EDGE 6

EDGE 6 — February 17, 1997

THE THIRD CULTURE

"PARALLEL MEMORIES: PUTTING EMOTIONS BACK INTO THE BRAIN"
A Talk with Joseph LeDoux

We have to put emotion back into the brain and integrate it with cognitive systems. We shouldn't study emotion or cognition in isolation, but should study both as aspects of the mind in its brain

THE REALITY CLUB

Reuben Hersh's Rejoinders to Rebuttals by Simonyi and Dehaene

I do not believe and would never say that mathematics is merely the pursuit of abstract beauty, or a game not rooted in the human mind and the physical world. But first of all, it's a problem-solving, theory-building activity, carried out on the basis of a given theory, and elaborated in the judgment of peers. Rigorous logic and physical application are usually far in the dim distance. To try and understand this activity, this world, out of its social context, is to make it incomprehensible, or comprehensible only by a falsification.

Paolo Pignatelli on Stanislas Dehaene

I hope that the next generation of computers will rely on some chemical at first, then neuronal architecture, for then not only will we have vastly superior computers, but we will also be solving many evolutionary and philosophical ones too.

Smolin, Hillis, and Lanier on Hersh

Lee Smolin

There is a list of questions we are on the verge of solving, like the origin of life or the nature of space, that require twentieth century physics and mathematics, and that a nineteenth century person could not even have gotten started on.

W. Daniel Hillis

Mathematics is not just a game or a poem with its own set of internal rules: it also has a striking correspondence to the real world. If I follow the rules it tells me things about the real world.

Jaron Lanier

There is room for a pseudo-humanistic philosophy of mathematics that seems more true to me than either Hersh's approach or strict Platonism as he presents it.

Coming Events

Jared Diamond, Murray Gell-Mann, J. Doyne Farmer, Stewart Kauffman, Seth Lloyd, Doug Rowan, Lee Smolin, Charles Simonyi , Joseph Traub

(9,875 words)

John Brockman, Editor and Publisher | Kip Parent, Webmaster

THE REALITY CLUB
Reuben Hersh's Rejoinders to Rebuttals by Simonyi and Dehaene

From: Reuben Hersh
Submitted: 2/16/97
It's interesting that even though Platonism is the most popular philosophy among mathematicians, all four respondents readily reject it.

I have not solved all the philosophical dilemmas of mathematics. Especially, Wigner's dilemma of "unreasonable effectiveness." I have tried simply to give an honest account of mathematical activity in real life, without any dogmatic preconception. That includes recognizing the science like aspect, the reproducibility and near unanimity of mathematical results, including the value of pi.

How does a socially shared concept system possess these science-like qualities? Kant asked, "How is mathematics possible?"

His ingenious answer doesn't work. There are no universal intuitions of time and space. But contemporary work in neurophysiology, like that described by Dehaene, may in a sense revive Kant's idea, now based on empirical science rather than pure speculation. If we do have brain structures for counting and for certain elementary spatial properties, we have them because they have survival value. They correspond to physical reality.

Now, certain full grown mathematical theories that physicists use also can be said to have survival value. Not biological survival, but social survival. (It has been argued that Newton's celestial mechanics gave England an edge in marine navigation, which would have been an edge in naval and commercial rivalry.) Can this parallel explain why mathematics is what it is?

If you look into the problem in more detail, this explanation is not so convincing. For instance, Heisenberg found matrix theory ready to hand for his version of quantum mechanics. Matrix theory was originated long before by Cayley, who found it a nice way to think about his algebraic transformations. Do you believe the military or commercial survival of England was really behind Cayley's thinking?

Complex numbers come in handy for a lot of things, like alternating current calculations and a scalar field for Hilbert space in q.m. Survival value, yes, if a.c. current and quantum mechanics do increase somebody's survival chances. But only after the fact! Originally, they were just unwanted "false solutions" for certain quadratic equations. Later, for cubics, they came in, uninvited, in the formula for real solutions. Were Ferrari and Cardano unconsciously getting ready for Steinmetz's electrical calculations?

This reminds me of anthropic discussions in cosmology. How in Heaven's name could it happen that the values of the fundamental constants are just what they need to be to make human life possible?

How is it that by solving problems, and inventing tools and concepts to solve those problems, and then solving the new problems about those new tools and concepts—mathematicians often give physics a hand?

Naturally it's no surprise that mathematicians working on questions from physics may give physics a hand. But that's not where fiber bundles and connections came from.

It's a mystery. I haven't tried to solve it. Is it more fruitful to be hung up on this mystery, or to accept it and go ahead?

You can't explain why there is matter rather than nothing. You don't wait to answer that before you do a little physics.

I can't explain (in a detailed, rather than vague and general way) why the social activity called mathematics is possible. I can recognize that since it exists, it is possible, without Platonistic ghosts or formalist devaluation. I can try to watch it with understanding, try to see what it does and how it works.

I agree largely with Dehaene, in particular with his psychological explanation of the origin of Platonism. I think the difference between his view and mine is largely a
matter of emphasis.

I do not believe and would never say that mathematics is merely the pursuit of abstract beauty, or a game not rooted in the human mind and the physical world. But first of all, it's a problem solving, theory-building activity, carried out on the basis of a given theory, and elaborated in the judgment of peers. Rigorous logic and physical application are usually far in the dim distance. To try and understand this activity, this world, out of its social context, is to make it incomprehensible, or comprehensible only by a falsification.

There are three serious objections to accepting the status of mathematics as part of the socio cultural-historic level of reality.

1) Numbers are part of physical reality ("there were 9 planets before there were people so 9 existed before there were people."

2) Concepts from pure math repeatedly have been found useful in physics (Wigner, "the unexpected usefulness of mathematics" I don't remember the exact title.)

3) Concepts and methods from the social world are never as exact, reliable, verifiable, and near-unanimously consensual as in math.

Point 1 was taken up in my original interview. But I will repeat, briefly. "Nine" or any other number word has two usages, as adjective (describing a physical or other collection) and as noun, describing something independent of any particular realization, a general concept, usually referred to as part of an abstract structure, the natural numbers as governed by certain rules (axioms). The number-adjectives have no least upper bound, yet it is easy to write down a numeral for a number greater than any of them—that is, that will ever be counted or
observed. In the shared concept of abstract theory if you will, that is a contradiction ---

(To Charles Simonyi:) There is a 2-volume collected papers of Lakatos, but his masterpiece is a separate book, "Proofs and Refutations."

Comments and arguments are invited

Reuben Hersh

Paolo Pignatelli on Stanislas Dehane

From: Paolo Pignatelli
Submitted: 2/17/97
To: Stanislas Dehaene

As always, very smart guests, fascinating subjects. I was especially fascinated by Stanislas Dehaene's comments on the existence of dedicated neuronal circuits in the brain for processing numbers, since I believe that present computational theories may have to be extended, in the way that Einstein extended Newtonian mechanics, so that computers may begin to approach the richness of ways of *computing* that the human brain possesses today.

First, a mild disagreement regarding the sentence "That this 'number sense' is also present in animals, and hence that it is independent of language and has a long evolutionary past." Starting out at the level of neuronal polymorphism, and looking at the interesting clinical examples of such, (case Alex for example in language acquisition), perhaps the division drawn between mathematics and language is one of hierarchy rather than a "logical" one. At what point of ancestral connectedness would you say that there is "independence? Naturally, this may all be in your book, which I eagerly await.

Have you found a connection between linguistic ability and mathematical one that is closer to that of either one to intelligence itself? But back to the dedicated circuits in the brain to process numbers, to what extent does this eminent group gathered here by John Brockman see these discoveries affecting the computing industry? Personally, I am hopeful that engineers who followed the experiments of the so-called "chemical computers" (Oliver Steinbock, Agota Toth and Kenneth Showalter at West Virginia University...) may see, as did Showalter (see Peter Coveney and Roger Highfield in their book Frontiers of Complexity), the implications of path optimization in neuron networks.

My interest is in the path optimization implications as related to a machine that I will call the "entropy pump machine". An entropy pump machine acts as a semi-permeable membrane between regions of different relative entropy, always maximizing the gradient between the two. This is obviously a path optimization phenomena. I hope that the next generation of computers will rely on some chemical at first, then neuronal architecture, for then not only will we have vastly superior computers, but we will also be solving many evolutionary and philosophical ones too. -

Paolo Pignatelli

PAOLO PIGNATELLI, a cyber-entrepreneur, is proprietor of the virtual Corner Store. He is a linguist, translator and scientist who previously worked in image processing algorithms at Bell Labs.

Lee Smolin, W. Daniel Hillis and Jaron Lanier on Reuben Hersh

From: Lee Smolin
Submitted: 2/10/97
I have to say that I disagree with almost everything that Reuben Hersh says. I can start with what I agree with, which is that the platonist and formalist schools of philosophy of mathematics do not capture what mathematics is. It must be said, however, that they are not stupid, or obviously wrong, I think that my disagreement with Platonism comes from two things: first from my philosophical commitment to the idea that the world we see is all there is, and that everything we see must be explained in terms of a network of relationships among real things. This leaves no place for a realm of real, eternal forms that transcends the particulars of the world, as well as no place for a view of the universe as if from outside of it. It also leaves no place for a platonic realm of mathematical form.

The second reason I disagree with platonism is that I think it is insufficient to make sense of the mathematical structures that arise in biology. It is one thing to speak of every possible platonic solid, but should we think that every possible biological species, or every possible niche, or every possible ecology exists eternally in some realm of ideal platonic biology? What about every possible way of earning a living in human society? Stuart Kauffman has been arguing that it may not be possible to list these kinds of things in advance, and I am tempted to think he may be right. There are lots of things that apparently cannot be classified in mathematics, like algebras or knots or four dimensional manifolds. For this reason, I suspect that Platonism will eventually come to be seen as insufficient to encompass the variety of possible mathematical structures. The point is that I believe that novelty is both possible and important, there are novel structures being discovered all the time, both by natural selection and human intelligence, and some of these are mathematical.

Formalism is easier to put away; it was basically killed by Godel's theorem. So what then is mathematics? I believe I understand the reasons why Hersh makes the move he does: that it is a shared construction of human beings, for that is some of it. There are an infinite number of possible mathematical structures, why some have been intensively thought about, while others were either thought uninteresting and most have not even been conceived of is a historical question. So historical and social questions may plausibly play some role in understanding why mathematics is as it is now. But this is not the same thing as to ask, what is a number, or what is mathematics.

I do not have an answer to this question that satisfies me, although I have thought a lot about it. In my opinion it is one of the really hard questions, like consciousness, or whether time might have begun, or might end. There are questions that I believe we cannot even conceive of satisfactory answers to given what we know presently. This does not mean they may not someday be solved-I think they may. There is a list of questions we are on the verge of solving, like the origin of life or the nature of space, that require twentieth century physics and mathematics, and that a nineteenth century person could not even have gotten started on.

Having said this, there are two thoughts that I find interesting when I try to think about what mathematics is. The first is the observation that time may play an essential role because a mathematical paradox can become a feedback loop when time is introduced. Something cannot be both true and not true eternally, but it can be alternatively in time. The second is the possibility that category theory may have profound implications for the question of what mathematics is, because it puts the emphasis exactly on relationships between different things. One might have looked down on category theory some years ago, but given the profound insights it has introduced into the relationships between different mathematical structures such as algebra and topology it seems very worth thinking about.

LEE SMOLIN is a theoretical physicist; professor of physics and member of the Center for Gravitational Physics and Geometry at Pennsylvania State University; author of The Life of The Cosmos, forthcoming (Oxford).

From: W. Daniel Hillis
Submitted: 2/12/97

I certainly cannot argue with Hersh's premise mathematics is a part of human culture and human history, but surely it is also something more. Mathematics is not just a game or a poem with its own set of internal rules: it also has a striking correspondence to the real world. If I follow the rules it tells me things about the real world. A calculation can tell me where a ball will go next, what shape the bubble will be, or when the train will arrive. Often a mathematical construct is invented long before the corresponding reality is even noticed. Dirac, for instance, suggested the existence of anti-matter just because the equations of quantum mechanics also allowed for a negative solution. As far as I know, this magical connection between the abstract operations of mathematics and the real world remains entirely unexplained, but is surely an important part of what make mathematics special.

W. DANIEL HILLIS is vice president of research and development at the Walt Disney Company and a Disney Fellow. He was cofounder and chief scientist of Thinking Machines Corporation.

From: Jaron Lanier
Submitted: 2/13/97

There is room for a pseudo-humanistic philosophy of mathematics that seems more true to me than either Hersh's approach or strict Platonism as he presents it. In this philosophy, the particulars of math would be understood as platonically mandatory and eternal, but the range of possible areas of math to study and know would be understood to be breathtakingly large. So large that two different cultures undertaking mathematical study might not necessarily come across any common material. It is hard for us to imagine aliens not thinking about integers, but it is not logically impossible. There are some elements of logic itself that would have to crop up in some form, but the notion of what form would be most elegant and normal could be so variable as to leave room for a universe of virtually disjoint cultures of mathematics. Cultural diversion in math is more pleasant than in other areas, since it will never lead to authentic contradiction. This approach gets out from under the common, but false, implication of determinism in the history of mathematical inquiry that weighs down the teaching of a subject that should be joyous like music. It allows educators to treat math as culture, but at the same time avoids relativising the one area of human activity in which we can know truth.

JARON LANIER, a computer scientist and musician, is a pioneer of virtual reality, and founder and former CEO of VPL.

THE THIRD CULTURE

We have to put emotion back into the brain and integrate it with cognitive systems. We shouldn't study emotion or cognition in isolation, but should study both as aspects of the mind in its brain.

"PARALLEL MEMORIES: PUTTING EMOTIONS BACK INTO THE BRAIN"
A Talk with Joseph LeDoux

Neuroscientist Joseph LeDoux seeks a biological rather than psychological understanding of our emotions. He explores the differences between emotional memories (implicit--unconscious--memories) processed in pathways that take information into the amygdala, and memories of emotion (explicit--conscious--memories) processed at the level of the hippocampus and neocortex.-

JB

JOSEPH LEDOUX is the Henry and Lucy Moses Professor of Science at the Center for Neural Science, New York University, and author of the recently published The Emotional Brain: The Mysterious Underpinnings of Emotional Life. He is the author with Michael Gazzaniga of The Integrated Mind and is editor with William Hirst of Mind and Brain: Dialogues in Cognitive Neuroscience. LeDoux's home page contains additional information about his lab's research:

"PARALLEL MEMORIES: PUTTING EMOTIONS BACK INTO THE BRAIN"

JB: Emotions and the brain? Isn't this something new for scientists?
LEDOUX: Twenty years ago no one cared about emotions and the brain, but it seems in the last couple of years there's been a flurry of activity. One reason for this may be that the topic was ignored for so long, and the vacuum is being filled. Another, though, is that there have been some successes in approaching the problem, and these have changed peoples' minds about the feasibility of studying emotions in the brain.

The most successful efforts have come from the study of fear. Fear is a relatively tractable emotion, unlike love or hope which are difficult to pin down. It's always easier to study brain functions that involve clearly defined stimuli and responses than those that don't. For fear, you can easily create experimental situations where the onset of a simple stimulus that warns of impending danger elicits a set of stereotyped responses in an animal, like a rat, that are very similar to the kinds of responses that occur in a human facing danger. By following the flow of the stimulus through the brain from the stimulus processing pathways to the response control networks, it's possible to identify the basic neural circuits involved. We've done this for fear.

JB: How did you get into this?

LEDOUX: I got interested in emotions while I was studying something completely different. I was doing split-brain research as a graduate student with Mike Gazzaniga. Mike and I were studying how information is transferred between the hemispheres of these patients. One of the questions we asked was what happens when we put information in the right hemisphere. Remember, it's the left hemisphere that usually does the talking, so information in the right hemisphere can't ordinarily be talked about in these patients. We put emotional information in the right hemisphere, and the left hemisphere couldn't tell us what it saw, but it could tell us how it felt about it. That led us to the idea that emotional information and information about the content of what a stimulus is, are processed by different pathways in the brain. That seemed very interesting, and I decided I wanted to pursue it.

At the time, I felt that the only way to go about studies of the pathways of emotional processing was to turn to an animal model, where you can do experimental lesions, cell recordings, pathway tracing and so on. The reason you want to do these kinds of studies is not to satisfy some reductionistic urge, but because they can help you see how emotion is put together in the brain, and this can tell you about how the function itself works. Today, there are more sophisticated ways of studying the human brain, such as functional imaging. These can give you a picture of the brain in some emotional state, but you can't then ask the next question. You want to know how the activated region fits into a larger system. You really can't get to those kinds of questions in humans and have to turn to the animal models for answers. The animal work, in other words, gives the framework for interpreting the snapshots we get from human imaging studies. Without the animal studies, though, many of the human studies probably never have been done, and if they had, they wouldn't be so readily interpretable.

So I left the world of human neuropsychology and went into animal research after finishing my PhD and a short post doc. Mike and I had moved to Cornell Medical School and after a year or so I hooked up with Don Reis in the Neurobiology Lab there. The lab's mission was to study the brain's control over the autonomic nervous system, and basically I was told that I could do whatever I wanted as long as I recorded blood pressure. So I developed a blood pressure model of conditioned fear.

I used conditioned fear because it seemed like a relatively straightforward technique: you give a meaningless tone followed by a mild shock a few times, and pretty soon the tone starts eliciting a blood pressure response. It was a good way to create an emotional reaction to the tone on the spot in an animal that wasn't afraid of the tone and didn't have any emotional reaction to it to begin with. Since the tone gets to the brain by way of the auditory system and the response comes out of the brain through the autonomic nervous system, the trick was to figure out how the auditory system is linked up with the autonomic system. By using a combination of brain lesions, neural recordings, and pathway tracing techniques, we were able to figure this out. The answer, in short, is that the amygdala turned out to be a necessary and sufficient link between the auditory system and the autonomic nervous system. However, in a more general sense, the amygdala is the link between all sensory systems and all fear responses systems. It's the part of the brain involved independent of how the stimulus gets into the brain and how the response comes out.

JB: I find it interesting that the first emotion you studied was fear.

LEDOUX: When I first began this work in the early 1980s, I was using fear conditioning techniques because they were convenient. As I said, you can take the stimulus, pair it with the shock one or two times and, as a result, create an emotional reaction that's relatively profound in the animal. I thought at the time that this was going to be a way of identifying a universal emotional system in the brain, something akin to the limbic system. I no longer feel that way. I think that the study of the limbic system, or more generally the idea that there is an emotion system in the brain, is misguided. I came to this conclusion empirically. Once we had outlined a neural circuit for fear responses, it was obvious that the limbic system had little to do with it. The only so-called limbic area involved was the amygdala. And the hippocampus, the centerpiece of the limbic system, had been implicated in non-emotional processes like memory and spatial behavior. It seemed clear that the limbic system, if it existed at all, was not systematically involved in any clear way. I decided I didn't need the limbic system concept to think about how fear works in the brain. But that still doesn't wholly justify the focus on fear to the exclusion of other emotions.

I've come to think that emotions are products of different systems, each of which evolved to take care of problems of survival, like defending against danger, finding mates and food, and so forth. These systems solve behavioral problems of survival. Detecting and responding to danger requires different kinds of sensory and cognitive processes, and different kinds of motor outputs, different kinds of feedback networks, and so on, than finding a mate or finding food. Because of these unique requirements, I think different systems of the brain are going to be involved in the different kinds of emotions.

A related point is that emotion systems, like the fear system, didn't come about to create feelings (like the feeling of being afraid when in danger). I think feelings came much later in evolution. All animals have to be able to detect and respond to danger, regardless of the kind of cognitive architecture they have. This is as true of bees and worms and snails, as it is of fish, frogs, birds, rats, and people. Fear conditioning, by the way, occurs in all animals. And in all those that have an amygdala, the amygdala appears to be the key. The list at this point includes reptiles, birds, and a host of mammals, including humans. I think it's safe to say fear behavior preceded fear feelings in evolution. If so, feelings are probably the wrong thing to focus on when we study emotions. In this sense, animals were unconscious, unfeeling, and non-linguistic before they were conscious, feeling, and linguistic. It's too bad that we define the more basic processes as the negation processes that typify the human brain. It's possible that once consciousness and feelings came along that new kinds of emotions specifically tied to these evolved. But I'm trying to understand the things about emotions that are similar in humans and other animals so that I can work on emotions through the brain.

I tend to agree with theorists who say there are basic emotions that are hard-wired into the brain's architecture, and that one of the advantages of having an extra big cortex is that we can blend different hard-wired emotions together to create softer emotions, where cognitions come into play in a major way. For example, while detection and responding to danger may be built into the brain, the capacity to be afraid of falling in love is something that requires the cognitive integration of the system for finding mates and the system for defending against predators. While I'm sympathetic towards the basic emotions view, I don't really ascribe to it. It requires that you state what the different emotions are. That just leads to arguments. I'd rather spend my time worrying about one well accepted emotion and its organization in the brain than fighting over whether this or that mental process is an emotion or not.

JB: So what about feelings? What are they?

LEDOUX: The study of emotion has focused on conscious feelings almost to the exclusion of everything else. Emotion researchers, for some reason, seem to be carrying the burden of the mind-body problem on their shoulders. In other words, I think the problem of feelings is one and the same as the problem of consciousness, and that emotion researchers have no more or less of an obligation to solve this before anybody else. Take vision. Philosophers have worried about where the redness we experience comes from when we see an apple. But vision researchers figured out that they could study how we process red without having to first figure out how we experience red. The same can be done in the study of emotion. We can study how the brain detects and responds to danger, even if we don't know how it experiences danger. So the feelings of fear that come about in dangerous situations are in a sense no different from any other kind of conscious experience. The only difference is in the system that consciousness is paying attention to the danger processing system, the color processing system, the language processing system, and so on. So emotional feelings come about when we become consciously aware of the activity of an emotional system, which does its work for the most part outside of consciousness.

JB: What's the difference between an emotional and a cognitive memory?

LEDOUX: By cognitive memory I'm going to assume you mean explicit conscious memory, the kind of memory we usually have in mind when we use the word memory in everyday speech. Emotional memory and explicit memory happen at the same time, but separately. For example, the amygdala mediates emotional memory and the temporal lobe memory system mediates explicit memory.

Here's an example. Imagine driving down the road and having an accident. You hit your head on the steering wheel and the horn gets stuck on. You're bleeding and in pain. It's awful. Sometime later, you hear the sound of a horn. The sound goes to your amygdala and activates your autonomic nervous system (raising your blood pressure and heart rate, making you sweat), tenses your body muscles, releases stress hormones into your blood, and so on. The sound also goes to the temporal lobe system and reminds you of the accident, of who you were with and where you were going. It also reminds you that it was awful. But these are all just facts about the situation. They are memories of the emotional experience rather than emotional memories. In general, one difference between emotional and cognitive processing is that emotional processing often leads to bodily responses, whereas cognitive processing leads to more cognitive processing. Cognitions are seldom characterized by specific kinds of responses, but emotions usually are. It's important that we understand as much as we can about the biology of these systems.

Many people have problems with their emotional memories; psychologists' offices are filled with people who are basically trying to take care of and alter emotional memories, get rid of them, hold them in check. If anything, emotional memory is more basic than explicit conscious memory. For example, it takes place at an earlier age. It's conceivable, and in fact seems very likely, that a child could be abused very early in life and develop unconscious emotional memories through the amygdala prior to the point where the temporal lobe memory system has kicked in. If that's true then emotional memories are being formed for things that will never be consciously understood, because the system that mediates conscious memory isn't available to encode the experience and can therefore never retrieve it.

We need to understand how unconscious emotional memories are formed-- not only because they occur in early childhood, but because emotional memories are created throughout our lives. And it appears that these memories are indelible. They can be extinguished in the laboratory or treated in the psychiatrist's office, but they can usually be brought back. And recently we've been able to find a mechanism in the amygdala that might be responsible for this. It's sort of complicated, but the finding goes like this. We record neural activity in the amygdala before and after conditioning. Cells fire more to the tone afterwards. With extinction the firing rate goes back to baseline. However, in addition to measuring these stimulus-evoked responses, we measure the correlation in the time when different cells fire spontaneously (no stimulus present). After conditioning, some cells that were not correlated become correlated. And for some of the cells the correlations remain past extinction. In other words, the feared stimulus no longer elicits activity in the amygdala, but the amygdala cells continue to be functionally coupled. It's as if extinction (and therapy) doesn't erase the memory, it just weakens the ability of the stimulus to activate the memory. So in order for the stimulus to again be effective all you have to do is change the synaptic strength of the connection between the stimulus and the memory rather than recreate the memory.

This is relevant to phobia, where the phobia can be in remission (the sight of a snake no longer elicits paralyzing anxiety) and then the patient's mother dies and snakes regain their propensity for producing terror. Phobia is also a good way to illustrate the difference between cognitive memory of emotion and emotional memory. We aren't born with phobia. Somehow they are acquired through experience and stored in the brain as links between stimuli (like snakes or heights) and fear responses. Once a phobia is successfully treated (the snake no longer elicits overt fear responses) the patient still has the explicit memory of having had the snake phobia. In other words, the therapy inhibited the amygdala's pathological response to the sight of snakes, but the therapy didn't eliminate the temporal lobe memory system's memory of having had a snake phobia.

JB: How can you talk about unconscious emotional memories? Why is this different than inventing a concept like "repressed memories?"

LEDOUX: I'm not talking about memories that have been repressed, they're just not consciously available. I'll give you a simple example. Patients who have damage to the temporal lobe memory system are unable to remember what happened to them five minutes ago. If you take those patients and give them a sound, pair it with a shock, and you later give them the sound again, their autonomic nervous system responds, but they have no conscious memory of the experience that led to that. The memory is in the brain, having an effect on systems that we can measure, including autonomic and behavioral systems, but the patient has no conscious memory of it. In everyday usage, the term memory usually refers to conscious memory, but as scientists we use the term in a more general sense to mean changes in the nervous system that reflect past experiences. By this definition, we can see all sorts of memories that have no conscious counterpart. This is the idea of implicit, or procedural memories that are in the brain's systems, but not reflected in consciousness.

There's a famous case from the early days of this century that beautifully illustrates this point. The patient had a pretty severe amnesia. Each day she had to be reintroduced to her doctor, as she didn't recognize him. One day the doctor put a tack in his hand, and he walked in and shook her hand. When their hands met, her finger was pricked. He then walked out of the room, walked back in, and asked whether she'd ever seen him before. She said she hadn't. But when he stuck out his hand to shake her's, she held back. Although we don't really know what was going on in her brain, it seems likely that the implicit memory that the handshake was dangerous was burned into her amygdala, and that allowed her to protect herself from getting stuck again. She knew this implicitly--but she couldn't tell you why because she couldn't remember the experience that led to it. The amygdala was forming its memories, but the temporal lobe memory system was not.

Normally, these systems work in parallel to give rise to our conscious memories about emotional experiences, and unconscious emotional memories. In this sense, emotional memories are by definition unconscious. But they aren't unconscious because they've been repressed. They're unconscious because they are not formed by the conscious memory system. The conscious memory system forms memories about emotions, but doesn't form the emotional memories that have direct access to emotional response systems.

JB: Where do you expect your research will take you?

LEDOUX: Right now my work is headed deeper and deeper into cellular-molecular events underlying how emotions are learned and stored. We are trying to understand as much as we can about how these memories are formed at the cellular level which has taken us into studies of synaptic plasticity, how changes happen at the level of individual synapses when this kind of learning takes place. We are asking questions about what neurotransmitters are involved and what sort of molecular changes take place to stabilize these memories over the long run. These studies are just beginning and they will take us well through into the next century. At the same time it's important not to lose sight of the fact that we're dealing with a psychological problem with important behavioral consequences. We need to study the behavior as well as the molecules. We try to work at all these levels; at the level of the behavioral system as well as the cellular and molecular systems.

JB: What has your work made explicit?

LEDOUX: There are ways that the brain can produce emotional responses in us that have very little to do with what we think we're dealing with or talking about or thinking about at the time. In other words, emotional reactions can be elicited independent of our conscious thought processes. For example, we've found pathways that take information into the amygdala without first going through the neocortex, which is where you need to process it in order to figure out exactly what it is and be conscious of it. So, emotions can be and, in fact, probably are mostly processed at an unconscious level. We become conscious and aware of all this after the fact. Conscious feelings of fear are thus not a necessary step in the link between a dangerous stimulus and emotional responses. We're probably not as in control of our emotions as we sometimes think we are, or wish to be.

Emotional reactions that occur in this quick and dirty way are really reactions that are important in survival situations. The advantage is that by allowing evolution to do the thinking for you at first, you basically buy the time that you need to think about the situation and do the most reasonable thing. For example, freezing is often the first thing people and other animals do when sudden danger appears. Predators respond to movement, so freezing is overall probably the best single thing to do first, at least it was for our distant ancestors. If they had to think about what to do first, they'd have been so caught up in the thought process they'd probably fidget around and then get eaten.

The Atlanta Olympic Bombing is a nice illustration of this. The bomb goes off and everyone hunches over in the freezing posture for a couple of seconds, and then they take off running. You can almost see the cognitive gears turning while they're freezing. Although we're not in direct control of these rapid fire unconscious emotional responses, I don't think that they are necessarily going to be things that someone can use as a legal defense, for example, for having carried out a very detailed crime, a murder or a rape or something of that nature. These quick and dirty systems produce relatively simple rapid responses (like freezing) in life-threatening situations. They're more likely to be used by the victim than by the perpetrator.

JB: What about therapy?

LEDOUX: The connectivity of the amygdala with the neo-cortex is not symmetrical. The amygdala projects back to the neo-cortex in a much stronger sense than the neo-cortex projects to the amygdala. David Amaral has made this point from studies of primate brains. The implication is that the ability of the amygdala to control the cortex is greater than the ability of the cortex to control the amygdala. And this may explain why it's so hard for us to will away anxiety; emotions, once they're set into play, are very difficult to turn off. Hormones and other long-acting substances are released in the body during emotions. These return to the brain and tend to lock you into the state you're in at the time. Once you're in that state it's very difficult for the cortex to find a way of working its way down to the amygdala and shutting it off.

This is why therapy is probably such a long and difficult process, because the neocortex is using imperfect channels of communication to try and grab hold of the amygdala and control it. It's like trying to find your way from New York to Boston by way of country roads rather than superhighways. The amygdala can control the neocortex very easily, because all it has to do is arouse lots of areas in a very non specific way. But for the cortex to then turn all of that off is a very difficult job. The evolution of the brain is at a point where we don't have the connectivity that would be necessary for cognitive systems to more efficiently control our emotions. But it's not clear to me that would necessarily be a good thing, because Mr. Spock is not necessarily an ideal kind of human that we'd like to become.

Designer drugs could be really practical, and I'm surprised that the drug companies are not knocking at my door to find out how to make drugs that could do more specific things than the drugs that are available. We know the circuit through which fear is elicited, and we know the specific points in that circuit that are involved. As we begin to identify the neurotransmitters that are involved in the elicitation of fear, it seems that we could probably come up with a chemical profile of fear in the amygdala. A particular drug could be developed that attacks that profile. For example, if you take Valium, it might make you sleepy and reduce your sex drive in addition to making you less anxious because it affects GABA transmission throughout your entire brain. But if you could develop a Valium that only acted in the amygdala, then you would have a drug that works at the particular sites involved in fear. That's pie in the sky at this point, but it's something they should be thinking about.

JB: How is your work being received today?

LEDOUX: I've been amazed that almost all areas of psychology have not only been sympathetic, but are reaching out and trying to find out as much about my work and the work of people like me. It's really surprising that this extends into psychoanalysis as well. I have received a number of invitations to speak to psychoanalytic groups and to attend meetings to try and understand how concepts about the emotional brain could help them understand that psychoanalysis and might take them into the 21st century. Psychoanalysis is in relatively bad shape right now. Young psychiatrists are not going into the field, so the elders are trying to figure out a way to make the field more appealing. I think they see neuroscience as a possible bridge.

JB: Who else?

LEDOUX: Developmental psychologists and social psychologists have been very open to the work on the emotional brain. The developmental psychologists are interested because of the early development of the amygdala before conscious memories kick into play. Social psychologists are interested because the amygdala seems to do its work unconsciously. There's a whole industry of social psychology dealing with unconscious emotional perception, how you use subtle cues that are given off even when you don't know you're giving them off, and how these are picked up by your unconscious mind, so your unconscious mind and my unconscious mind are talking back and forth to each other without our conscious minds knowing anything about it. They're interested in all this work on the amygdala and the possibility that it's an unconscious emotional processor.

Cognitive scientists previously banned emotion from their field, but are beginning to realize that they don't really have a science of mind as such, but instead a science of a part of the mind. They now want to bring emotion and cognition back together, and that's a good thing. Lots of AI modeling of emotion, and some connectionist modeling, is also going on.

JB: How does a Dan Dennett or a Steve Pinker relate to your work?

LEDOUX: In Pinker's recent talk, "Organs of Computation", I noticed that he did talk about emotions; he was talking about how passions fit into the mind. I think we'd agree on a lot, say about the evolutionary aspects of emotions and their unconscious nature as processes in the mind. I'm more interested in how evolution has kept emotional systems the same in man and other animals, whereas he seems to be more interested in what makes the human brain capacity for language a unique function not present in other species. Where we'd probably differ the most is in how we approach the problem. I want to do it from the brain, so that I know that my theories are tied to the hardware in a biologically plausible way, but I think he wants to do it without depending on the brain. I think both approaches have their strengths and weaknesses, and both are needed.

What I talk about is compatible with Dennett's views in some ways, because I'm dealing with emotions not as conscious feelings but instead as computational functions of the nervous system. The way I talk about emotions puts them at the level of what some people in cognitive science call the sub-symbolic level. In this sense, emotional systems are among many systems that operate in parallel at an unconscious level. In Dennett's view, there's a symbolic system sitting on top of all these sub-symbolic systems. This is where consciousness comes from, loosely speaking.

The symbolic system has some access to the outputs of the unconscious emotion systems as well as all the other perceptual and other subsymbolic systems and the one that grabs hold of the symbolic system at the moment is what we are conscious of. So we can be conscious of emotional events or mundane events. So I'd say there's not a special system for emotional experience separate from other kinds of conscious experiences. There's one mechanism of consciousness and it can be occupied with mundane events or highly charged emotions. I think my view of the mind is not incompatible with Dennett's. That's not to say that I agree that Dennett's explained consciousness. Instead, I agree that most of the mind doesn't work through consciousness.

JB: How would you describe yourself?

LEDOUX: I was recently called a radical behaviorist disguised as a neuro-scientist. I thought that was an interesting twist. It's true that I try to deal with emotions as unconscious processes as far as I can, but I don't deny the importance of consciousness. I just think that it's gotten in the way in the study of emotions. I'm not really a radical behaviorist. I realize that I am simplifying and probably oversimplifying emotion to study it the way I do, but I hope to build up to complex issues from a solid understanding of the simple stuff rather than have to reach down from confusion to try and account for the simpler processes.

JB: Can you say more regarding the difference between repressed memory and a sub-symbolic system?

LEDOUX: There are several things that are important to pull into this topic. One is the newly emerging data on the effects of stress on memory. The basic finding is that in periods of intense stress the explicit memory functions of the temporal lobe memory system can break down. Stress is usually defined physiologically by the amount of so-called stress hormones from the adrenal gland. When this stuff is released it floats around in your blood stream and gets into the brain. The hippocampus and amygdala are targets. These hormones adversely affect the hippocampus. They make it very difficult, for example, to induce long-term potentiation in the hippocampus, so the hippocampus begins to shut down physiologically. Also, spatial learning is interfered with. If the stress continues, dendrites begin to shrivel up, and if the stress continues even longer the cells die and the hippocampus itself begins to shrink in size. Bruce McEwen and Robert Sapolsky have done a lot of this work on stress and the hippocampus. There have also been some recent studies of patients with post traumatic stress disorder, Vietnam vets, for example, who have a greatly reduced volume of their hippocampus, and they have all these memory disturbances.

In contrast, stress seems to potentiate the amygdala. Stress will make the amygdala do what it's doing but even better. Let's say you get mugged or raped. The stress system releases all of its hormones (probably as a result of the amygdala detecting the threat and activating the stress hormone system). The hormones get into the brain and the hippocampus is adversely affected to the point where it can't consciously form a memory of this experience. But your amygdala is potentiated, so it's not only forming a memory unconsciously, but it's doing it better than before. So the exact conditions that can lead to hippocampal memory impairment (an inability to form a conscious memory) can lead to a facilitation of unconscious emotional memories through the amygdala.

Now you're a person who's walking down the street with no conscious memory of having been traumatized. There are witnesses that tell you it happened but you deny it--there is in fact often denial in situations like this. You carry unconscious traumatic memories but no conscious understanding of what happened. I don't know that something like this actually happens, but the biology is very plausible. It's totally conceivable that someone can be traumatized in this way and have no conscious memory of it. I believe that. And it fits with all the science that we have about all of this.

Now the next question is, can you then, through psychological tricks, comforting, therapy, whatever, bring these memories back in a person who never had them? And the answer to that is a clear No. It's not possible to take a memory that was not coded through the hippocampus and turn it into a hippocampal memory. So the amygdala has its memory; it doesn't then share it with the hippocampus, because they do things differently. The amygdala does its business, the hippocampus does its business. They communicate with each other, but their coding and representation is different. So you can't just get information out of the amygdala and turn it into content that the hippocampus can read. I think this kind of work tells us a lot about the psychology of memory and emotion, not just the biological details.

JB: What do you want to accomplish in the next five years?

LEDOUX: I want to understand several aspects of emotion that we have very poor understanding of now. The first part we're beginning to understand pretty well, which is how the initial aspect of an emotional reaction is elicited. In other words, how you jump back from a bus as it's approaching, and only afterwards consciously realize that you've jumped back, and only then feel afraid. We understand that reactive system in pretty good detail. But what we don't understand is the system for emotional action. How do you voluntarily make decisions and control your emotional behavior once you've reacted in this unconscious way. What circuits in the brain are involved in what psychologists call coping, the cognitive and behavioral effects that follow the arousal of emotion and one's attempts to deal with emotional arousal? Probably the basal ganglia and cortex area involved. The question of what makes us emotional actors as well as reactors really interests me.

That takes us to another issue, which is where do conscious feelings come into emotions? How do we get a deeper understanding of emotional feelings? We all want to know where feelings come from and how they work. So much of the work in the past started with feelings and tried to back into the problem and didn't get anywhere, which is why I start at the bottom and work up to feelings. I also want to know a lot more about emotional memory. Most of the things that make us emotional are learned through experience. So a key part of an emotion system is how it learns and stores information. Overall, I would summarize all this by simply saying I want to try to understand more about cognitive-emotional interactions. We have to put emotion back into the brain and integrate it with cognitive systems. We shouldn't study emotion or cognition in isolation, but should study both as aspects of the mind in its brain.