 Let me call your attention to three special events that are available for your enjoyment this evening. These are all described more completely in your program. At 7 p.m. the Nobel concert, which will be held in Christchapel, at 8 p.m. the Nobel art exhibit in Schaefer Gallery, and at 8.30 p.m. the Nobel firing lines, which will be in Girling Concert Hall, Anderson Theater, and Nobel Auditorium. I also want to thank David Fienin, the chair of the music department, and his colleagues for putting together the concert, and especially Stan Schetka of the art department for constructing the art exhibit. Now I'll ask Tim Robinson to return to introduce our third speaker of the day. It's a pleasure to welcome Dr. Brenda Miller here to Gustavus. Dr. Miller was born in England, educated at Cambridge, where she studied experimental psychology. While working with Frederick Bartlett, she developed an interest in the study of memory, which remains to this day. After World War II, she emigrated to Canada, where she put her ability to speak French to good use, teaching at the Institute of Psychology, the University of Montreal. While still teaching at the university, she also found time to go across town to McGill University to take some classes from a dynamic young psychologist by the name of Donald Heb. Unlike many of his contemporaries, Heb was quite willing to speculate on the topic of higher brain function. In his landmark book called The Organization of Behavior, he spoke of cell assemblies and phase sequences and other postulated activities in what he called the conceptual nervous system with sufficient vigor and persuasiveness to convince Dr. Miller to register as a graduate student at McGill, where she finally completed her doctoral program. Her dissertation was entitled Intellectual Deficits of Temporal Lobe Damage in Man. She accepted the position in the Department of Neurology and Neurosurgery at McGill a year later and has been there ever since. Understanding the operation of the human brain is very difficult. The sheer size and complexity of it is enough to send most investigators looking for simpler species. Furthermore, the research techniques which can be employed on humans are understandably very limited. As a result, the systematic study of those unfortunate individuals who have suffered some type of brain damage has become one of the most important sources of information about the function of the normal human brain. Over the years, the Montreal Neurological Institute, which has been staffed by such world-famous neurosurgeons, as Wilder Penfield and Theodore Rasmussen, has been recognized as a leading center for the treatment of such varied neurological disorders as gunshot wounds and intractable epilepsy. For 30 years now, Brenda Miller has had the opportunity to study the effects of this kind of surgery on the cognitive functioning of these patients. The most startling of these patients was a young motorwinder who underwent surgery for bilateral removal of the hippocampus in the early 1950s. Dr. Milner described the profound amnesia, which ensued and has followed the progress of this unusual patient ever since. She was also a pioneer in the study of the function of the two halves of the brain. Today, she's convinced that there is just as many interesting differences between the front and back of the brain as there are between the right and the left. Here to talk to us about the topic of brain and memory is Dr. Brenda Milner. Thank you very much. Mr. President, Dr. Schafter, Mrs. Lund, fellow speakers, colleagues and friends, I'd like, first of all, to express my appreciation and gratitude for being invited to talk to you to take part in this Nobel conference at Gustavus. It's my second visit here, and again, I feel extreme pleasure to be with you. I took rather a general title, Memory and the Human Brain, not how we know, but I hope it fits a little into that context, to talk really about those dissociable or partially dissociable aspects of memory that I have been privileged to observe in patients, as you just heard, in patients who were undergoing brain operations, planned brain operations for the relief of long-standing epilepsy. However, my interest in memory really goes back much earlier. I'm a student from my old undergraduate days in Cambridge of the late Sir Frederick Bartlett, and so I was very happy to hear Dr. Edelman say this morning that the memory, at least in the sense of, I remember, memory is a reconstructive process, a process which requires a lot of imagination, a process that is not just a strict reproduction of something in the external world. This is indeed the lesson that we were taught in those faraway days. I have a plan in this lecture. I have a plan to try to cover three themes. I don't know what will happen to the third one because I do want to dwell a little on the first. The three themes also go a little historically, I'm afraid, because the first I'm going to talk about is probably the oldest and best known, but it is probably still crucial for the kind of argument I want to develop today. I want to begin by talking about the characteristics of this grave amnesic syndrome, this gross disturbance in everyday memory for events as they occur, that one sees in patients who are unfortunate enough to have bilateral damage in structures deep in the two temporal lobes of the brain. The most famous patient is not a patient from Montreal, is the patient H.M., who is actually a patient of Dr. William Scoville in Hartford, Connecticut, and he is the patient that we have followed over so many, many years. But I have to say that there have been other patients that time does not allow me to discuss who show essentially this same pattern of memory loss and preserved learning of certain kinds to which I would like to address myself initially. Now I have a wonderful pointer, tailor-made, custom-made. So I think before looking at the slides and talking to the patients, maybe we should reflect for a moment of the kind of memory that we take for granted, the fact that we can remember what it was like to be listening to the last lecture, what it was like to be at that library lunch, what lovely sunny weather it is out there. We couldn't forget those things if we tried. Yet at every moment the focus of our attention is shifting, life distracts us constantly, and yet we can depend upon some system in the brain to keep that continuity going, and I believe also by inner reflection on subjective rehearsal of events past, we do update our memory, and I will try to say more than our personal memories our knowledge of the world. From now on it's going to be the slides. Could I have the lights down on the first slide, please? What you see here in, I suppose one should say, a cartoon, a sketch of the medial surface of the right hemisphere with your imagination. This half of my brain removed and you are looking in, and you will note this shaded region around the inner core of the hemisphere, circling and therefore called the limbic lobe, and I would like to direct your attention, particularly to this region here, which is where you have the anchors, the pericampal gyrus, and underneath where you don't see very well this critical structure, the hippocampus, and buried there also the amygdala, the amygdala so-called for its almond shape, the hippocampus for its resemblance to a seahorse, and everybody gives me little seahorses as presents, and I have them everywhere as the sign of the house. Now in the typical operation for the treatment of epilepsy, in which the part of one temporal lobe is removed, the other functioning normally on the other side, it is usual to have to include at least the amygdala and often part of the hippocampus in that removal, but since the other half is functioning normally in the other hemisphere, the patients do not lose their memory of ongoing events. In the case of HM, the next slide please, the operation carried out by Dr. William Scowlo, I'm afraid you can't see the top of this slide, I think it's one of the few that are going to be like that. He, back in the early 50s, carried out what was frankly an experimental operation long thought about before decided upon in this young man who was totally incapacitated by major seizures and was on mere toxic doses of only available anticonvulsant medication of that day. The operation, which was not like the ones carried out in Montreal, differed in two ways. First that it was bilaterally symmetrical, which is something that always, I think, should worry a surgeon, bilaterally symmetrical and did not involve the temporal neocortex. It involved just those structures I've illustrated to you, the amygdala hippocampus and parahippocampal gyrus, going back a considerable way along the medial surface of both temporal lobes. It was a radical procedure. Now, Dr. Scowlo had carried out operations less drastic in seriously ill schizophrenic patients in those days and had found no impairment in a general sense and no evident impairment of memory in removals that were limited to the amygdala. But I think it's very important now for those of us who are interested in the brain and the cerebral organization of the brain in memory to re-emphasize the fact that the critical operation that had this devastating effect on the memory of the patient did not involve only the hippocampus but did also include the amygdala. I say this now because we are getting evidence from the monkey laboratory, notably from the work of Mortimer-Michigan at NIH, that in the monkey bilateral removal of amygdala and hippocampus does indeed produce something like an analogue of the human amnesic syndrome, whereas removal of either structure alone does not. So I think we have to take seriously that what we're seeing may be the effect of this combined lesion. Now, what about the effect? Well, the most striking thing, of course, for anybody seeing the patient is, and what his family and friends would talk about and still talk about to this day, is his tremendous anterior grade amnesia, his inability to remember new happenings, and this is indeed a tremendous devastating interruption in your normal life and your normal relationship with others. We were hearing in the previous lecture today about the computer that doesn't remember in that sense, and I think that this is very critical to the system in the brain that we're talking about, that there is certainly not this updating of personal memories that are still, and yet the past knowledge acquired, the past knowledge of facts remains untouched. We'll come back a little more to what else may remain untouched. But I would like perhaps just to illustrate with a simple experiment that what is affected when the system is damaged is very much what William James writing so many years ago called a secondary memory. You remember, James said that a primary memory, one can hardly call memory at all, it hasn't gone out of consciousness, it's the most realward part of consciousness, but it hasn't actually left consciousness, so we talk about immediate or primary memory, but it's a quaint use of the word. Whereas memory purpose, as James, or secondary memory as it might be styled, is what has just previously been in your consciousness, which is out of mind and has already dropped from consciousness, but which we have available as a sort of backcloth, I like to think, in the wings, and this is what we are so constantly dependent upon in our everyday remembering. Now I give you a very simple little recent experiment carried out with HM to illustrate this point, really, of interruption in his processing, sorry, not processing, I shouldn't say, in his preservation of ongoing events. The next slide, please. This is a simple absolute judgment task, so-called, carried out by one of my former students, Don Reed, quite recently with HM. That shows you that HM has not changed, incidentally. I have followed him, I've worked with him over the years on and off, and he still doesn't recognize me as a familiar person. This is really no change. In this little task you see these rectangles of various widths. But as they are presented in the experiment, they are on a continuous loop, so that you only see one rectangle at a time. And what you have to do is to assign the correct number, from one being the narrowest rectangle to six being the broadest, to give the right name to the rectangle as it appears. And if you say four, when it's really number three, you're looking at the examiner says no, and then you guess again, and you probably get four. Now, in the particular instance that I have shown you, this is a ludicrously easy task, even for patients with extensive brain lesions in one temporal lobe, say, or one frontal lobe. But for HM, this became an impossible task. It was not because he had to remember the instructions, because there's always a sign up in front of HM saying, one is the narrow, and six is the large. The next slide, please. And he did indeed seem in some absolute way, these are his errors, in some absolute way to recognize one as an absolutely narrow, and therefore always called it one. But you see that he never got straight, three, four, and five, and was also making bad errors, not necessarily saying the neighbor, but maybe even two away. Now, I think that what this really means is that in your ordinary memory of something like that, we shouldn't perhaps call it absolute judgment, because we very rapidly establish a frame of reference of this set, and so we recognize the individual within the set. But this, in the case of HM, as his attention shifts from one item to another, he does not build up this back cloth, really James's secondary memory, this comforting continuity of experience on which we constantly rely. Now, I'd like to really to go back in time, and here I'll do this very quickly, because it's very long ago. When I first went to meet HM, I went down from Montreal, and in those days there were no planes from Montreal to Hartford. I went by the overnight train, and I took with me a couple of experiments, a couple of learning tasks from the psychology lab to see what HM could learn. And I was lucky in taking two things which gave different results. It was pure luck, and I just show you them quickly. The next slide, please. This is a simple stylus maze in which you have to learn to find the path from the start to the finish, and here the path is shown to you with a line, but of course it isn't visible to the subject. Who must discover it by trial and error and remember it? Every time he goes out of the correct path, a buzzer sounds telling him he's made a mistake, he goes back and tries something else, and of course makes many mistakes on the first trial. But normal subjects and patients with various brain lesions can learn this route to a good learning criterion in 19 or 20 trials. With HM it was a disaster. The next slide, please. We worked with him over three days. The next slide, please. Thank you. And averaging his errors over blocks of five trials with time out for lunch and so on, and sleeping overnight. You see that the three days of testing, he made absolutely no progress, and at the end I had to leave and go back to Montreal. In retrospective scope, I realised what a silly thing it was to expect a patient who couldn't remember anything once his immediate span of attention, which was normal, was overloaded, was passed. How can he learn something with 28 choice points? And subsequently we shortened this maze a great deal, and he managed to learn a very simple version of it in 155 trials. But the point I would like to make is that this extremely slow learning of a route is paralleled in his daily life. This patient takes literally years, years, to memorise the route around a small house in which he lives. He did achieve this in the years that followed the operation. His family moved to a new house, and for a long, long time he had no idea of the route within it, and he still of course didn't know the address and would give the address of his former house and guide us to it if you ask for instructions. He had no difficulty with spatial layout as such. It's strictly remembering and learning the new layout. And then now he has learnt the second house, but unfortunately had to move again, and it is taking him again three or four years to learn the layout once more. Now I think this new learning is presumably taking place in a slow, incremental way. We don't know anything about the locus or the representation of these very long and stable memories. We're sure that they must indeed involve large populations of cells. They must indeed be distributed widely in the brain. The thing is that without the help of this medial temporal, and of course it's not just medial temporal, but the deeper diencephalic system, this system in the core of the brain, without the help of that system, the learning by increment is so extraordinarily slow. This is a sobering thought in telling us the importance of this medial temporal system. Now the other task that I took, the next slide please, was a very different kind of task. Again, one loved by psychologists of those days, having to learn to draw, to put your pencil within the confines of this star at the point S and trace a line within the confines of the star, never touching the sides, but the difficulty is that you only see your hand and the star as reflected in a mirror. And so we all make mistakes when we come to turning points in the star, but gradually improve. And I thought, oh dear, when I saw HM do this the first time, this is going to be so difficult, and I'm going to be wasting all these precious time with him doing this task he won't learn. But again, I was totally wrong. The next slide please. The next slide please. Shows you his learning during exactly the same period when he was failing the maze, his learning over this three-day period with a completely perfect performance at the end, but with no sense of familiarity whatsoever. So I think that here is the next message that you can have learning systems going on in the brain, which presumably have nothing to do, because his learning of this kind of thing is normal, not using this medial temporal system which is related to our conscious memory of things and events. This is the kind of learning, and in those days back in 1962 I speculated that many kinds of motor skill would be learned independently of this system. By this I meant learning to dance and swim, learning to pronounce the foreign language, though of course not learning its syntax or its vocabulary, that many things of this kind, the things you learn early in life with a lot of practice, the things that you cannot introspect on how you do and indeed trying to impair your performance, and that are indeed very stable, once learned and carry over from one season to the next. That I think is true, that generalization I made, where average is being tested seems to have held up, but it was too narrow. Work subsequently by Elizabeth Warrington and Wisecrance in England showed that there was also a remarkable facilitation of the ability to recognize again something that was, not recognize as having seen before, to perceive more accurately a fragmented drawing, you show somebody a fragmented drawing of an airplane and it's really just a few fragments, we can't guess what it is. Then you add a little more contour and you say, oh gosh it's an airplane, and so on with other objects. A child's test, very easy. You then show this again to the patients months later, months later, and they show better, faster recognition, require less contour information in order to identify the object. Now they're not learning something new, this is something that these are very stereotyped drawings of things in a child's vocabulary, in a child's picture knowledge, but they are being, this early activation seems to leave the nervous system ready to recognize, not to say I've seen it before, they have no idea they've seen it before, they can identify more readily at the second showing. This then is another kind of memory which doesn't seem to depend at all upon the medial temporal system. More recently Larry Squire and Neil Cohen have shown that the Tower of Hanoi problem, Dr. Simon has told us all about it before, is something on which you show steady learning in an amnesic condition, again learning without awareness and so the examples multiply and people have tried to characterize what is preserved in this case using the terminology of the present day as procedural, I don't know whether I like these dichotomies of procedural versus declarative and so on but there are indeed different kinds of memory, some very critically dependent upon the integrity of the medial temporal lobe system. Now, before leaving HM to talk of other things but still with the temporal lobes, I have to address one more issue which is that Professor Tolving, Toronto, has distinguished between episodic memory, what he calls episodic memory which is the memory of actual episodes and events of one's own life. I had lunch at the library today, I ate such a thing last night and so on. There is no question that that memory is completely disrupted by lesions of this kind, although the patients are still able to remember the events of their childhood, not a lot of course, none of us do, but to remember with vividness, HM will say I can remember the first time I had a seizure, I can remember the first cigarette, I smoked, it was on such and such a highway going to Florida and did I cough? It was apparently vivid for him but things going back for a year or two before the surgery and that's a very long time seem not to be recalled and of course new events are not recalled. There isn't any argument about that but what Professor Tolving has called semantic memory, he means knowledge of facts and I don't agree that knowledge of facts is preserved or our acquisition of facts, the knowledge of the old facts is more or less preserved but if you look at your ability to acquire new knowledge in personal knowledge like vocabulary, you do indeed find an impairment. John Gabriele at MIT has been studying this in HM. He has taken advantage of the fact that apparently in the Webster dictionary there is a new one brought out every five years in a lot of new words and so John Gabriele has been able to check HM's knowledge of the vocabulary as compared to the control group. I should point out that HM is not in an impoverished environment these days, he seems to spend most of his time at MIT being tested by generations of bright graduate students and post-doctoral fellows and otherwise visited a great deal and so I don't think we can just think that he's in such a sheltered hospital environment that perhaps it's not fair to expect him to know these things. He also loves to watch television though he can't tell you anything about it afterwards and he will do things over and over again as we heard about the computer without getting bored. I have never seen him bored. So he has this opportunity to learn the new vocabulary and it's clear that when he is compared with age matched controls that he has acquired really very little. He has acquired something more than he has of his memory of the events of his life but there is certainly an impairment in the way in which this kind of learning is acquired. Again, I think it's this business that with a great deal of repetition the cortical systems involved will become modified by exposure to this information but it is an extraordinarily slow process. So we have here to postulate and here I would love to know something of the underlying mechanisms but I am certainly not going to venture any guesses because I really believe in that point even more than Dr. Adelman because it's starting from a different vantage point that one has to be very conservative but what one requires here is a system that is modified very quickly within one trial this memory that we take for granted as we walk around our lives is not the memory of sitting down to study very hard it is the things that automatically stay with us it is a great gift and this system then has to be modified very rapidly it has to be modified and stay modified for a surprisingly long time that is what the facts of retrograde amnesia tell us the long period of events before such a surgery that seem to be irretrievable there must be something going on this medial temporal system seems to be participating in the act or seems to have its role to play for a surprisingly long time however I don't believe it's forever I think that most of the rather trivial happenings that are so important for us to keep to be able to depend upon remembering for the continuity of our day to day lives are not in the least important to remember in the long range sense and that ultimately if they are not things that are especially interesting or relevant to other things that we are doing do indeed get lost but as far as the memory for the individual events what one had for lunch five years ago in the cafeteria at the Neuro I think that's really gone I would believe however if we look at the well preserved intelligence of patients who suddenly have this kind of surgical insult if we look at the well preserved vocabulary and general knowledge and knowledge of how to behave and so on we have to think that at some point the system drops out and enough has been perhaps by a great deal of inner rehearsal enough has been left to maintain the system without any longer the participation of what I call some kind of well no I won't call it a name well this medial temporal lobe system which is so much related to our conscious awareness of events and happenings as they occur so that is a little bit the challenge of trying to find out the cerebral substrate of such a system and the role it plays in memory now I would like to move on to considering the effects of unilateral temporal lobe lesions could I have the next slide please Dr Edelman referred to the speech areas this is a sketch by Dr Rasmussen of the left cerebral cortex of the typical and I agree with Dr Edelman that there is tremendous individual variation in as far as he is talking at a much more basic level than I can but I think all the clinical data fit it but this is the generalized left cerebral cortex of the typical right handed person adult and you see in the stippled areas here which represent the posterior parietal posterior temporal speech zone and here brokers area in the frontal lobe these are areas which are sacrosanct in any operation that is elective it is better to have epilepsy than to not to be able to talk so the surgeon has to map these areas out during an operation in conscious subjects by electrical stimulation this was a lot of the pioneer work of Dr Penfield many years ago and by electrical stimulation in these critical zones you can interfere with ongoing speech interfere with the patient's ability to name and speak and when you do that you know that is an area that you should keep out of because if you damage it you will have some lasting trouble probably with language it is very interesting how different that kind of difficulty in naming is from the conceptual difficulty Dr Penfield has some beautiful examples in his very carefully kept operating room notes of patient for example who was being shown a picture of a shoe I'm sorry a picture of a butterfly and he said tried as he said after the electrode was removed as he said to Dr Penfield I tried to think of the name for butterfly and I couldn't and then I tried to think of the word for math so you see there is this really the awareness of these relationships without the awareness or access to the names now we don't then work with patients who have these areas damaged nor with patients who have speech difficulty permanently we work with patients who have the anterior temporal region removed for the treatment of epilepsy as you saw on the medial surface this means removing the amygdala often but not always part of the hippocampus and if you do this if you have this kind of difficulty in your dominant hemisphere for speech you find that you have a difficulty a partial memory disorder I called a material specific disorder because you have trouble with memory for words, for verbal material what you read and what you hear but you don't have any difficulty in remembering faces and places and so you can help your faulty verbal memory by a method of imagery by trying to picture objects in relation to one another if you want to remember to associate words all these techniques that of course don't work with a patient like HM because HM can make an image quite well and currently you say nail lettuce which are two very improbable associations as my colleague Marlon Gottman did and he said oh yes I can picture that and I have the nail sticking out of the lettuce so I won't bite it but then later he doesn't remember a nail lettuce or the image because it is not just a verbal difficulty but a generalized difficulty patients with a purely verbal difficulty of course can be helped by this kind of imagery corresponding lesions in the right hemisphere give you no verbal difficulties but give you difficulty in memory for pictures and faces and so on now one of the things that has interested us has been to study the relationship of the hippocampus to its overlying temporal to the hemisphere of which it is a part since we have groups of patients the next slide please if we are sketching here the removals these white areas are the removals and remember that from now on I am not talking about patients like HM I'm talking about patients who have one good hemisphere and who are having a removal of a damaged area in one temporal lobe and in the left hemisphere on the left you could have a removal as you see at the top which would take the amygdala but spare the hippocampus and you could have a removal which is more extensive and this of course it depends upon the individual patient's difficulties and what part of his brain is damaged and these are the corresponding representative removals in the right hemisphere and so for many of our memory tasks we have looked not just at the effects of left temporal lesions but on verbal memory let us say but on comparing the effects of a left temporal lesion that spares the hippocampus with one that destroys it extensively and of course we have the control groups of lesions in the right cerebral hemisphere they are very young patients and they are very comparable in other ways and I will illustrate just briefly the role of the left hippocampal region in verbal memory which means of two very similar tasks which gave as you'll see slightly contrasting results they are what are called word generation tasks carried out by Donald Reed it's based on a paradigm developed by Slamica and Graf essentially what you do is produce your own list of words I say tell me a word that rhymes with nice and begins with M and you generate mice and so on and you go and you produce a whole list of words that you are just generating yourself on the basis that it has to rhyme with the word you're given and it has to begin with a certain letter and then at the end of that task you say now tell me as many of the words as you can remember it's not especially easy it's easier if you generate them yourself and if you just listen passively to somebody else but it's still not very easy and you produce a certain proportion of these words the next day you do a synonym task a word generation task following exactly the same procedure except that now you say tell me a word that means the same as big and begins with L and you have large and you say now tell me what you find now psychologists have long ago described what they call a serial curve of recall if you're recalling a list of items you remember the most recent very well they're right there in your mind really James's primary memory it's right there you remember the first ones because unless you're somebody like HM you've probably had a chance to think over them to rehearse them a little you don't remember so well the ones in the middle because there is more interference now what do we get when we look at rams first of all I should tell you that patients with right temporal oblesions give you a completely normal curve for both these tasks and so their data are not presented on the slide just for clarity the next slide please this shows you the results for the rimes test this is the normal control group with the recency and the primacy effect when the words are grouped according to their position in the original list this is that these two curves which you can't distinguish one from another a bit irregular are the two left temporal subgroups in other words even if you have a left temporal lobectomy that spares the hippocampus you have a very little primacy effect on this rhyming task this is one of about three studies which tend to lead us to conclude that the left anterior temporal neocortex has some specialization in the evocation of the sounds of words which is probably something that is useful in getting back these rhyming sounds the next slide please if we look at the results for the synonym task it's quite different here is the left temporal neocortical group with hippocampus spared and they're completely normal no significant difference in the normal group here is the group with the hippocampal lesion on the left and you see a normal very good recency what's right in their mind comes out but they're completely wiped out at the rest of the curve and I think what you see here is the specialization which is really due to the extrinsic connections of these groups of cells that in a sense the left hippocampus is and the neighboring structures are in the service of that hemisphere of which they are apart they are linked by many connections through the parahippocampal gyrus linked reciprocally with widespread areas of the neocortics of that hemisphere and so you do see when with a lesion there a very specialized difficulty this is a little bit like HM's difficulty that something just out of mind in the wings can't be brought back but in this case it is specific to a certain kind of it's to verbal material and not to pictures and so on and as you will see we have somewhat different results for the other side of the brain the next slide please now I always get more excited about the right cerebral hemisphere on the left it's partly I think because we can draw some parallels from animal work it is much easier to link our thoughts about the brain mechanisms if we can I think tie this in with animal work this is just a reminder of what I said earlier in the talk that damage to the right temporal neocortics interferes with your ability to remember to identify complicated visual patterns this means you have a bad memory for faces you have a bad memory for nonsense designs as Doreen Kimura showed many years ago this is the kind of thing you see with bilaterally symmetrical lesions in the macaque but because of the functional asymmetry of the human brain a lesion on the right temporal lobe will make this deficit appear whereas a lesion on the left will not this doesn't mean it would be not logical to conclude the left temporal lobe doesn't add anything I think bilaterally symmetrical lesions give you much more but rather that because of the asymmetry you get the effect or some clue to function from the unilateral lesions however this was not the question we were asking was it the question we were asking is what does the addition of a hippocampal lesion do it doesn't do anything to this because we think that we know that the temporal neocortex is specialized for this kind of learning this kind of memory and so if the patients can't do it because they have a temporal neocortical lesion adding the hippocampal lesion isn't going to make on one side isn't going to make their memory worse however there are other kinds of tasks not to do with language that are affected by right temporal lobe lesions that include the hippocampus and not by right temporal lobe lesions that spare it and these are concerned with the memory for the position of objects in space and here again I would remind you of the remarkable achievements of normal memory and in this case we can go from rat to monkey to human being no doubt to a lot of other fellows in between and say that a very salient part of one's adaptation to the world is in noting where things are remembering where things are objects in space and so on this is something that we do very efficiently even those of us who say we have bad memories and so on we can find we're really rather good at that if somebody says where did you put such in such a reprint but I have a very quick image of the reprint under a shoebox somewhere near the blackboard or something like that and it's probably correct we just do this and normal human beings do this apparently without any special effort to do so they do just as well if they're not trying to remember as if they're trying to remember so it seemed very interesting to know how patients with right temporal lesions would fare on such tasks we do have evidence now from a lot of spatial memory tasks of impairment after right temporal abectomy that includes the hippocampus and no impairment after left but I want to illustrate this just from one task an incidental learning task carried out by Mary Lou Smith this is based on adapted from a procedure devised by Jane Mandler what you see here is an array of toy objects on a table and what you have to do you're not told it's a memory task we deceive you what you have to do is to point to each object in turn and you wait 10 seconds to make sure you really look at the object and at the end of the 10 seconds you have to be able to tell the examiner the price of an object of that kind of average quality the price in the world today and so you think it's a pricing task you won't be surprised to know incidentally that H.M. underestimates wildly prices of everything because again he hasn't kept his knowledge up his knowledge structure up about inflation and what has happened to the prices of things and more than he has about the appearances of things and so on this is why I say his semantic memory leaves something to be desired however come back to the patients with unilateral lesions they're asked to have the price of the objects then they are given in the original version of this they're asked to recall as many of the objects as they can this is not an easy task I mean the thing is covered up and they're asked to recall as many it's not an easy task, people remember about half of them but both temporal lobe groups can do that, they've just seen the objects and thought about them and it's immediately afterwards and then Mary Lou presented them with a completely blank sheet and asked them to place the objects back where they had originally been and measured the distance the displacement of the objects from their original position get the same results if you do it in a fancy mathematical way looking for relative position, fortunately it's just as valid to use the simple-minded measure the next slide please and this is the only slightly complicated slide what we have plotted is the effect of removals from the left hemisphere on the left of removals from the right hemisphere on the right and what we are interested in is the points on the left these are performance 24 hours later when there's more noise in the data essentially the same results but let us concentrate on what we are calling here immediate recall we see that the normal control group and the two left temporal groups and the left frontal lobe group are all performing normally there is no effect of a left hemisphere lesion on your ability to do this in the right hemisphere there is no effect of a frontal lobe lesion you can have a very big right frontal excision and have a very good memory for incidental observation of where those objects were but the right temporal group with a large hippocampal removal this is their mean error right up here I can't reach it with my pointer and this group with less of the hippocampus out is not significantly impaired there so that we have a very clear evidence of the important contribution there in terms of its connectivities with the presumably with the right parietal cortex the right temporal cortex doesn't have much to do with position in space probably but the spatial representation is by a superior pathway out of the visual cortex and that we are there seeing the representation of the memory of an object in a particular place affected by damage to the right hippocampal region well we have a lot of replications of that there is one point to add which is that although we have called this immediate we perhaps should not have done so because there was in fact a minute or two in which the subject was recalling the names of the objects Merleau has now controlled for this and we can say that it is not the deficit is not apparent at zero seconds or three seconds just bang like that but is apparent after two or three minutes unfilled interval so we have there very clearly I think the idea that in the sort of overall economy and memory systems of the hemisphere we do see this formal this role it's important contribution of these structures again I would say getting modified in some way very quickly and rather stably but not I think forever well I'm coming now to the third bit which is really going to be an addendum as I see that time is passing but I would like to talk about another kind of memory and another part of the cerebral cortex could I have the next slide please this is a not a very serious slide for you it is a reminder to me of this marker this change into a different topic which is to say a little about the frontal lobes the frontal cortex we have very large removals from left and right frontal cortex for the treatment of epilepsy due to scarring of the brain in the left or right hemispheres always sparing broker's speech area in the left hemisphere but otherwise they can be quite extensive removals or they may be small ones we saw that despite these large removals these patients had for example completely normal memory incidental observation and ability to recall and reproduce the position of an array of objects in space they perform normally on many standard intelligence tests though that is another story that tells us more about the failings of standard intelligence tests than about the fact that the frontal lobes have nothing to do with intelligence because I think that what more and more one finds deficits on what one can loosely call fluency tasks ability to generate a variety of responses to a single question not to give you one quick answer but to be able to to conceive of of things used in different ways words used in different ways, concepts used in different ways but I promise myself to stick to memory what can we say about it well this little slide is to make the point that a frontal lobes lesion may impair your performance on a memory task is certainly if that task gives you the opportunity to organize and structure the material to be recalled there is a vast literature a lot of it going to George Mandler and his colleagues pointing to the importance of so called organization is an aid to recall if you're asked to cluster material words, pictures what have you into meaningful categories you will remember that better than if you don't or if the material doesn't lend itself to that if you bring an informed and educated mind if you bring your own idiosyncratically modified nervous system to bear to select if you like from the stream of new things coming to you though that's part that can be meaningfully organized into a structure has a better chance of other things being equal are being recalled we all know this organizing being helpful to memory now this is a very silly little experiment really done by one of my students but it makes the point you have actually twice as many as this pictures of common objects, animals, household tools you're given all this rather crowded array it's twice that size as I said and just ask to group these objects into groups as you wish according to your own pleasure and then afterwards you are tested for your ability to recall the objects your skill in categorizing is judged by independent judges who know nothing about the group to which the patient belongs and so on and the student who did this, Antonio in Chisa showed that as he had predicted that patients with temporal lesions had no difficulty in categorizing but still had their impairment in recall whereas patients with frontal lesions showed impairment in the categorization they were items that were left out they were items that were falsely classified or in not sufficiently precisely analyzed in their categories and they also had an impairment in recall but that is a little illustrative experiment I would prefer if I could take a few minutes to talk briefly about two other complementary or related experiments one old, the other fairly new which deals with the role of organizing and judging temporal order of external events or temporal order of one's own responses let me insert though one picture of a monkey brain at this point the next slide please this is a slide lent me by Dr. Nauta in which you see the schema of the cortex of the macaque and you note here in the frontal lobe the famous area of the principal sulcus there is a rather silly little task which as you probably know many of you monkeys with bilateral damage in that region of the frontal lobes fail abysmally it is called a delayed response task or another version of it, a delayed alternation in which all you have to do is if you're a monkey that likes peanuts you're sitting here and somebody taps smartly and puts a peanut underneath a can here there is another identical can obviously if you're just there watching you grab the peanut but if the examiner doesn't allow you to do this he drops the screen and you wait a variable interval and this is the delay then the screen comes up and you have to find the peanut but the catch is that it varies at random from trial to trial you can't do this by thinking oh it's the one on the right or thinking in monkey terminology it's the one on the right that would be too easy and you can't do it by saying it's the green can not the blue can because both the cans are identical and under this condition what you have to do is to suppress you have to forget and I think that it's this kind of disciplined forgetting which is difficult for the frontal lesion you have to forget all the previous trials and just give salience to that moment at which you are watching that peanut go in forget the other peanuts you've eaten today forget the other trials and give salience to that immediate one and this is beyond monkeys with bilateral lesions there now of course a human patient could do this very easily because he could just say to himself it's on the left and remember by verbal mediation so one tries to trick them with things that are a little more complicated and many years ago I developed the hypothesis that there would at frontal lesions were giving patients difficulty knowing what trial they were working on but they were blurring the distinctions between trials that they couldn't sort of suppress and concentrate on what was in hand and use it and move on but there was this confusion in the temporal ordering of their memory and I set out to test it and I'll just show you the kinds of tasks and then tell you the results I set out to test it the next slide please this is going to be enormous I apologize the verbal form of the task you look at a series of cards and read the words out like hand, cuff, peanut next slide please next slide thank you then ashtray, keyhole and so on imagine you're reading a series of pairs of words like that and then next slide please you see these two and you have to say which one you saw more recently because you saw both of them but which one you saw more recently and my hypothesis was that this would be difficult for patients with frontal lobe lesions the next slide please because we always use verbal and pictorial forms of our tasks if we can to bring out subtle hemispheric differences we also had objects where you see a picture here the hand and pictures of people and representative drawings and when there is a question mark which one did you see more recently and finally the next one please you have this abstract art this is of course difficult for patients with right temporal lobe lesions to recognize but they can recognize they don't have difficulty with the order as such the but the idea now is when there is a question mark which of these two paintings that you see more recently you can't do it by naming the artists because there are too many instances of a few artists and then on this task Phil Corsi and I showed that there is indeed an abysmal performance after right frontal lobectomy that the patients just cannot do this they can recognize the paintings but they cannot order them in time there is a lesser impairment with the pictures and just missing impairment with the words after left frontal lobectomy on the other hand there is only an impairment on the words so we got an effect and the discrimination of recency which was related to the side of the lesion but on which the overall greater effect was from the right side of lesion on the left now before commenting on that this is really my final point I'd like to show you a similar but subtly different experiment carried out by Michael Petridis much more recently the next slide please this is a task in which again following the idea that perhaps one of the difficulties, one of the things that the frontal among many things that this part perhaps of the frontal cortex is important for is keeping track of events as they happen keeping track of your own responses so Michael carried out a variance of the recency if you like he called it a self-ordered task in which you are really given a stack of cards and you move with and you are asked to touch one of the cards it doesn't matter which here you have pictures the next slide please you have abstract drawings and we had a verbal form but let us say that on this task you just have to touch one of these pictures, whichever you like and then you turn the page and on the next page you'll find all the same pictures but they will be arranged differently and you have to touch another one and you go on through these 12 cards the only instruction is that you mustn't repeat yourself so you have to monitor what you are doing you have to look for the displays changed each time so you can't just go through the page and no rule of thumb like that and you have to keep track of what you are doing and the hypothesis was that this would be difficult for patients with frontal lobe lesions and again we hoped for a hemispheric difference and what Dr. 3D's found was indeed a deficit in patients with frontal lobe lesions but we got an interesting difference from the findings for recently because now we found that the most impaired group on all the tasks was the left frontal group, the group with frontal lobe excision in the dominant hemisphere whereas the patients with right frontal lesions were impaired only on this very difficult pattern, abstract pattern and my last slide illustrates the contrast between these tasks, the last slide please showing you that for concrete words for representative drawings and for abstract designs that for the recency discrimination you have a greater effect of right frontal lesions than of left and the reverse for the self-ordered pointing with the same kind of material and we think that perhaps and this is really so fanciful I obviously have to stop we think that if you look at those two tasks apart from temporal order judgments the recency one you are very passively looking at an inexorable stream of pictures or words that come at you and what you have to do is be very watchful and signal to keep track of the events in the order in which they happen and maybe this right hemisphere I don't really like hemispheric dichotomies because I am so aware of hemispheric overlap but nevertheless we do know that the right parietal cortex the right posterior cortex seems to be especially watchful it quick off the mark its organization to what is happening in that external world and maybe therefore the right frontal lobe within that hemisphere has a little special edge when on the other hand we come to look at the monitoring of one's own actions we cannot help but remember which is the left hemisphere notably the left parietal region which is so important for the organizing and programming of one's own movements perhaps therefore it is not by chance or trivial that we do get this subtle difference between the two sides well I must thank you for your patience to summarize I have talked first and I would like to mention last because perhaps really the most important message is the great significance of this medial temporal and cephalic core system in the brain for ensuring our ongoing identity and building up and updating of our memories, our personal memories first of all but then indirectly our knowledge of facts and of the world around us and if we then we look at the left and the right side separately we see that there is this importance of the extrinsic connectivity of the regions to the overlying cortex which gives them their material specific specialization and finally this great and challenging problem of the frontal cortex that we are just beginning to explore I think trying to look at how it plays its role in the control processes of memory, thank you we'll take a five minute break and then reconvene for the panel discussion you may hand your question cards to an usher who will bring them up and deliver them to chaplain Elvie or myself even though our panel is still in the process of reassembling I think we'll go ahead and start with some of the questions from the audience for the panelists that are here hello can you hear me out there can you hear me test first of all lost it again, testing first of all we'll hear a response from Dr. Dennett responding to Dr. Edelman's lecture this morning Professor Edelman one of your remarks this morning was to distinguish rather sharply between two approaches to the study of the brain one of which you call the information processing approach the other you call the population approach now this puzzled me because as I look at the information what I would consider the information processing approach I see a vast range of different approaches some of them preposterously profoundly unbiological some of them earnestly biological and some of them are neutral on this score and so I didn't see the conflict that you saw between these two approaches could you expand a bit on what you meant by the conflict between a population thinking and information processing well first of all I would like to know a little more about what you think is biological but I'll leave that for after my comment perhaps those are not the best solicitous choice of terms but what I really meant is something like this the contrast between what was called essentialism in the problem of taxonomy during evolutionary theory and its development and what Darwin came up with it has to do with the name and the name and I agree with you that the information processing approach for example is entirely compatible at the level that Dr. Schenck discussed it and you have discussed it with described facts the kinds of things Dr. Milner talked about but at the level prior to language at the confrontation of the organism with the echo niche if you will with its environment surely you would agree that there's no name or unnamed in the strict sense it is in that sense that I use the information processing model you'd be surprised how many scientists I've met and how many neurophysiologists I've met who actually act as if the world comes in nice little packages with categories and labels on them they talk about information going down one neuron stimulating a neuron neuro transmitter and being transferred to the other neuron and when you actually investigate that you find it's an implicit assumption and it is this travesty I would say the information processing model that I had in mind not a judgment on the larger term here's a question for Dr. Milner when and how does present memory become historic memory maybe when I can't hear it I'll read the question and then I'll give it to Dr. Milner when or how does present memory become historic memory is this an interesting question or one whose answer is well known I don't think that memory has to go through stages and filters necessarily there's quite a bit of evidence that that information can be built up as it were in parallel systems although maybe it benefits from interaction I don't think that memory has to go through stages and filters necessarily there's quite a bit of evidence that interaction but how the long-term the really enduring memories I think there has to be a great deal of inner reflection my goodness and rehearsal and maybe some things get damped down and other things get really over rehearsed and linked into our own idiosyncratic systems one of the interesting things has been following patients who had say normal correct answers on an information test right after the surgery that gave them the memory loss and then when you see them a year or two later they have more imprecise they have lost their precise notions of common knowledge as though we are normally giving boosts by this reactivation when we read the newspapers, when we rehearse and so on do you want me to answer all these I had an earlier question about why does an excessive amount of alcohol sometimes cause a memory loss and I'm not sure there are two possible meanings to this question if you mean as I think you do why is it sometimes if you go out and really drink a lot of alcohol you don't remember the next day what happened and I think that alcohol that amount is pretty toxic to the activity of the cortex as a whole we don't have to say it's just knocking out the hippocampal memory systems if you test the intelligence of people while they are imbibing all this so you test their skills or anything there are very grave generalized impairments but if somebody has in mind the further question of why there is this Korsakov memory impairment among some people who are taking alcohol to excess over many many years I think now it's to recognize that this is a vitamin deficiency a thymine deficiency and it's not a direct consequence of the alcohol do you want me to go on there's so many maybe somebody else should answer a question I think Dr. Edelman has one you want to read the question yes I have a question that says if your population model is correct is it not true that the theory of relativity all of Shakespeare's plays and indeed all past present and future human knowledge is buried within the mind of each of us and that tapping this knowledge is only a matter of quote pressing the correct combination or permutation of buttons now this gives me an opportunity to come back to what Professor Dennett was saying it is exactly this that is not implied by the model nor is it implied by evolution either what is implied by the model is the potential or the capacity upon interaction with the appropriate environments to do all of these things not the precise information as if some large observer was saying let there be Shakespeare plays and then the only question is to tickle them out if in fact I could make the analogy to evolution it was exactly this idea that accounted for species before Darwin a tiger was a tiger and if a man said but this tiger has a long nose and queer spots the reply would be well that's just a sport it's noisy it doesn't matter tigers are tigers but it's just the fundamental point about selectionistic models that number one they do not have information built except ex post facto after the selections been made and adaptations occurred is information defined not before and second of all anything that works will do I think it's extremely important to imply and to infer that evolution means that there is no progress there might be increasing complexity but there's no value judgment about what's a better fit or not if it works it works in other words you don't design a bird's wing by aerodynamics what you do is you extend a leg if that works so I think this is a rather important point to make because it's commonly confusing to people who hear about selectional models they think it's stacked away like a kind of prior dictionary that isn't the case it is the potential for making a good enough choice that is built into biological systems Dr. Shank has quite a stack of questions from the audience I'm going to let him pick his favorite one and read it and then answer it I don't know if my favorite one should be ones that are funny or ones that I like to answer the first one says specific suggestions please if possible to encourage and stimulate learning in children particularly preschool age I picked that one as the one I wanted to answer partially because I actually have a company that produces educational software some of which is available to be shown at some place in this university at the moment but I'm not sure where but anyway there's some place where there are computers and somebody sitting there showing software and the reason I think it's worth mentioning in fact the reason I'm doing it I think about the kinds of things that I was talking about you realize that a lot of what intelligence is about is the ability to explain ourselves to self-reflect to have to ask the right questions at the right time it's possible to take advantage of the fact that there are personal computers all over the place these days and in the schools for no good reason to actually have them be in the schools for some real good reason and what you can do is you can design games that teach reflectivity or creativity or teach the ability to explain or bear upon the ability to explain and so what we've been doing is exactly that and we have games that teach reasoning ability and inferencing ability it's not that you in fact need AI to build those games you don't but the idea is that AI generates can help you have a sense of what there is to teach so I do think that although the current educational software available is awful that it's possible that in the near future there'll be some very good stuff on the market which reflects the view of intelligence that we have in AI let's see another one, do you ever foresee that you will ever be able to have computers that express emotion I want to read that one at the same time as another one that I have which is when computers at present generalize and achieve new insights isn't this generalization not real understanding not even mundane understanding but only something vaguely and crudely analogous to it this is a standard question that comes up suppose I got a computer to cry at the right moments or laugh at the right moments or be angry at the right moments in response to a story it's possible to build programs that behave that way and in fact express the appropriate emotion at the appropriate time I have often also known people who behave that way who have learned to express the right emotion at the right time and one often wonders if they're really feeling it and you would certainly be right to wonder that about a computer the issue is at what point would it in some sense really be feeling it I know that a piece of hardware will actually feel something so there's a sense in which that may be a silly question but there's a real important question underlying all that namely if a computer has a set of goals that it's trying to achieve and those goals are frustrated it might in trying to reorganize its information react in some way I'm not suggesting react angrily or happily but just to reorganize, to work harder to learn from the negative experience to that extent I think that computers will express and feel emotions they won't be the kind that we feel and express because I think that as living breathing organisms our emotions have other forms than pure cognitive form they have body forms bodily forms for example and computers being basically minds will only have these cognitive forms of emotions but they will be able in principle to be able to react to the frustration of their goals or the success of their goals by learning and that's a level of emotion which I think is important and with respect to this issue of understanding it's a very hard question to assess when you talk to another human being whether he's in fact understanding you I see all these faces out there and I have no idea whether you're understanding me it's a good guess that some are and some aren't but I assume that you're understanding me because you sort of look like people now if one of you turned out to actually be a machine to me I would still make these same assumptions about you if I found out it was you or a machine I because I'm a human being would immediately assume that well you're not really understanding because you're a machine but that's unfair that's a very sort of human chauvinistic view of life I do believe that machines will understand and are understanding to some extent whether they're understanding to an extent that makes me feel understood which is what I was trying to get at in my distinction about complete empathy that's another issue there's a lot of time with each of you individually to be able to feel in my own mind that you really understood what I was saying at that same level in other words I think understanding is a very subjective affair and there'll be those of us who never want to say computers are doing it and there'll be those of us who never want to say that any other person is doing it like nobody understands me I think that in fact there's a whole range of that kind of understanding that computers will get pretty close to what people do but probably you wouldn't want to marry one I'd like to ask Dr. Milner to answer a couple and then we'll come back to Dr. Edelman again about this question the question I received the question I just received the one you really have more questions about HM first of all a very simple one that how old he is he's 58 I used to go around saying this young man and I stopped saying that studying him all those years but the question that does HM know that he doesn't know i.e. is he aware of the nature of the problem this is an interesting question and the answer is yes it's very characteristic of these patients that they have insight the patients with the bilateral medial temporal lesions now HM's operation had been discussed for quite a while before he actually had it because it was really an experimental operation he knows he's had that operation but he knows he has bad memory he will a good example is that his mother used not to be able to leave him alone in the house because a stranger might come to the door like a brush salesman or something and HM would reason that he didn't recognize this man but he had a bad memory therefore it might be a friend of his mother's so he would invite this man in and the mother would come home and find a stranger that she didn't particularly want to see sitting in her living room so this is very clear and as long as he's working on a problem that is fine it has its concentration but that little moment when you are moving the test material and putting something else in place he becomes anxious, visibly anxious and he says well right now everything is clear but what I went before I have this trouble with my memory what have I done have I said something inappropriate so there is really the continuous awareness that he is not he says it's like waking from a dream I have been asked something else which is actually the negative findings are a little surprising I have been asked what does the change in brain function would you expect after a bilateral surgery and one might indeed think that as part of the system one would have some tenuated version at least but in fact Dr. Susan Corkin at MIT and all her group studied a whole series of patients who underwent singulotomy for the treatment of pain and they went to study these patients with a very strong prejudice that they would find all kinds of cognitive damage of various sorts and they really were not able to find anything with a very that is no proof I mean that doesn't mean the absence of evidence isn't evidence of no function of course but rather that what this does is presumably it would have to see in combination with some other other damage to get a clue in people as to what was going on I have been asked something which I frequently ask which is whether it's valid the question isn't quite in those words but whether it's valid to extrapolate from findings in patients with seizures obviously all patients populations the different kinds of patient populations on studies have certain built in disadvantages in terms of interpretation but I think all I would like to say now is that patients have been observed with spontaneously occurring illness and cephalitis affecting these regions of the brain they have been observed with trauma with damage to blood vessels with things that occur suddenly in what has been up to that point a healthy brain and there is concordance of the findings from different etiologies of lesions so I don't myself feel we have to worry too much about that Dr. Simon has an opportunity given a question we should like to hear from him please sir I guess it's my chance to let you hear my voice today do my hard work tomorrow someone asks what areas are suited for the development of expert systems we hear a lot about expert systems today in the area of applied AI what areas are not so suitable and what's necessary for the first list to grow larger I think the criteria for suitability of a system of building an expert system for some practical task is that the set of knowledge the range of knowledge required for the performance of that task have a reasonably definable boundary around it for example I think Roger Schenck illustrated today lots of tasks that would not be appropriate tasks for expert systems because they call for a very undefined set of everyday knowledge for their performance so when we look around the expert systems that have been built today which perhaps the best known example are medical diagnosis systems they are built on the foundation of a very specific and highly developed body of professional knowledge a good deal of which although not all of which can be verbalized by the experts and specialists in that field so I guess the question of what is necessary for the list of expert systems to grow is that the number of fields in which there are human experts who have that kind of knowledge should grow any of those fields is a candidate I think for expert system treatment Dr. Edelman I have two questions here that are related I'm not an expert in the field but I did mention the subject so I might expatiate a little on it the first one is how many compounds or different compounds are found in the area of a synapse that's a very good question indeed I mentioned only the word neurotransmitter and the chemical that was released when the electrical signal reached the area which one nerve touches the other in fact what's being discovered is an extraordinary number of so-called neurotransmitters I wouldn't say one every day but the classes of them are truly expanding at a fantastic rate the way I would answer the question is this that there are really two main kinds of transmitters one is called excitatory and the next neuron fire another is called inhibitory and decreases its tendency to fire but there are a whole set of compounds in fact some of them which are the peptide or protean analogs in terms of function of opium and morphine and things of that kind the so-called enkephalins and endorphins which act as modulators of the system so you not only have one kind but most of kinds and they differ in the brain in different regions the second thing that I think is worth saying is that people are studying the region in the second neuron that receives the transmitter the so-called receptor or channel now it's that channel which is changed by binding this neurotransmitter and makes the next neuron fire the interesting thing about that is that such channels are found not only in nerve cells but they lay on down in evolution and the number of them is absolutely staggering the one that amuses me the most is the one that makes the paramecium reverse its direction it's a so-called calcium channel when a paramecium bunks into something the calcium channel changes its conformation or shape in such a way as to reverse the direction of the hairs that make the paramecium swim evolution is simply incredible the major idea that you might take from this is this that the brain as Dr. Milner would describe it is in a tight bony box and after you develop all these billions of connections the fact is you can't really grow much more well then nature comes along with a very clever device you don't have just one neurotransmitter you change the grammar by having a whole bunch and use the neuronal circuits in different ways by changing the chemistry a beautiful example of that has been recently found by Eve Marder in the so-called stomatogastric ganglion of the lobster of all things the ganglion that controls how the stomach contracts in a lobster and that ganglion has different transmitters and depending on the ratio the same wiring will do two different things or three different things or four different things so on top of the billions of possibilities I mentioned this morning we have countless others I finally have a question which in fact was asked to me privately by Professor Dennett and I think is worth mentioning Dr. Edelman I'm reminded of your automaton model of the so-called perceptron in which the components had adaptive and selective properties my recollection is that it failed to behave quote intelligently unquote because of an overly simple organization would you comment please perceptron was an early pioneering attempt by Rosenblatt at Cornell to develop automata that did not have inherent programs but were responsible to limited environments the automaton that we have built differs in a great number of technical details from the perceptron one of the major ones is that it is so constituted because it's a selective machine and a perceptron is not has to be completely independent of the position of an object or whether it's rotated or not perceptron was simply a wiring scheme which could or could not be varied in which the response was reinforced either internally or by the investigator if the right response came along but if you move the object just half an inch that was the end of that you couldn't turn it it wasn't recognized in fact as a separate object our automata take account of that by a variety of devices that I won't mention finally someone asked me um if I have time um would I share another joke with you well well I'm at a loss in that respect but perhaps to summarize some of the feelings I have about some of the questions and answers I would just tell you the little tale of the lady who went to the psychiatrist and this is dedicated Dr. Shank in his marvelous stories and he said you say you have trouble making up your mind and she said yes and no Dr. Shank do you want to pop a few off the stack alright if a particular instance of failure expectancy which I suppose he means expectation failure must be understood at a higher level of generalization in order to match up with a new instance like the stake in a haircut example what is learned when the instances are compared the higher level generalization seems present already present to make the match if I wasn't clear on that it's a good thing to clear up what happens is that when you have an expectation fails you remember it I have a lot of cases of people remembering it for as long as 30 years and it lies around looking for a mate to compare it to so when you compare the stake in a haircut what you're doing is you're trying to see if the generalization you made on the haircut example namely that somebody didn't want to cut your hair that extreme in fact has some relationship to reality finding another one allows you to scope it and the stake example and that's also somebody doing something extreme it allows you to say the parameters on the people acting and the parameters on the doing are so the parameters on the people acting are people in a service position and the things they're doing are things that they are easily in the range of capability of doing and this verifies for you a hypothesis you've now had for 30 years sometimes you can hold this for only a week it verifies hypotheses for you so essentially what you're doing in learning is the creating of hypotheses on the basis of explanations of expectation failures and then the verifying of hypotheses by your next encounter with this situation so it isn't that it is in some sense sitting around in memory but it's waiting for verification we tend to learn I argue from two instances not from one another question that's here that I found amusing do you agree or disagree with the contention that a chess playing computer will never beat Bobby Fisher please defend your answer and then underneath that it says the answer we'll never know because Bobby Fisher would never agree to play this machine is unacceptable well okay I won't give that answer I think the whole idea of the development of a chess playing computer has been very interesting it has an interesting history in the early stages it was a that's something that computers that people do and that's smart people do so therefore if we had a computer that did it it would be smart so let's work on it later on it involved into a what can we get a computer to do to play chess and anything will do approach and the brute force approaches actually made chess playing programs that were fairly good but all their relationship to human performance and human intellectual activity was gone it was just a question of high speed computation later on particularly under the direction of Professor Simon we got much more intelligent approaches to the problem of chess playing where we were actually trying to learn what kinds of expertise a chess master might have and get that kind of information inside the machine but I think that there's a fourth stage in the development of chess playing computers that we haven't yet reached and that is the deriving of new rules by first principles such that the machine itself could be creative in its own approaches that is to say the invention of new strategies the thing that ultimately any really expert of any variety has like a grand master in chess or an expert football coach who is inventive or anything in between the ability to create something new in relationship to circumstances the problem with expert systems I don't know that they all work they work all that well but the extent to which they can never really work very well under their current conception is that they don't have the ability to remember what they didn't get right and improve on it by themselves they always require the human to come back in there and fix it for them I'm fond of pointing out that an expert medical diagnosis program that couldn't plan a wedding is an odd combination you can be an expert diagnostician but not be able to do something so common as plan a wedding why because the baseline simple information for planning isn't built into these expert systems it's only the expert information nothing of what really ought to underlie it so what I'm suggesting is that a chess playing program although it can be built in to put in all the rules we can figure out what the master might have and that's certainly a good strategy that the best strategy is get a chess playing program that behaves the way Bobby Fischer behaved right at the beginning namely he played a lot of chess he reasoned about playing chess he invented things himself which in fact may have been invented by others but he had to invent them for himself and this allowed him to continue to invent for himself so ultimately I think that it's creativity that's the issue in AI and that we really have only begun to look at that I have a few projects like that I do have a chess playing project where somebody's trying to have a program play other chess playing programs and learn how to play chess better by doing that I have a football coach program which is trying to invent new football plays I have a cooking program which is trying to invent new recipes and in each case these programs invent things and then they find out they didn't work so well expectation failures and go back and remember various expectation failures combine and make generalizations and then invent something new that people will have to solve I'm only just started on it I don't know how many other people have started on it but we're a long way from the solution I think maybe we should head over head over to dinner thank you for your questions they provide valuable feedback to us and thank you for your answers panelists