Memory, p.10
Memory, page 10
So can we conclude that we have found the engram – or at least identified the processes whereby engrams are constructed? The suggestion that memories are encoded in terms of changed synaptic connectivities has certainly proved attractive to a new breed of researcher, who call themselves computational neuroscientists, interested in making mathematical and computer models for how learning might occur in a distributed neural network connected by a mesh of synapses. In such a theoretical network, each memory (or association) is represented by activity in a specific set of synapses, a unique pattern, but any one synapse can be involved in many different such associations. On this basis, and estimating the number of synapses that it contains, Edmund Rolls has calculated that the hippocampus can store some 36,500 memories.
But a calculation of this sort is based on a prior set of assumptions: that biological memories can be decomposed into isolated monads and measured in terms of the bits and bytes with which computer people calculate the power of their machines. It is this that is so unrealistic; how many bits of information does the variety of memories listed by St Augustine require? For that matter, how many bits of information do my chicks need to remember to avoid pecking at a small red bead but know that it is safe to continue pecking at a yellow one? The chick categorises the experience of pecking the bitter bead in terms of the colour, shape and size of the bead, the context in which it was pecked, its own past experience of pecking other beads, and probably many other features as well, any one of which may provide the cue for its subsequent behaviour. I am far from sure that, for the chick, this complex of meanings within which any subsequent sight of and response to the bead is embedded is simply decomposable into information theory’s bits.
Indeed, this theoretical concern is rapidly confirmed by experiment. The linear cascade that the biochemical and pharmacological experiments demonstrate, leading from transient changes in the release of neurotransmitters to seemingly permanent structural changes in the synapse, was no sooner established than paradoxes began to appear in the data. Memory traces apparently firmly located in one brain region seem over the subsequent hours and days to migrate to others – as indeed might have been suspected from HM’s experience. His hippocampal damage did not erase old memories, only prevent new ones being formed. Furthermore, there is no single site for ‘the memory’ as if it constituted a discrete entity. The MEG experiment I referred to above shows that many brain regions are involved in the dynamic process of recalling and responding to prior experience. And even for my chicks we have been able to show that different aspects of the memory of the bitter bead – its colour, shape, size – engage different ensembles of nerve cells and synapses distributed in different regions of the brain.
Furthermore, memory involves more than just synapses, or even just brains. How well a person or a chick learns and remembers depends on many other aspects of body state. Alertness and attention depend on physiological processes such as blood flow and hormonal level. Memory involves emotion as well as cognition, and hormones produced outside the brain, notably adrenaline and its neurotransmitter relative noradrenaline, are engaged in determining what is remembered. When chicks peck a bitter bead, there is a surge of steroid (corticosterone, the chick’s equivalent of cortisol in humans) release into the bloodstream. Too little or too much corticosterone, and the chick will not remember the experience, and will peck the previously bitter bead when tested later. In this sense learning and remembering – memory – is a property not of individual synapses or nerve cells or brains, but of the entire organism, the person.
Nor is this all. Hebb’s model is one of learning: what happens when an animal, or human, registers some new experience. Implicit in it is also a theory of recall: that remembering the experience involves reactivating the novel pathways that learning has generated. Memories are stored as in computer files, and remembering would seem to be no more than pulling these files out of deep store and reopening them. But this mechanical model won’t do. Each act of recall is itself a new experience. Reactivated memories are subtly changed each time we recall them. Classroom experiments beautifully illustrate what we all know to be the case. Thus in the aftermath of the disaster which destroyed the US Challenger rocket and killed the astronauts in it, a group of psychology students were encouraged to write down their recollections of the event. The records were stored, and a year later they were asked to write the account again. The huge discrepancies between their first and second accounts indicated just how labile memories of quite dramatic events are. Far from passively recording the past, we in our memories actively reconstruct it.
Very recently, neurobiology has begun to catch up with common experience and the psychologists. Many labs, including our own, have now shown that when an animal is given a reminder of a previously learned experience, the memory becomes labile once more, and can be disrupted by drugs and biochemical inhibitors rather as it can be during initial consolidation. Some researchers have begun to speak of this as ‘reconsolidation’. However, the temporal dynamics of reconsolidation are rather different from those during consolidation; different brain regions are involved, and the biochemical changes do not exactly recapitulate those of consolidation.
Of course, being reminded of a past experience is itself to some extent a novel experience. We don’t step into the same stream twice, and memory depends on history. That neurobiologists have only so recently come to realise this shows just how blinkered and reductionist their – our – paradigms have been. We are trapped by the experimental need for simple and reproducible designs, for operationalising our definitions of ‘learning’, ‘memory’ and so on as if these complex processes could be trapped within small boxes, sealed off from everything else that is going on in a living, behaving, learning and remembering organism throughout every moment of its existence. Our experiments capture only a small part of such complexity, and we are at fault if we mistake this small part for the whole.
Half a century ago, neuroscience saw the brain as composed of discrete centres, regions responsible for vision, audition, pain, memory and so forth. Superimposed on all these different regions was a super-coordinating centre, the association cortex. Separate regions reported upwards to this coordinator, which assembled them and instructed the motor regions of the brain how to respond. This homunculus inside the head was the source of identity, individuality, the ‘I’ located a few centimetres behind our eyes.
Alas for simplicity, there is no such homunculus. As Gertrude Stein said about, I believe, downtown Los Angeles, there is no there there. Brains don’t have a central processor, a super-manager controlling everything. Rather they are distributed networks of cellular ensembles, richly interconnected, which between them create the illusion of coherent experience that we all in our normally functioning moments share. The enigma of memory, as with so many aspects of brain processes, seems to be that it is both localised and non-localised. Remembering is at once sure and certain, as when we recall the names of the days of the week, or mount and ride off on a bicycle for the first time for many years, and as evanescent and elusive as a soap bubble, as when we try to remember the first moment we saw a lover and compare our own memory of that event with his or hers.
Fanny Price was surely right. Which is why we neurobiologists of memory must from time to time come out of our labs, reflect on our own varied procedural, declarative, episodic and autobiographical memories, and turn to the work of those philosophers, poets and novelists who can illuminate and interpret our experience so much more richly and meaningfully than can the most ingenious experimenter.
Patrick Bateson, ‘Memory and Evolution’
If you are lucky enough to visit one of the great game parks of East Africa and witness a cheetah stalking a gazelle, you may see something very surprising. The gazelle on which the crouching cheetah has fixed her gaze may suddenly jump into the air. The cheetah relaxes and turns her head, searching for other gazelle. The curious leap of the gazelle when approached by a predator is known as ‘stotting’. The leap indicates to the cheetah, or so we suppose, that she is less likely to catch that gazelle than one that does not jump or does not jump so high.
What has this to do with memory and evolution? Anyone brought up on popular accounts of biological evolution will have been told that the record of what made ancestors successful lies in the genes and that the genes, being immortal, are the drivers of change. In this case, however, the cheetah does not have to change genetically. All that individual animals have to do in each generation is to profit from their experiences. Chasing gazelle that leap into the air is a waste of breath. Genetic change in the gazelle will occur because the important thing is to stand out from the crowd. Those animals that signal ineffectively to the predator that they have spotted the predator will be less likely to survive. By degrees, the standard for jumping will be raised even though at any one time the young, the sick and the lame will always be prime targets for the cheetah. The evolutionary pressure on the gazelle to make their jumps ever more conspicuous than those of their peers will be driven by the cheetah’s preferences and these will be driven by her memory of past failures. I should make clear at this point that by memory I do not mean ‘represented in the genes’ in some metaphorical sense. I mean memory in its generally accepted sense of experience represented in the nervous system. The experienced cheetah’s choice of prey depends on such memory.
The role of choice in evolution was clearly recognised by Charles Darwin in his principle of sexual selection.fn1 His idea was that the female peacock, say, chose the male with the biggest tail. She would have sons with bigger tails. These sons in their turn would subsequently be chosen by females with similar preferences for males with the largest tails. The evolutionary pressure would be for bigger and bigger tails since the males with the biggest tail would mate with the most females. Moreover the females that made such choices would have more grand-offspring than those that did not. Eventually the trend to produce ever greater tails would halt when the male could no longer carry it safely. Males with tails that were too large could no longer rise from the ground quickly enough to escape the jaws of a fox. Most biologists suppose that the evolutionary process of sexual selection, as Darwin called it, involves two genetic changes, one in the male giving rise to a bigger tail (or whatever) and one in the female giving rise to a preference for an ever more striking adornment. If that is the case, then no memory is involved. However, another possibility is that the female’s choice is affected by her memory, as in the case of the cheetah.
The activities of animal breeders can be used to illustrate the effects of memory in mate choice. The appearance of the Siamese cat has changed astonishingly in a hundred years as the result of human intervention. From being an apple-headed animal with a robust body it has become long-muzzled and cadaverously thin. My guess is that this change was not the result of a conscious desire to alter the breed with a particular pre-determined goal in mind. Rather, judges at cat shows grew bored with the appearance of ever more perfect apple-heads and awarded prizes to animals that were slightly different from what they had seen so often. Cat breeders were quick to note the judges’ preferences and picked from among their kittens those that most nearly matched the new fashion. In fashion, things never stand still. As with the plunging hemline, the Siamese cat’s nose became longer as the breeders attempted to catch the judge’s eye with their latest would-be champions. What is helpful about the analogy is that trend over time only requires genetic changes in the cats. The drivers of the change are the judges’ memories and their interest in slight novelty.
Just as genetic changes in the judges are not required to explain the astonishingly rapid changes in pedigree cats, so genetic changes in female peahens are not required to explain the extremely large tails of the males. Females may not know who their fathers are but they certainly will recognise their brothers. When they become sexually mature, they will do what so often happens in mate choice, preferring mates who are a bit different but not too different from the brothers who are most familiar to them.
The interest in slight novelty need not lead to a consistent trend over generations. However, if the attention of the female is drawn to the male who is not only slightly different from her brothers but is also more conspicuous, then an evolutionary trend is likely. In the case of the peacock the more successful male is characterised not only by a larger tail but also by a tail with more eye spots on it. Birds find eye spots especially attention-grabbing and, indeed, some of their prey such as moths use the sudden flashing of eye spots on their wings as a means of deterring an attack by startling their insectivorous avian predators. The essence of the argument is that a startling evolutionary change in male form can be driven by female memory. To be sure, genetic change is required to produce ever more conspicuous males, but the evolutionary process is rapid because improvements in their ornaments do not need to be locked into genetic changes in the females’ preferences.
Whether or not my conjectures about the role of memory are correct, biologists are generally agreed that animals make active choices and the results of their choices have consequences for subsequent evolution. Three additional proposals have been made for the ways in which an animal’s behaviour could affect the subsequent evolution of its ancestors. First, by their behaviour, animals change the physical or the social conditions with which they and their descendants have to cope and thereby affect the subsequent course of evolution – sometimes referred to as ‘niche construction’. Secondly, by their behaviour animals often expose themselves to new conditions that may reveal heritable variability and open up possibilities for evolutionary changes that would not otherwise have taken place. Finally, and perhaps most interestingly, animals are able to modify their behaviour in response to changed conditions; this allows evolutionary change that otherwise would probably have been prevented by the death of the animals exposed to those conditions. The last one is most relevant to my general theme of the role of memory in biological evolution.
Modern thinking about the importance of behavioural plasticity in evolution is usually thought to stem from James Baldwin, but the Russell family’s tutor, Douglas Spalding, advanced very similar ideas twenty-three years before him,fn2 and in 1896 two others, Lloyd Morgan and Osborn, independently developed ideas about ‘organic selection’, as the subject was called at the time, at the same time as Baldwin.fn3 Lloyd Morgan’s account of the process was particularly clear and may be paraphrased as follows:
(a) Suppose that a group of organisms that are capable of change in their own lifetimes are exposed to new environmental conditions.
(b) Those whose ability to change is equal to the occasion survive. They are modified. Those whose ability is not equal to the occasion are eliminated.
(c) The modification takes place generation after generation in the changed environmental conditions, but the modification is not inherited. The effects of modification are not transmitted through the genes.
(d) Any variation in the ease of expression of the modified character which is due to genetic differences is liable to act in favour of those individuals that express the character most readily.
(e) As a consequence, an inherited predisposition to express the modifications in question will tend to evolve. The longer the evolutionary process continues, the more marked will be such a predisposition.
(f) Thus plastic modification within individuals might lead the process and a change in genes that influence the character would follow; the one paves the way for the other.
It is obvious, from this outline of the proposed process, that Lloyd Morgan was not suggesting a Lamarckian genetic inheritance of acquired characters as a mechanism. Nor, less obviously and less consistently, were Osborn or Baldwin. The crucial postulate is a cost of operating the original process of phenotypic adaptation, a cost that can be subsequently reduced by genotypic change, enabling Darwinian evolution to occur. Adaptability to new conditions might be physiological, such as coping with high altitudes by enhancing the oxygen-carrying capacity of the blood. Alister Hardy, a much-loved Professor of Zoology in the middle of the twentieth century, did more than most in the intervening years to pick up on this point and stress that the process could be of great significance. He put it as follows:
If a population of animals should change their habits (no doubt often on account of changes in their surroundings such as food supply, breeding sites, etc., but also sometimes due to their exploratory curiosity discovering new ways of life, such as new sources of food or new methods of exploitation) then, sooner or later, variations in the gene complex will turn up in the population to produce small alterations in the animal’s structure which will make them more efficient in relation to their new behaviour pattern; these more efficient individuals will tend to survive rather than the less efficient, and so the composition of the population will gradually change. This evolutionary change is one caused initially by a change in behaviour.fn4
Hardy envisaged a cascade of changes flowing from the initial behavioural event. Even without structural change, control of behavioural development might alter over time. A requirement for this to happen is that adapting to the new conditions be more costly than doing it easily without active modification. One instance might involve differential responsiveness to particular types of food. Many cases of choice of an unusual food for a given species are probably not due to genetic changes, but to the functioning of normal mechanisms in unusual circumstances. A group of animals might be forced into living in an unusual place after losing their way, but they cope by changing their preferences to suitable foods that are locally abundant. Later, those descendants that didn’t need to learn so much when foraging might be more likely to survive than those that could show a fully functional behavioural repertoire only by learning. A cost was incurred in the time taken to learn. As a consequence, what started as a purely learned difference between animals of the same species living in different habitats became a difference that developed without learning.
