, DH lesions made 1 day after training produced a
near-complete deficit in contextual fear, while sparing
tone freezing. In contrast, lesions made 28 days after
lesions produced a selective and time-limited RA of con-
textual fear (Fig. 1A). This led us to suggest that the
hippocampus plays a temporary role in the formation of
some aspects of contextual fear memory, which must be-
come independent of the hippocampus over some time
period after training.
Grant sponsor: NSF; Grant number: IBN 9723295; Grant sponsor: UCLA
C.M. Kernan Dissertation Year Fellowship.
*Correspondence to: Michael S. Fanselow, Department of Psychology,
University of California at Los Angeles, 1285 Franz Hall, Los Angeles, CA
90095-1563. E-mail: firstname.lastname@example.org
Accepted for publication 1 August 2000
HIPPOCAMPUS 11:8–17 (2001)
© 2001 WILEY-LISS, INC.
In a second study, Maren et al. (1997) examined RA of contex-
tual fear after excitotoxic N-methyl-d-aspartate (NMDA) lesions
28, or 100 days prior to an NMDA lesion of DH or sham surgery.
NMDA lesions appeared somewhat larger than the electrolytic
produced a near total RA of contextual fear if made 1 day after
training, but those made 100 days after training produced only a
mild deficit. These lesions also produced mild tone deficits. Thus,
although the time-course for consolidation in this study appeared
to be longer, RA was also temporally graded, as 100-day-old con-
textual memory was much more resistant to hippocampal damage
than 1-day-old memory (Fig. 1B).
In both Kim and Fanselow (1992) and Maren et al. (1997),
temporally graded RA was examined between groups, i.e., ani-
mals had only recent or remote memory. In a third study, we
wanted to examine if temporally graded RA of contextual fear
could be demonstrated within subjects. Within-subjects exam-
ination offers a number of advantages, including better control
over performance effects and a reduction in the number of
animals required. Moreover, it more closely approximates the
way temporally graded RA is examined in humans. Thus, An-
agnostaras et al. (1999) gave animals 10 tone-shock pairings in
one context (remote memory), followed by 10 tone-shock pair-
ings in a highly distinct context (with a different tone) 50 days
later (recent memory). One day after recent training, the ani-
mals received electrolytic lesions of the DH (as in Kim and
Fanselow, 1992) or sham surgery. In this study, DH lesions
produced a severe deficit in recent contextual fear memory,
reducing it by two thirds, while sparing remote contextual fear
as well as remote and recent tone fear. Thus, it is apparent that
unambiguous time-limited RA of contextual fear can be dem-
onstrated within subjects (Fig. 1C).
Based on these findings, we offered a mnemonic account of
contextual freezing deficits after hippocampal damage which
accords well with other views of this structure’s role in memory
(Squire and Alvarez, 1995; O’Keefe and Nadel, 1978; Suther-
land and Rudy, 1989; Squire, 1992). In this account, the hip-
pocampus plays a role in acquiring and consolidating the uni-
fied (spatial, configural, multimodal) representation of the
contextual CS (Young et al., 1994; Maren et al., 1998). The
specific role we have proposed is the construction and tempo-
rary maintenance of the unified representation of the contextual
CS, rather than the context-shock association or other mne-
monic aspects of the fear conditioning experience. The bases for
this hypothesis are discussed in detail below. Very generally, the
view is that simple elements such as a tone CS or simple details
that comprise a context are available to animals regardless of the
integrity of the hippocampus. But animals without a hippocam-
pus cannot establish or consolidate a memory of an integrated
representation of the context as a unified whole (Fanselow,
from our laboratory. A: From Kim and Fanselow (1992). Rats were
DH 1, 7, 14, or 28 days after training. After recovery, contextual
observed when lesions were made 1 day, but not 28 days, after train-
ing. B: From Maren et al. (1997). Rats were given three tone-shock
100 days after training. After recovery, contextual freezing (peak %
time, mean ? SEM) was assessed. A severe deficit was observed in rats
that received lesions 1 day after training, but only a very mild deficit
Summary of studies examining RA of contextual fear
was observed 100 days after training. C: Anagnostaras et al. (1999).
Unlike A and B, temporally graded RA of contextual fear was exam-
ined within subjects in this study. Rats received 10 tone-shock pair-
ings in one context, followed by 10 tone-shock pairings in a different
context (with a different tone) 50 days later, followed by electrolytic
acquired 1 day prior to the lesion, but no significant deficit of that
acquired 50 days before. Thus, these three studies show that RA of
contextual fear is temporally graded after electrolytic or excitotoxic
DH lesions, when examined between groups or within subjects. See
original publications for additional information.
HIPPOCAMPUS AND CONTEXTUAL FEAR CONDITIONING
1980, 1990; Young et al., 1994; Maren et al., 1998; Anagnost-
aras et al., 1999).
This role for the DH is in contrast to that proposed for the
amygdala. After NMDA lesions to the basolateral amygdala,
Maren et al. (1996a) found a complete loss of contextual and tone
fear even when lesions were made 28 days after training, and there
is no evidence of a temporal gradient for retrograde amnesia in
amygdala-dependent Pavlovian fear conditioning (Lee et al.,
1996). Thus, the amygdalar complex plays a general role in con-
textual and tone fear for some time after acquisition. In contrast to
the proposed role for the hippocampus in the acquisition and
temporary maintenance of the contextual CS, the amygdala may
mediate CS-US associations, the shock US representation, and
perhaps even the production of fear responses (i.e., URs and CRs)
as well (Maren and Fanselow, 1996; Lee et al., 1996; Campeau et
al., 1992; Rogan et al., 1997).
Nonetheless, several recent controversies have emerged which
pal lesions are made prior to training (e.g., Maren et al., 1997;
Frankland et al., 1998; Gerlai, 1998; Cho et al., 1999). Second, a
dramatic, temporally stable, and nonspecific deficit in freezing has
been observed after large kainate-colchicine lesions of the hip-
freezing deficits after hippocampal damage was recently offered as
an alternative to the mnemonic account (Good and Honey, 1997;
McNish et al., 1997).
In this review we will address how these controversies may still
be addressed by the standard mnemonic account of contextual
freezing deficits. We will then offer a more detailed description of
this mnemonic account.
ABSENCE OF ANTEROGRADE AMNESIA OF
CONTEXTUAL FEAR AFTER
Although there is considerable evidence of anterograde amnesia
LeDoux, 1992; Kim et al., 1993; Maren and Fanselow, 1997;
Young et al., 1994; Maren et al., 1998), there is also evidence that
contextual fear acquisition can sometimes be spared. For example,
Maren et al. (1997) found that, after NMDA lesions of DH, ret-
rograde amnesia was severe, but contextual fear acquisition was
completely spared. Second, Phillips and LeDoux (1994) reported
an absence of AA from electrolytic DH lesions after unsignaled
reported only mild impairment in contextual fear after electrolytic
acid lesions in mice (Cho et al., 1999). These findings are not that
surprising, however, because anterograde impairments when ob-
served were also milder than those observed when lesions were
made 1 day after training (e.g., Kim and Fanselow, 1992; Kim et
et al, 1998).
An explanation has been offered for this discrepancy between
anterograde and retrograde amnesia of contextual fear, based on
alternate solutions that may be adopted in contextual fear acquisi-
tion (Maren et al., 1997, 1998; Frankland et al., 1998). We have
argued that fear of contextual cues may be mediated by the unified
representation of the context and/or the myriad simple individual
elements that make up the context (Fig. 2). By the unified repre-
sentation solution (Fig. 2B), the rat first associates all static, low-
contingency (weakly predictive) contextual elements into a single
representation, which is then associated with the shock. We have
argued that the hippocampus, because of its role in spatial and
al., 1985). However, contextual fear conditioning is not uniquely
representational in nature, because an alternative solution exists.
Since the individual elements of the context are perceived by an
animal even without a hippocampus, an elemental solution (Fig.
directly associated with shock, and these associations would not
require the hippocampus.
In the intact animal both strategies would be available, while an
animal with a hippocampal lesion could only use the elemental
strategy. For reasons elaborated above, animals with a functioning
hippocampus will favor the unified representational solution,
while hippocampally impaired animals will favor the elemental
solution because that is all that is available to them. Therefore, AA
will be mild if the elemental solution sufficiently solves the task
demands. An intact animal that learned the unified representa-
tional solution would be severely impaired if it subsequently lost
(i.e., retrograde amnesia would be severe).
It is important to note that the relatively weak anterograde am-
nesia and robust, but time-limited retrograde amnesia appear to
conflict with data from amnesics indicating profound anterograde
amnesia (e.g., Scoville and Milner, 1957). However, evidence pre-
sented here indicates that hippocampal animals may acquire the
classically conditioned fear task in a manner that normal animals
would not, using an elemental strategy. Anterograde deficits may
tal solutions are not efficient. Frankland et al. (1998) reported
evidence supporting this view. Although DH-lesioned animals
more severely impaired in contextual fear discrimination, where
mice had to discriminate two similar chambers, only one of which
had been paired with shock. Frankland et al. (1998) posited that
this discrimination is more difficult to solve using an elemental
Nonetheless, robust deficits in contextual fear (trained accord-
ing to the standard paradigm) were observed when lesions were
hippocampus to acquire the contextual representation (see also
Maren et al., 1997).
ANAGNOSTARAS ET AL.
TEMPORALLY STABLE AND NONSPECIFIC
LOSS OF FEAR AFTER KAINATE-
COLCHICINE HIPPOCAMPAL LESION
Unpublished results of Weisend et al. (1996) indicate a tempo-
kainate-colchicine hippocampal lesions. According to the authors,
these lesions affect the entire hippocampus, and the data suggest
that the hippocampus has a permanent role in the storage of fear
memory or performance of the freezing response; studies examin-
ing hippocampal function, it was argued, should produce large,
rather than restricted damage to the hippocampus. Based on these
data, Nadel and Moscovitch (1997, 1998) concluded that spatial
memories become more efficiently represented within the hip-
pocampus over time; thus, partial lesions become less effective as
the hippocampus comes to encode the memory more efficiently.
However, this conclusion may need to be qualified.
First, the possibility of hippocampal lesion-induced extrahip-
pocampal damage, particularly to the amygdala, has not been ad-
dressed. Kainic acid injection in one region is known to produce
damage in distal areas that the region projects to (e.g., Mintz and
Knowlton, 1993). Because we have reported a similar complete
loss of context and tone conditional fear after basolateral amygdala
damage (Maren et al., 1996a), exploration of this possibility is a
necessity. Indeed, Jarrard (1983, 1989) abandoned the use of
kainic acid for selective hippocampal lesions because of cell loss in
areas a considerable distance from the site of injection. Substantial
damage to the amygdala and cortex were reported after intrahip-
mice, kainate-colchicine lesions of the hippocampus resulted in a
50% shrinkage of the cortex (Logue et al., 1997). Thus, the assess-
ment of extrahippocampal damage, particularly in the putative
permanent stores for contextual fear memory (i.e., the amygdala
and cortex), is critical for the interpretation of these results. It is
likely that even minimal damage to the amygdala may produce
significant retrograde amnesia of fear (Maren, 1998).
Second, there exists the possibility that ventral hippocampal-
the amygdala. The loss of these afferents might cause significant
ventral hippocampal and ventral angular bundle lesions (e.g.,
Maren and Fanselow, 1995) must be further examined before
functional disruption of the amygdala can be dismissed. Indeed,
Maren (1999) reported a severe and nonspecific disruption of
freezing after restricted damage to the subicular region.
Third, the excitotoxic technique may be unsuitable for the lo-
calization of memory in retrograde amnesia paradigms. The mas-
sive and sustained excitatory discharge emanating from the hip-
pocampus may be sufficient to cause “catastrophic interference” of
memories already stored in other structures (e.g., McClelland et
al., 1995), i.e., structures which receive afferents from the hip-
persistent and lasting noisy output. This may severely disrupt
memories in these structures, perhaps especially those that were
recently encoded. Thus, studies examining the effects of excito-
toxic lesions of the DH are potentially confounded by disruption
of memories stored outside the hippocampus.
This could also qualify the findings of Maren et al. (1997), in
which we examined the effects of NMDA lesions of the DH. In
100 days, and mild tone impairments were evident. This is in
contextual fear. A: Elemental solution. As the result of pairing, the
tone comes to be strongly associated with the shock representation
(i.e., shock memory). Individual elements of the context are relatively
poorly paired with the shock and form weak associations with the
shock. During tone testing, the tone strongly arouses the shock rep-
resentation, eliciting strong fear. In context testing, each individual
element only weakly arouses memory of the shock, but these may
summate to elicit strong fear. B: Unified representational solution. As
before, the tone becomes strongly associated with shock. However, in
this case, in addition to having access to the elements of the context,
the hippocampus forms a unified representation of the context. This
is a strong predictor of shock and forms a strong association with the
Alternate solutions that may be used in acquiring
shock representation. During testing, either the tone or context can
strongly arouse the shock memory and elicit strong fear. It is assumed
that in normal animals both solutions are available, but the unified
superior predictive power and/or salience. Thus, posttraining hip-
pocampal lesions produce a severe retrograde amnesia of contextual
associated with shock during training. However, when lesions are
made prior to training, the animal may learn an elemental solution,
because the unified representation is not available and therefore does
not compete with the elements. Thus, during testing contextual fear
may appear normal, if the task demands are such that the elemental
strategy can support sufficient conditioning.
HIPPOCAMPUS AND CONTEXTUAL FEAR CONDITIONING
contrast to those studies that used electrolytic lesions of the DH,
which are argued to also damage fibers of passage, but produced
gradients lasting less than 50 days and no tone deficits (Kim and
Fanselow, 1992; Anagnostaras et al., 1999). There seems to be a
relationship between RA to remotely trained contexts, impair-
ments in tone conditioning, and the potential for distal damage.
The extension of the temporal retrograde gradient for contextual
fear is accompanied by tone conditioning deficits and is produced
by lesions of the kind most likely to produce distal damage and/or
catastrophic interference (complete/kainate-colchicine lesions).
These combined tone and remote contextual deficits are least ap-
parent with electrolytic lesions.
Thus, substantial additional research remains before any con-
clusion can be reached about the effects of complete hippocampal
damage on RA for contextual fear. However, the possibility re-
mains that ventral hippocampal neurons, perhaps because of their
connections with the amygdala, play a more general role in the
production of the fear response than dorsal hippocampal neurons.
However, temporally stable retrograde amnesia of fear after intra-
hippocampal kainate-colchicine injection is not good evidence for
this technique also likely produces disruption in the putative per-
manent storage sites for memory, i.e., the standard consolidation
model can accommodate a temporally stable RA after this kind of
PERFORMANCE ACCOUNT OF
CONTEXTUAL FREEZING DEFICITS AFTER
Despite the specificity of the deficit we have reported for recent
contextual fear (and not remote contextual fear or tone fear), a
performance-based alternative to the mnemonic account of con-
textual freezing deficits after hippocampal lesions has been offered
(Good and Honey, 1997; McNish et al., 1997; Blanchard et al.,
1977). According to this view, the deficits observed in contextual
freezing may be due to locomotor hyperactivity, an established
effect of hippocampal lesions (Teitelbaum and Milner, 1963;
hippocampal animals fail to exhibit normal freezing because hy-
peractivity caused by the lesion disrupts freezing and behavioral
inhibition directly (Good and Honey, 1997; Blanchard et al.,
1977; Douglas, 1967). This is a performance account, because in
this view hippocampal lesions disrupt the freezing response rather
for recent contextual fear, several assumptions have been made
(McNish et al., 1997). First, it is assumed that higher levels of fear
are less susceptible to hippocampal lesion than low levels of fear.
Second, it is assumed that remote contextual fear is stronger than
recent contextual fear because of “incubation” of fear over time.
Third, tone fear is assumed to be greater than contextual fear
because of presumed greater CS-US contingency (McNish et al.,
1997). We have responded to this view in detail elsewhere (Maren
et al., 1998); in brief, two direct lines of evidence fail to support
First, the amnesic deficit in contextual fear is unrelated to the
aras et al. (1999), remote and recent contextual fear were equiva-
lent (see Fig.1), and in Anagnostaras et al. (1999), tone fear was
weaker than context fear. Because remote and recent contextual
and tone fear were examined within subjects in the latter study, it
is quite apparent that DH-lesioned rats can in fact freeze at the
same levels as sham animals. Indeed, because of the substantial
which retrograde amnesia of contextual fear has been reported are
Second, even when lesions are made before training, where def-
icits are less substantial, hyperactivity cannot predict the level of
the freezing deficit (Maren et al., 1998; Maren and Fanselow,
1997). Because hyperactivity is proposed to directly disrupt freez-
ing, according to this view, there should be a strong, negative,
within-group activity-freezing correlation, and there is not. In a
between hyperactivity and the contextual freezing deficit, even
when hyperactivity was measured in the same conditioning cham-
adapted from Maren et al. (1998). Hipp rats, which received electro-
lytic lesions of the DH prior to testing, were split into Low (below or
at activity median; n ? 27) or High (above activity median; n ? 21)
activity groups and compared with Sham rats (n ? 49). A: Activity.
Full cage crossovers (mean ? SEM) were scored for the 3-min period
prior to the first shock on the training day. There were significant
group differences (ANOVA, F(2,94) ? 47, P < 0.0001). High Hipp
rats were significantly more active than Low Hipp (Fisher’s PLSD,
P < 0.0001) or Sham rats (P < 0.0001), which did not differ signif-
icantly from each other (P > 0.2). B: Contextual fear. The animals
were returned to the conditioning chambers at least 1 day after train-
ing, and % time freezing (mean ? SEM) was assessed. There were
significant group differences (F(2,94) ? 12, P < 0.0001). High Hipp
both groups exhibited substantially less freezing than Sham rats (P <
0.001). Thus, a robust deficit in contextual freezing can be demon-
strated even in rats with DH lesions that are not significantly more
active than shams. Indeed, even though High Hipp rats exhibited
twice as much activity as Low Hipp or Sham animals, their freezing
deficit was not much more pronounced than that of Low Hipp rats,
suggesting that the contribution of hyperactivity to freezing deficit is
Hyperactivity and contextual freezing deficit. Data
ANAGNOSTARAS ET AL.
In this analysis, the sample of hippocampal animals was split (me-
controls (Low Hipp) and a group that was more than twice as
active as controls (High Hipp; Fig. 3A). However, these animals
did not differ in terms of amnesic severity (Fig. 3B). Thus, it
appears that AA can be demonstrated even in a sample of rats that
is not significantly hyperactive. Indeed, the level of freezing deficit
is only slightly different in these animals from those that are over
twice as active. Thus, the freezing deficit and hyperactivity do not
appear to be directly related.
A number of indirect lines of evidence also fail to support the
response competition view. For example, local infusion of APV,
muscimol, or scopolamine into the hippocampus at the time of
training selectively blocks the acquisition of contextual fear, even
though the drugs are not present during testing (Young et al.,
1994; Bellgowan and Helmstetter, 1995; Gale et al., 1998). In-
deed, posttraining infusion of an RNA synthesis inhibitor locally
into the hippocampus blocks the consolidation of contextual fear
while sparing tone fear (Bailey et al., 1997).
showing that context-potentiated startle was not affected by DH
lesions made immediately after training. This suggests that the
deficits in contextual fear we have observed may be specific to the
ever, require qualification of this conclusion. First, because the
measure of context-potentiated startle compared presurgery base-
line and postsurgery testing, it is possible that hippocampal lesions
elevated startle unconditionally, as some evidence exists that this
occurs (Coover and Levine, 1972; Tilson et al., 1987; Lee and
Davis, 1997). This elevation of baseline may have occluded any
amnesic effect of the lesions. Second, the startle apparatus used is
quite small (and under dark conditions), which may promote ele-
mental associations. Whereas the typical chamber we use is many
times larger than the rat, the startle apparatus snugly fits only the
rat. Moreover, in our preparation the rats have a view of the distal
features of the room, whereas in the startle apparatus used by
McNish et al. (1997), the rats were in a light and sound attenua-
tion chamber. These conditions may not be ideal for promoting
poverty in complexity of the context, and the rat may instead be
acquiring associations that are elemental in nature (see Maren et
(1997) did not find good contextual control over freezing. In their
studies where rats were trained in one context and tested in an-
other, they still exhibited considerable freezing. With the very dif-
ferent and larger contexts used in our laboratory, such context-
shifts between training and testing eliminate freezing (e.g., Kim
and Fanselow, 1992; Maren et al., 1997; Anagnostaras et al.,
1999). Thus, considerable further research is required to fully ap-
preciate how the findings of McNish et al. (1997) will expand our
understanding of the hippocampus’ role in contextual fear condi-
may be useful in resolving these issues.
Nonetheless, the hyperactivity account has raised an important
ulation (e.g., lesion, drug, mutation) has the possibility of disrupt-
ing performance of the response, rather than the learning and/or
memory that mediates the behavior. This is why it is critical to
include controls for such performance effects and to provide con-
fear and freezing will provide a useful example in this regard.
Thus, three controversies have complicated the view that the
hippocampus has a specific role in the acquisition and storage of
the contextual CS in Pavlovian fear conditioning. These contro-
versies have substantial merit and, we believe, have actually
strengthened our understanding of the hippocampus’ role in con-
anterograde and retrograde amnesia of fear after hippocampal le-
sions is now apparent. Second, the substantial, and perhaps non-
pocampal damage in this region) on contextual fear is becoming
apparent. Third, an important performance confound, the degree
to which hippocampal lesion-induced hyperactivity may have dis-
rupted freezing, has been examined and, we believe, ruled out.
WHAT IS THE ROLE OF THE
HIPPOCAMPUS IN CONTEXTUAL FREEZING?
In Pavlovian fear conditioning, contextual and tone CSs come
to be associated with a shock US as the result of pairing. Based on
several classic experiments, it is generally believed that the animal
forms internalized neural representations of the contextual CS,
tone CS, and shock US, and that conditioning results in the asso-
ciation of these representations. After pairing, when the animal is
response. It is believed that CSs elicit fear by arousing the repre-
sentation (i.e., the memory) of the shock US (Rescorla, 1973,
1974), i.e., when confronted with a place or signal that was previ-
ously paired with threat, the animal remembers that these predict
the threat and exhibits a defensive response that may protect it
from the threat. For example, in the natural environment, freezing
(Fanselow and DeOca, 1998). A signal for threat (the presence of
a number of adaptive responses. For example, freezing is accompa-
the threat and may enhance survival (Fanselow, 1986a, 1998;
Fanselow and Bolles, 1979).
Deficits in conditioning have traditionally been interpreted as
grab the attention of associative processes), US processing (the
US’s ability to reinforce associative learning), or CS-US associa-
tion. Whereas the processing of the tone or shock may be simple
because of their unimodal and discrete nature, the processing of a
contextual CS may be more complex because it is multimodal and
temporally diffuse (Fanselow, 1986b, 1990). To provide signifi-
HIPPOCAMPUS AND CONTEXTUAL FEAR CONDITIONING
cant information content, the multiple elements of the contextual
CS may need to be bound into a unified representation (Fig. 2B).
It is specifically in the formation and consolidation of this repre-
sentation that we believe the hippocampus is involved, because we
shock or tone) can sometimes protect rats from anterograde am-
nesia (Young et al., 1994).
Recently we found that remote context preexposure can protect
rats from RA after hippocampal damage as well (Fig. 4). In this
out shock or tone) to one context, and then 35 days later given 10
ings in the same context and a highly distinct, novel context (1 day
apart; order counterbalanced; for a complete description of the
procedures, see Anagnostaras et al., 1999). On the next day, they
received sham (n ? 12) or electrolytic DH lesions (n ? 11). After
a 10-day recovery period, they were given two 6-min remote and
a novel context (on separate days). The results of the context tests
are depicted in Figure 4. There was a group ? preexposure inter-
action (F(1,21) ? 7.7, P ? 0.01). Although RA of the not-preex-
significant deficit to the remotely preexposed context (F(1,21) ?
0.8, P ? 0.3), even though both had been conditioned 1–2 days
before the lesion. Moreover, fear of the not-preexposed and preex-
posed contexts differed for DH (F(1,10) ? 11, P ? 0.01) but not
sham rats (F(1,11) ? 0.1, P ? 0.9). The rats also did not exhibit
deficits to either of the two tones used during conditioning (data
not shown; group ? context preexposure interaction, F(1,21) ?
remote context preexposure appears to be sufficient to protect rats
from the otherwise devastating RA of recent contextual fear that
DH lesions produce.
Hippocampal LTP may be required for assembling these stim-
5-phosphonovaleric acid (APV) into the hippocampus is sufficient
to block the acquisition of contextual fear (Young et al., 1994).
There is evidence that the other mnemonic processes (shock
ing are supported by the amygdala (Maren and Fanselow, 1996;
Rogan et al., 1997; Campeau et al., 1992; Fanselow, 1998). For
example, local infusion of APV into the basolateral amygdala
blocks the acquisition of tone or contextual fear (Fanselow and
Kim, 1994; Maren et al., 1996b; Campeau et al., 1992), and
lesions of this region produce a severe anterograde and temporally
stable retrograde amnesia (Maren et al., 1996a; Lee et al., 1996).
Thus, the amygdala supports the association of the unified contex-
tual representation (acquired in the hippocampus) with the shock
US in a way similar to which it supports tone-shock associations.
Context-shock association may be via hippocampal-amygdala
LTP, as high-frequency stimulation of the pathways carrying in-
formation from the hippocampus to the amygdala produces LTP,
and lesions of these projections selectively block the acquisition of
contextual fear (Maren and Fanselow, 1995; Maren, 1996).
representational solution actively competes with an alternative el-
impairment in contextual conditioning, posttraining lesions are
devastating. This suggests that even though animals with hip-
fear, intact animals do not. This may result from the unified rep-
resentation competing with multiple elemental representations for
association with the shock.
ing experiments that it is at the core of many models of condition-
ing (Sutherland and Mackintosh, 1971; Rescorla and Wagner,
1972; Wagner and Brandon, 1989). There are two factors that
favor a particular stimulus in this competition for associative
strength. One is that better predictors win over poorer predictors.
The other is that more salient stimuli win over weaker stimuli.
Both of these factors should favor the unified representation over
the elemental stimuli. Because a rat could only perceive a subset of
elements of the context at any instant, the unitary representation
will be more reliably paired with shock than any constituent ele-
cue that is paired with shock on 50% of its trials will condition
strongly if it is the best predictor available, but not at all if a better
predictor is available (Wagner et al., 1968). Additionally, it seems
the unified representation of these elements. Another well-estab-
lished finding in the literature, i.e., overshadowing, speaks to this
issue. In overshadowing, a weak stimulus conditions well on its
own, but poorly if presented in compound with a more salient
pocampal lesion-induced RA of recent contextual fear. Rats were pre-
days later were given Pavlovian fear conditioning in the same context
and a highly distinct novel context. One to 2 days after conditioning,
the animals received either sham or electrolytic DH lesions. Plotted
(% time freezing, mean ? SEM). DH lesions produced a severe RA of
fear to the not-preexposed context, but only a mild, nonsignificant
impairment of fear to the preexposed context.
Remote context preexposure protects rats from hip-
ANAGNOSTARAS ET AL.
pus is intact, the unified representation may overshadow the single
elements. Note that learning theories explain the overshadowing
and the predictive validity effects with the same algorithm (e.g.,
Rescorla and Wagner, 1972). Therefore, a fundamental assump-
tion from theories of Pavlovian conditioning is that if there are
both elements of a context (this would have to be true) and the
elements can be unified into a contextual representation, the ele-
ments and representation must compete. Fortunately, this a priori
prediction from Pavlovian theory also explains a number of sub-
tleties of the effects of hippocampal lesions on contextual fear
For example, Phillips and LeDoux (1994) reported that hip-
pocampal lesions produce AA of contextual fear when initial train-
ing is signaled (with a tone) but not when it is unsignaled (without
a tone). When a tone (that has high salience and predictive power)
the unified contextual representation and the elemental stimuli.
While we have argued that hippocampal-lesioned animals can
sometimes acquire contextual fear by elemental associations, the
addition of a competing cue, the tone, which is both more salient
and a better predictor, could weaken elemental contextual associ-
ations enough to reveal a deficit. However, the relative lack of
and tone signal suggests that they have similar salience and predic-
With regard to the processes of consolidation, the study of con-
recent memory traces (engrams) actually migrate from the hip-
pocampus to distal locations such as the cortex over the consolida-
tion gradient (months to years). In another view, engrams are
established initially at “slow-learning” permanent locations, and
the hippocampus only plays a role in temporarily keeping these
engrams stable (e.g., McClelland et al., 1995). We believe that
of these views, see Squire and Alvarez, 1995). For example, if the
recently acquired unified contextual CS (located in the hippocam-
pus) migrated during consolidation to another location, it is un-
representation (located in the amygdala). This would not be a
problem if the hippocampus was only involved in temporarily
stabilizing a nonmigratory trace elsewhere. These speculations re-
main to be tested, of course, but contextual fear may offer an
The assembly of the unified contextual CS is similar to the role
the hippocampus is argued to play in many paradigms. One im-
portant advantage is that because a functional hippocampus is not
required for the performance of contextual freezing, time-limited
to the spatial Morris water and Olton radial maze tasks, in which
of performance, perhaps because of navigational or working mem-
ory demands (Knowlton and Fanselow, 1998). One weakness of
contextual fear in examining the anterograde influence on hip-
pocampal function is that because acquisition of contextual fear
can sometimes appear normal even in animals with hippocampal
lesions, evidence of acquisition of contextual fear cannot be taken
as unequivocal evidence that hippocampal function is normal
(Gerlai, 1998). This is in contrast to maze learning tasks, which
apparently show more robust anterograde amnesia. While RA for
contextual fear provides clear evidence, we need to refine the an-
terograde amnesia tests in a way to selectively control the unified
vs. elemental solutions. Nonetheless, taken together, evidence
from contextual freezing and maze learning experiments, which
can often be done in the same animals, may be a good indicator of
functional disruption of the same neural system that mediates hu-
man declarative memory.
comments on an earlier version of this manuscript. This research
Anagnostaras SG, Maren S, Fanselow MS. 1999. Temporally graded ret-
rograde amnesia of contextual fear after hippocampal damage in rats:
within-subjects examination. J Neurosci 19:1106–1114.
Bailey DJ, Sun W, Kim JJ, Helmstetter FJ. 1997. Inhibition of RNA
synthesis in the amygdala and hippocampus selectively blocks acquisi-
tion of Pavlovian fear conditioning. Soc Neurosci Abstr 23:1609.
Bellgowan PSF, Helmstetter FJ. 1995. Effects of muscimol applied to the
dorsal hippocampus on the acquisition and expression of cued versus
contextual fear conditioning. Soc Neurosci Abstr 21:1219.
Blanchard DC, Blanchard RJ, Lee MC, Fukunaga KK. 1977. Movement
arrest and the hippocampus. Physiol Psychol 5:331–335.
ory reactivation in rats with ibotenate lesions to the hippocampus or
subiculum. Q J Exp Psychol [B] 47:129–150.
Bolles RC. 1970. Specifies-specific defense reactions and avoidance learn-
ing. Psychol Rev 77:32–48.
the N-methyl-D-aspartate receptor antagonist AP5 blocks acquisition
but not expression of fear-potentiated startle to an auditory condi-
tioned stimulus. Behav Neurosci 106:569–174.
Cho YH, Kesner RP. 1996. Involvement of entorhinal cortex or parietal
cortex in long-term spatial discrimination memory in rats: retrograde
amnesia. Behav Neurosci 110:436–442.
Cho YH, Beracochea D, Jaffard R. 1993. Extended temporal gradient for
the retrograde and anterograde amnesia produced by ibotenate ento-
rhinal cortex lesions in mice. J Neurosci 13:1759–1766.
Cho YH, Friedman E, Silva AJ. 1999. Ibotenate lesions of the hippocam-
pus impair spatial learning but not contextual fear conditioning in
mice. Behav Brain Res 98:77–87.
Coover GD, Levine S. 1972. Auditory startle response of hippocampec-
tomized rats. Physiol Behav 9:75–77.
Douglas RJ. 1967. The hippocampus and behavior. Psychol Bull 67,
Douglas RJ, Isaacson RL. 1964. Hippocampal lesions and activity. Psy-
chonomic Sci 1:187–188.
Fanselow MS. 1980. Conditioned and unconditional components of
post-shock freezing. Pavlov J Biol Sci 15:177–182.
HIPPOCAMPUS AND CONTEXTUAL FEAR CONDITIONING
Fanselow MS. 1986a. Conditioned fear-induced opiate analgesia: a com-
peting motivational state theory of stress-analgesia. Ann NY Acad Sci
Fanselow MS. 1986b. Associative vs. topographical accounts of the im-
mediate shock-freezing deficit in rats: implications for the response
selection rules governing species-specific defense reactions. Learn Mo-
Anim Learn Behav 18:264–270.
Fanselow MS. 1998. Pavlovian conditioning, negative feedback, and
blocking: mechanisms that regulate association formation. Neuron
Fanselow MS, Bolles RC. 1979. Naloxone and shock-elicited freezing in
the rat. J Comp Physiol Psychol 93:736–744.
Fanselow MS, DeOca BM. 1998. Defensive behaviors. In: Greenberg G,
Haraway MM, editors. Comparative psychology: a handbook. New
York: Garland Publishing. p 653–655.
Fanselow MS, Kim JJ. 1994. Acquisition of contextual Pavlovian condi-
tioning is blocked by application of an NMDA receptor antagonist
D,L-2-amino-5-phosphonovaleric acid to the basolateral amygdala.
Behav Neurosci 108:210–220.
for contextual conditioning. Behav Neurosci 112:863–874.
Gale GD, Anagnostaras SG, Fanselow MS. 1998. Cholinergic modula-
tion of Pavlovian fear conditioning: selective disruption of contextual
rosci Abstr 24:1904.
ing in mice: a strain comparison and a lesion study. Behav Brain Res
Good M, Honey RC. 1997. Dissociable effects of selective lesions to
and spatial learning. Behav Neurosci 111:487–493.
Jarrard LE. 1983. Selective hippocampal lesions and behavior: effects of
kainic acid lesions on performance of place and cue tasks. Behav Neu-
components of the hippocampal formation. J Neurosci Methods 29:
Kim JJ, Rison RA, Fanselow MS. 1993. Effects of amygdala, hippocam-
pus, and periaqueductal gray lesions on short- and long-term contex-
tual fear. Behav Neurosci 107:1093–1098.
Kim JJ, Clark RE, Thompson RF. 1995. Hippocampectomy impairs the
memory of recently, but not remotely, acquired trace eyeblink condi-
tioned responses. Behav Neurosci 109:195–203.
on-line memory. Curr Opin Neurobiol 8:293–296.
Lee Y, Davis M. 1997. Role of the septum in the excitatory effect of
corticotropin-releasing hormone on the acoustic startle reflex. J Neu-
Lee Y, Walker D, Davis M. 1996. Lack of temporal gradient of retrograde
assessed with the fear-potentiated startle paradigm. Behav Neurosci
ing deficits in inbred mice in the Morris water maze and conditioned-
fear task. Behav Neurosci 111:104–113.
of stimuli with reinforcement. Psychol Rev 82:276–298.
Maren S. 1996. Synaptic transmission and plasticity in the amygdala. An
emerging physiology of fear conditioning circuits. Mol Neurobiol 13:
J Neurosci 18:3088–3097.
produce deficits in acquisition and expression of fear conditioning in
rats. Behav Neurosci 113:289–289.
dala induced by hippocampal formation stimulation in vivo. J Neuro-
Maren S, Fanselow MS. 1996. The amygdala and fear conditioning: has
the nut been cracked? Neuron 16:237–240.
pus, fimbria-fornix, or entorhinal cortex produce anterograde deficits
Maren S, Aharonov G, Fanselow MS. 1996a. Retrograde abolition of
conditional fear after excitotoxic lesions in the basolateral amygdala of
rats: absence of a temporal gradient. Behav Neurosci 110:718–726.
Maren S, Aharonov G, Stote DL, Fanselow MS. 1996b. N-methyl days-
aspartate receptors in the basolateral amygdala are required for both
acquisition and expression of conditional fear in rats. Behav Neurosci
Maren S, Aharonov G, Fanselow MS. 1997. Neurotoxic lesions of the
dorsal hippocampus and Pavlovian fear conditioning in rats. Behav
Brain Res 88:261–274.
Maren S, Anagnostaras SG, Fanselow MS. 1998. The startled seahorse: is
the hippocampus necessary for contextual fear conditioning? Trends
Cogn Sci 2:39–42.
McClelland JL, McNaughton BL, O’Reilly RC. 1995. Why there are
complementary learning systems in the hippocampus and neocortex:
insights from the successes and failures of connectionist models of
learning and memory. Psychol Rev 102:419–437.
lesions of the hippocampus: a disruption of freezing but not fear-
potentiated startle. J Neurosci 17:9353–9360.
Milner B, Squire LR, Kandel ER. 1998. Cognitive neuroscience and the
study of memory. Neuron 20:445–468.
Morris RGM. 1983. An attempt to dissociate “spatial mapping” and
“working memory” theories of hippocampal function. In: Seifert W,
editor. The neurobiology of the hippocampus. London: Academic
Press. p 405–432.
Nadel L. 1968. Dorsal and ventral hippocampus lesions and behavior.
Physiol Behav 3:891–900.
Nadel L, Moscovitch M. 1997. Memory consolidation, retrograde amne-
sia and the hippocampal complex. Curr Opin Neurobiol 7:217–227.
Nadel L, Moscovitch M. 1998. Hippocampal contributions to cortical
plasticity. Neuropharmacology 37:431–439.
Nadel L, Willner J. 1980. Context and conditioning: a place for space.
Physiol Psychol 8:218–228.
Nadel L, Willner J, Kurz EM. 1985. Cognitive maps and environmental
context. In: Balsam PD, Tomie A, editors. Context and learning.
London: Erlbaum. p 285–406.
O’Keefe J, Nadel L. 1978. The hippocampus as a cognitive map. Oxford:
Olton DS, Becker JT, Handelmann GE. 1979. The hippocampus, space,
and memory. Behav Brain Sci 2:313–365.
Pavlov IP. 1927. Conditioned reflexes. Anrep GV, translator. London:
Oxford University Press.
hippocampus to cued and contextual fear conditioning. Behav Neu-
ANAGNOSTARAS ET AL.
Phillips RG, LeDoux JE. 1994. Lesions of the dorsal hippocampal forma- Download full-text
tion interfere with background but not foreground contextual fear
conditioning. Learn Mem 1:34–44.
Reed JM, Squire LR. 1998. Retrograde amnesia of facts and events: find-
ings from four new cases. J Neurosci 18:3943–3954.
of enduring memory impairment after bilateral damage limited to the
hippocampal formation. J Neurosci 15:5233–5255.
Rescorla RA. 1973. Effect of US habituation following conditioning.
J Comp Physiol Psychol 82:137–143.
Rescorla RA. 1974. Effect of inflation on the unconditioned stimulus
value following conditioning. J Comp Physiol Psychol 86:101–106.
Rescorla RA, Wagner AR. 1972. A theory of Pavlovian conditioning: Varia-
sical conditioning II: current theory and research. In: Black AH, Prokasy
WF, editors. New York: Appleton Century Crofts. p 65–99.
Rogan MT, Staubli UV, LeDoux JE. 1997. Fear conditioning induces asso-
ciative long-term potentiation in the amygdala. Nature 390:604–607.
Scoville WB, Milner B. 1957. Loss of recent memory after bilateral hip-
pocampal lesions. J Neurol Neurosurg Psychiatry 20:11–21.
Squire LR. 1992. Memory and the hippocampus: a synthesis from find-
ings with rats, monkeys, and humans. Psychol Rev 99:195–231.
Squire LR, Alvarez P. 1995. Retrograde amnesia and memory consolida-
Sutherland NS, Mackintosh MJ. 1971. Mechanisms of animal discrimi-
nation learning. New York: Academic Press.
Sutherland RJ, Rudy RJ. 1989. Configural association theory: the role of
the hippocampal formation in learning, memory, and amnesia. Psy-
Teitelbaum H, Milner P. 1963. Activity changes following partial hip-
pocampal lesions in rats. J Comp Physiol Psychol 56:281–289.
RS. 1987. Time daysependent neurobiological effects of colchicine
administered directly into the hippocampus of rats. Brain Res 408:
Wagner AR, Brandon SE. 1989. Evolution of a structured connectionist
editors. Contemporary learning theories: Pavlovian conditioning and
the status of learning theory. Hillsdale, NJ: Erlbaum. p 149–189.
Wagner AR, Logan FA, Haberlandt K, Price T. 1968. Stimulus selection
in animal discrimination learning. J Exp Psychol 76:171–180.
Weisend MP, Astor RS, Sutherland RJ. 1996. The specificity and tempo-
Neurosci Abstr 22:1118.
Wiig KA, Cooper LN, Bear MF. 1996. Temporally graded retrograde
amnesia following separate and combined lesions of the perirhinal
cortex and fornix in the rat. Learn Mem 3:313–325.
hippocampal or dorsomedial thalamic lesions. Behav Brain Res 38:
Young SL, Bohenek DL, Fanselow MS. 1994. NMDA processes mediate
anterograde amnesia of contextual fear conditioning induced by hip-
sure. Behav Neurosci 108:19–29.
Zola-Morgan SM, Squire LR. 1990. The primate hippocampal forma-
tion: evidence for a time-limited role in memory storage. Science 250:
HIPPOCAMPUS AND CONTEXTUAL FEAR CONDITIONING