THE AMERICAN JOURNAL OF PSYCHIATRY

From Metapsychology to Molecular Biology:

Explorations Into the Nature of Anxiety

Eric R. Kandel, M.D.

other now makes it possible to confront one of the last

Through the use of animal models, specific forms frontiers of science. Within a theoretical framework
of mentation can now be explored on the cellular based largely on insights from experimental psycholo-
and molecular levels. Chronic anxiety and gy and psychiatry, biologists are beginning to study
anticipatory anxiety in humans are closely paralleled successfully elementary aspects of mentation and to
by two forms of learned fear in the sea snail Aplysia: address a number of central questions: What function-
sensitization and aversive classical conditioning. In al changes must take place in nerve cells for learning
Aplysia’s simple nervous system it is possible to and memory to occur? Are there unifying cellular and
delineate how the two forms are acquired and molecular principles that relate one form of learning to
maintained. Both rely on the mechanisms of another? That relate short-term to long-term memory?
presynaptic facilitation. An augmented form of Can experience lead to enduring structural changes in
presynaptic facilitation accounts for the associative the nervous system? Do these structural changes in-
component of conditioning. These findings suggest volve alteration of gene expression, and, if so, is
that a surprisingly simple set of cellular and psychotherapy successful only when it induces such
molecular mechanisms in various combinations may changes? These questions, although originating from
underlie a wide range of both adaptive and different behavioral and neurobiological perspectives,
maladaptive behavioral modifications. are increasingly converging on a common ground. In
(Am J Psychiatry 140:1277-1293, 1983) this essay I shall try to illustrate how the independent
contributions of psychiatry, psychology, and neurobi-
ology can be combined in animal models to yield
D espite important progress during the last decade, insights into mentation that extend from the behavior-
the cell-biological mechanisms of mentation have al to the molecular level and that promise to apply to
until very recently eluded analysis. However, growth experimental animals and humans alike.
in the conceptual and technical power of cognitive To emphasize the relevance of these new develop-
psychology on the one hand and neurobiology on the ments for psychiatry, I shall concentrate on two
learned abnormalities of behavior: anticipatory anxi-
ely and chronic anxiety. I hope to document how
Expanded version of the first John Flynn Memorial Lecture, cognitive psychology, which has shown that the brain
presented to the Department of Psychiatry, Yale University Medical
School, New Haven, Conn., April 1982. Received Nov. 4, 1982;
stores an internal representation of experiential events,
revised April 7, 1983; accepted May 13, 1983. From the Center for converges with neurobiology, which has shown that
Neurobiology and Behavior, College of Physicians and Surgeons, this representation can be understood in terms of
Columbia University; and the New York State Psychiatric Jnstitute, individual nerve cells, so as to yield a new perspective
New York, N.Y. Address reprint requests to Dr. Kandel, Center for
in the study of learned anxiety. In a larger sense, I shall
Neurobiology and Behavior, College of Physicians and Surgeons,
Columbia University, 722 West 168th St., New York, NY 10032. try to illustrate that mentation loses none of its power
The original work described in this paper was supported by or its beauty when the approach is moved from the
Career Scientist Award MH-185S8 from NIMH, grant MH-26212 domain of metapsychology into the range of molecular
from NIMH, and a grant from the McKnight Foundation.
biology. On the contrary, the combined developments
The author thanks Morton Rieser, Ethel Person, E. Terrell Wal-
in cognitive psychology and in neurobiology promise
ters, Tom Carew, Sally Muir, and James H. Schwartz for comments
on an earlier draft of this paper. to renew interest in aspects of mentation that until
Copyright : 1983 American Psychiatric Association. now have been out of experimental reach. Although

Am J Psychiatry 1 40: 1 0, October 1 983 1277

THE NATURE OF ANXIETY

behaviorist psychology has been content to explore insights into anxiety have come from clinical observa-
observable aspects of behavior, advances in cognitive tion. One key insight was the appreciation, first by
psychology indicate that investigations which fail to Freud and then by others, that anxiety is not unitary
consider internal representations of mental events are but is manifested in a variety of forms. Thus, in his
inadequate to account for behavior, not only in hu- later writings, Freud (1) distinguished actual (automat-
mans but-perhaps more surprisingly-also in simple ic) anxiety, an automatic, inborn response to external
experimental animals. This recognition of the impor- or internal danger, from signal anxiety, an acquired
tance of internal representations, a conclusion intrinsic fear response in anticipation of danger, either internal
to psychoanalytic thought, might have been scientifi- (unconscious) or external. (Actual anxiety is referred
cally disappointing as recently as 10 years ago, when to as fear by some investigators [4, 51. For a clear
internal mental processes were essentially inaccessible discussion of the evolution of Freud’s writings on
to experimental analysis. However, subsequent devel- anxiety see Strachey’s editorial introduction to Freud’s
opments in cell and molecular biology have made it essay in the Standard Edition [1] and Brenner [6]).
feasible to explore elementary aspects of internal men- Subsequent work (for reviews see 7-9) has shown that
tal processes. Thus, contrary to some expectations, acquired anxiety can further be subdivided into three
biological analysis is unlikely to diminish the interest forms on the basis of clinical characteristics and re-
in mentation or to make mentation trivial by reduc- sponse to psychopharmacological agents. These forms
tion. Rather, cell and molecular biology have merely are panic attacks, anticipatory anxiety, and chronic
expanded our vision, allowing us to perceive previous- anxiety.
ly unanticipated interrelationships between biological Panic attacks are brief, spontaneous episodes of
and psychological phenomena. terror without manifest or clearly identifiable precipi-
The boundary between behavior and biology is tating cause. The attacks are characterized by a sense
arbitrary and changing. It has been imposed not by the of impending disaster accompanied by a sympathetic
natural contours of the disciplines, but by lack of crisis: The heart races; breath is short and unsteady.
knowledge. As knowledge expands, the biological and This form of anxiety often responds to tricyclic antide-
behavioral disciplines will begin to merge at certain pressants and to MAOIs (9-11). Anticipatory (or
points, and it is at these points that the ground on signaled) anxiety also is typically brief. Unlike panic
which modern psychiatry is based will become particu- attacks, anticipatory anxiety is triggered by an identifi-
larly secure. able signal, real or imagined, that has come to be
associated with danger. This form of anxiety tends to
respond to benzodiazepines and 3-receptor blocking
THE CLINICAL SYNDROMES OF ANXIETY agents such as propranolol (7-9). Chronic anxiety is a
persistent feeling of tension that cannot be related to
Anxiety is a normal inborn response either to obvious external threats; it may or may not be reduced
threat-to one’s person, attitudes, or self-esteem--or by benzodiazepines (12).
to the absence of people or objects that assure and Panic attacks, occurring suddenly and without an
signify safety (1-3). Anxiety has subjective as well as apparent trigger, are not under obvious stimulus con-
objective manifestations. Subjective manifestations trol. In contrast, anticipatory and chronic anxiety are
range from a heightened sense of awareness to deep to some degree under stimulus control. This feature
fear of impending disaster. The objective manifesta- suggests that both forms are at least partly learned.
tions of anxiety consist of increased responsiveness, That is, each form involves learning a relationship (or
restlessness, and autonomic changes (for example, the absence of a relationship) between a neutral and a
changes in heart rate and blood pressure). Anxiety can threatening stimulus.
be adaptive: It prepares us for potential danger and The idea that anxiety is inborn and that a neutral
can contribute to the mastery of difficult circumstances stimulus can be associated with it through learning has
and thus to personal growth. On the other hand, come from two sources. First, work in comparative
anxiety can become dysfunctional, by either being and evolutionary biology beginning with Darwin (13)
inappropriately intense or being displaced by associa- and Romanes (14, 15) has shown that most animals,
tion with neutral events that neither are dangerous like humans, have a repertoire of inborn defensive
themselves nor indicate danger. Thus, anxiety is patho- behaviors. Aware of the contributions of Darwin and
logical when it becomes inappropriately severe and Romanes, William James (16) proposed in 1893 that
persistent or no longer serves only to signal danger. in animals and in humans these built-in defensive
The biological mechanisms that give rise to feelings behaviors are triggered by anxiety, an inborn tendency
of anxiety represent a central problem in the neurobi- to react with fear to dangerous situations. Experimen-
ology of normal affective behavior. Anxiety is also an tal support for the notion that anxiety can be learned
important component of neurotic and psychotic ill- came from Pavlov’s discovery at the turn of the century
nesses. Yet despite the importance of anxiety, little is that defensive reflexes can be modified by experience
known of its underlying cellular and molecular mecha- and can be elicited by a previously neutral stimulus.
nisms. Thus, in 1927 Pavlov noted the utility of such associa-
As in other areas of behavior, most of the initial tive learning for an animal’s survival:

1278 Am J Psychiatry I 40: 1 0, October 1 983

 ERIC R. KANDEL

It is pretty evident that under natural conditions the humans. (For earlier discussions of animal models, see
normal animal must respond not only to stimuli which 4, 23, 24, 27, 28.)
themselves bring immediate benefit or harm, but also to Since we think of anxiety as characteristically hu-
other physical or chemical agencies . which in them- . .
man, it is important to review the evidence that simple
selves only signal the approach of these stimuli; though it
animals can learn anxiety and that conditioned fear in
is not the sight or the sound of the beast of prey which is
these animals approximates certain forms of anxiety in
itself harmful to smaller animals, but its teeth and claws.
(17, p. 14) humans. I shall argue that classically conditioned fear
and long-term sensitization provide models of antici-
A similar proposal was made independently by patory and chronic anxiety, respectively (see figure 1).
Freud. Because painful stimuli are often associated In classical (Pavlovian) conditioning an animal
with neutral stimuli, symbolic or real, Freud postulat- learns to associate two stimuli, a conditioned (or cue)
ed that repeated pairing of a neutral and a noxious stimulus (CS) and an unconditioned (or reinforcing)
stimulus can cause the neutral stimulus to be perceived stimulus (US). The US, by definition, elicits from the
as dangerous and to elicit by itself the anxiety re- animal an effective reflex, or instinctive response, that
sponse. Placing this argument in a biological context, is called the unconditioned response because it is
Freud wrote in 1926: present before conditioning. In contrast, the CS need
not elicit a reflex response before conditioning takes
The individual will have made an important advance in place. After repeated pairing, the animal learns to
his capacity for self-preservation if he can foresee and
associate the CS with the US and as a consequence will
expect a traumatic situation of this kind which entails
helplessness, instead of simply waiting for it to happen. Let
show reliable conditioned responses to the CS that
us call a situation which contains the determinant for such often resemble the inborn unconditioned response to
expectation a danger situation. It is in this situation that the US. In Pavlov’s classic experiment, food (meat
the signal ofanxiety is given [italics added]. (1, p. 166) powder) served as a US, eliciting an inborn response,
reflex salivation. After several trials in which the food
Pavlov and Freud not only appreciated that anxiety was paired with a neutral stimulus, a tone, the tone
can be learned, but each also had the important insight reliably elicited salivation. For pairing to be effective,
that the ability to manifest anticipatory defensive Pavlov found that he had to present the tone and the
responses to danger signals is biologically adaptive. food in a precise sequence: The tone had to precede the
Anxiety as a signal prepares the individual for fight or food on each training trial. This, as we shall see later, is
flight if the danger is external. For internal danger, because what the animal actually learned during classi-
Freud suggested that defensive mental mechanisms cal conditioning is that tone predicts food. The animal
substitute for actual flight or withdrawal. I would only salivates after the tone to prepare for food. Thus, if the
make a cautionary comment here: Simply because pairing sequence is reversed (backward conditioning),
aspects of anxiety may be learned and thus acquired the animal does not respond to the tone by salivating.
does not exclude the possible contribution of a genetic Pavlov further found that by varying the nature of
predisposition to anxiety. In fact, what might be the unconditioned stimulus he could produce different
inherited is the predisposition to learn certain stimulus types of learning. Unconditioned stimuli that satisfied
relationships (8, 9, 18-22). the animal’s needs or that enhanced survival gave rise
to appetitive learning, leading to satisfaction and ulti-
mately satiation. Unconditioned stimuli that threaten-
ANXIETY CAN BE STUDIED IN ANIMAL MODELS ed survival, such as a painful shock, produced aversive
learning, leading to conditioned fear (4, 23, 29).
In people suffering from anticipatory anxiety, a cue Is there a relationship between aversive conditioning
stimulus is thought to predict the occurrence of an in animals and specific anxiety in humans? One of the
aversive stimulus (4, 17, 23, 24). By contrast, chronic first experiments to apply aversive conditioning to
or unsignaled anxiety is thought to occur when people humans illustrated how classical conditioning could
learn either that danger is associated with a wide range give rise to anticipatory anxiety. In 1920 Watson and
of ever-present environmental cues or that danger is Rayner (29) found that an infant they were studying
always present and not signaled by any cues (25, 26). cried readily. Loud and sudden noise proved a particu-
As a result, chronic anxiety is triggered in a less larly effective US for eliciting the unconditioned re-
discriminating way. sponse, crying. They then added a neutral CS, a white
With the recognition of distinct types of acquired rat that initially did not elicit crying. After several
anxiety, experimental interest turned to the develop- pairings of CS and US, the infant started to cry the
ment of animal models for studying each type. From instant the white rat was presented: The previously
work with experimental models it soon became clear neutral CS produced the conditioned response.
that animals can learn to manifest aspects of anticipa- Aversive conditioning can lead to one of two forms
tory and chronic anxiety. This evidence has strength- of anxiety, depending on whether the aversive stimulus
ened the initial clinical distinction between anticipa- is presented in a signaled or an unsignaled manner
tory and chronic anxiety and supports the belief that (figure 1). As emphasized by Mowrer (4) and Miller
aspects of these forms of anxiety are also learned in (24), the presence of a CS, as a cue that predicts the

Am J Psychiatry I 40: 1 0, October 1983 1279

THE NATURE OF ANXIETY

FIGURE 1. Two Forms of Learning That Give Rise to Two Forms of animals depends on mechanisms by which animals
Anxietya process information about the predictive interrelation-
ships among various environmental events. A useful
ANTICIPATORY (SIGNALED) ANXIETY CHRONIC (UNSIGNALED) ANXIETY perspective on these problems has been provided by
Neutral Signal Predicts Danger No Signal to Predict Danger cognitive psychologists, who have shown that learning
(Anxiety) Therefore involves considerable mental processing and elabora-
No Signal to Predict Safety tion of sensory information, with the result that hu-
Lack of Neutral Signal Predicts Safety (Chronic Anxiety)
mans and animals develop internal representations
(No Anxiety)
(“cognitive structures”) of environmental events that
allow flexible behavioral decisions (for reviews see 35-
AVERSIVE CONDITIONING SENSITIZATION 38).
Neutral Signal (CS) Predicts No Neutral Signal (CS) to Predict If learning in humans and other higher animals
Aversive Stimulus (US) Aversive Stimulus (US) involves the establishment of certain cognitive struc-
tures, why are aspects of such cognitive mechanisms
Lack of CS Predtcts No US
likely to be similar in humans and in simple animals
like the snail Aplysia? One good reason for believing
aComparison of anticipatory anxiety and its animal model, aversive condition-
that this would be the case lies in the consequences of
ing, to chronic anxiety and its animal model, long-term sensitization.
adaptation to evolutionary pressure. Animals that
differ greatly in habitat and heritage nonetheless face
occurrence of the aversive stimulus (US), allows the common problems of adaptation and survival, prob-
animal to learn to focus its anxiety on a particular lems for which learning and flexible decision making
event in time (see also 17, 23). By contrast, repeated are useful. When different species face a common
exposure to the aversive stimulus without a cue stimu- environmental constraint, they often manifest homolo-
lus produces chronic anxiety (long-term sensitization), gous patterns of adaptation because a successful solu-
as pointed out by Seligman (25) and by Pinsker, tion to an environmental challenge, first evolved in a
Hening, Carew, and me (30) (for a review see 31). common primitive ancestor, will continue to be inher-
Seligman (25) has used the following analogy to ited as long as it remains useful and the selective
illustrate the distinctions between the two forms of pressure is present. As the physicist Weisskopf (39) put
anxiety in terms of biological adaptation. Imagine a it: “Nature likes to use the same old tricks again.” In
world in which each aversive stimulus capable of addition, common environmental pressures often lead
causing pain, and therefore fear, is predicted accurate- to the independent evolution of functionally analogous
ly and invariably by a brief neutral stimulus so that the processes in distantly related species.
presence of this neutral stimulus comes to produce a What constraint might have shaped or maintained a
brief episode of anxiety. As long as the cue stimulus is common cognitive learning mechanism in a wide vane-
not present the animal can relax and do what it wants. ty of species? Testa (40) and Dickinson (41) have
A consequence of traumatic events being predictable is argued that to function effectively, animals need to
that the absence of traumatic events is also predictable. recognize certain key relationships between events in
When, however, aversive events are unpredictable, their environment. They must be able to recognize and
safety also is unpredictable: No reliable event exists to mate with their own species and to avoid even closely
indicate that the trauma will not occur. Lacking a related species; they must distinguish animals that are
safety signal, organisms remain in a state of chronic prey and learn to avoid those that are predators; they
anxiety. According to this view, people and animals must search out food that is nutritious and avoid food
seek safety signals (25). They look for predictors of that is poisonous. There are two ways in which an
danger because these also provide information about animal arrives at such knowledge: The correct infor-
safety (32-34). mation for every choice can be preprogrammed in the
Thus, both in people and in experimental animals, animal’s nervous system, or the ability to choose
what distinguishes signaled (anticipatory) from unsig- correctly among alternatives can be acquired through
naled (chronic) anxiety is that signaled anxiety is learning. Genetic and developmental programming
predictive with respect to its cause, whereas unsignaled may suffice for all of the behavior of simple parasites
anxiety is completely unpredictive. The two variants of and certain free-living forms such as the nematode
aversive training that we have already considered worm C. elegans, which exists in a limited and relative-
therefore model the essential difference between antici- ly invariant environment in the soil. But for more
patory and chronic anxiety. complex animals, extensive learning is probably re-
quired to cope efficiently with varied or novel situa-
tions. Complex animals need to maximize their ability
MECHANISMS UNDERLYING ANXIETY ARE LIKELY to order the world. An effective way to do this is to be
TO BE GENERAL THROUGHOUT PHYLOGENY able to learn about predictive relationships between
related events.
An implication of these observations is that the Given that humans and experimental animals
development of anticipatory and chronic anxiety in should be capable of learning predictive relationships,

1280 Am J Psychiatry I 40: 1 0, October 1983

 ERIC R. KANDEL

do such relationships have properties in common that ing, and perhaps all forms of associative learning, so
could constitute a universally selective evolutionary readily because the brain has evolved to enable animals
pressure? They do. First, predictors always precede the to distinguish events that reliably and predictively
signaled event and, second, they are highly correlated occur together from those which are unrelated.
with it; they provide optimal information about the
probability of its occurrence. As Dickinson (38, 42)
and Testa (40) have argued, these distinctive properties BEHAVIORISM, COGNITIVE PSYCHOLOGY, AND
of predictive relationships are of such importance that RENASCENCE OF PSYCHOANALYTIC PERSPECTIVE
they probably form the basis of widespread adapta-
tional and evolutionary pressures that have acted on These several arguments indicate that explanations
all animals and enhanced the survival of those species of learning based solely on temporal contiguity are
capable of taking them into account (for a related limited. The behaviorist position, which has empha-
discussion see also 43). Some psychologists therefore sized temporal contiguity (49, 55, 56), has also run
believe that common associative mechanisms of learn- into difficulty in addressing questions central to other
ing exist in all species capable of learning and that areas of behavior and learning (for critiques of the
these common mechanisms are designed to recognize behaviorist position see 26, 36, 57-59). Objective
and store information about predictive relationships in measurements of behavior through analysis of stimuli
the environment (36, 38, 44-46). As we have already and responses are clearly important for the study of
seen, this issue is not new but was first raised by behavior. They are the only indices of behavior that
William James in 1892 when, following Darwin, he can be manipulated experimentally and the only ones
argued with his usual prescience that mental processes that can be measured objectively. Indeed, the most
evolved to serve adaptive functions for animals in their useful definition of behavior-observable movement-
struggle with a complex environment: “Mental facts derives from traditional behaviorism. Nonetheless, de-
cannot be properly studied apart from the physical spite the great technical and conceptual debt that
environment of which they take cognizance. . . Our. psychology owes to behaviorism, there is a substantial
inner faculties are adapted in advance to the features of difference between the view of mental life held by
the world in which we dwell, adapted, I mean, so as to behaviorists such as Watson and Skinner and that
secure our safety and prosperity in its midst. . . Mind
. found useful by most current students of behavior
and world in short have evolved together, and in (59-62). The extreme behaviorist view (63, 64) is that
consequence are something of a mutual fit” (47, p. 4). observable behavior is synonymous with mental life.
What is the evidence that animals are particularly This view narrowly defines a larger reality, psychic life,
adept at learning predictive relationships? Actually, in terms of the scientific techniques available for
until quite recently, animal psychologists thought that studying it. By so doing, this approach denies the
classical conditioning depended only on temporal con- existence both of consciousness and of unconscious
tiguity: A conditioned stimulus had only to precede the mentation, feelings, and motivation merely because
reinforcing unconditioned stimulus by a certain critical they cannot be studied objectively. Broader perspec-
period to be effectively conditioned (48, 49). However, tives such as those used by cognitive psychology are
this simple idea appears to be inadequate. If animals necessary to account for the behavioral capabilities
learned to derive predictive information simply from both of people and of animals (26, 38, 65, 66).
the occurrence of two events in close temporal contigu- Cognitive psychologists have emphasized the richness
ity, they might acquire a variety of erroneous notions of the internal representations that intervene between
about signals in the environment and begin to act stimuli and response. Even the acquisition of simple
maladaptively. The world is full of chance, and events associative tasks by invertebrates involves the learning
sometimes occur together without being highly come- of surprisingly complex predictive relationships, which
lated or causally related. suggests that many animals may form “cognitive”
Analyses of learning by Prokasy (50), Rescorla (44), representations of relationships among events in their
Kamin (Si), Mackintosh (52), Wagner (53), and their environment (67).
colleagues have shown that classical conditioning de- In the past, ascribing a particular behavioral feature
velops best when in addition to contiguity there is also to an unobservable mental process essentially excluded
a contingency-a truly predictive relationship-be- the problem from direct biological study because the
tween the conditioned (or cue) stimulus and the uncon- complexity of the brain posed a barrier to any comple-
ditioned stimulus (36, 38, 42, 45, 54). Classical condi- mentary biological analysis. But the nervous systems of
tioning works not simply because the CS and US are invertebrates are quite accessible to a cellular analysis
temporally paired but also because there are time of behavior, including certain internal representations
intervals between successive pairings within which the of environmental experiences that can now be ex-
US does not occur. Thus, in addition to being paired in plored in detail; This encourages the belief that ele-
time, the signal and reinforcer need to be positively ments of cognitive mentation relevant to humans and
correlated; the signal must indicate an increased prob- related to psychoanalytic theory can be explored di-
ability that the reinforcer will occur. It therefore rectly and need no longer be merely inferred.
appears likely that animals learn classical condition- The psychoanalytic perspective has been devalued

Am J Psychiatry 1 40: 1 0, October 1983 1281

THE NATURE OF ANXIETY

recently and, in the United States, is in decline. Its I emphasize at the outset that I do not, even in my
propositions have relied heavily on intervening con- most optimistic moments, believe that the mechanisms
structs: on a mental apparatus for unconscious and for anxiety in simple animals are likely to be identical
conscious mental activity and on postulated libidinal to those in humans. However, I would argue that, at
and aggressive drives. Some of these ideas are vague; this early stage in our understanding of the biological
all are difficult to quantify. As a result, the exploration mechanisms of anxiety, the precision of the fit is not
of psychoanalytic theories has been hampered by a critically important. What is important initially is that
lack of opportunities for experimental verification. we learn something meaningful about the mechanisms
Nevertheless, psychoanalytic thought has been panic- by which any form of anxiety arises in any animal, no
ularly valuable for its recognition of the diversity and matter how simple the animal (or the anxiety). In view
complexity of human mental experience, for discern- of the biological adaptiveness of anxiety and its appar-
ing the importance both of genetic and learned (social) ent conservation across diverse species, it seems likely
factors in determining the mental representation of the that a rigorous analysis of the mechanisms of anxiety
world, and for its view of behavior as being based on in any animal will prove instructive for understanding
that representation. By emphasizing mental structure human anxiety. An analogy from recent developments
and internal representation, psychoanalysis served as a in molecular biology underscores this point. The regu-
source of modern cognitive psychology, a psychology lation of gene expression in eukaryotes (animal cells)
that has stressed the importance of the logic of mental has recently been shown to be more complex than in
operations and of internal representations. Just as I prokaryotes (bacteria) (for reviews see 68-72). Never-
believe that the vigor now evident in cognitive psychol- theless, almost everything that has been learned from
ogy will be strengthened by its contact with the cellular bacteria applies to animal cells, and without the foun-
neurobiology of behavior through work on simple dation provided by the earlier work with bacteria, our
systems such as invertebrates, I also think that the understanding of gene regulation in animals would still
emergence of an empirical neuropsychology of cogni- be fairly primitive.
tion based on cellular neurobiology can produce a As with humans, Aplysia manifests behavioral states
renascence of scientific psychoanalysis. This form of resembling anticipatory anxiety (or fear) in response to
psychoanalysis may be founded on theoretical hypoth- a classical aversive conditioning paradigm and chronic
eses that are more modest than those applied previous- anxiety in response to a long-term sensitization para-
ly but that are more testable because they will be closer digm (figures 2 and 3; references 73, 74). This is most
to experimental inquiry. clearly demonstrated by using the same aversive, un-
In the remainder of this paper I shall try to docu- conditioned stimulus (strong shock to the head) during
ment these ideas by describing studies directed at de- training and the same test pathway (escape locomotion
veloping an animal model for both anticipatory anxi- in response to a weak tail shock) to assess learning in
ety and chronic anxiety in Aplysia that makes it possi- both paradigms. The degree of learned anxiety is
ble to explore their cellular and molecular mechanisms. assayed by measuring the amount of escape locomo-
tion an animal displays following training. The only
difference between sensitization training and classical
APLYSIA SHOWS ASPECTS OF ANTICIPATORY AND conditioning is the presence in the latter of a cue, the
CHRONIC ANXIETY conditioned or cue stimulus, specifically paired with
the head shock (figure 3). A particularly effective cue is
Because of language, we know that in humans a neutral chemical stimulus, extract of shrimp. Aplysia
anxiety is to an important extent subjective. In assess- is a herbivore; it eats only seaweed. Although it
ing anxiety in animals, we must rely exclusively on normally ignores shrimp, its chemosensitivity can
inferences derived from objective manifestations. Al- readily detect the presence of shrimp. Thus, shrimp can
though the correspondence of the two types of anxiety be an effective signal.
to the two objectively defined laboratory forms of The two training procedures produce two forms of
aversive learning is imperfect, the analogy is useful conditioned anxiety that differ primarily in their speci-
because it allows aspects of acquired anxiety to be ficity to the neutral signal. To demonstrate this differ-
explored experimentally. Being able to study aspects of ence, Walters and associates (73) tested the condi-
anxiety in animals like rats and monkeys is not surpnis- tioned and sensitized groups twice (figure 4), first in
ing. But it is surprising-at least it surprised my the absence and then in the presence of the CS. The
colleagues Temrell Walters and Thomas Carew and animals trained with a paired stimulus showed no
me-to find that even simpler animals, such as the increase in escape locomotion when tested in the
marine snail Aplysia, manifest behavioral changes that absence of the CS; when the signal was present,
by inference resemble anxiety and that can be used as a however, this group exhibited significantly more es-
model of anxiety in higher forms. The advantage of cape locomotion than it had either to the same test
invertebrate animals over monkeys and rats is that stimulus in the absence of the signal stimulus or before
their nervous systems are much simpler, which offers a training. For animals trained with a cue signal, the
chance to explore the cellular and molecular mecha- neutral signal is required for anxiety to be expressed;
nisms that contribute to anxiety. they thus show a form of anticipatory anxiety. In the

1282 Am J Psychiatry 140:10, October 1983

 ERIC R. KANDEL

FIGURE 2. Experimental Arrangements for Studying Aversive Con- FIGURE 4. Comparison of Responses of Conditioned (Paired) and
ditioning and Long-Term Sensitization in Aplysiaa Sensitized Animals After Traininga

A B A ______
Paired B io - E Sensitized
Aversive Conditioning (Signaled Anxiety( Sensitization lUnsignaled Anxiety)

_JL Shrimp (CS) 90 seconds 8

__.L.Head shock (US) _.__n_Head shock (US)
30 seconds 30 seconds
6
shock

2
h
Siphoo

-S Test probe Test probe 1

aln both cases a strong noxious stimulus to the head serves as the uncondi-
rioned stimulus, and a weak test probe to the tail elicits escape locomotion
and other defensive reflexes.

-I
FIGURE 3. Experimental Protocols Used on Aplysia for Condition- Test Without CS Test With CS Test Without CS Test With CS

ing and Sensitizationa

aDifferences between test and pretest scores are shown for two administrations
PRETEST (Day 1) of the test
stimulus: one in the absence and the other (3 hours later) in the
...,I..,,Tail shock
presence of the CS. Zero on the scale represents the mean number of steps of
__1;;._J5est period,,,,,,,_ 15 seconds
minutes
escape locomotion taken in the pretest. Paired animals showed significantly
__D__ Head shock (US) more escape locomotion after training than before when tested in the
30 seconds presence of the CS but not when tested in the absence of the CS. Sensitized
animals took significantly more steps to escape than they did before training
..JL Shrimp (CS) in both the presence and absence of the CS. (Modified from a figure in
TRAINING (Days 2-3)
90 seconds reference 73.)
3 hours
Paired __F-i ‘r-i i-i__________CS
(Conditioning) US

cognitive psychology, assumed that actual danger pro-

US Alone ,
duces an internal state, which, as we have seen, he
called “actual anxiety.” A neutral signal, he argued,

(Sensitization)
. US that comes to be associated with danger may elicit any
of a variety of responses, which may differ dramatical-
ly from the response to the actual trauma that the
signal predicts. Indeed, the immediate reaction to the
TEST)Day 4)
-‘.-Test period-.-j danger signal is not an overt response but an internal
state of tension, an augmented preparedness for ac-
_J i-CS
I Test tion, which Freud called “signal anxiety.” By signaling
fear, anxiety motivates behavioral response systems
aThe test stimulus was a train of electrical pulses to the tail. The US was a designed to reduce danger. In Freud’s view, anxiety is a
shock to the head. The CS was 1.5 ml of crude shrimp extract. These
motivational state, a defensive arousal, similar to other
procedures were identical for paired and sensitized groups except that the
latter was not exposed to the CS during training. motivational states stemming from hunger, thirst, or
the need for sex (figure 5).
In humans and other mammals, Freud’s view ap-
absence of the cue, the animals show no apprehension. pears accurate. Both chronic and anticipatory anxiety
In contrast, the chronically sensitized animals, trained represent a motivational (defensive) state in prepara-
without a signal stimulus, show a generally heightened tion for expected danger, a preparation that is not
responsiveness that is unaffected by the presence or necessarily expressed in motor activity. As such, these
absence of the cue and thus show a form of chronic forms of anxiety have two characteristics. They act not
anxiety. just on a single response but are likely to engage a
Human anticipatory anxiety and anxiety in Aplysia repertory of responses, some components of which are
are also comparable in the pattern of effects produced enhanced while others are suppressed, and the effects
by the anxiety. Some behaviorists, including Pavlov are motivationally consistent: Defensive responses are
(17), assumed that anticipatory anxiety results from enhanced and appetitive responses suppressed (figure
stimulus substitution: The learned response to the 5).
previously neutral stimulus comes to produce the same To test the possibility that simple invertebrates can
overt response as does the painful stimulus. In con- also learn to associate a neutral stimulus with a
trast, Freud, in some ways the founder of modern motivational state analogous to anxiety during aver-

Am J Psychiatry 1 40: 1 0, October 1983 1283

THE NATURE OF ANXIETY

FIGURE 5. Comparison of Behaviorist’s and Freud’s Views of FIGURE 6. Conditioning of Anticipatory Anxiety in ApIysiab
Anticipatory Anxietya

I. BEHAVIORIST’SVIEWOF ANTICIPATORYANXIETY
CS--.._
r APPETITIVE RESPONii
[eding j

US Response

r DEFENSIVERESPONSES

2. FREUD’SVIEW OF ANTICIPATORYANXIETY IHead Withdrawal

I Siphon Withdrawal
I Inking
j Appetitive
Repertoire Locomotion
(Depressed)
aThe conditioning is consistent with a cognitive interpretation whereby the US
4 Defensive elicits a motivational state (defensive arousal), which then becomes associated
I Repertoire
with the CS after repeated pairing. This motivational state suppresses ap-
(Enhanced)
(“Automatic anxiety” petitive behavior and augments defensive responses.
or trauma) bReprinted by permission from “Associative Learning in Aplysia: Evidence for
Conditioned Fear in an invertebrate,” by E.T. Walters, Ti. Carew, ER.
aThe behaviorists view anxiety as a form of stimulus substitution. Initially, Kandel, Science, volume 211, pages 504-506, 1981 (74). © 1981 by the
American Association for the Advancement of Science.
defensive responses
are produced only in response to the US. After pairing of
the CS and US, the CS produces the same defensive responses elicited by the
US (a conditioned response, or CR). According to Freud’s view (and that of
modern cognitive psychology), the learning of anticipatory anxiety involves FIGURE 7. Neural Circuit Mediating Sensitization of the Gill-
the endowment of a neutral stimulus with the ability to trigger an internal Withdrawal Reflex in ApIysia
state-anxiety or defensive arousal-which then modulates in a motivation-
ally consistent way not just a single response but an entire family of responses
or a behavioral repertoire.

sive conditioning, Walters and associates (74) exam-

med the effect of the CS, after aversive conditioning
(pairing of shrimp extract and head shock), on three
defensive responses in addition to locomotion: two
graded reflex acts (head and siphon withdrawal) and
an all-or-none fixed act (inking). They also examined
the suppressive effects of the CS on an appetite behav-
br, feeding. They found that conditioning modulates
these responses in a manner that is motivationally
consistent with anxiety in mammals: Defensive me-
sponses are enhanced and appetitive responses are
inhibited (figure 6).
6 Motor

The interneurons and the motor cells are all unique, identified cells.
A SIMPLE FORM OF CHRONIC ANXiETY CAN NOW
BE UNDERSTOOD IN TERMS OF ITS CELL-
BIOLOGICAL MECHANISMS touched, the siphon and gill contract vigorously, with-
drawing into the mantle cavity. The gill-withdrawal
Knowing that Aplysia shows elementary forms of reflex to stimulation of the siphon is analogous to
both anticipatory and chronic anxiety, we are in a simple defensive responses in humans such as with-
position to explore the cellular mechanisms of each drawing a hand from a hot object.
and to examine the relationships between them. I shall Most of the nerve cells making up the neural circuit
begin by considering sensitization, the animal model of (or wiring diagram) of the gill-withdrawal reflex have
chronic anxiety, because this form of anxiety has been now been identified (figure 7). The siphon skin is
analyzed in detail at both the cellular and molecular innervated by 24 sensory neurons, and there are six
levels. I shall focus on a very simple defensive system motor neurons. There are also several intemneurons,
modulated by anxiety-siphon and gill withdrawal- one of which produces inhibition and five others which
because it is understood most fully (figure 7). produce excitation. The sensory neurons that carry
As is the case with other snails, Aplysia has a tactile input from the siphon skin connect to the
respiratory chamber called the mantle cavity that intemneurons and to the motor neurons; the motor
houses the gill. This cavity is covered by a protective neurons connect directly to the muscles of the gill that
sheet, the mantle shelf, which terminates in a fleshy effect the behavior. By examining this neural circuit
spout, the siphon. When the siphon or mantle shelf is during sensitization, Castellucci and associates (75,

1284 Am J Psychiatry I 40: 1 0, October 1983

 ERIC R. KANDEL

76) found that a stimulus which produces chronic FIGURE 8. Experimental Application of Serotonin and Cyclic AMP
to Simulate Presynaptic Facilitation in ApIysia
anxiety in Aplysia leads to an enhancement of the
connections made by the sensory neurons on their
A CONTROL SEROTONIN
target cells: the intemneurons and the motor neurons.
This enhancement, called presynaptic facilitation,
works as follows. A noxious stimulus to the head
activates a group of cells-the L29 cells-that are
thought to use a transmitter closely related to seroto-
Motor Neuron
nm as their chemical transmitter (77-79). This group
of facilitating cells acts as a defensive arousal system.
The cells impinge on the synaptic terminals of the
sensory neurons of the reflex system for gill withdraw- Sensory Neuron
al, and they amplify the strength of the connections
that these sensory synapses make onto the motor 2 mV
neurons and intemneurons (78-80). Semotonin simu- I 20
lates all the actions of this defensive arousal and 50 msec

produces its amplifying action by increasing the intra-

cellular messenger cyclic AMP in the sensory neurons B CONTROL CYCLICAMP
(81). The increase in intracellular cyclic AMP in turn
strengthens the connections of the sensory neurons by
facilitating transmitter release from their terminals
Motor Neuron
(figure 8; references 76, 82).

Senso Neuron
A MOLECULAR
ANXIETY
EXPLANATION FOR CHRONIC
I 5 mV
50 mV
50 msec
On the basis of pharmacological and biochemical
studies, we have been able to piece together a coherent am part A, Serotonin, a sensory neuron was stimulated 15 times, at the rate of

sequence of biochemical steps that take place in the once every 10 seconds, and produced a monosynaptic excitatory postsynaptic
potential in a gill orsiphon motor neuron, as seen in the set of electrical
sensory neurons when the behavior is altered by recordings on the left. Between the 15th and 16th action potentials, the
anxiety (83, 84). As an action potential propagates ganglion was bathed with iO M serotonin, which produced presynaptic
facilitation (right). In part B, Cyclic AMP, as in part A, a sensory neuron was
toward the synaptic terminals of the sensory neurons, stimulated once every 10 seconds for 15 stimuli (left). Between the 15th and
it begins to depolarize the terminals and open up the 16th action potentials, cyclic AMP was injected into the cell body of the
sodium channels, thereby producing more depolariza- sensory neuron and produced presynaptic facilitation (right). (Modified from
a figure in reference 82.)
tion and generating an action potential in the terminal.
The depolarizing component of the action potential in
the terminal then opens up calcium channels and model, serotonin released by the facilitating neurons
allows a certain amount of calcium to come into the acts on a serotonin receptor in the membrane of the
cell. The depolarizing component of the action poten- presynaptic terminals of the sensory neuron; the recep-
tial also opens up potassium channels; the resulting tom then activates a serotonin-sensitive adenylate cy-
influx of potassium repolarizes the action potential clase. The adenylate cyclase increases cyclic AMP
and turns the calcium channels off. Thus the activation within the terminals, which activates a protein ki-
of the sodium and potassium channels not only genem- nase-the enzyme thought to be the common site of
ates the action potential and determines its duration action for cyclic AMP in all eukaryotic cells. Protein
but also activates the calcium channels and determines kinases are enzymes that phosphorylate proteins; that
how long they remain open. Entry of calcium into the is, they add phosphoryl groups to certain amino acid
terminals is critical for transmitter release. Calcium is residues in the protein. The addition of a phosphoryl
thought to allow the vesicles that contain the transmit- group changes the charge of the protein, making it
ter to bind to discharge sites-a necessary step for more negative. This in turn changes the three-dimen-
transmitter release. Serotonin and cyclic AMP work to sional shape and therefore the functional state of the
prolong the action potential and thus enhance calcium protein. We have found that the activated protein
influx into the sensory neuron terminals. When the kinase phosphorylates a certain species of potassium-
action potential is prolonged, the calcium channels channel protein or a protein that is associated with the
stay open longer, and more calcium is available to potassium channel. Phosphorylation in effect closes
allow more transmitter-containing vesicles to bind to this species of potassium channel and thereby reduces
release sites. the potassium currents that normally repolarize the
Klein and I (83) have outlined a molecular model for action potential. Reduction of these currents prolongs
sensitization based on a series of biophysical and the action potential, allowing more calcium to flow
biochemical experiments (figure 9). According to this into the terminals. Consequently, more vesicles bind to

Am J Psychiatry 1 40: 1 0, October 1983 1285

THE NATURE OF ANXIETY

FIGURE 9. Molecular Model of Presynaptic Facilitation Underlying

of ways, and we have found that we can either trigger
Sensitization in ApIysia
or block the enhancement of transmitter release by
perturbing any one of the several steps in the biochemi-
PRESYNAPTIC TERMINAL
cal cascade. Thus, there now is compelling evidence
that serotonin increases the level of cyclic AMP in
individual sensory cells, that cyclic AMP activates a
protein kinase, and that kinase activation leads to
closing of a certain species of potassium channel
G Protein (figure 9). Indeed, recently Siegelbaum, Camardo, and
I (87) have been able to record the activity of a single
Adenylate Cyclase
potassium channel and the conformational changes in
a single protein molecule, and we have shown that
phosphorylation of this species of potassium channel
either directly or indirectly (by means of a regulatory
Phosphodiesterase protein that affects the channel) decreases the proba-
bility that the channel will open.
Thus, we have been able to take the analysis of a
form of anxiety from the behavior of the intact animal
Protein Kinase to the neural circuit of the behavior and to some of the
critical cells involved. Within these critical cells (the
K Channel I
sensory neurons of the reflex), we localized the change
to a particular component of the neuron, the presynap-
tic terminals, and demonstrated that the expression of
anxiety involves enhancement of transmitter release.
We found that the molecular mechanism of this en-
hancement is protein phosphorylation, which leads to
a broadening of the action potential and a greater
Ca
influx of calcium. We are now able to focus on the
individual protein molecules modulated by learning
and explore them in a behavioral as well as a biochem-
ical context.

THE MAINTENANCE OF CHRONIC ANXIETY

INVOLVES STRUCTURAL CHANGES

( Postsynaptic Cell Sensitization

defensive arousal
is a form of chronic
system is activated
anxiety
and increases
whereby
the
a

aSerotonin or a related amine released from the facilitating interneurons release of transmitter from particular identified synap-
reaches a serotonin receptor in the presynaptic terminal of the sensory
ses. We can therefore ask: Does the maintenance of this
neuron, where it activates an adenylate cyclase. The receptor does not
activate the cyclase directly but, as indicated in the diagram, through another learned anxiety involve a morphological change? To
membrane protein called the G protein. Once activated, the adenylate cyclase answer this question, Bailey and Chen (88) have
causes an increase in cyclic AMP, which in turn activates a protein kinase (an
enzyme that is composed of separate regulatory and catalytic subunits). This
visualized the synaptic terminals of the sensory neuron
kinase phosphorylates a protein associated with the potassium channel so electron-microscopically using the electron-dense
that the channel closes. With the potassium channel closed, the inflow of marker horseradish peroxidase. Their evidence sug-
potassium ions that would normally repolarize the action potential is
reduced. Consequently, the action potential is prolonged and calcium remains gests that, as in other neurons, synaptic vesicles, the
free to continue entering the cell, where it binds vesicles of transmitter (large likely storage sites for transmitter, are released at
circles at bottom) to their release sites. Thus, in presynaptic facilitation the
varicose expansions of the presynaptic terminal of the
amount of time available for transmitter release from the sensory neuron
terminal is extended. Eventually, the enzyme phosphatase dephosphorylates axon. The varicosities contain specialized regions
the potassium channel, which causes it to reopen, thereby terminating the called active zones where the vesicles are loaded into
action of the cascade activated by serotonin. (Modified from a figure in
reference 83.)
release sites from which they subsequently discharge
their contents. Comparing sensory neurons from
chronically sensitized and control animals, these me-
searchers have analyzed the changes in the number and
release sites, more transmitter is released, the function- distribution of the synaptic vesicles and in the size and
al output of the cell increases, and the animal shows extent of the active zones. They found that in normal
the enhanced responsiveness that characterizes chronic animals not all varicosities contain active zones. Rath-
anxiety in Aplysia. er, the incidence of active zones and the average size of
My colleagues Schwartz, Castellucci, Hawkins, and each active zone can be modified by anxiety. In sensory
Klein and I (85, 86) have tested this model in a variety neurons from naive animals, only 41% of varicosities

1286 Am J Psychiatry 1 40: 1 0, October 1983

 ERIC R. KANDEL

FIGURE 10. Morphological Correlates of Long-Term Sensitization CHRONIC ANXIETY MIGHT INVOLVE
in ApIysia ALTERATIONS IN GENE EXPRESSION

A B C
How is this structural change achieved? We do not
Control
yet know the answer to this question. But recent
x = Sensitized
“vs progress in the molecular genetics of animal cells
70
Q I suggests a possible mechanism. Each somatic cell in the
60
body contains all the genes present in every other cell.
50
What makes a liver cell a liver cell and a brain cell a
40 n}
?:2 4- brain cell is that during development from a single
20F fertilized egg cell, the various cells of the body differen-
Em _J.1
1O- tiate by shutting off the activity of certain genes while
,-
01 ‘-0
allowing others to be expressed. This developmentally
ALong-term sensitization produces an increase in both the number (A) and size determined repression and activation of organ- and
(B) of sensory neuron active zones. These complementary changes are even tissue-specific genes occurs during certain critical pen-
more apparent when viewed together, as illustrated in part C. The value for
the average number of varicosities per sensory neuron has been taken from
ods in development and is then self-maintained
total reconstructions of simple horseradish peroxidase-injected sensory neu- throughout the life of the differentiated cell (for review
rons in untrained animals (N=2, mean±SE). (Based on data reported in see 71, 72, 89-91). As a result, in any given cell most
reference 88.)
genes are closed; only some are open and available for
transcription.
In addition to these relatively permanent changes in
gene expression produced by differentiation, the genes
FIGURE 11. Ionic and Morphological Mechanisms of Long-Term
Sensitization in Aplysiaa that are available for expression within a given cell
type can be regulated. For example, a gene’s rate of
CONTROL LONG-TERM SENSITIZATION activity (its rate of transcribing messenger RNA) can
be transiently enhanced or depressed (92, 93) by a
variety of molecules, such as hormones that act direct-
ly on the genes or on proteins that regulate the genes.
In contrast to differentiation, these forms of gene
modulation can be either rapid and readily reversible
or self-maintained and enduring. Because learning
produces enduring changes in the structure and func-
tion of synapses, Schwartz and I (84) have proposed
that learning is likely to involve enduring, self-main-
tamed alterations in gene expression. This idea is
consistent with the suggestion that new protein synthe-
sis is required for long-term memory (94, 95). If this
speculation proves correct, it would provide a new
perspective on the nature of normal learning and
The ionic mechanism leads to the closing of a species of potassium channel,
which results in an increase in calcium influx due to a broadening of the thereby on the nature of certain learned neurotic
action potential. This contrasts with the situation in controls, in which illnesses such as chronic anxiety. Specifically, the pos-
potassium flowing in through open channels leads more quickly to a closure
of the calcium channels (filled squares). In addition, a morphological sta-
sibility of gene regulation by experience suggests a
bilization of the varicosities occurs during long-term sensitization whereby class of molecular regulatory defects that might be
there are increases in both the number of varicosities containing active zones caused by learning.
(filled triangles) and the number of active zones per active varicosity.
Consequently, more synaptic vesicles can be released. To put this view into perspective, let me illustrate
one way of looking at the relationship between psy-
chotic and neurotic illness. There is substantial evi-
dence that the major psychotic illnesses, such as
have active zones; the rest do not. In contrast, in schizophrenia and depression, are heritable. The ill-
animals that have been sensitized, the incidence of nesses presumably represent mutations-alterations in
active zones is increased to 65%. In addition, the the nucleotide sequence of the DNA-leading to ab-
average size of each zone is larger in sensitized than in normal messenger RNA and abnormal protein. The
naive animals (figure 10). hereditary information of a cell is carried in its nucleic
Thus, simple forms of anxiety produce profound acid, DNA. The strands of DNA contain one of four
morphological as well as functional changes (figure characteristic bases: adenine, guanine, thymine, and
11). The normal set of varicosities serves as a mere cytosine. The information carried by a gene is defined
scaffolding for behavior. Learning experiences, such as by the sequence of bases along the strand. Consecutive
the acquisition of chronic anxiety, can build upon this triplets of bases serve as code words called codons;
scaffolding by altering the functional expression of with some exceptions, each codon specifies an amino
neural connections. acid, and a string of 100 or more codons provides the

Am J Psychiatry 1 40: 1 0, October 1983 1287

THE NATURE OF ANXIETY

FIGURE 12. Comparison of Mutation of the DNA Sequence by Disease, Leading to the Expression of an Altered Gene, and Modulation of
Gene Expression by Environmental Stimuli, Leading to the Transcription of a Previously Inactive Genes

A ALTERATIONIN GENEEXPRESSIONBY INHERITEDDISEASE B ALTERATION IN GENE EXPRESSION BY ACQUIRED DISEASE

1 NORMALGENE 1 GENE IS NOT EXPRESSED
Promoter Structural Gene Promoter Structural Gene

I f]AICICIGITICIA] I t#{232};;: ITICIAITICIGIAJ

I I Transcription
Regulatory Protein #{174}
Regulatoty Protein kh’-P mRNA,

Translation
I Protein1

2 MUTATION 2 GENE IS EXPRESSED

Promoter Structural Gene Promoter Structural Gene

igenetic FactJ
I JAICIIJGtT1CIA1 I cTP ITICIAITICIGIAJ

_:y..-P
Transcription
mRNA2
Anxiety -
5-HTCycIicAMPf I Transcription

mRNA

r Translation I Translation
I ] Protein2 I I Protein

For simplicity, a specific example is illustrated. The gene is illustrated as having two segments, a structural gene (that is transcribed by a messenger RNA and in
turn is then translated into a specific protein) and a regulatory or promoter segment. The promoter is located upstream from the structural gene and regulates the
initiation of the transcription of the structural gene. In this example, the promoter segment can be activated (and made accessible to transcription) only when a
regulatory protein binds to the promoter. To bind, the regulatory protein must first be phosphorylated. Thus, in part A, the phosphorylated regulatory protein
binds to the promoter, thereby activating the transcription of the structural gene leading to the production of a gene product: protein 1 . in part A, 2, a mutant
form of the structural gene is illustrated in which a single base change has occurred; a thymidine (T) has been substituted for cytosine (C). As a result, an altered
messenger RNA is transcribed and an abnormal protein (protein 2) is produced, giving rise to the disease state. This alteration in gene structure is present in the
germ line and is inherited. Part B illustrates a specific example of alteration in expression of a normal structural gene that is not heritable. The regulatory protein is
indicated in its dephosphorylated state; it therefore cannot bind to the promoter site and gene translation cannot be initiated. An (epigenetic) learning experience,
such as learned anxiety, acting through serotonin and cyclic AMP, activates a protein kinase enzyme. This enzyme has both a regulatory unit component (R) and a
catalytic unit component (C). The increase in cyclic AMP removes the regulatory unit, thereby activating the catalytic unit. The catalytic unit phosphorylates the
regulatory protein, which can now bind to the promoter and consequently initiate gene transcription.

genetic code for the assembly of a protein chain. The biogenic amine that serves as a transmitter in the
sequence of amino acids in a protein chain determines presynaptic neuron or by affecting the expression or
how the chain folds and therefore how it assumes the onset of receptors on the postsynaptic cell (for a review
three-dimensional structure necessary for its biological see 96-98). I would now suggest that whereas the
activity. Alteration by mutation in only one nucleotide major psychotic illnesses (that do not respond to
subunit-one base-of one codon will be sufficient to psychotherapy because the disease is not fundamental-
alter the amino acid sequence of the protein and ly acquired or altered by learning) may involve alter-
thereby alter the protein properties, possibly even ation in the structure ofspecific genes, certain neurotic
making it inactive. illnesses such as chronic anxiety (that are acquired by
The information of DNA is not translated directly learning and that can respond to psychotherapy) might
into a protein. Rather, the sequence of bases that codes involve alterations in the regulation of gene expres-
for a protein is transcribed into a complementary sion_ According to this speculative view, schizophrenia
strand of RNA called messenger RNA because it and depression would be due primarily to heritable
carries the information for the sequence of amino acids genetic changes in synaptic function in a substrain or
necessary to construct the protein. The messenger mutant population; neurotic illnesses would not be.
RNA in turn is translated into protein. Thus, altered Rather, neurotic illnesses might represent alterations in
genes give rise to altered messenger RNAs, which synaptic function produced by environmentally in-
produce altered proteins. duced modulation of gene expression (figure 12). Even
How the genetic abnormalities of schizophrenia and though the learning mechanisms are inherited, neurot-
depression are manifest in the brain is still not known, ic individuals would be neurotic only if experience
but it is thought that they lead to changes in synaptic taught their genes to be pathologically expressed. A
function either by altering the release process of the corollary to this argument is that insofar as psycho-

1288 Am J Psychiatry 1 40: 1 0, October 1983

 ERIC R. KANDEL

FIGURE 13. A Model for the Biochemical Basis of Long-Term Memoryab

A
SHORT-TERMMEMORY Cyclic AMP

.A%:#{149}:..
1
_________
<__
__
_______
IiII3 Regulatory
Catalytic
Subunit
2
: iicz:i C C ) . . .#{149}#{149}#{149}

K4 Channel L AOAJ) Membrane AQA JJ

B
LONG-TERM MEMORY Cyclic AMP
Functional Morphological

AOAOJ

In short-term sensitization (part A, 1), the cyclic AMP-dependent protein kinase is proposed to have a normal regulatory subunit (RN) and no particular
orientation with respect to a substrate membrane protein associated with the K channel. In naive terminals, relatively high concentrations of cyclic AMP are
needed to activate the catalytic subunit (C) (part A, 2) to phosphorylate the membrane protein (part A, 3). This phosphorylation brings about enhanced release of
transmitter, the neurophysiological event underlying sensitization. The memory is brief because the concentration of cyclic AMP diminishes soon after stimulation
with serotonin. In trained neurons (part B, 1), a new class of regulatory subunit (RL) has been induced. As a result, the protein kinase differs from the naive enzyme
in being site specific and thus being advantageously oriented both to the channel and to the mechanism that governs the organization of dense projections at the
active zone (filled triangles), where synaptic vesicles line up to release transmitter. In addition, this new kinase has higher affinity for cyclic AMP. Consequently,
lower concentrations of cyclic AMP are required to phosphorylate these target proteins (part B, 2). In part B, 3, functionally, as in part A, 3, the K channel is in-
hibited as long as the channel protein remains phosphorylated. Morphologically, protein phosphorylation leads to the stable enlargement of the synapse. In this
form of sensitization, the memory persists for longer periods of time because it is embodied in R1, a protein molecule.
bReprinted by permission from “Molecular Biology of an Elementary Form of Learning: Modulation of Transmitter Release by Cyclic AMP,” by ER. Kandel, J.H.
Schwartz, Science, volume 218, pages 433-443, 1982 (84). © 1982 by the American Association for the Advancement of Science.

therapy works and produces long-term learned sisting of two classes of subunits, a catalytic subunit
changes in behavior, it may do so by producing and a regulatory subunit. The free catalytic subunit
alterations in gene expression. Needless to say, psy- carries out the phosphorylation. The regulatory sub-
chotic illness, in addition to partaking obligatorily of unit binds to the catalytic subunit and prevents it from
alterations in gene structure, may also involve a sec- acting. The function of cyclic AMP is to cause the
ondary disturbance in gene expression. regulatory unit to dissociate from the catalytic unit and
free it for action.
Serotonin, acting repeatedly on the terminals of the
A MOLECULAR GENETIC MODEL FOR THE sensory neuron (as a result of repeated aversive stimu-
MAINTENANCE OF ANXIETY lation), might activate a gene able to produce a novel
class of regulatory subunit for the protein kinase. The
How might one envision the alteration of gene specific inducer that activates the gene for the regula-
expression in learning? The model in figure 9 accounts tory subunit might be cyclic AMP. The prolonged
only for the immediate (short-term) effects of sensitiza- elevation of cyclic AMP that occurs in short-term
tion, that is, for the acquisition of anxiety. It is sensitization may allow cyclic AMP to enter the nude-
attractive to think, however, that this model might be us of the cell and there to cause the gene for a new
more general and might account for the long-term regulatory subunit to be expressed by cyclic AMP-
maintenance of anxiety, including the structural dependent phosphorylation (of perhaps one or more
changes. Indeed, Schwartz and I (84) recently extended proteins). Activation of a gene for a new class of
this model to account for the long-term maintenance subunit could be permanent or it could slowly decay if
of anxiety by positing a specific kind of alteration in not reinforced by subsequent aversive training.
gene expression. According to this theory, a gene is Schwartz and I (84) posited that this regulatory sub-
induced to produce a new protein kinase that ensures unit would have two novel features: It would have
prolonged phosphorylation of the potassium channel greater sensitivity to cyclic AMP, thereby allowing the
(figure 13). catalytic subunit to dissociate more readily; and it
The cyclic AMP-dependent kinase is a protein con- would be site specific, allowing the kinase to be bound

Am J Psychiatry 1 40: 1 0, October 1983 1289

THE NATURE OF ANXIETY

to the presynaptic membrane near the potassium chan- FIGURE 14. Relationship of Classical Conditioning to Sensitization
nels that are to be modulated (figure 13, part B). The in ApIysia
synthesis of a regulatory subunit with greater affinity
for cyclic AMP would allow the cyclic AMP-dependent
protein kinase to work at relatively normal concentra-
tions of the cyclic nucleotide. Slight elevations above
the normal concentrations of cyclic AMP of the sort
ASSOCIATIVE COMPONENT
that accompany the arousing stimuli of everyday life
are inadequate to evoke sensitization in the untrained
terminal. However, with the new regulatory subunit
facilitating the work of the protein kinase, such slight
elevations in cyclic AMP would now be sufficient to
provide, by modification of ion channels, the enhanced
influx of calcium required to increase transmitter
release and thus sustain the learned anxiety reaction
(figure 13, part B, 2 and 3). MODULATORY
COMPONENT
A subunit that would position the protein kinase
optimally could allow it to trigger a family of parallel
cyclic AMP-dependent
For example,
in the potassium-channel
it could
changes
1) produce
protein
in the sensory
the functional
and 2) alter
neuron.
change
the
11 Gill

assembly of the protein components that constitute the Activity-Independent Presynaptic Facilitation

active zone (figure 13, part B) and thereby initiate the

striking change in morphology of sensory terminals According to the model this form of learning involves two components: an
associative component that accounts for the timing or temporal specificity of
observed by Bailey and Chen (88). These two molecu- the modulation and a modulatory component that enhances or reduces a
lam changes, both caused by the same cyclic AMP- number of behavioral responses. It is attractive to think that presynaptic
facilitation, the mechanism underlying sensitization, could be used in each
dependent kinase, could operate together to bring
component, although in different ways, conventional and amplified, to
about enhanced transmitter release from the long-term modulate the strength of a particular connection. The key point of the model
sensitized neuron. is that the two components represent distinct neuronal processes: 1) Activity-
dependent presynaptic facilitation could yield the temporal discrimination
Although obviously premature because of lack of necessary to achieve associative specificity, and 2) conventional presynaptic
experimental support, this speculative explanation for facilitation could provide the modulatory component responsible for enhanc-
the maintenance of anxiety can be tested. The large ing the strength of the response.

nerve cells of Aplysia offer special experimental advan-

tages for molecular genetic studies of the nervous
system, since nuclei of individual cells can be isolated ficity is an augmented form of presynaptic facilitation
by hand dissection (99, 100). In addition, recombinant called activity-dependent enhancement of presynaptic
DNA techniques have recently been successfully ap- facilitation. The invasion of the sensory neuron termi-
plied to Aplysia, allowing genes of known function to nals by action potentials resulting from activation of
be isolated (101). But the primary reason I have the CS pathway makes these terminals more respon-
engaged in this speculative digression is to illustrate sive to the effects of serotonin released by the facilitat-
that the molecular regulatory processes are likely to ing neurons in the US pathway. Thus, classical condi-
prove important for understanding long-term modifi- tioning uses an amplification of the molecular machin-
cation in behavior produced by natural experience and ery used by sensitization, suggesting that there may be
by psychotherapeutic intervention. a molecular alphabet to learning, whereby complex
forms of learning use components found in simple
forms. In signaled anxiety, these presynaptic facilitat-
ANTICIPATORY ANXIETY SHARES MOLECULAR ing mechanisms appear to be used for two components
COMPONENTS WITH CHRONIC ANXIETY of the learning at two points in the neural circuit: an
associative component to provide for the temporal
We do not yet know in cellular detail the mecha- specificity of the modulation and a modulatory com-
nisms underlying the aversive conditioning that is used ponent to enhance defensive reflexes.
as a model for anticipatory anxiety in Aplysia, but The first component, enhancement, is a modulatory
there is already good evidence that the cellular mecha- component identical to the presynaptic facilitation that
nisms of aversive conditioning are related to those of accounts for sensitization. As is the case with sensitiza-
long-term sensitization (102-104). Analyses of classi- tion, the modulatory component enhances defensive
cal conditioning of simple reflexes in Aplysia suggest reflex responses (such as gill withdrawal) through the
that the learning of signaled anxiety involves a modi- serotonergic defensive arousal cells (figure 14).
fled form of the same cellular and molecular mecha- The second component, called the associative corn-
nisms-those of presynaptic facilitation-that underlie ponent, consists of a modified and augmented form of
chronic anxiety. The mechanism for associative speci- the same mechanism and gives classical conditioning

1290 Am J Psychiatry 1 40: 1 0, October 1983

 ERIC R. KANDEL

its temporal specificity. Using a simple reflex pathway different ways (conventional and amplified) to achieve
that can be associatively conditioned, Hawkins and chronic (unsignaled) anxiety or anticipatory (signaled)
associates (103) and Walters and Byrne (104) found anxiety. This model of classical conditioning further
that, after a series of pairing trials in which action suggests that we are on the threshold of understanding
potentials in the sensory neuron of a CS pathway our ability to learn predictive relationships. I would
immediately precede activity in the US pathway, the argue that the ability to learn predictive relation-
sensory neuron releases more transmitter than when ships-an ability critical to our mental life-lies in the
action potentials in the sensory neuron are not paired precise temporal requirement of a signal produced by
with the US. This evidence suggests that classical spike activity (perhaps calcium) for the adenylate
conditioning is essentially an amplified form of the cyclase system of certain neurons.
mechanism of presynaptic facilitation. It produces a
more profound depression of the potassium channels
and a larger increase in the duration of the action AN OVERALL VIEW
potential than does conventional presynaptic facilita-
tion. The models that I have considered emphasize the
According to this model for anticipatory anxiety, cellular interrelationship of chronic and anticipatory
when head shock is paired with shrimp, serotonergic anxiety in Aplysia. Both forms of anxiety involve the
cells in the head ganglia produce a highly robust strengthening of connections by modulating synaptic
presynaptic facilitation of the sensory neurons that transmission. Both lead to enhancement of transmitter
respond to shrimp. In this augmented form of facilita- release by depressing a potassium channel and thereby
tion, the ability of the serotonergic neurons to produce increasing influx of calcium. Thus, these studies sug-
presynaptic facilitation is substantially enhanced be- gest that there is a basic molecular grammar underly-
cause an action potential in the neurons of the CS ing the various forms of anxiety, a set of mechanistic
pathway that responds to shrimp immediately pre- building blocks that can be used in different combina-
cedes the action of the serotonergic cell (which is tions and permutations. They further suggest that a
activated by the aversive US). Activity dependence of variety of mental processes that appear phenotypically
classical conditioning explains why the CS must pre- unrelated may share a fundamental unity on the
cede the US during pairing for anticipatory anxiety to cellular and molecular levels.
be acquired. I have in this discussion purposely gone beyond the
facts in pointing arguments based on animal studies
toward human behavior because I have wanted to
A MOLECULAR MODEL FOR ANTICIPATORY emphasize two conceptual points that I believe will be
ANXIETY fundamental for the future study of the cellular mecha-
nisms of anxiety. The first is the power of experience in
How does the action potential in the neurons of the modifying brain function through altering synaptic
CS pathway lead to the enhanced presynaptic facilita- strength and regulating gene expression. The second is
tion that underlies the association of the anxiety state the utility and promise of animal models for the study
with a specific environmental signal? Clearly, one or of anxiety. Unlike schizophrenia, which does not exag-
more aspects of the action potential produce the gerate a normal adaptive process and is therefore a
amplification. Four events occur during the action characteristically human mental illness, fear or anxiety
potential: depolarization, sodium influx, calcium in- is a general adaptive mechanism found in simple as
flux, and potassium efflux. One attractive possibility is well as complex animals. There is good reason to
that calcium influx is the signal, since calcium affects believe that some of the cellular mechanisms of anxiety
the activity of cyclic AMP in a number of ways (105). may also be general.
The activity-dependent enhancement produced by cal- Moreover, I have suggested that normal learning,
cium could be achieved by modulating one or more the learning of anxiety and unlearning it through
steps in the cyclic AMP cascade. For example, calcium psychotherapeutic intervention, might involve long-
might modulate the activation of the adenylate cyclase term functional and structural changes in the brain
in the sensory cells by serotonin. According to this that result from alterations in gene expression. Thus,
idea, the action of the serotonergic facilitating neuron we can look forward, in the next decade of research
would be more effective in classical conditioning than into learning, to a merger between aspects of molecu-
in sensitization because activation of the cyclase by lam genetics and cellular neurobiology. This merger, in
serotonin is preceded by an influx of calcium within turn, will have important consequences for psychia-
the sensory neuron. Calcium, perhaps in association try-for psychotherapy on the one hand and for
with calmodulin, could bind to a site on the adenylate psychopharmacology on the other.
cyclase, producing a conformational change that leads
to its greater activation by serotonin.
Thus, a basic presynaptic regulatory mechanism REFERENCES
involving enhancement of transmitter release by cyclic 1. Freud S: Inhibitions, symptoms and anxiety (1925-1926), in
AMP-dependent phosphorylation could be used in Complete Psychological Works, standard ed, vol 20. London,

Am J Psychiatry 140:10, October 1983 1291

THE NATURE OF ANXIETY

Hogarth Press, 1959 Hamburg DA, Brodie HKH. Arieti S, editor-in-chief. New
2. Bowlby J: Attachment and Loss, vol 1 : Attachment. London, York, Basic Books, 1975
Hogarth Press, 1969 32. Badia P, Suter S, Lewis P: Preference for warned shock:
3. Nemiah JC: Anxiety neurosis, in Comprehensive Textbook of information and/or preparation. Psychol Rep 20:27 1-274,
Psychiatry, 2nd ed, vol 1. Edited
by Freedman AM, Kaplan HI, 1967
Sadock BJ. Baltimore, Williams
& Wilkins Co, 1975 33. Badia P, Culbertson S: Behavioral effects of signaled versus
4. Mowrer OH: A stimulus-response analysis of anxiety and its unsignaled shock during escape training in the rat. J Comp
role as a reinforcing agent. Psychol Rev 46:553-565, 1939 Physiol Psychol 72:216-222, 1970
5. Kimmel HD, Burns RA: The difference between conditioned 34. Weiss JM: Somatic effects of predictable and unpredictable
tonic anxiety and conditioned phasic fear: implications for shock. Psychosom Med 32:397-408, 1970
behavior therapy, in Stress and Anxiety, vol 4. Edited by 3S. Bindra D: How adaptive behavior is produced: a perceptual-
Spielberger CD, Sarason 1G. Washington, DC, Hemisphere motivational alternative to response reinforcement. Behavioral
Publishing Corp, 1977 and Brain Sciences 1:41-91, 1978
6. Brenner C: An Elementary Textbook of Psychoanalysis, re- 36. Rescorla RA: Some implications of a cognitive perspective on
vised ed. New York, International Universities Press, I 973 Pavlovian conditioning, in Cognitive Processes in Animal
7. Klein DF: Anxiety reconceptualized, in Anxiety: New Re- Behavior. Edited by Hulse SH, Fowler H, Honig WK. Hills-
search and Changing Concepts. Edited by Klein DF, Rabkin dale, NJ, Lawrence Eribaum Associates, I 978
JG. New York, Raven Press, 1981 37. Bolles RC, Fanselow MS: A perceptual, defensive, recuperative
8. Goodwin DW, Guze SB: Psychiatric Diagnosis, 2nd ed. Ox- model of fear and pain. Behavioral and Brain Sciences 3:291-
ford, Oxford University Press, 1979 323, 1980 -

9. Sheehan DV: Current concepts in psychiatry: panic attacks 38. Dickinson A: Contemporary Animal Learning Theory. Cam-
and phobias. N Engi J Med 307:156-158, 1982 bridge, Cambridge University Press, 1980
10. Klein DF, Fink M: Psychiatric reaction patterns to imipramine. 39. Weisskopf VF: Bicentennial address: frontiers and limits of
AmJ Psychiatry 119:432-438, 1962 physical sciences. American Academy of Arts and Sciences
1 1 . Klein DF: Delineation of two drug-responsive anxiety syn- Bulletin 35:4-23, 1981
dromes. Psychopharmacologia 5:397-408, 1964 40. Testa TJ: Causal relationships and the acquisition of avoidance
12. Mayer-Gross W: Clinical Psychiatry, 3rd ed. Edited by Slater responses. Psychol Rev 8 1 :491-SOS, I 974
E, Roth M. Baltimore, Williams & Wilkins Co, 1969 41. Dickinson A: Appetitive-aversive interactions: supercondition-
13. Darwin C: The Expression of the Emotions in Man and ing of fear by an appetitive CS. Q J Exp Psychol 29:71-83,
Animals. New York, D Appleton and Co, 1873 1977
14. Romanes GJ: Animal Intelligence. New York, D Appleton and 42. Dickinson A: Conditioning and associative learning. Br Med
Co, 1883 Bull 37:165-168, 1981
Is. Romanes GJ: Mental Evolution in Man: Origin of Human 43. Staddon JER, Simelhag VL: The “superstition” experiment: a
Faculty. London, Paul, 1888 reexamination of its implications for the principles of adaptive
16. James W: The Principles of Psychology, vols 1, 2. New York, behavior. Psychol Rev 78:16-43, 1971
Holt, 1893 44. Rescorla RA: Probability of shock in the presence and absence
17. Pavlov IP: Conditioned Reflexes: An Investigation of the of CS in fear conditioning. J Comp Physiol Psychol 66:1-5,
Physiological Activity of the Cerebral Cortex. Translated and 1968
edited by Anrep GV. London, Oxford University Press, 1927 45. Rescorla RA: Second order conditioning: implications for
18. Cohen ME, Badal DW, Kilpatrick A, et al: The high familial theories of learning, in Contemporary Approaches to Condi-
prevalence of neurocirculatory asthenia (anxiety neurosis, tioning and Learning. Edited by McGuigan FJ, Lumsden DB.
effort syndrome). Am J Hum Genet 3:126-158, 1951 Washington, DC, VH Winston & Sons, 1973
19. Sargant W, Slater E: An Introduction to Physical Methods of 46. Rescorla RA: Conditioned inhibition and extinction in mecha-
Treatment in Psychiatry, 4th ed. Edinburgh, Livingstone, 1963 nisms of learning and motivation, in Mechanisms of Learning
20. Slater E, Shields J: Genetical aspects of anxiety. Br Psychiatry and Motivation: A Memorial Volume of Jerzy Konorski.
Special Publication 3:62-71, 1969 Edited by Dickinson A, Boakes RA. Hillsdale, NJ, Lawrence
21. Crowe RR, Pauls DL, Slymen DJ, et al: A family study of Erlbaum Associates, 1979
anxiety neurosis. Arch Gen Psychiatry 37:77-79, 1980 47. James W: Psychology: Briefer Course. New York, Holt, 1892
22. Pauls DL, Bucher KD, Crowe RR, et al: A genetic study of 48. Guthrie ER: The Psychology of Learning. New York, Harper
panic disorder pedigrees. Am J Hum Genet 32:639-644, 1980 and Brothers, 1935
23. Estes WK, Skinner BF: Some quantitative properties of anxi- 49. Gormezano I, Kehoe EJ: Classical conditioning: some method-
cry. J Exp Psychol 29:390-400, 1941 ological conceptual issues, in Handbook of Learning and
24. Miller NE: Studies of fear as an acquirable drive: fear as Cognitive Processes: Conditioning and Behavior Theory, vol
motivation and fear-reduction as reinforcement in the learning 2. Edited by Estes WK. Hillsdale, NJ, Lawrence Erlbaum
of new responses. J Exp Psychol 38:89-101, 1948 Associates, 1975
25. Seligman MEP: Helplessness: On Depression, Development, so. Prokasy WF (ed): Classical Conditioning: A Symposium. New
and Death. San Francisco, WH Freeman and Co, 1975 York, Appleton-Century-Crofts, 1965
26. Kandel ER: Cellular Basis of Behavior: An Introduction to Si. Kamin U: Predictability, surprise, attention, and conditioning,
Behavioral Neurobiology. San Francisco, WH Freeman and in Punishment and Aversive Behavior. Edited by Campbell BA,
Co, 1976 Church RM. New York, Appleton-Century-Crofts, 1969
27. Dollard J, Miller NE: Personality and Psychotherapy. New 52. Mackintosh NJ: The Psychology of Animal Learning. London,
York, McGraw-Hill, 1950 Academic Press, 1974
28. Hammond U: Conditioned emotional states, in Physiological 53. Wagner AR: Priming in STM: an information-processing
Correlates of Emotion. Edited by Black P. New York, Academ- mechanism for self-generated or retrieval-generated depression
ic Press, 1970 in performance, in Habituation: Perspectives From Child
29. Watson JB, Rayner R: Conditioned emotional reaction. J Exp Development, Animal Behavior, and Neurophysiology. Edited
Psychol 3:1-14, 1920 by Tighe TJ, Leaton RN. Hilisdale, NJ, Lawrence Eribaum
30. Pinsker HM, Hening WA, Carew TJ, et al: Long-term sensiti- Associates, 1976
zation of a defensive withdrawal reflex in Aplysia. Science 54. Rescorla RA, Wagner AR: A theory of Pavlovian conditioning:
182:1039-1042, 1973 variations in the effectiveness of reinforcement and nonrein-
3 1. Kandel ER: Perspectives in the neurophysiological study of forcement, in Classical Conditioning II: Current Research and
behavior and its abnormalities, in New Psychiatric Frontiers: Theory. Edited by Black AH, Prokasy WE New York, Apple-
American Handbook of Psychiatry, 2nd ed, vol VI. Edited by ton-Century-Crofts, 1972

1292 Am J Psychiatry 1 40: 1 0, October 1983

 ERIC R. KANDEL

55. Guthrie ER: Association by contiguity, in Psychology: A Study 82. Brunelli M, Castellucci V, Kandel ER: Synaptic facilitation
of a Science, vol 2. Edited by Koch S. New York, McGraw- and behavioral sensitization in Aplysia: possible role of seroto-
Hill, 1959 nm and cyclic AMP. Science 194:1178-1181, 1976
56. Skinner BF: Verbal Behavior. New York, Appleton-Century- 83. Klein M, Kandel ER: Mechanism of calcium current modula-
Crofts, 1957 tion underlying presynaptic facilitation and behavioral sensiti-
57. Tolman EC: Purposive Behavior in Animals and Men. New zation in Aplysia. Proc Nati Acad Sci USA 77:6912-6916,
York, Appleton-Century-Crofts, 1932 1980
58. Chomsky N: A review of BF Skinner’s Verbal Behavior. 84. Kandel ER, Schwartz JH: Molecular biology of an elementary
Language 35:26-58, 1959 form of learning: modulation of transmitter release by cyclic
59. Neisser U: Cognitive Psychology. Englewood Cliffs, NJ, Pren- AMP. Science 218:433-443, 1982
tice-Hall, 1967 85. Castellucci VF, Kandel ER, Schwartz JH, et al: Intracellular
60. Seligman MEP, Meyer B: Chronic fear and ulcers in rats as a injection of the catalytic subunit of cyclic AMP-dependent
function of the unpredictability of safety. J Comp Physiol protein kinase simulates facilitation of transmitter release
Psychol 73:202-207, 1970 underlying behavioral sensitization in Aplysia. Proc Nat! Acad
61. Posner MI: Cognition: An Introduction. Chicago, Scott, Fores- Sci USA 77:7492-7496, 1980
man & Co, 1973 86. Castellucci VF, Nairn A, Greengard P, et al: Inhibitor of
62. Klatzky RL: Human Memory: Structures and Processes, 2nd adenosine 3 ‘ :5 ‘-monophosphate-dependent protein kinase
ed. San Francisco, WH Freeman and Co, 1980 blocks presynaptic facilitation in Aplysia. J Neurosci 12:
63. Watson JB: Psychology as the behaviorist views it. Psychol Rev 1673-1681, 1982
20:158-177, 1913 87. Siegelbaum SA, Camardo JS, Kandel ER: Serotonin and cyclic
64. Watson JB: Behaviorism. New York, WW Norton & Co, I 925 AMP close single K channels in Aplysia sensory neurones.
65. Dickinson A, Mackintosh NJ: Classical conditioning in ani- Nature 299:413-417, 1982
mals. Annu Rev Psychol 29:587-612, 1978 88. Bailey CH, Chen MC: Morphological basis of long-term
66. Griffin DR (ed): Animal Mind-Human Mind: Report on the habituation and sensitization in Aplysia. Science 220:91-93,
Dahlem Workshop. New York, Springer-Verlag, 1982 1983
67. Sahley CL, Rudy JW, Gelperin A: An analysis of associative 89. Gurdon JB: The Control of Gene Expression in Animal
learning in a terrestrial mollusc, I: higher-order conditioning, Development. Cambridge, Mass, Harvard University Press,
blocking and a transient US-pre-exposure effect. Journal of 1974
Comparative Physiology (Part A) 144:1-8, 1981 90. Davidson EH: Gene Activity in Early Development, 2nd ed
68. Gilbert W: Why genes in pieces? Nature 271:501, 1978 New York, Academic Press, 1976
69. Crick F: Split genes and RNA splicing. Science 204:264-271, 91. Watson JD: Molecular Biology of the Gene, 3rd ed. Menlo
1979 Park, Calif, WA Benjamin, 1976
70. Darnell JE Jr: Transcription units for mRNA production in 92. Brown DD: Gene expression in eukaryotes. Science 211:667-
eukaryotic cells and their DNA viruses. Prog Nucleic Acid Res 674, 1981
Mol Biol 22:327-353, 1979 93. Darnell JE Jr: Variety in the level of gene control in eukaryotic
71. Lewin B: Gene Expression: Eucaryotic Chromosomes, 2nd ed, cells. Nature 297:365-371, 1982
vol 2. New York, John Wiley & Sons, 1980 94. Barondes SH: Multiple steps in the biology of memory, in The
72. Stryer L: Biochemistry, 2nd ed. San Francisco, WH Freeman Neurosciences: Second Study Program. Edited by Schmitt FO,
and Co, 1981 Q uarton GC, Melnechuk T, et al. New York, Rockefeller
73. Walters ET, Carew TJ, Kandel ER: Classical conditioning in University Press, 1970
Aplysia californica. Proc Natl Acad Sci USA 76:6675-6679, 9S. Agranoff BW, Burrell HR, Dokas LA, et al: Progress in
1979 biochemical approaches to learning and memory, in Psycho-
?4#{149}
Walters ET, Carew TJ, Kandel ER: Associative learning in pharmacology: A Generation of Progress. Edited by Lipton
Aplysia: evidence for conditioned fear in an invertebrate. MA, DiMascio A, Killam KE New York, Raven Press, 1978
Science 211:504-506, 1981 96. Kety 55: Disorders of the human brain. Sci Am 241:202-2 14,
75. Castellucci V, Pinsker H, Kupfermann I, et al: Neuronal 1979
mechanisms of habituation and sensitization of the gill-with- 97. Sachar EJ: Psychobiology of schizophrenia, in Principles of
drawal reflex in Aplysia. Science 167:1745-1748, 1970 Neural Science. Edited by Kandel ER, Schwartz JH. New
76. Castellucci V, Kandel ER: Presynaptic facilitation as a mecha- York, Elsevier-North Holland, 1981
nism for behavioral sensitization in Aplysia. Science 194: 98. Sachar EJ: Psychobiology of affective disorders. Ibid
1176-1178, 1976 99. Lasek RJ, Dower WJ: Aplysia californica: analysis of nuclear
77. Bailey CH, Hawkins RD,
Chen MC, et al: Interneurons DNA in individual nuclei of giant neurons. Science 172:278-
involved in mediation and modulation of the gill-withdrawal 280, 1971
reflex in Aplysia, IV: morphological basis of presynaptic 100. Schwartz JH, Castellucci VF, Kandel ER: Functioning of
facilitation. J Neurophysiol 45:340-360, 1981 identified neurons and synapses in abdominal ganglion of
78. Hawkins RD, Castellucci VF, Kandel ER: Interneurons in- Aplysia in absence of protein synthesis. J Neurophysiol
volved in mediation and modulation of gill-withdrawal reflex 34:939-953, 1971
in Aplysia, I: identification and characterization. J Neurophys- 101. Scheller RH, Jackson JF, McAllister LB, et al: A family of genes
iol45:304-314, 1981 that codes for ELH, a neuropeptide eliciting a stereotyped
79. Hawkins RD. Castellucci VF, Kandel ER: Interneurons in- pattern of behavior in Aplysia. Cell 28:707-719, 1982
volved in mediation and modulation of gill-withdrawal reflex 102. Duerr JS, Quinn WG: Three Drosophila mutations that block
in Aplysia, II: identified neurons produce heterosynaptic facili- associative learning also affect habituation and sensitization.
tation contributing to behavioral sensitization. J Neurophysiol Proc Nati Acad Sci USA 79:3646-3650, 1982
45:315-326, 1981 103. Hawkins RD, Abrams TW, Carew TJ, et al: A cellular
80. Hawkins RD: Interneurons involved in mediation and modu- mechanism of classical conditioning in Aplysia: activity-dc-
lation of gill-withdrawal reflex in Aplysia, III: identified pendent amplification of presynaptic facilitation. Science
facilitating neurons increase Ca2 current in sensory neurons. 219:400-405, 1983
J Neurophysiol 45:327-339, 1981 104. Walters ET, Byrne JH: Associative conditioning of single
81. Bernier L, Castellucci VF, Kandel ER, et al: Facilitatory sensory neurons suggests a cellular mechanism for learning.
transmitter causes a selective and prolonged increase in cAMP Science 219:405-408, 1983
in sensory neurons mediating the gill- and siphon-withdrawal 105. Cheung WY: Calmodulin plays a pivotal role in cellular
reflex in Aplysia. J Neurosci 2:1682-1691, 1982 regulation. Science 207: 19-27, 1980

Am J Psychiatry 1 40: 1 0, October 1983 1293