Behavioural Brain Research, 58 (1993) 69-79

© 1993 Elsevier Science Publishers B.V. All rights reserved. 0166-4328/93/$06.00 69

BBR 1510

Emotional memory systems in the brain

J o s e p h E. L e D o u x *
Center for Neural Science, New York University, 6 Washington Place, New York, NY 10003 (USA)

(Received 5 January 1992)

(Accepted 2 October 1993)

Key words: Emotional memory system; Thalamo-hippocampal pathway; Long-term potentiation; Emotional learning function

The neural mechanisms of emotion and memory have long been thought to reside side by side, if not in overlapping structures, of the lim-
bic system. However, the limbic system concept is no longer acceptable as an account of the neural basis of memory or emotion and is being
replaced with specific circuit accounts of specific emotional and memory processes. Emotional memory, a special category of memory involving
the implicit (probably unconscious) learning and storage of information about the emotional significance of events, is modeled in rodent
experiments using aversive classical conditioning techniques. The neural system underlying emotional memory critically involves the amygdala
and structures with which it is connected. Afferent inputs from sensory processing areas of the thalamus and cortex mediate emotional learn-
ing in situations involving specific sensory cues, whereas learning about the emotional significance of more general, contextual cues involves
projections to the amygdala from the hippocampal formation. Within the amygdala, the lateral nucleus (AL) is the sensory interface and the
central nucleus the linkage with motor systems involved in the control of species-typical emotional behaviors and autonomic responses. Studies
of cellular mechanisms in these pathways have focused on the direct relay to the lateral amygdala from the auditory thalamus. These studies
show that single cells in AL respond to both conditioned stimulus and unconditioned stimulus inputs, leading to the notion that AL might be
a critical site of sensory-sensory integration in emotional learning. The thalamo-amygdala pathway also exhibits long-term potentiation, a form
of synaptic plasticity that might underlie the emotional learning functions of the circuit. The thalamo-amygdala pathway contains and uses the
amino acid glutamate in synaptic transmission, suggesting the possibility that an amino-acid mediated form of synaptic plasticity is involved
in the emotional learning functions of the pathway. We are thus well on the way to a systems level and a cellular understanding of at least one
form of emotional learning and memory.

INTRODUCTION The amygdala is a key component of the brain sys-

tem involved in a variety of emotional func-
For the past several decades, the limbic system tions 1'2'23'24'32'33'14"15'37'38'40. Although the amygdala
concept 53-54 has dominated thinking about the neural was in the past included, together with the hippocam-
basis of emotion. However, during the same time, nu- pus, as part of the temporal lobe memory system 56 it
clei and pathways of the so-called limbic system have now appears that the amygdala is not involved in this
emerged as key components of the brain system under- system 57'85. While the amygdala may contribute to some
lying memory and other cognitive f u n c t i o n s 56'59'77'8°. cognitive processes (see ref. 1) its contribution to cog-
This indicates either that the limbic system concept is nition would seem to be considerably less than its con-
useful as an account of both memory and emotion or tribution to emotion. It might even be possible to ex-
that it is not very useful as an account of either. For plain amygdala contributions to cognition, when they
several reasons, detailed below, it seems that the latter occur, in terms of the loss of the affective qualities of
is more likely and that we should look beyond the cognition. The survival of the limbic system concept as
limbic system for a conceptual framework for under- an account of the neural basis of emotion may be di-
standing questions about the neural organization of rectly related to the inclusion of the amygdala in the
emotion and memory. limbic system 39.
First, those limbic areas most important for emotion Second, those limbic areas that have been implicated
are less directly involved in cognitive processes. in memory and other cognitive processes seem to have
less significant roles in emotion. For example, the hip-
* For all correspondence. pocampal formation is now believed to be an essential
70

component of the brain system underlying the con- ferent kinds of learning and memory capacities in the
scious recollection of past events 8° and the processing brain (e.g. refs. 58, 76, 80, 84). Most typically, the term
of spatial, configural, contextual and/or relational in- memory refers to the ability to consciously remember
formation 59'58'73'16. In general, damage to the hippoc- past experiences or previously learned information.
ampus does not produce clinical alterations of emotion This kind of memory, which is dependent upon the
or interfere with emotional responsivity in experimen- hippocampal system, is often referred to as declarative
tal tasks; the hippocampus is mainly involved in emo- or explicit (conscious) memory 1~'76'8°. There are a va-
tional processes when the non-emotional (especially riety of non-declarative or implicit (unconscious) me-
cognitive) aspects of the task require the hippocam- mory systems v6'a°. Each of these has its own neural
pus 3v. A prime example of this secondary role of the representation. The one of most interest here can be
hippocampus in emotion comes from studies showing referred to as an emotional memory system. This sys-
that electrical stimulation of the hippocampus is ca- tem is required for the acquisition and storage of infor-
pable of modulating the expression of emotional re- mation about the emotional significance of expe-
sponses elicited by hypothalamic stimulation but riences 14"15'32"33'36-3~'41. Thus, networks involving the
hippocampal stimulation itself does not elicit such re- hippocampal formation and associated cortical areas
sponses vS. The hippocampus thus seems to provide are involved in storing declarative or explicit memory
important inputs to emotional circuits but is not itself traces of experiences, regardless of whether the expe-
a primary link in those circuits. Gray's 25 hippocampal riences are emotional or not. But emotional experiences
theory of anxiety would seem to be at odds with this also lead to storage of information through the
view but it may be that contributions of the hippocam- amygdala. When memories stored through both the
pus to anxiety are more in the area of cognitions that amygdala and hippocampus are retrieved, they have a
trigger and maintain anxiety than in the emotional different flavor than when only the hippocampal system
aspects of anxiety 37'39. is involved. This dual activation of the amygdala and
Third, the limbic system concept itself has been chal- hippocampus may be what gives emotional memories
lenged, as there are no objective criteria that can be their special quality.
universally applied to decide whether a given structure Two hypotheses can be suggested as to why amygdala
belongs to the limbic system 7'37'39"82.While areas of the activation might add subjective affect to experiences
limbic system, however it is defined, are involved in and memories. The first is an hedonic hypothesis. That
emotional functions, not all limbic areas are involved, is, there is something special (emotional) about the in-
emotion requires non-limbic as well as limbic areas, formation processing performed by the amygdala.
and non-emotional functions are also represented in When the amygdala processes information, a central
the limbic system. The limbic system concept, though hedonic states results. This hedonic state accompanies
highly innovative and important in its time, has outlived dual activation of the amygdala and hippocampus but
its usefulness as an account of the neural basis of emo- not activation of the hippocampus alone. The second
tion and suffers from many of the same problems as an is an anhedoic theory of emotional memory. When the
account of memory or other cognitive processes. amygdala circuits are activated, a host of behavioral,
The best way to figure out circuits of emotion and autonomic, and endocrine responses are either elicited
memory is to study them empirically rather than to (or the systems are at least readied for action), and
impose a vague anatomical concept on the functions. recalled memories are charged with a degree of arousal
As described above, research in this vein suggests that that does not accompany activation of the hippocam-
the neural systems underlying memory and emotion pal system alone. Somehow this arousal is translated
are, at least in part, separately localizable. While these into positive and negative hedonic tone, depending on
networks interact in important ways, they are suffi- the nature of the situation. These two hypotheses will
ciently distinct that they should be considered separate be readily recognized as restatements of the centralist
systems. vs. peripheralist theories about the nature of subjective
If memory and emotion are separately represented, emotional experiences (see refs. 36, 74). Neither theory
how does the brain form memories about the emotional is acceptable or rejectable at this point. Other theories
significance of a stimulus? Is such information stored may also be possible. Failure to resolve this contro-
through the hippocampal system? What is the role of versy is an embarrassment to the field but is not a
the emotional system involving the amygdala? Before stumbling block that impedes empirical research on the
answering these questions, it is necessary to consider neural basis of emotional memory.
what the term "memory" refers to? Thus far, ! have emphasized the fact that both the
It is now widely recognized that there are many dif- amygdala and hippocampus are involved in the storage
 71

and retrieval of information about the emotional sig- US. The associative nature of the responses is demon-
nificance of experienced emotional events. In this view, strated by studies comparing the responses elicited after
emotional experiences lead to parallel storage in emo- CS-US pairings with the responses elicited after ran-
tional (amygdala-based) and non-emotional (hippo- dom presentations of the CS and US65 (Fig. 1). Acti-
campal-based) systems. When memories are activated vation of the stored association by the CS elicits re-
in this way, we have conscious knowledge of why emo- sponses similar to those elicited by actual threat8.
tion is aroused, However, it is also conceivable, if not
likely, that the amygdala system can store information
that is not processed by the hippocampal system. These A
1 [ZE
20
“implicit” (unconscious) emotional memories, when
activated, would lead to the same kind of aroused con-
dition as when explicit emotional memories are acti-
vated, but in the absence of explicit (conscious) knowl-
edge of why the arousal occurs. In such situations of
emotional ambiguity, conscious systems may seek ex-
planations for the arousal in terms of the social envi-
ronment or one’s own behavior4,37,74,75.
It is obviously important to understand how the brain
goes about the business of forming emotional memories
and to tease apart the mechanisms involved in the es-
tablishment of explicit as well as implicit memory traces
of emotional experiences. Although we are a long way
from this ambitious goal, considerable progress has
been made in understanding some basic facts about
how memories of the emotional significance of events
are stored in the brain through the amygdala and its 012345676 9 10

neural connections. In the following, the approach CS (set)

taken by my laboratory and the major findings that
have resulted will be described. Summaries of the work
of other investigators who also work on this general
topic can be found in recent reviews14315*32,33. B ‘“1

THE BEHAVIORAL MODEL

Our studies have involved the use of aversive clas-

sical conditioning techniques to examine the neural
mechanisms through which environmental events are
transformed into emotional signals in rats. We have
primarily focused on the use of auditory stimuli, the
meanings of which are modified by associative pairing
with footshock. In the language of conditioning, the
auditory stimulus is the conditioned stimulus (CS) and
the footshock is the unconditioned stimulus (US). After
a small number of pairings, the CS comes to elicit emo- 1

PAIRED RANDOM NAIVE

tional responses (increases in arterial pressure and heart
Fig. 1. Autonomic and behavioral responses elicited by an acoustic
rate; freezing behavior) when presented alone (Fig. 1). conditioned stimulus (CS). All responses were measured outside of
These are species-typical responses that are elicited the conditioning context (in the home cage) 24 h after conditioning
when rats are exposed to threatening stimuli, including or control procedure. During conditioning, the paired group experi-
stimuli with innately programmed significance as well enced the 10 s CS in a paired relation with a 0.5 s footshock US, the
random group experienced the CS and US in a random relation, and
as CSs that have been associated with aversive
naive group experienced neither stimulus. During testing 3 CS alone
uss 5317318,38.Thus, the rat does not learn the responses trials was presented to measure arterial pressure responses and one
but instead learns the association of the CS with the 120 s CS alone trial was presented to measure freezing.
72

Understanding how the brain processes conditioned to thalamo-cortical systems. Second, the projections
fear stimuli, therefore, allows to approach questions from auditory cortex to amygdala arise not in the
about: (1) how hardwired, innate emotional programs primary auditory cortex, but in secondary fields. The
are activated and controlled by environmental events, secondary fields receive direct thalamic proiections as
(2) how novel environmental events gain access to these well as projections from primary auditory cortex. Third,
innate systems through learning, and (3) how informa- the projections to the amygdala from both the auditory
tion about aversive experiences in the environment are association cortex and the auditory thalamus converge
stored and used adaptively in future encounters with in amygdala.
threatening events. Although the work involves rats, it The lateral nucleus of the amygdala is the sensory
involves an emotional system, the fear system, that is receptive area 42-44. It receives the convergent inputs
probably conserved across mammals better than any from auditory thalamus and auditory cortex (Fig. 2).
other. This no doubt explains the extremely successful Lesions of the lateral nucleus interfere with fear con-
use of rats in the development of drugs for the control ditioning 42. In contrast, the central nucleus is the re-
of fear and anxiety in humans. sponse expression area 14'15'32'33'38'47. Lesions of this
structure also interferes with conditioning. The lateral
nucleus is connected with the central nucleus directly
THE BASIC NEURAL CIRCUIT and by way of the basal (basolateral) and accessory
basal (basomedial) nuclei 35"64'81 (Fig. 2). The contribu-
The neural pathways involved in the association of tion of specific intraamygdala pathways to conditioning
an auditory CS with a footshock US, and thus in the has not been examined.

L~
2[SENSORY
elicitation of defensive responses by the CS in the The central amygdala projects to a variety of brain-
absence of the US, have been identified. These will be
described in terms of sensory, integrative, and motor
PRIMARY UNIMODAL t
systems. NEOCORTEX ASSOCIATION
The auditory CS is relayed through the auditory sys-
tem to the thalamus. Lesions of the auditory thalamus
completely eliminate the ability of the animal to asso-
ciate the auditory CS with the US 5]. From the thala-
mus, the signal is transmitted both to the auditory SENSORY I Ll~m I I Extra'l
THALAMUS ~
cortex and to the amygdala. Lesions of the auditory
cortex have no effect on conditioning 51'68, but lesions of
the amygdala prevent conditioning 14"15'28'32"33'37'38"5°. r
These results indicate that the thalamo-amygdala path-
way is a sufficient conditioning pathway, but do not
indicate whether it is also a necessary pathway. Indeed,
we have recently shown that neither thalamo-amygdala
EMOTIONAL
STIMULUS "L
@
or thalamo-cortico-amygdala pathways are necessary EMOTIONAL MEMORY
PATI-hNAYS BEHAVIORAL AUTONOMIC
and that each is sufficient as a conditioning pathway, 1: T h a l a m o A m y g d a l a EMOTIONAL
2: Cortico A m y g d a l a
at least for a simple, single auditory CS 67. RESPONSES
Anatomical tracing studies have characterized the
Fig. 2. Emotional memory pathways. Emotional memory, as under-
pathways through which the auditory thalamus and stood through studies of fear conditioning, can be established by
auditory cortex are connected with the amygdala. The thalamo-amygdala and thalamo-cortico-amygdala pathways. The
details are described in relevant publications 43- lateral nucleus of the amygdala (Lat Amyg) is involved in the con-
46,49,69,70. The most important points can be summa- ditioning of fear responses to modality-specific stimuli processed
thorough both thalamo-amygdala and cortico-amygdala projections.
rized as follows. First, the main thalamic inputs to the
Through pathway one (1), which links extra-lemniscal (Extra-Lem)
amygdala originate in the medial areas of the medial sensory areas of the thalamus with Lat Amyg, undiscriminated
geniculate body and associated regions of the posterior acoustic stimuli gain access to fear control systems. More complex
thalamus. These thalamic areas also projects to audi- sensory stimuli require transmission to the cortex over the lemniscal
tory cortex, but as part of the extralemniscal auditory (Lem) and Extra-Lem projections and then connections to Lat Amyg
through pathway two (2). Lat Amyg projects to the central nucleus
system and not part of the tonotopically organized lem-
of the amygdala (Ce) by way of the basal (B) and accessory basal
niscal system that makes possible veridical perceptions. (AB) nuclei. Ce then has connections with brainstem areas invovled
The extent of sensory representation is thalamo- in the regulation of species typical behaviors (central gray, CG) and
amygdala systems is thus relatively weak as compared autonomic adjustments (lateral hypothatamus, LH)~
 73

stem areas involved in the expression of behavioral, though neither of these questions can be answered with
autonomic, and endocrine responses and lesions of dif- the detail that is available for auditory aversive condi-
ferent projections selectively interfere with different re- tioning, partial answers are available. Aversive condi-
sponse modalities 47. Thus, lesions of the lateral hypo- tioning with a visual CS requires the integrity of the
thalamus interferes with the expression of autonomic amygdala and does not depend on visual cortex48. Sub-
but not behavioral responses, whereas lesions of the cortical visual pathways to the amygdala must therefore
central gray region interferes with behavioral but not be involved, although these have not been clearly iden-
autonomic responses. Lateral hypothalamic lesions in tified at this point. Unfortunately, little information is
primates have similar effects 79. These findings, summa- available for appetitive conditioning tasks that are ex-
rized in Fig. 2, provide an interesting contrast to studies actly comparable to fear conditioning. The only work
showing that both defensive behavior and associated directly relevant comes from Gallagher and Holland,
autonomic responses can be elicited by electrical or showing that the central nucleus of the amygdala is
chemical stimulation of the central gray region 3'24. involved in some aspects of aversive conditioning 22.
However, the stimulation studies show the full range of Some other appetitive tasks have implicated the
responses that can be elicited when the entire system is amygdala 9'21'26"3°, while others have not ~°. Thus, it
directly activated, whereas our studies show the path- seems that visual aversive conditioning is probably or-
ways actually used by the brain in producing defensive ganized similar to auditory aversive conditioning but
responses to an environmental stimulus. However, it is it is not clear whether appetitive conditioning uses
also possible that both behavioral and autonomic re- similar pathways as aversive conditioning.
sponses might be mediated by the central gray path-
ways in response to environmental stimuli in other
situations. AN EXTENDED EMOTIONAL SYSTEM
Together, the various findings described above pro-
vide an input-output description of the fear condition- Recently, we have begun to ask how the fear condi-
ing pathways from sensory to motor neurons. The path- tioning circuits involving sensory projections to the
ways involve transmission through the auditory system amygdala from the thalamus and cortex fit into larger
to the thalamus and then to the lateral amygdala, either cerebral networks. Thus far, we have concentrated on
directly or by way of the auditory cortex. After in- the question of whether the hippocampus might con-
traamygdala relays, the central nucleus transmits to tribute to fear conditioning.
brainstem areas involved in emotional response con- It has been known for some time that lesions of the
trol. These pathways are mainly relevant for the con- hippocampal formation have no effect on the acquisi-
ditioning of fear responses to simple, undiscriminated tion of fear reactions to a CS paired with an aversive
auditory stimuli like a pure tone. US 66. However, during conditioning, in addition to de-
Studies by Jarrel et al. 29 have shown that if a dis- veloping fear reactions to the CS, the animals also ac-
criminated conditioning task is used, where one CS is quire fear reactions to the chamber in which C S - U S
paired with the US and another is not, the auditory pairings take place 5'~7'18. The hippocampus has long
cortex is required for accurate responding to the paired been thought of as play some role in spatial/contextual/
CS. Interestingly, auditory cortex lesions interfere with configural/relational p r o c e s s i n g 16'58'59"73. W e 62 and
this task not by preventing responding but instead by others 34"78 therefore asked whether lesions of the hip-
preventing the differential responding to the paired pocampus might interfere with contextual conditioning.
stimulus; thus, the animals respond to both stimuli. In Hippocampal damage indeed impaired this so-called
this situation, the fear responses are elicited through contextual fear conditioning without affecting condi-
thalamo-amygdala systems, which appear unable to tioning to a C S 34'62. Contextual fear conditioning was
distinguish between the two stimuli. Studies of animals also interfered with by amygdala damage 62. The
with auditory cortex lesions will provide valuable in- amygdala is thus required for fear conditioning, regard-
formation into the limits of the thalamo-amygdala less of the kind of CS (phasic CS, such as a tone, or
system. continuously present CS, such as the conditioning box)
Two important questions are whether this basic pat- associated with the US. W e 62 proposed that projec-
tern of connectivity involving projections to the tions from the subiculum of the hippocampal formation
amygdala from both the sensory thalamus and sensory to the amygdala61 are the essential pathway to the
cortex holds for emotional conditioning in other (non- amygdala in contextual fear conditioning and that this
auditory) CS modalities and whether it holds for con- projection functions as a high-level CS processing path-
ditioning in other (non-aversive) US modalities? A1- way, much as the sensory projections to the amygdala
74

do for phasic stimuli. An important challenge is to un- latencies starting from 12 ms. These cells had a variety
derstand just what context is. Context differs from a of frequency receptive field tuning functions, with some
standard CS in important ways. First, contextual showing surprisingly narrow receptive fields for a struc-
stimuli are continuously present and are thus not ex- ture outside of the auditory system. Other cells rapidly
plicitly paired with the US in a precise, temporal rela- habituated to auditory stimuli. Most cells had relatively
tionship. Second, contextual stimuli may not be stimuli high thresholds. These properties suggest that AL neu-
at all but instead may be complex modailtity indepen- rons are hard wired to respond to auditory stimuli in
dent representations. The hippocampus could be in- three ways that are particularly relevant to emotional
volved because it is needed for attending to continu- coding: first, the short latencies are consistent with di-
ously present stimuli that have no obvious temporal rect transmission from the thalamus (thalamic cells re-
parameter, because it is needed for processing modal- spond to auditory stimuli in 7-10 ms and thalamo-
ity independent information, or for a number of other amygdala transmission takes 4-8 ms) and may allow
reasons that are implied by one of the various theories the amygdala to produce rapid emotional responses to
of hippocampal function 16'58'59'73. acoustic events; second, the high thresholds suggest
Our study of contextual conditioning may have re- that AL cells may serve as intensity filters, allowing the
vealed a general feature of cortico-amygdala function. amygdala produce defensive, protective reactions to
The amygdala receives inputs from many high level loud stimuli; third, the rapidly habituating cells may be
association areas in the frontal and temporal lobes. novelty detectors. Learning in this system then could
These may each be playing roles in fear conditioning, involve a lowering of threshold, while maintaining the
but in the absence of clear tasks that are sensitive to same receptive field; modification of the receptive field;
these roles, we are unable to notice the effects of lesions or reduction in habituating tendencies. Recent work
of these structures. As we learn more about the func- shows that most of the acoustically responsive cells are
tions of various cortical areas, hypotheses about the also responsive to somatosensory stimulation 71. These
contributions of these areas to fear conditioning, and physiological studies thus suggest clues into the nature
possibly to emotional processing in general, may of processing in the thalamo-amygdala system. Future
emerge. work will be aimed at determining how C S - U S inter-
actions in the amygdala might account for emotional
CELLULAR PROCESSING MECHANISMS learning and memory, as expressed behaviorally.

Once key aspects of the neural system involved in the Synaptic plasticity
formation of emotional memories through fear condi- Long-term potentiation (LTP) has been used exten-
tioning were identified, we began efforts to uncover sively to study synaptic plasticity in the hippocampus
some basic cellular mechanisms in the hope of gaining and other areas of the nervous system 8'52'83. LTP in-
insights into the nature of information processing in volves the delivery of high frequency electrical stimuli
these circuits. To date, this work has focused on the to a fiber system and recording of synaptic responses
thalamo-amygdala system and three major issues have in a terminal region. LTP occurs when the synaptic
been researched: sensory processing capacities of single response to a fixed stimulus is increased after high
neurons, synaptic organization and neurotransmission, frequency tetanization. Tetanization of the thalamo-
and synaptic plasticity. amygdala pathway produces an increase in a field
potential recorded in A L 12. This is indicative of an
Processing capacities of single neurons amplified synaptic response. LTP has also been elicited
The goal of this line of work has been to record from in AL by stimulation of cortico-amygdala pathways 1~.
single cells in the thalamo-amygdala pathway in order Many questions remain unanswered about thalamo-
to characterize the neuronal codes underlying the pro- amygdala LTP 4~. However, at a minimum it shows that
cessing of CS and US information by these cells. As a the circuit is plastic. An important step would be to
first step, we recorded unit activity in AL in response show that stimulation of CS and US inputs to AL can
to electrical stimulation of the auditory thalamus 13. produce LTP. This would provide a realistic model of
These studies demonstrated that cells in AL can be synaptic plasticity in a circuit with a welt-defined learn-
synaptically activated by thalamic stimulation. The ini- ing and memory function and could help to further our
tial response latencies were in the range of 4-8 ms. understanding of the relation of LTP to memory as well
Next, we recorded unit activity in AL in response to as our understanding of how emotional memories are
stimulation with auditory CS-like events 6. Many cells formed. Such studies will have to await the identifica-
in AL were acoustically responsive and had initial on set tion of the US pathway to AL.
 75

Synapses and transmitters microscopic studies demonstrated that many terminals

Considerable progress has been made in identifying in AL contain glutamate immunoreactivity19. The
the kinds of synapses formed and the neurotransmitter glutamate-labeled terminals, like thalamo-amygdala
used by thalamo-amygdala projections. Dual labeling terminals, tend to contact dendritic spines and at least
light microscopic studies have been performed to de- some of the thalamo-amygdala terminals contain
termine possible transmitter substances present in the glutamate immunoreactivity2°. Moreover, intraven-
cells of origin of the thalamo-amygdala projection. The tricular or iontophoretic injections of excitatory amino
dual labeling involved retrograde transport to selec- acid antagonists interferes with the synaptic responses
tively label the cells of origin of the thalamo-amygdala recorded in amygdala after thalamic stimulation63.
projection and immunocytochemistry to determine the These studies thus begin to characterize the morpho-
presence of the transmitter substance 45. Such studies logical and physiological bases of neurotransmission in
show that the excitatory amino acid glutamate is a likely the thalamo-amygdala projection. Such information
candidate transmitter. Electron microscopic studies provides initial clues to the local circuit organization of
have shown that the thalamo-amygdala projection the projection and suggests hypotheses for additional
mainly forms asymmetric synapses in A L 46. Since physiological and behavioral studies. Given the role
asymmetric synapses are usually indicative of excita- of glutamate in synaptic plasticity in the hippocam-
tory transmission, these findings are consistent with the pus 8'52"83 and the observation that blockade of excita-
presence of the excitatory substance glutamate in the tory amino acid transmission in the amygdala interferes
cells of origin. Most of the terminals contact dendritic with fear conditioning TM, we begin to see the possibil-
spines or the distal shafts of dendrites. Further electron ity of a cellular understanding of fear conditioning, a
possibility that was once believed to be mainly obtain-
able in simple, invertebrate preparations 31.
PRIM/~
NEOCORTEX SENS(

EMOTION AND MEMORY: CONCLUSIONS

The emotional memory system revealed by this re-

SENSORY
THALAMUS
search is depicted in Fig. 3. This Figure emphasizes the
extensive degree of parallel processing that takes place,
even in this highly oversimplified illustration of the rel-
evant pathways. For purposes of this discussion, con-
sider three kinds of inputs that reach the amygdala in
an emotional learning situation. First, the amygdala
EMOTI
STIMULUS receives multiple inputs about the CS: inputs from au-
ditory thalamus and various stages of auditory asso-
EMOTIONALMEMORY PATHWAYS
1: undiscrirninated sensory cues
t ciation cortex. These provide the amygdala with vari-
ous levels of CS representation. The representations
BEHAVIORAL AUTONOMIC
2: dlscrfminated sensory cues
3: contextual cues EMOTIONAL vary in temporal and spatial domains, with faster rep-
4: higher order inputs (thoughts?) RESPONSES
resentations being less complete. Second, the amygdala
receives inputs from the US pathway, the origin of
Fig. 3. Extended emotional memory pathways. Thalamo-amygdala
projections (Pathway 1) are involved in conditioning to simple (un- which is still unknown. These make conditioning pos-
discriminated) modality-specific inputs, whereas cortico-amygdala sible by modifying the neural processing in any or all of
projections (Pathway 2) are required for conditioningto stimuli that the various levels of CS representation. Third, the
must be disriminated on the basis of perceptual fearture. Lat Amyg amygdala receives inputs from the hippocampal forma-
also receives inputs from the subiculum, a major output of the
tion about the context in which conditioning is taking
hippocampal formation (Pathway 3). Hippocampo-amygdala pro-
jections have been implicated in contextual conditioning. Complex place. This contextual representation may also interact
association cortex (e.g. multimodal areas of the frontal lobes) also with the US in the amygdala to make contextual con-
project to Lat. Amyg. (Pathway 4). These connections may be in- ditioning possible. Alternatively, the context represen-
volved in the formation of emotional memories on the basis of com- tation may instead simply modulate the C S - U S rep-
plex cognitions (thoughts). AB, accessory basal (basomedial) nucleus
resentation at the level of the amygdala, allowing the
of the amygdala; B, basal (basolateral) nucleus of the amygdala; Ce,
central nucleus of the amygdala; CG, central gray; Extra-Lem, extent of fear expressed and conditioned to be adjusted
extralemniscal areas of sensory thalamus; Lem, lemniscal areas of to the context.
sensory thalamus; LH, lateral hypothalamic area. It must be recognized that this basic scheme is only
76

A B
MGm
PIN

+FpAL AL
CS '
Pathway
@ + ABL
" ABM
ACE
MGm
PIN
ax

G D
MGm
PIN sp
AL
@
I PAL

+ ABL
ABM
ACE
CS
Pathway
A +
~+

US . . . . . . . i~ l
9
Pathway
@
Fig. 4. Synaptic mechanisms in the lateral nucleus of the amygdala (AL). A: auditory conditioned stimulus (CS) pathway from the auditory
thalamus (medial areas of medial geniculate body, MGm/PIN) to AL. The pathway forms asymmetric (excitatory) synaptic contacts with :the
dendritic spines of projection neurons. Projection neurons send their axons to the basal (basolateral, ABL), accessory basal (basomedial, ABM),
and central (Ce) nuclei of the amygdala. B: thalamo-amygdala axon (ax) terminals (term) arising in MGm/PIN appear to use glutamate (Glu)
as a neurotransmitter. Glu interacts with N M D A (N) and AMPA (A) postsynaptic receptors. C: recurrent collateral ihnibition in AL. Excita-
tory inputs to AL terminate on projection neurons that give rise to excitatory projections to target areas but also give rise to local collateral
projections to GABA inhibitory interneurons. These then contact the projection neuron and inhibit its further activation. D: one way in which
the unconditioned stimulus (US) might mediate conditioning is by removing the influence of the inhibitory interneuron in the recurrent collat-
eral feedback loop. This would allow the CS to exert a stronger excitatory impact on target areas. Many other hypotheses are possible. The
origin of the US pathway is unknown.

part of the story. It is surely an important part, but hypothalamus and basal forebrain systems involved in
some of the factors not accounted in much detail are the motivational control and basal motivational states may
following. First, the system contains many reciprocal have important influences on emotional system func-
and back projections, so that most of the areas that tioning. For example, a hungry animal may be less fear-
project to the amygdala also receive projections from it ful than a satiated one. Finally, the amygdala also re-
as well. This allows for highly complex interactions not ceives inputs from a variety of diffuse projecting systems
only between the amygdala and each of its inputs but within the brain and from the periphery that allow the
also between the inputs by way of the amygdala. Thus, modulation of all of the processing described so far.
the amygdala is a site where information from auditory Such modulation can have powerful effects on long-
cortex can be integrated with information from hippo- term storage 55. All of these aspects, and many more.
campus and prefrontal cortex. Second, the amygdala is need to be accounted for.
but one of several possible sites of plasticity in the We have thus come a long way towards an under-
circuit. Plasticity occurs in most of the areas in this standing of the neural system underlying fear condi-
circuit during auditory fear conditioning and condition- tioning and have made some progress at the cellular
ing expressed at the level of behavior reflects the inte- level as well. It is important to keep in mind, though,
gration of these various plastic changes. Third, the that fear conditioning is a very elemental kind of emo-
amygdala is also reciprocally interconnected with the tional learning. In this respect, two important chal-
 77

lenges face us. The first is to convert the fear condi- potentiated startle paradigm. In G.H. Bower (Ed.), The Psychol-
ogy of Learning and Motivation, Vol. 21, Academic Press, San
tioning system into a more general understanding of
Diego, 1987, pp. 263-305.
emotional memory, one which accounts for more com- 16 Eichenbaum, H., The hippocampal system and delclarative
plex kinds of fears and their learning, as well as one that memory in animals, J. Cogn. Neurosci., 4 (1992) 217-23l.
extends from fear to other emotions. The second is to 17 Fanselow, M.S., Conditional and unconditional components of
integrate the emotional memory system into a more postshock freezing, Pavlov. J. Biol. Sci., 15 (1980) 177-182.
18 Fanselow, M.S. and Bolles, R.C., Naloxone and shock-elicited
general understanding of memory. Progress in these
freezing in the rat, J. Comp. Physiol. PsychoL, 93 (1979) 736-744.
areas will require the joint efforts of researchers now 19 Farb, C., Aoki, C., Milner, T., Kaneko, T. and LeDoux, J.,
working on topics that are often perceived as separated. Glutamate immunoreactive terminals in the lateral amygdaloid
But now that so much progress has been made in some nucleus: a possible substrate for emotional memory, Brain Res.,
of the individual areas, the way is paved for beginning 593 (1992) 145-158.
20 Farb, C.R. and LeDoux, J.E., 1992, unpublished.)
the difficult task of integration across areas.
21 Gaffan, D., Amygdala and the memory of reward. In J.P. Aggleton
(Ed.), The Amygdala: Neurobiolgical Aspects of Emotion, Memory,
and Mental Dysfunction, Wiley-Liss, New York, 1992, pp. 471-
483.
REFERENCES
22 Gallagher, M., Graham, P.W. and Holland, P.C., The amygdala
central nucleus and appetitive Pavlovian conditioning: lesions
1 Aggleton, J.P., The Amygdala: Neurobiological Aspects of Emotion, impair one class of conditioned behavior, J. Neurosci., 105 (1990)
Memory, and Mental Dysfunction, Wiley-Liss, 1992. 1906-1911.
2 Aggleton, J.P. and Mishkin, M., The amygdala: sensory gateway 23 Gloor, P., Amygdala. In J. Field, H.W. Magoun and V..E. Hall
to the emotions. In R. Plutchik and H. Kellerman (Eds.) Emo- (Eds.), Handbook of Physiology: Neurophysiology, Vol. II, Ameri-
tion: Theory, Research and Experience, Vol. 3, Academic Press, can Physiological Society, Washington, DC, 1960, pp. 1395-
Orlando, 1986, pp. 281-299. 1420.
3 Bandler, R., Carrive, P. and Zhang, S.P., Integration of somatic 24 Graeff, F.G., Brain defence systems and anxiety. In M. Roth,
and autonomic reactions within the midbrain pereaqueductal grey: G.D. Burrows, and R. Noyes (Eds.), Handbook of Anxiety, Vol.
viscerotopic, somatotopic and functional organization, Prog. 3, Elsevier, Amsterdam, 1990, pp. 307-354.
Neurobiol., 87 (1991) 269-305. 25 Gray, J,A., The Neuropsychology of Anxiety, Oxford University
4 Bem, D.J., Self-perception: an alternative interpretation to cog- Press, New York, 1982.
nitive dissonance phenomena, Psychol. Rev., 74 (1972) 183-200. 26 Hiroi, N.M. and White, E.L., The lateral nucleus of the amygdala
5 Blanchard, R.J. and Blanchard, D.C., Crouching as an index of mediates expression of the amphetamine conditioned place pref-
fear, J. Comp. Physiol. Psychol., 67 (1969) 370-375. erence, J. Neurosci., 11 (1991) 2107-2116.
6 Bordi, F. and LeDoux, J., Sensory tuning beyond the sensory 27 lwata, J., Chida, K. and LeDoux, J.E., Cardiovascular responses
system: an initial analysis of auditory properties of neurons in the elicited by stimulation of neurons in the central amygdaloid
lateral amygdaloid nucleus and overlying areas of the striatum, nucleus in awake but not anesthetized rats resemble conditioned
J. Neurosci., 12 (1992) 2493-2503. emotional responses, Brain Res., 418 (1987) 183-188.
7 Brodal, A., Neurological Anatomy, Oxford University Press, New 28 lwata, J,, LeDoux, J.E., Meeley, M.P., Arneric, S. and Reis, D.J.,
York, 1982. Intrinsic neurons in the amygdaloid field projected to by the me-
8 Brown, T.H., Chapman, P.F., Kairiss, E.W. and Keenan, C.L.,
dial geniculate body mediate emotional responses conditioned to
Long-term synaptic potentiation, Science, 242 (1988) 724-728.
acoustic stimuli, Brain Res., 383 (1986) 195-214.
9 Cador, M., Robbins, T.W. and Everitt, B.J., Involvement of the 29 Jarrell, T.W., Gentile, C.G., Romanski, L.M., McCabe, P.M. &
amydala in stimulus-reward associations: interaction with the Schneiderman, N., Involvementof cortical and thalamic auditory
ventral striatum, Neuroscience, 30 (1989) 77-86.
regions in retention of differential bradycardia conditioning to
10 Cahill, L. and McGuagh, J.L., Amygdaloid complex lesions dif-
acoustic conditioned stimulii in rabbits, Brain Res., 412 (1987)
ferentially affect retention of tasks using appetitive and aversive
285-294.
reinforcement, Behav. Neurosci., 104 (1990) 532-543. 30 Jones, B. and Mishkin, M., Limbic lesions and the problem of
11 Chapman, P.F., Kairiss, E.W., Keenan, C.L. and Brown, T.H.,
stimulus-reinforcement associations, Exp. Neurol., 36 (1972)
Long-term synaptic potentiation in the amygdala, Synapse, 6
362-377.
(1990) 271-278.
31 Kandel, E.R. and Spencer, W.A., Cellular neurophysiological
12 Clugnet, M.C. and LeDoux, J.E., Synaptic plasticity in fear con-
approaches to the study of learning, Physiol. Rev., 48 (1968) 65-
ditioning circuits: induction of LTP in the lateral nucleus of the
134.
amygdala by stimulation of the medial geniculate body, J. Neu-
32 Kapp, B.S., Whalen, P.J., Supple, W.F. and Pascoe, J.P.,
rosci., 10 (1990) 2818-2824.
Amygdaloid Contributions to Conditioned Arousal and Sensory
13 Clugnet, M.C., LeDoux, J.E. and Morrison, S.F., Unit responses
Information Processing. In J.P. Aggleton (Ed.), The Amygdala:
evoked in the amygdala and striatum by electrical stimulation of
Neurobiological Aspects of Emotion, Memory, and Mental Dysfunc-
the medial geniculate body, J. Neurosci., 10 (1990) 1055-1061.
tion, Wiley-Liss, New York, 1992, pp. 229-254.
14 Davis, M., The role of the amygdala in conditioned fear. In J.P.
33 Kapp, B.S., Wilson, A., Pascoe, J., Supple, W. and Whalen, P.J.,
Aggleton (Ed.), The Amygdala: NeurobiologicalAspects of Emotion,
A neuroanatomical systems analysis of conditioned bradycardia
Memory, and Mental Dysfunction, Wiley-Liss, New York, 1992,
in the rabbbit In M. Gabriel and J. Moore (Eds.), Learning and
pp. 255-306.
Comoputational Neuroscience: Foundations of Adaptive networks,
15 Davis, M., Hitchcock, J.M. and Rosen, J.B., Anxiety and the
MIT Press, Cambridge, MA, 1990, pp. 53-90.
amygdala: pharmacological and anatomical analysis of the fear-
78

34 Kim, J.J. and Fanselow, M.S., Modality-specific retrograde anl- 54 Maclean, P.D., The triune brain, emotion and scientific bias. ha
nesia of fear, Science, 256 (1992) 675-677. F.O. Schmitt (Ed.) The Neurosciences: Second Study Program.
35 Krettek, J.E. and Price, J.L., A description of the amygdaloid Rockefeller University Press, New York, 1970, pp. 336-349.
complex in the rat and cat with observations on intra-amygdaloid 55 McGaugh, J.L., Significance and remembrance: the role of neu-
axonal connections, J. Comp. Neurol., 178 (1978) 255-280. romodulatory systems, P~tvchol. Sci., 1 (19903 15--25.
36 LeDoux, J.E., Neurobiology of emotion. In J.E. LeDoux and W. 56 Mishkin, M., A memory system in the monkey, Phih)s. Trans~ R.
Hirst (Eds.), Mind and Brain, Cambridge University Press, New Soc. Lond. B, Bilo. Sci.. 298 (1982) 85-95.
York, 1986. 57 Murray, E.A., The functional effects of amygdala lesions in
37 LeDoux, J.E., Emotion. In F. Plum (Ed.) Handbook of Physiol- humans: a comparison with findings from monkeys. 111 J.P.
ogy. 1: The Nervous System. Vol. V. Higher Functions of the Brain, Aggleton (Ed.), The Amygdala: Neurobiological A~pects of Emotion,
American Physiological Society, Bethesda, 1987, pp. 419-460. Memory, and Mental Dysfunction, Wiley-kiss, New York, 1992,
38 LeDoux, J.E., Information flow from sensation to emotion: plas- pp. 453-483.
ticity in the neural computation of stimulus value. In M. Gabriel 58 Nadel, L., Multiple memory systems: what and why, Ji Cogn.
and J. Moore (Eds.) Learning and Computational Neuroscience, Neurosci., 4 (1992) 179-188.
Bradford Books/MLT Press, Cambridge, 1990, pp. 3-51. 59 O'Keefe, J. and Nadel, L., The Hippocampus as a Cognitive Map,
39 LeDoux, J.E., Emotion and the limbic system concept, Concepts Clarendon Press, Oxford, 1978.
Neurosci.. 2 (1991) 169-199. 60 Phillips, R.G. and LeDoux, J.E., Different brain systems medi-
40 LeDoux, J.E., Emotion and the amygdala. In J.P. Aggleton (Ed.) ate explicitly cued and contextual fear conditioning, Soc. Neuro-
The Amygdala: Neurobiological Aspects of Emotion, Memory, and sci. Abstr., 17 (1991) (Abstract).
Mental Dysfunction. Wiley-Liss, Inc., New York, 1992, pp. 339- 61 Phillips, R.G. and LeDoux, J.E., Overlapping and divergent pro-
351. jections of cal and the ventral subiculum to the amygdala, Soc.
41 LeDoux, J.E., Brain mechanisms of emotion and emotional Neurosci. Abstr.. 18 (1992) 518.
learning, Current Opinion in Neurobiology, 2 (1992) 191-198. 62 Phillips, R.G. and LeDoux, J.E., Differential contribution of
42 LeDoux, J.E., Cicchetti, P., Xagoraris, A. and Romanski, L.M., amygdala and hippocampus to cued and contextual fear condi-
The lateral amygdaloid nucleus: sensory interface of the amygdala tioning, Behav. Neurosci., 106 (1992) 274-285.
in fear conditioning, J. Neurosci., 10 (1990) 1062-1069. 63 Phillips, R.G., LeDoux, J.E. and Clugnet, M.C., 1992, unpub-
43 LeDoux, J.E., Farb, C. and Romanski, L., Overlapping projec- lished.
tions to the amygdala and striatum from auditory processing 64 Pitk~men, A. and Amaral, D.G., Demonstration of Projections
areas of the thalamus and cortex, Neurosci. Lett., in press. from the lateral nucleus to the basal nucleus of the mnygdala: a
44 LeDoux, J.E., Farb, C.F. and Ruggiero, D.A., Topographic PHA-L study in the monkey, Exp. Brain Res., 83 (1991) 465-470.
organization of neurons in the acoustic thalamus that project to 65 Rescorla, R.A., Pavlovian conditioning and its proper control
the amygdala, J. Neurosci., 10 (1990) 1043-1054. procedures, Psychol. Rev., 74 (1967) 71-80.
45 LeDoux, J.E. and Farb, C.R., Neurons of the acoustic thalamus 66 Rickert, E.J., Bennett, T.L., Lane, P.L. and French, J., Hippo-
that project to the amygdala contain glutamate, Neurosci. Lett., campectomy and the attenuation of blocking, Behav. Biol., 22
134 (1991) 139-144. (1978) 147-160.
46 LeDoux, J.E., Farb, C.R. and Milner, T.A., Ultrastructure and 67 Romanski, L.M. and LeDoux, J.E., Equipotentiality of thalamo-
synaptic associations of auditory thalamo-amygdala projections amygdala and thalamo-cortico-amygdala projections as audi-
in the rat, Exp. Brain Res., 85 (1991) 577-586. tory conditioned sitmulus pathways, J. Neurosci.. 12 (1992) 4501-
47 LeDoux, J.E., Iwata, J., Cicchetti, P. and Reis, D.J., Different 4509.
projections of the central amygdaloid nculeus mediate autonomic 68 Romanski, L.M. and LeDoux, J.E., Bilateral destruction of neo-
and behavioral correlates of conditioned fear, J. Neurosci., 8 (1988) cortical and perirhinal projection targets of the acoustic thalamus
2517-2529. does not disrupt suditory fear conditioning, Neurosci. Lett., 142
48 LeDoux, J.E., Romanski, L.M. and Xagoraris, A.E., Indelibility (1992) 228-232.
of subeortical emotional memories, J. Cogn. Neurosci., 1 (1989) 69 Romanski, L.M., LeDoux, J.E., Information cascade from pri-
238-243. mary auditory cortex to the amygdala: cortico-cortical and
49 LeDoux, J.E., Ruggiero, D.A. and Reis, D.J., Projections to the cortico-amygdaloid projections of temporal cortex in the rat,
subcortical forebrain from anatomically defined regions of the Cerebral Cortex, 3 in press.
medial geniculate body in the rat, J. Comp. Neurol., 242 (1985) 70 Romanski, L.M. and LeDoux, J.E., Organization of rodent au-
182-213. ditory cortex: anterograde transport of PHA-L from MGv to
50 LeDoux, J.E., Sakaguchi, A., lwata, J. and Reis, D.J., Interrup- temporal neocortex, Cerebral Cortex, 3 in press.
tion of projections from the medial geniculate body to an archi- 71 Romanski, L.M., Clugnet, M.C., Bordi, F. and LeDoux, J.E.,
neostriatal field disrupts the classical conditioning of emotional Somatosensory and auditory convergence in the lateral nucleus
responses to acoustic stimuli in the rat, Neuroscience, 17 (1986) of the amygdala, Behav. Neurosci., 107 (1993)444-450:
615-627. 72 Rosen, J.B., Hitchcock, J.M., Sananes, C.B., Miserendino,
51 LeDoux, J.E., Sakaguchi, A. and Reis, D.J., Subcortical efferent M.J.D. and Davis, M., A direct projection from the central
projections of the medial geniculate nucleus mediate emotional nucleus of the amygdala to the acoustic startle pathway: antero-
responses conditioned by acoustic stimuli, J. Neurosci., 4 (1984) grade and retrograde tracing studies, Behav. Neurosei., 105 (1991 )
683-698. 817-825.
52 Lynch, G., Synapses, Circuits, and the Beginnings of Memory, The 73 Rudy, J.W. and Sutherland, R.J., Configural and elemental as-
MIT Press, Cambridge, MA, 1986, pp. vii-122. sociations and the memory coherence problem, J. Cog. Neurosci..
53 Maclean, P.D., Some psychiatric implications of physiological 4(3) (19923 208-216.
studies on frontotemporal portion of limbic system (visceral 74 Schachter, S., Cognition and centralist-peripheralist controver-
brain), Electroencephalogr. Clin. Neurophysiol., 4 (1952) 407-418. sies in motivation and emotion. In M.S. Gazzaniga and C.B.
 79

Blakemore (Eds.) Handbook of Psychobiology, Academic Press, 80 Squire, L.R., Declarative and nondelclarative memory: multiple
New York, 1975, pp. 529-564. brain systems supporting learning and memory, J. Cogn. Neuro-
75 Schachter, S. and Singer, J.E., Cognitive, social, and physiologi- sci., 4(3) (1992) 232-243.
cal determinants of emotional state, Psychol. Rev., 69 (1962) 379- 81 Stefanacci, L., Farb, C.R., Pitkanen, A., Go, G., LeDoux, J.E.
399. and Amaral, D.G., Projections from the lateral nucleus to the
76 Schacter, D.L., Implicit memory: history and current status, J. basal nucleus of the amygdala: a light and electron microscopic
Exp. Psychol. Learn. Mere. Cog., 13 (1987) 501-518. PHA-L study in the rat, J. Comp. Neurol., 323 (1992) 2-17.
77 Scoville, W.B. and Milner, B., Loss of recent memory after 82 Swanson, L.W., The hippocampus and the concept of the limbic
bilateral hippocampal lesions, J. Neurol. Psychiatry, 20 (1957) system. In W. Seifert (Ed.), Neurobiology of the Hippocampus,
11-21. Academic Press, London, 1983, pp. 3-19.
78 Selden, N.R.W., Everitt, B.J., Jarrard, L.E., Robbins, T.W., 83 Teyler, T.J. and DiScenna, P., Long-term potentiation, Annu.
Complementary roles for the amygdala and hippocampus in aver- Rev. Neurosci.. 10 (1987) 131-161.
sive conditioning to explicit and contextual cues, Neuroscience, 42 84 Thompson, R.F., The neurobiology of learning and memory,
(1991) 335-350. Science, 233 (1986) 941-947.
79 Smith, O.A., Astley, C.A., DeVito, J.L., Stein, J.M. and Walsh, 85 Zola-Morgan, S., Squire, L.R., Alvarez-Royo, P. and Clower,
R.E., Functional analysis of hypothalamic control of the cardio- R.P., Independence of memory functions and emotional behav-
vascular responses accompanying emotional behavior, Fed. Proc., ior: separate contributions of the hippocampal formation and the
29 (1980) 2487-2494. amygdala, Hippocampus, (1990).