Memory, p.11
Memory, page 11
The influential child psychologist Jean Piaget, who began his career as a biologist and was much influenced by Baldwin, provided an early example of how the process might work.fn5 He studied pond snails in different Swiss lakes that differed in how much wave action occurred. When he transferred snails from one lake to another, the morphology of the shell was changed in response to the musculature required for the wave action of that lake. The descendants of the transferred snails eventually inherited the changed morphology without needing to be exposed to the wave action of that lake. In the terms used by Lloyd Morgan, the initial change could involve adaptability by the individual snail; the adaptability is won at some cost so that descendants expressing the character more efficiently would be more likely to survive. Piaget believed that he had found a clear example of the inheritance of acquired characteristics and never seems to have realised how his supposed Lamarckian effect could be Darwinised.
A well-known palaeontologist, George Gaylord Simpson, coined ‘the Baldwin effect’ as a term for organic selection, but he did so in order to disparage it.fn6 He and many others who forged the new synthesis of Darwinism did not think that behaviour played an important role in evolution and this became the standard line of neo-Darwinists. The dispute is about whether individual adaptability provided the leading edge for evolutionary change or whether it was both unimportant and, if it occurred, involved no new principles. As one critic of organic selection put it: ‘If learned behaviors are so effective in getting a useful trait passed from generation to generation at the cultural level, there will be presumably be no selection pressure for the spread of genetic factors favoring the trait.’fn7 In other words, if learning is so useful, why dispose of it? My view is that this criticism is based on an impoverished understanding of how behaviour is changed and controlled. The answer to those who think that the proposed evolutionary change would lead to a generalised loss of ability to learn is to state quite simply that it would not. Learning in complex organisms consists of a series of sub-processes. At the very least it involves detectors that respond to many different features such as lines, movement and more complex patterns such as eyes and faces; on the other side it involves particular motor systems that execute functionally coherent patterns of behaviour; and in between is a parallel array of intermediate systems that allow features to be combined in a host of different ways and connected to any of the executive systems controlling behaviour.fn8 Even in the simplest systems an array of detectors is linked directly to an array of executive mechanisms as well as indirectly through an intermediate layer and all connections are plastic. With such networks a particular feature detector can become non-plastically linked to an executive system in the course of evolution without any further loss of plasticity. The effect is that after organic selection, automating behaviour that was previously learned would have a trivial and unnoticeable effect on the animal’s capacity to learn new things.
The critics of organic selection also argued that if a characteristic form of behaviour were valuable then the species would acquire it anyway by natural selection, so invoking adaptability to set the process in motion was unnecessary. Once again the critics miss the point. The existence of a new pattern of behaviour, acquired by learning, sets an end-point against which the same pattern, developed without learning, must be compared. Dan Dennett surmised that he could artificially select a population of parrots to say ‘Boo Chomsky’ without instruction, but throughout the evolutionary process the goal would have to be specified.fn9 The chances that all the necessary mutations and genetic recombinations would arise at the same time are very small indeed. In the natural world, if an unlearned pattern of behaviour is not as good as the learned one in the sense that it is not acquired more quickly or at less cost, then nothing will happen. If it is better, evolutionary change is possible. The question is whether the unlearned behaviour could evolve without the comparison. If learning involves several subprocesses or several sequences then the chances against an unlearned equivalent appearing in one step are very small. However, with the learned behaviour as the standard, every small step that cuts out some of the plasticity with a simultaneous increase in efficiency is an improvement.
As an example of how the setting of an end-point might work, suppose that the ancestor of the Galapagos woodpecker finch, that pokes sharp sticks into holes containing insect larvae, did so by trial and error and its modern form does so without much learning. In the first stage, a naïve variant of the ancestral finch, when in foraging mode, was more inclined to pick up sharp sticks than other birds were. This habit spread in the population by Darwinian evolution because those behaving in this fashion obtained food more quickly. At this stage the birds still learn the second part of the sequence. The second step is that a naïve new variant, when in foraging mode, was more inclined to poke sharp sticks into holes. Again this second habit spread in the population by Darwinian evolution. The end result is a finch that uses a tool without having to learn how to do so. Simultaneous mutations increasing the probability of two quite distinct acts (picking up sticks and poking them into holes in the case of the woodpecker finch) would be very unlikely. Learning makes it possible for them to occur at different times. Without learning, having one act but not the other has no value. It goes without saying that learning involves memory, so memory becomes a crucial part of what I call ‘the adaptability driver’ of evolution.fn10
Before exploring further the importance of the adaptability driver I shall briefly discuss a concept called ‘genetic assimilation’fn11 often confused with organic selection. Frequent references are made in the biological literature to learned behaviour becoming an instinct by genetic assimilation. The claims are made casually and without much thought being given to the way in which the evolutionary process is meant to work. How is it supposed that a shake-up of development in one generation leaves the developmental process more likely to be shaken up in the next? The nature of this cumulative process is not explained. By contrast, the adaptability driver of organic selection involves necessary compensation for the effects of a new set of conditions and immediate response by the individual to the challenge. The accommodation is not inherited and differential survival of different genotypes may arise from subsequent differences in the ease with which the new adaptation is expressed. The empirical findings on which genetic assimilation was based also involve expression of a novel character in a new environment, but the character is not an adaptation to the triggering condition, even though in the regime of artificial selection used in the experiments, the character conferred some advantage on its possessor. The novel characters do not bear any functional relation to the conditions that disrupted normal development. Nor need there be such a relationship under natural conditions. All that is required initially is that the environmental conditions trigger the expression of a character that can be repeated generation after generation so long as the environmental conditions persist. The initial response of the animal in organic selection is fast whereas the developmental effects of exposing animals to abnormal environment were not seen until they were adult. Since most individuals will be adaptable, most will survive in the initial stages of organic selection. In the experiments giving rise to the idea of genetic assimilation those animals expressing the novel character, a subset of the total population, were artificially selected for further breeding. Finally, in the case of organic selection described by Lloyd Morgan, fresh phenotypic variation presumably arises by mutation that allows the adapted character to be expressed more easily and thence leads to differential survival. In Waddington’s experiments mutation was neither postulated nor needed.
By contrast, organic selection involves necessary compensation for the effects of a new set of conditions and immediate response by the individual to the challenge. The accommodation is not inherited and differential survival of individuals with different genes may arise from subsequent differences in the ease with which the new adaptation is expressed. The evidence offered for genetic assimilation also involves expression of a novel character in a new environment, but the character is not necessarily an adaptation to the triggering condition, even though it may confer some advantage on its possessor. In the case of organic selection most individuals will be adaptable and will survive through the initial stages of the evolutionary process. In the examples of genetic assimilation only a subset of the total population develops the new character, and these were artificially selected for further breeding. Finally, in the case of organic selection fresh phenotypic variation presumably arises by genetic mutation, allowing the adapted character to be expressed more easily and thence leading to differential survival. In the experiments used to discuss genetic assimilation mutation was neither postulated nor needed.
Returning to the adaptability driver, it is worth considering its relevance to human evolution. Why did the hominoid line double the relative size of the brain in two million years? Why is the linguistic capacity of a modern human so much greater than that of a chimpanzee? Where did consciousness come from (if such a question is amenable to a coherent answer)? The critical issues are not so much whether the various conjectured links between adaptability and evolution could work, but whether they could have driven evolution so hard and so fast as seems to be the case with the hominoid line. The combination of existing characteristics to provide something new may provide an answer.fn12 Non-human primates are known to make signals that are reliably linked to their internal state. A variant might have arisen, producing new signals when in a certain state not previously associated with a signal. This could matter a lot in situations where the success of the group depended on co-operation and trust. In these circumstances, a powerful upward evolutionary drive could be generated. If members of a group were able to link the signal by an individual to the context in which it was produced and then could predict the behaviour of the signaller, the capacity to generate new signals linked to the signaller’s state could spread in the population. This enhanced capacity could then have the emergent effect that individuals would be able to create signals in contexts that were both novel and abstract. At each stage, social learning could spread the benefits within the population, thereby accelerating the evolutionary process. New signals could be copied and linked to context and thence to internal state. Here again learning and memory are crucial in driving the evolutionary ratchet.
Another line of thought also linking memory to evolution is that play boosted the evolution of cognitive capacity. The playful use of objects can lead to discovery and result in their proving useful as tools. In what is metaphorically called the adaptive landscape, a species may be trapped on a low hill when a higher one (and better, in terms of the image) can only be reached by going down before starting to climb to the more beneficial place. The exuberance of play may enable an individual to discover the route because play will often involve leaving the hilltop on which the species is trapped. Once discovered, the behaviour patterns that led to the beneficial outcome can then be automated piecemeal along the lines suggested for the tool-using in the Galapagos woodpecker finch. On this argument those aspects of play that are creative or break out of local traps are especially promising candidates for driving evolution. Once again memory is the crucial link in the driving process. When complex sequences of behaviour appear, their components are gradually automated by Darwinian selection. The spontaneously expressed improvements in what could be readily perceived as cognitive ability would not have evolved by genetic recombination or mutation since the probability of the simultaneous occurrence of all the rare necessary events is vanishingly small. Thus aspects of play can increase the total sum of spontaneously developing behavioural structure that serves to solve complex problems. The adaptability driver of play increases the cognitive tool kit available to the young animal.
As with other aspects of evolution, no necessary link need exist between adaptation and the predictability of behaviour. The evolutionary process can establish rules that affect how the individual changes its behaviour in response to new conditions. To understand the opportunities that this regularity opens up, consider a rule-governed game like chess. It is impossible to predict the course of a particular chess game from a knowledge of the game’s rule. Chess players are constrained by the rules and the positions of the pieces in the game, but they are also instrumental in generating the positions to which they must subsequently respond. The range of possible games is enormous. The rules may be simple but the outcomes can be extremely complex. Once the route to a particular outcome has been found, it can be automated.
The most important conclusion here is that the adaptability driver involving memory provides a ladder in evolution. Clearly, complex structures can develop without such a process, but the driver is important when intermediates provide no benefit and a combination of simultaneous mutations needed to provide a functional whole is improbable. Whole organisms survive and reproduce differentially and the winners drag their genes with them. This is the engine of Darwinian evolution and the reason why it is so important to understand the role of learning and memory in biological evolution of complex behaviour.
Ulric Neisser, ‘Memory with a Grain of Salt’
In a tradition that goes back to ancient Greece, popular accounts of memory tend to be metaphors. Plato himself likened it to the writing on a wax tablet, which was the information technology of his time. His tablet metaphor is just one version of memory as writing, which we still use today when we speak of remembering something ‘literally,’ i.e. exactly, by the letters. But real remembering is almost never literal: people soon forget the exact language they may have heard. At best they remember the ‘gist,’ and even that is far from certain.
The next significant memory metaphor to appear was based on photography, invented early in the nineteenth century. Photographs reproduce scenes accurately and completely, but also with a curious indifference: every blade of grass that was projected to the lens ends up in the picture, whether the photographer was interested in it or not. Does anybody have photographic memory in that sense? I doubt it. The term is occasionally used, but I think that’s only because the metaphor is so readily available. Before photography nobody described remembering as if it were a kind of looking at mental pictures, or claimed that they could recall every blade of grass in a scene. To be sure, some people had good memories. Some people have good memories today too, and they may even claim that they can see past scenes in their ‘mind’s eye.’ But no matter how vivid and clear those images may be, the reports that go with them are not necessarily accurate. Even the best memory image is nothing like a photograph.
The next relevant bit of technology, a century or so after the camera, was the tape recorder. Is there perhaps a recorder in the brain, such that our every experience is somehow preserved on its tape? The Canadian neurosurgeon Wilder Penfield, working in the 1940s, became convinced of this by observations he made while electrically stimulating the brains of conscious epileptic patients. When a certain cortical locus was stimulated, such a patient might say that she heard a familiar piece of music playing, or perhaps a child calling. Despite years of study, none of those experiences was ever validated as being an accurate recall or even as having the characteristics of a remembered episode (as opposed to, say, a dream image). Penfield nevertheless attributed them to an automatic neural tape recorder, the ‘permanent record of the stream of consciousness.’fn1 The newly available tape-recorder metaphor was irresistible.
Nowadays, of course, we have videotapes as well as audiotapes. Professional hypnotists, occasionally hired by the police in criminal investigations, may tell hypnotised witnesses that they need only ‘zoom in’ on the relevant ‘frames’ of their memories to see again what they saw before. But in fact there are no videotapes in our brains, just as there are no wax tablets. What such witnesses may say is just their idea of how to respond in a manner driven by the metaphor. Happily, most American jurisdictions no longer allow hypnotically obtained testimony to be used in court.
All these metaphors compare memory to some permanent medium of storage: written documents, photographic film, magnetic tape. Such a comparison seems harmless enough, but once the metaphor is in play we tend to endow memory itself with properties that only the medium really has: permanence, detail, incorruptibility. Cautious skepticism would be an appropriate response when the witness gives a highly detailed report of what the perpetrator was wearing, describes a once briefly seen object in great detail, or recalls the special tone of someone’s voice on a given occasion. Unfortunately the usual response is just the opposite: jurors are much more likely to believe a richly elaborated recall than a bare-bones account of what happened. They say ‘Listen to all those details! He couldn’t be making it all up!’ But he could: data show that there is no consistent relation between the amount of detail reported in a memory and its accuracy.
Misleading as they may be, there is still something comforting about such metaphors. Our past experiences are the very life we lived; how wonderful if they were really on tape somewhere! This line of thought is encouraged by the occasional occurrence of unexpected recollections. We say, ‘I haven’t thought of that in years! If one long-lost memory like that is still somewhere in my head, couldn’t everything else be in there too?’ Rhetorical questions are difficult to answer; sometimes the best reply is just another metaphor. Imagine buying an old house, tearing up the floorboards, and unexpectedly finding a copy of a newspaper dated forty years earlier. You would probably be amazed and interested, read every page, tell your friends. But would you conclude that every issue of that newspaper ever printed was somewhere in your house? I think not.
