Scientist Stories: Richard Axel, Representations of the Olfactory World in the Brain

I would like to welcome all of you to this evening's Stavros Niarchos Foundation Brain Insight Lecture. My name is Tom Jessell. I'm one of the co-directors of the Zuckerman Institute and the lucky beneficiaries of the Jerome L. Green Science Center, which will open soon, sometime next year. And the purpose for which we gather tonight is to listen to a lecture by Columbia's own Richard Axel. Before I talk about Richard, I want to say a few thank-yous, both to the Green Foundation, to Mort Zuckerman, for their collective generosity, and in particular for sponsoring this lecture series, as well as the Teacher-Scholar Program. Andreas Dracopoulos, the other members, Stelios Basileakis, the other members of the Niarchos Foundation Board, for their continuing support of these efforts at Columbia to explore the workings of brain and mind. How can one introduce Richard Axel? The issue here, I think, is not so much how to do Richard justice, but to decide just which elements of his multifaceted character to focus on this evening. And from the word go, Richard has been all about science. Science. And so in order to illustrate Richard's presence in the world scientific, I want to go back to the first, first author paper that Richard ever published and just draw a few inferences from that scientific snippet. So here's the snippet, and the first inference that I want to draw from this You can read what the title is. You can see the date. The first thing I want to draw is that Richard, from the word go, has been and always will be a molecular biologist. So beware those of the wolf in sheep's clothing if he has the affect of a systems neuroscientist about him, because he's a molecular biologist at core. And what I mean by that is he wants clean answers to well-defined questions. And so We're gonna hear tonight from him in a different arena than what you're seeing on the board, his ability to define problems in a precise way and obtain clean solutions to that. The second thing I want to point out is Richard as Columbian. Richard came to Columbia as an undergraduate in 1963. So he has been more or less without interruption under the Columbia mask for some 52 years. And this is a remarkable achievement. Richard breathes, lives, exists in this Columbian world. The third thing I want to draw from this is Richard as precocious talent. So this paper, which you see before you, was published in 1967. When Richard was a sophomore undergraduate working in Bernie Weinstein's lab. 2 years before that, as a teenager, he'd been asked by Bernie to present the first results of this study to a meeting in Chicago. So Richard just exploded into the scientific world in a remarkable, remarkable way. And the fourth thing I want to extract from what is before you now is the idea that Richard manages to elicit enthusiasm and support and endorsement from people worldwide. The paper that you see before you was contributed by the Columbia grump Irwin Chargaff. And for those of you who might not know who Chargaff was, he first defined and observed the unity of purine and pyrimidine bases in DNA, which was an important clue that let Jim Watson and Francis Crick decipher the structure of DNA. The second thing that Chargaff is known for is more the grump. So when he famously uttered, "How late in the day has it become?" when giant shadows are cast by such pygmies. This was in reference to Crick and Watson for their DNA structure. So Richard's scientific accomplishments, I would hold, stand tall in the face of such Chagaffian scrutiny. And anyone who knows Richard maybe will appreciate that in— if you've ever been the object of Richard's lacerating prose and opinion, you will see that there's a little bit of Chagaff in Richard. So this Columbia influence has writ large. Richard's role as a neuroscientist emerged a little bit later in the late '70s and '80s through a momentous encounter with the other person on this image, Eric Kandel. And early on, they performed an elegant series of papers that described the relationship between patterns of gene expression and individual animal behaviors that still sets the standard for this molecular dissection of behavior. This was the beginning, the real beginning of neuroscience at Columbia, so that now 30 years on, everything good that is happening about Colombia came from the fusion of intellect and standards set by these two. So 30 years on, we have now the Zuckerman Institute and the Jerome L. Green Science Center. All of that can be traced directly back to this first encounter between Axel and Kandel. And I would argue that Richard continues to exert an influence, remarkable influence on all things bioscientific that relate to Columbia. The other big deal at Columbia at the moment is this initiative in precision medicine, which is led by Tom Maniatis. And Richard has been colleagues and friends of Tom for since time immemorial, and I think was largely instrumental in persuading Tom to leave the close-knit world of Harvard and venture up to New York or down to New York some 6 years ago. And so what we have now is that— and on the left, we're looking at a much more youthful Tom Maniatis and a somewhat more youthful Richard Axel, together with this idea that precision medicine is together with Mind, Brain, and Behavior, one of the key efforts in biomedical sciences that the university is pursuing. So Richard, the seismic epicenter of things neuroscience at Columbia, we are delighted that you've agreed to give this lecture, and we'd like you to come up now and tell us about sense and sensibility. Richard Axel.

Photo of Don Green with the mayor and the president responsible— whose generosity is responsible for the Jerome Green Science Center. And we also have tonight another generous donor and a new and very good friend, Mort Zuckerman, whose generosity has allowed the formation within the center of the Mort Zuckerman Mind-Brain Behavior Institute. And here is Mort. Announcing his generous gift, and here is Mort after he signed the check. So it's actually Mort in a movie. He's an actor as well, as you know, and Thank you very much. It really is a pleasure. Am I wired or do I have to— Can you hear me? This beautiful painting by the German Renaissance artist Lucas Cranach depicts the Earth's first lovers with a snake. 500 years later, the master of the zip, Barnett Newman, painted the very same image. Barnett Newman's Adam and Eve has Adam and Eve reflected by lines, colors, forms that bear no resemblance to reality. This is an abstraction, and meaning must be imposed upon this image by prior knowledge or experience. This is precisely what the brain must do. Perception, emotion, cognition, memory, action, all in the— Okay, there you go. Where was I? Perception, memory, thought, cognition, all must be represented in a brain by the activity of large ensembles of neurons. And these ensembles, these neural representations, are reflecting thought. And so they too must be abstractions. And meaning must be imposed upon these abstract brain representations either by evolution or by experience. Let's consider the problem for sensory perception. How is it that the richness of the world, the frequent— which consists of discrete physical parameters, wavelengths of light in vision, frequencies of sound, of sound in hearing, the chemicals of smell and taste, all can be recognized and represented in a brain which simply consists of one thing: neurons. And neurons can only vary in two dimensions. They can only change their firing in time and space. This, I argue, is an astonishing problem. Now, this idea of an abstract representation of the sensory world in the brain immediately implies to me that different species, and at the extreme, different individuals within a species, will represent the world in different ways. Perception will depend upon the ability of an organism's sense organs to recognize information in the physical world and then transmit that information to the brain where it must create an internal representation of the physical world. This is what I call and is called a bottom-up process. It's a process that is often incomplete, and it comes without meaning, and meaning must be imposed by other brain structures by a process known as top-down processing. So meaning is imposed by brain structures upon this initial representation, and this meaning is imposed by bringing a stored record—experience, emotion, expectation—upon a neural process as what I would call a top-down process. Process. So let's consider the bottom-up process first. How sensory information is recognized in the world. Now there are many ways for an organism to probe the external world. Some smell it, others listen to it, many see it. Each species therefore lives in its own unique sensory world to which other species may be partially or totally unaware. Bees, for example, can recognize ultraviolet light and polarized light, which we cannot see. But in contrast, bees cannot recognize red light. The pigments for the recognition of red light are absent. This, for example, is how a human would view Edvard Munch's Scream, a painting that elicits enormous tension. And as Munch said, "And the sky turned a blood red, and I heard a great cry." eminating from nature. We are entranced. The painting is iconic. This is how a honeybee would see the very same painting. It cannot absorb red pigments, and the anxiety lines of the sky no longer entrance. A honeybee is simply not entranced. Left by Monk's scream. Another example, bats. Bats extract remarkably detailed information about the position, size, and velocity of predators and prey by virtue of employing a rather unique auditory mechanism, biosonar, or echo location. And so bats can indeed, through echolocation, identify predators and prey in the dark. Really quite remarkable. So these 6th, 7th, 8th senses which humans do not possess illustrate quite clearly that each species perceives but a meager image of the richness of the outside world. These differences in one's— in a species' sense organ and in the way this information is processed may separate the perceptual environment of different species or different individuals as radically as if the stimuli they sense were coming from different worlds. The brain functions then not by recording an exact image, but by creating its own selective and unique picture. Our perceptions, I argue, are not direct recordings of the world around us, Colors, tones, tastes, smells are active constructs created by our brains out of sensory experience. They do not exist as such outside of the brain. So let's consider how the brain, which consists solely of a collection of neurons, could represent olfactory sensory perception, this vast repertoire of odors, and then ask how these odors can elicit an appropriate behavior. And behavior comes in two forms: innate behaviors and learned behaviors. Olfactory sensory perception is initiated by an olfactory sensory epithelium consisting of neurons that reside in the posterior recess of your nose. And they're shown here. These neurons are rather simple sensory neurons, and they extend one process, a specialization of a dendrite, out to the surface of your nose where they are in direct contact with the air. Odorous molecules in air then bind to receptor molecules which reside on the surface of these processes. And the energy of binding of odors is translated into alterations in the electrical properties of these neurons. And these alterations are then transmitted through a second process which extends through the skull into the first relay station of the brain, a structure known as the olfactory bulb. And so there is a singular cell, a direct connection between the outside world and your brain. And that connection mediates signals the recognition of odors. Now, decades ago, a fellow in my laboratory, Linda Buck, identified the genes encoding the receptors that are capable of binding to odorants. And we experienced two surprises. The first is that there are about 1,000 genes encoding odorant receptors, whereas there are only 3 genes encoding the receptors allowing one to visualize the thousands and perhaps 1 million discrete colors in the eye. And as my colleague Charles has shown, there are only 29 genes in the tongue responsible for recognizing taste. So obviously, this vast repertoire of genes accommodates a unique need for olfaction, perhaps reflects the fact that olfaction is indeed the primal sense by which organisms detect food, predators, and mates. And moreover, olfaction is clearly the— evolutionarily the most primitive of sensory modalities. Now, 1,000 genes dedicated to the recognition of odors is a lot of genes. If there are only about 20,000 genes encoded in your chromosome. This means that about 5%, 1 in 20, of each of your genes is dedicated to the recognition of odors. And this feature, that is an enormously large gene family, the largest gene family in the chromosome, is not only true in vertebrates, it's true as one moves through the evolutionary tree. So it is a principle. And as I will show you, the identification of these genes not only surprised in their number, but they surprised in their explanatory power, for they allowed us to move into the brain. If there are 1,000 genes, you can all reason that different odors will activate a different combination of receptors. And so the question as to how it is that you know what you are smelling can be broken down to a question of how it is that a brain could know which of the 1,000 receptors had been activated or occupied by an odor. But that turns out not to be a trivial problem. But it was reduced by virtue of a simple explanation. I like to use Magritte above my images. This is the olfactory epithelium. And what this slide shows is we stain for only one type of the 1,000 receptors And what this slide shows is that a given nerve cell, a given olfactory neuron experiencing the world makes only one kind of receptor. Now, if there are 1,000 receptors and a cell makes only one kind, then there are 1,000 different cells. And so the problem as to how the brain knows which receptors are activated by a given odor reduces to a far simpler problem. How does the brain know which cells have been activated? These cells have to be identifiable in the brain. And in a set of experiments performed in the lab by Bob Vassar and then Peter Moombarts, We were able to discern this, and that is what we were— what we simply did was to use genetic chicanery so that we could decorate in blue cells making receptor 111, and we could also decorate cells making all of the other receptors. And by decorating those cells, we could trace the path of the projections of those neurons into the brain. And what we saw was really quite neat. So this is the nose. And you can see about 1,000th of the neurons are decorated blue. They make receptor M12. A second mouse, and we see the nose with the neurons making receptor P2. And what we observe is that the neurons in the epithelium appear to be distributed throughout the nose. But as their processes extend through the skull— this is the skull— into the first relay station in the brain, They converge on a fixed point. That point, or that locus, is called a glomerulus. And if each neuron that makes a different receptor converges on a different point, then what you have in the brain is an anatomic map. You have a map in the brain which defines each of the 1,000 receptors. And it should immediately follow then that a given odor will bind a combination of receptors different from the combination of a second odor, and that in turn should activate a different combination of points. So cherry, for example, the smell of cherry might activate this combinatorial of points, whereas lemon would activate this combinatorial of points. And this very simple model would argue that odor quality would be determined by distinct spatial patterns of neural activity in this very early brain structure. The olfactory bulb. We together? This is true. One can use one of the new and important technologies, that is optical imaging through a sophisticated sophisticated two-photon microscope and actually look at the activity of brain regions with a resolution about 1,000 times greater than functional magnetic resonance imaging that many of you are aware of, and with a sensitivity about 1,000 times greater. So we really can look at a subcellular level at neural activity in the brain. And that's what you see here. And we're looking in the olfactory bulb in a brain— in a mouse exposed to a fruit odor and a mouse exposed to fox urine. Mice do not, as we will see, like foxglove. So an appetitive odor and an aversive odor. And these patterns are so different, and they are different for every odor. And so I, an experienced observer, can look down upon this very defined pattern of neural activity in the olfactory bulb of a mouse, and I, because of my experience, can tell you what odor the mouse has encountered in nature simply on the basis of the pattern of neural activity. It's neat. Come on, it's neat. And— but there's a problem. There's a serious problem. While I can look down on the mouse olfactory bulb, the mouse brain, with a two-photon microscope and a pair of eyes and my brain and discern the nature of an odor an organism has encountered by virtue of the spatial pattern of neural activity, a mouse does not have a two-photon a photon microscope in its brain. And so how does a mouse look down on this pattern of neural activity and interpret it meaningfully to elicit an appropriate behavior? Problem. And so the way you address this problem is to move up in the brain and ask what these loci are connected to and how those connections might lead to reading, to interpretation. So let's begin to move upward in the brain. This very beautiful drawing is by a brilliant Spanish neuroscientist, Ramón y Cajal, well over 100 years ago. And he saw clearly what one had to do. He drew out the olfactory bulb. Here are sensory neurons arborizing in the olfactory bulb at these loci known as glomeruli. And each of these loci are connected to a set of neurons, projection neurons, that project into high brain. So we asked, where is this information that I can look at and interpret projected to in the brain? And we did this in a very simple way. We We genetically modified a mouse so that every one of these loci— so this is the olfactory bulb, this is the whole brain stretched, and each of these is a locus that receives input from the nose. And what a student in the lab, Dara Sosulsky, along with Bob Datta— I'm going to mention a number of my students' names because I'm a religious man. And it is written in the Talmud that the work of the righteous is done by others. I do nothing. And so what Dara did was to inject only one of these loci loci and follow their projections to the brain. And these projections from only 1 of 1,000 of these loci are very rich and very complicated. This information actually pentafurcates. It goes to 5 brain regions, but I want to focus on 2. One, a limbic structure, part of the emotive brain, a structure known as cortical amygdala, and a higher brain structure, a sensory cortex known as piriform cortex. And what we observe is that projections from a single point in the bulb to cortical amygdala are such that each glomerulus, each locus projects to a spatially segregated and stereotyped locus within the cortical amygdala. So this spatial pattern, this insular segregated spatial pattern that encodes odor identity in the bulb is, through transformations, also projected to cortical amygdala. These projections are fixed in space and identical in all individuals in a species. And when one sees these highly determined projections, one intuits that these projections might be responsible for encoding innate responses to odors. In sharp contrast, piriform cortex is far more distributive in the receipt of its information. This singular locus here projects in a distributive and unstructured fashion to piriform cortex. And this projection, as we'll see later, presages a participation not in innate behavior, but rather in learned behavior. And so we can test this. And we test it in very simple ways. So we make the argument that cortical amygdala is necessary for innate responses to odors. So as I said earlier, mice don't like fox urine. There is a major chemical component which is volatile and odorous —called TMT—and if you expose a mouse to that odor, it's out of here. Very robust behavioral response. So we can simply ask, using modern technologies, whether indeed this structure, this limbic structure in emotive brain, is responsible for innate behavior. And so we use something that has gotten a great deal of press and is extremely valuable, uh, a procedure developed, uh, at Stanford by, um, uh, Karl Deisseroth and Ed Boyden, in which you can genetically modify specific neuronal subsets in the brain in such a way that the activity of these neurons can be inhibited by light. Or alternatively, the activity of these neurons can be activated by light. And it's an extremely valuable tool. Let me very briefly show you how it works. You genetically modify a mouse, you introduce into a brain region an optical fiber— you see that here— And light exposure will, at your moment of choosing, alter the activity of neurons. If you do that here, if you put an inhibit— a light-activated inhibitor into these neurons and now light comes in, these neurons completely shut down. They are silenced. And so this turns out to be an extremely valuable new technology, and it lets us address this question very simply. What we do is we take a brain and we genetically modify it so that this structure, the olfactory bulb that I showed you projects to cortical amygdala, is expressing a light-activated silencer. Or you put a light-activated silencer right into cortical amygdala, and you develop an assay for avoidance. So here we have a very large chamber which has 4 quadrants with different odors, and in this quadrant you put fox urine. And you can see that compared to the other quadrants, this mouse doesn't go near this quadrant. But if I do the same experiment silencing cortical amygdala, then it experiences this quadrant as frequently as if this structure did not exist. These are simple experiments, experiments performed by Corey Root, and what they say straightforwardly is this this limbic structure, the cortical amygdala, is necessary for the response to innate odors. And there are innate attractive odors— peanut oil, rose oil— and you can do the same experiment with those odors and you see precisely the same thing, that if you silence cortical amygdala, the mouse will no longer show any apparent innate response to volatile odors. And we can expand this trickery in even more refined ways. I don't want to technically bore you, but the power of this technology is providing in many different arenas significant insight. You can alter this technology where you now introduce a light-activated molecule, a molecule that activates neurons only in those neurons in the brain that are activated by an odor like TMT. So I can put a light-activated— activatable molecule only in the subpopulation a population of neurons activated by TMT. And then I can shine light on this animal with no TMT present and ask, does the animal exhibit aversive behavior? Does it run away? And it does. So this experiment, taken together with the previous experiment, says that indeed this structure, the cortical amygdala, is both necessary and sufficient to elicit— for an innate response to odors. Moreover, as I argued earlier, there is spatial segregation here —and there is spatial segregation here. So those glomeruli that appear to be activated by aversive innate odors tend to project over here, and those activating loci responsive to appetitive odors activate here. And you can, in fact, test this by artifactually activating here, and you will get aversion. And if you artifactually activate here, you get appetitive behavior. So it seems then that this limbic structure is involved in innate behavior. But most odors do not elicit innate behaviors.— in fact, there are very few, if any, odors in humans that elicit an innate response. You can think of them, but they're very rare. And so what about learned odors? Things get a bit more complex. Now, I showed you earlier that this olfactory bulb projects to two structures— to five structures, but we're going to talk about two— the limbic structure, the cortical amygdala involved in innate behavior, and something called piriform or primary sensory cortex. And you're looking at the processes that come from a very small number of neurons in the bulb that receive input from from only one kind of receptor. So multiply these processes by 1,000 and you get a look at the complexity of sensory cortex. Sensory cortex discards that spatial order that I described to you in the bulb and that I described to you in cortical amygdala. You lose that insular segregation of the bulb. Moreover, this anatomic lack of structure is mirrored by a functional lack of structure. So in the previous experiments, we could show that TMT activates a discrete region in cortical amygdala different from peanut oil,— whereas in piriform cortex, every odor activates the whole cortex and is intermingled. The neurons responsive to rose are intermingled with the neurons responsive to alcohols. There is no apparent order. And these observations, taken together with physiologic measurements, have revealed something quite informative. And that is the input to these neurons in piriform cortex is random. That is, a given neuron gets input from about 100 of those loci in the bulb.. And a given neuron will receive input from a random 100 loci. So that's very different than this innate pathway. It's very different and it's very— it was very important to us because if this structure is built by unstructured distributed, seemingly random projections, what that means is that in the same odor in different people will activate different neurons. And if that is true, then this representation of an odor has no meaning whatsoever. Whatsoever. None. Now we're starting to go back to this idea of an abstract representation upon which meaning must be imposed by prior knowledge or experience. And this lack of— this distributive unstructured representation with no meaning recalled to us something that all of us knew in the field, And that is if you expose a mouse to an odor, to most odors that it has never encountered before, it will have absolutely no behavioral response. None. So any given odor is likely to activate a pattern like this and there will be no behavioral response. On the other hand, if I now take that odor and I associate it with an experience— so here I have an odor representation in piriform. First exposure, no behavior. If, however, I associate that odor with water or food, in a thirsty mouse, and now I come back a day later and I just give it the odor alone, in the absence of water, it will begin to lick interminably. And so this odor, which is called a conditioned stimulus, at the outset elicits no behavior, But if I now pair it, I give it pre-knowledge, I give it an experiential scenario, now that odor alone elicits a behavior absent reward. And the same is true for pairing an odor with an aversive stimulus or pairing an odor with a social or sexual stimulus. Sexual stimulus. So you can get a mouse to expect a female by pairing an odor with a female, then coming back with that odor, and the male begins to exhibit anticipatory delight. Hence perfumes. Now, We can play games with this and we can create what appear to be random ensembles of neurons that are activatable at will by us using this optogenetic trickery. And so if we make a random ensemble and shine light on olfactory cortex, we get no behavior. If we take this random exogenously created ensemble, has nothing to do with odor, created in the lab, and pair light with a water reward, the next time I come back in the absence of odor with light alone and activate these neurons, the animal licks, which provides real support for the idea that there is a distributive and abstract representation. That representation can differ in different individuals in the same species. But when you associate that representation with something meaningful to the individual, only then does that representation take on behavioral significance. I would argue that this is true not only for olfaction, but for vision, hearing, on out. OK. If this is indeed the case, we can now turn to an even more difficult question. We've generated a distributive and abstract representation, and association has imposed meaning on that representation. How is that meaning imposed? What is the circuit? What is the mechanism by which water, shock, social interactions, can lead to that behavioral output. I'll try and move rather quickly now, um, because now you've got it. There is a region of the brain, also in the limbic system, called basolateral amygdala. Many investigators— Joe Ledoux, Dan Saltzman— many investigators have demonstrated that this structure receives input from all of the sensory modalities, and this structure is responsive to many emotive cues, and the output of this structure can lead to defined behaviors. And so what Felicity Gore and Ed Schwartz in the lab were able to do with Dan Saltzman is to demonstrate that if I shock a mouse, I will activate a subset of cells in basolateral amygdala, and that allows me to decorate those cells, and I can demonstrate that activating those cells will— sorry, activating those cells will lead to aversive behavior. I've also just demonstrated to you that there is a representation of activity in response to odors that is specific for a given odor. What Felicity was able to demonstrate is that there indeed is a connection between this representation and that representation. But that connection, in response to a naive odor absent experience, is weak. If, on the other hand, I pair the activation of this representation with a shock, water, food, social interaction, this interaction strengthens, and by virtue of the strengthening, this representation can now elicit an output. And so we have now moved this circuit from primary cortex to yet an additional a structure upon which, in which, meaning is imposed on this representation. I won't go through this. Now, I've outlined for you then neural pathways that seem to be necessary and sufficient sufficient for behavioral response to the small number of odors that elicit an innate response. And I've told you, I've outlined a neural pathway that uses a very different conceptual logic, which takes an abstract representation and imposes meaning upon it. And then that meaning is imposed on that representation in amygdaloid structures. So I can elicit an innate response, I can teach a mouse to respond appropriately to odors, and I've done this with a couple of small brain regions. So why do we have such a big brain? What do we do with it? I've accommodated the kinds of behaviors we wish to study. The human brain has 86 billion neurons. And so what we are studying, it turns out, is a very simple, elemental form of of sensory learning. And I want to give you an example of that. High brain, higher brain, prefrontal cortex, thought to be responsible for cognition and executive function, receives a very strong input from piriform cortex. If I use optogenetic tinkering to silence prefrontal cortex, every behavior I've just described to you is left intact. What do I need this for? Well, it turns out you need it because you are doing far more sophisticated things, and we're beginning to be able to tease away at the sophistication of the multiple additional layers of cortex to which this primary olfactory cortex projects. And there are many such projections, one of which is prefrontal cortex. A very simple experiment that Christian Boboile did was simply to use optogenetic silencing to shut down prefrontal cortex, as I told you a moment ago, if I shut down prefrontal cortex, simple associative olfactory learning remains intact. Now, if however I ask the mouse to think, to modify, to update what it has learned, something I find very difficult to do myself, So I entrain an animal such that odor A is rewarded and odor B is not rewarded, and I silence prefrontal cortex, the animal will still be able to recognize these meanings and perform this task. If, however, I now do something called reversal learning, that is, I update, I change this information so that now odor A in the same mouse is not rewarded and odor B is rewarded. To do this, you need prefrontal cortex. So you can be— begin to see what you're doing is building out Hierarchical brain structures. You're building out hierarchical brain structures, each of which is contributing new and additional information to what has been learned by more primitive brain structures. And without getting too complex, what this allows you to do is to retain that which you have learned, even though the new knowledge countervenes the implications of what you have learned. That is, you do not want to discard learning. You do not want to unlearn. You want to add additional information as one moves hierarchically up the brain structure. And the higher those brain structures, the more dominance they may have over lower structures. The higher the organism evolutionarily, the more recent the organism, the greater these higher-order brain structures. Okay. So I've tried to describe experiments that reinforce the idea I discussed earlier, the using olfactory sensory information. Olfactory sensory information enters the brain. It enters the brain without meaning. By a process that I described as bottom-up. And it generates an abstract representation. This bottom-up process results in an incomplete and a selective image of physical reality, an image without meaning. This image must then be completed by other brain structures. Some higher, some lower, as I've just described to you with respect to the amygdala and prefrontal cortex. And so this image is completed by using information from these other brain structures by what I called earlier a top-down down process. And this process brings expectation, experience, pre-knowledge, emotion to our perceptions. In fact, many of our perceptions would have no meaning without them. If this is true, however, then perception is only a hypothesis, a best guess. That only asymptotically approaches reality. It's a guess. And if this is indeed correct, then our assumptions about what we perceive from the world are based in part from input through our sense organs and in part from the brain's stored record, a record nourished by our experience. Now, while I've tried to describe this to you in the olfactory system, let me provide for you perhaps more vivid examples from both the visual and auditory systems. If perception is really just an assumption, a best guess, then you can get it wrong. Consider this classic example, the Kanizsa triangle. All of you, I think, see an equilateral triangle. Keep looking. See it? Your brain has essentially connected the 3 Pac-Men to create an equilateral triangle, but those sides do not exist. They exist in your brain. Your brain is using pre-knowledge, expectation, preconceived notions of the visual world to create what is not. This is a visual illusion. Now, I was recently exposed— we're almost there— I was recently exposed to an even more compelling example when I learned of the dramatic improvements in hearing that can be accommodated by the new cochlear transplants. And the new transplants try and recapitulate the cochlea by adding multiple channels to the implant. And I want you to try and listen to this. This and see how quickly you can begin to comprehend the noise.

What kind of bait do you use to catch salmon? What kind of bait do you use to catch salmon? What kind of bait do you use to catch salmon? What kind of bait do you use to catch salmon?

So it took a while for you to hear it. Now listen.

You're experienced. What kind of bait do you use to catch salmon?

Same sound. You are hearing—

What kind of bait do you use to catch salmon? What kind of bait do you use to catch salmon?

You have heard the same set of sounds twice, and they have sounded extremely differently after having heard it once. The very same sounds. It seems to me that this is a really clear example of how profoundly experience has altered what you have heard. You have listened to the same sound and heard it differently differently after only one exposure. Very dramatic to me. This is a top-down process. And so what this shows is that experience dramatically shapes how you perceive the world. Now, allow me to conclude with a very brief discussion of olfactory memory. Smell, it has been argued, is the most evocative sense. It is that sense that brings forth memory and associations with a richness that is not elicited by other sensory stimuli, vision or hearing. Consider, for example, Marcel Proust, for whom the smell of a madeleine Combray brought forth 7 volumes of remembrances of things past. Why is smell so evocative? For the other senses, for hearing, for example, I would argue that it is possible to communicate sound to others. Either by speaking, by singing. Moreover, music can be written, and so music is communicable to the world. The same is true for vision. A visual image is very easy to describe, to communicate. Tall, thin, blue eyes. Got it. Moreover, we have photographs. There's a structure to sound and a structure to visual images that allows them to be readily communicated. There is, however, no easily discernible structure to olfaction, and it can only be described in the context of previous experience through association. Consider that I ask you to describe to someone the smell of an orange without ever mentioning an orange. They will have no idea what you are talking about. None. It is not possible. And so odor, I argue, can be recreated and communicated only by association to past experiences. And nowhere has this been more beautifully stated than in Proust's Remembrances of Things Past, when he says, "But from a long distant past, nothing subsists." After the people are dead, after the things are broken and scattered, taste and smell alone, more fragile but more enduring, more unsubstantial, more persistent, more faithful, remain poised a long time, like souls remembering, waiting, hoping amidst the ruins of all the rest. And bear unflinchingly in the tiny and almost impalpable drop of their essence the vast structure of recollection.

Thank you. So we have time for questions. There are people who will be providing you with microphones, so wait till you get a mic to ask a question. This is not meant to be a deterrent, the microphone.

Let's wait. There's a question down here. Based on your research, do you think there's a way to restore sense of smell in cases where it's been diminished or lost entirely? That depends on the reason for the loss. Now, one of the remarkable things about olfactory sensory neurons is with other very, very rare exceptions, olfactory olfactory sensory neurons are capable of being regenerated throughout the life of the organism. They only last 60 to 90 days and you're constantly generating new ones. So if you destroy the nose, the epithelium, you regenerate a functional epithelium within months. On the other hand, if you do damage to more central olfactory structures where neural regeneration is more difficult to conceive, you are less likely to be able to restore it. Quick question. I was wondering if you could use prefrontal cortex activation as an unconditioned stimulus like you did in the piriform cortex? Whether I use the prefrontal cortex— the piriform cortex as a conditioned stimulus, and I can also use prefrontal cortex as a conditioned stimulus, yes, and that works. So it says that indeed prefrontal cortex is hooked to behavioral output and it can be entrained. But if I get rid of it, I use the entrainment of the piriform cortex. Thank you, Professor. One of the things I think you've shown us is to what degree we are machines. L'homme, la machine. But I believe you've also said that you're religious. So my question is, what kind of machine are we? So, so I would question your premise. That is, I think our studies show that there is a machine-like component to innate deterministic responses, but rather a flexible individualistic character to sensory learning. I think we're hardly machines. Now, machines can be taught to do precisely what I described, but I would have to use my brain to teach that machine to do it.

Hi, so a lot of, sometimes before brain surgery, like functional MRI is used to, you know, map sensitive areas of the brain, but as you've shown that there are some areas that may not appear sensitive, but they impact learning. So do you think whenever brain resection occurs, you lose some function?

In the brain? Sorry, I missed the last phrase.

Whenever brain— Like, let's say the fMRI scan says, OK, surgery will be safe. You will not touch a sensitive area. But no matter what, does brain resection cause loss of brain function then?

No, I would not. First, I'm not a neurosurgeon. If we had more time, I'd explain why. And it's a good thing. But it is— indeed, it is true that the resolution of fMRI is limited. One cannot always know the significance of brain structures. And the neurosurgeons are well aware of that and simply do not resect based upon a thought that that region is safe. Rather, they take into consideration the reason that they are resecting it and weigh that onto the possibility possibility that brain function could be perturbed.

We have time for one last question. There's a question over there in the absence of any competition that I see.

Hi, thank you. So you displayed how using certain odors could cause a certain action with within mice. And so I'm curious as to whether or not you think it would be possible for that to occur within humans as well, seeing as people think of humans as so much higher on the cognition scale.

I tacitly assume that at the level of that at which I described representations in the mouse brain, that one would see something very similar in the human brain. We have all of those structures.

So one of the things that we aspire to in the Jerome L. Green Science Center and the Zuckerman Institute is the price of admission is that you're really good at one thing, and really audacious in your appetite for exploration of things far beyond your sphere of competence. Richard Axel is a molecular biologist, and he's done not too badly in going into this area of cognitive systems neuroscience, and I want you to join me in thanking him for a remarkable lecture this evening.

Thank you so much.