ANIMAL LEARNING

the alternation of behaviour as a result of individual experience. When an organism can perceive and change its behaviour, it is said to learn. That animals can learn seems to go without saying. The cat that runs to its food dish when it hears the sound of the cupboard opening; the rat that solves a maze in the laboratory; the bird that acquires the song of its speciesthese and many other common examples demonstrate that animals can learn. Yet what is meant by saying that animals can learn? What, in other words, is learning? This question proves exceedingly difficult to answer, and, in fact, some theorists propose that no single, all-encompassing definition of learning is at all possible. Moreover, a moment's reflection yields the realization that there exist different kinds of learning. The learning of number concepts, for example, surely seems to be of a different nature than the learning of the association between the sound of a cupboard door and the receipt of food. To explore animal learning, then, this article first considers what learning is and is not and then examines in detail some of the specialized types of learning that occur in animals. Additional reading General works Historical background is provided by Robert Boakes, From Darwin to Behaviourism: Psychology and the Minds of Animals (1984). See also T.R. Halliday and P.J.B. Slater (eds.), Animal Behaviour, vol. 3: Genes, Development, and Learning (1983); Robert A. Hinde, Animal Behaviour: A Synthesis of Ethology and Comparative Psychology, 2nd ed. (1970); J.E.R. Staddon, Adaptive Behavior and Learning (1983); and David McFarland, Animal Behavior: Psychology, Ethology, and Evolution (1985). Simple nonassociative learning Gabriel Horn and Robert A. Hinde (eds.), Short-Term Changes in Neural Activity and Behaviour (1970); and Harman V.S. Peeke and Michael J. Herz (eds.), Habituation, 2 vol. (1973). Associative learning and conditioning Anthony Dickinson, Contemporary Animal Learning Theory (1980); Michael Domjan and Barbara Burkhard, The Principles of Learning and Behavior, 2nd ed. (1986); N.J. Mackintosh, The Psychology of Animal Learning (1974), and Conditioning and Associative Learning (1983); and Barry Schwartz, Psychology of Learning and Behavior, 2nd ed. (1984). Biological functions of and constraints on learning Robert A. Hinde and J. Stevenson-Hinde (eds.), Constraints on Learning: Limitations and Predispositions (1973); Timothy D. Johnston, Contrasting Approaches to a Theory of Learning, Behavioral and Brain Sciences, 4(1):125173 (March 1981); and J.R. Krebs and N.B. Davies, An Introduction to Behavioural Ecology (1981). Physiological basis of learning Daniel L. Alkon and Joseph Farley (eds.), Primary Neural Substrates of Learning and Behavioral Change (1984); and Eric R. Kandel, Cellular Basis of Behavior: An Introduction to Behavioral Neurobiology (1976). Spatial learning and navigation P. Robin Baker, Bird Navigation: The Solution of a Mystery? (1984); John O'Keefe and Lynn Nadel, The Hippocampus as a Cognitive Map (1978); and K. Schmidt-Koenig and W.T. Keeton (eds.), Animal Migration, Navigation, and Homing (1978). Song learning and imprinting P.P.G. Bateson, The Imprinting of Birds, in S.A. Barnett (ed.), Ethology and Development (1973); Howard S. Hoffman, Experimental Analysis of Imprinting and Its Behavioral Effects, The Psychology of Learning and Motivation, 12:139 (1978); Donald E. Kroodsma, Aspects of Learning in the Development of Bird Song, in Gordon M. Burghardt and Marc Bekoff (eds.), The Development of Behavior: Comparative and Evolutionary Aspects (1978); and P.J.B. Slater, Bird Song Learning: Theme and Variations, in Alan H. Brush and George A. Clark, Jr. (eds.), Perspectives in Ornithology (1983). Comparative psychology and complex training E.M. Macphail, Brain and Intelligence in Vertebrates (1982); R.E. Passingham, The Human Primate (1982); and H.L. Roitblat, T.G. Bever, and H.S. Terrace (eds.), Animal Cognition (1985). Nicholas John Mackintosh Types of learning Spatial learning One of the major problems many animals must confront is how to find their way around their worldfor example, to know where a particular resource is and how to get to it from their present location, or what is a safe route home to avoid a predator. Such spatial learning may cover only the highly restricted confines of an animal's home range or territory, or it may embrace a migration route of several hundreds or even thousands of miles. Although some forms of navigational behaviour may be explicable in relatively simple terms, not necessarily requiring appeal to processes more complex than those of simple conditioning, others suggest some quite new principles. Maze learning In the psychologist's laboratory, the primary method of studying spatial learning has been to put a rat in a maze and watch how it finds its way to the goal box, where it is fed. As befits the analytic (some would say sterile) approach so popular in experimental psychology, the elaborate and complex mazes used in earlier studies (the very first published experiment used a scaled-down replica of the maze at Hampton Court, London) soon gave way to something very much simpler, a T-maze or Y-maze. A rat placed at the end of one arm must run to the central choice-point, from where it has to enter one of the two remaining arms. Although extremely simple, even this apparatus allows for a number of possible modes of solution. One possibility is that the rat learns to execute a particular response, a left turn or a right turn, at the choice-point, because that response is followed by food. A second possible solution is that the rat learns that the two alternative arms differ in some particular way and further learns to associate one of the arms with food and hence to choose it. The third and most interesting possibility is that the rat learns to define the rewarded arm not in terms of its own intrinsic characteristics but by its spatial relationship to an array of landmarks outside the maze. Thus the rat might learn that the correct arm is the one pointing to the left of a window and away from a table with a lamp on it. Experiments show that whenever such landmarks are available, this third solution mode is the one used. Perhaps the most convincing demonstration that rats can find their way to a particular locationone defined solely in terms of its spatial relation to various external landmarkshas been provided by experiments in which the animals are placed in a large circular tank of water and must swim to a transparent platform submerged somewhere in the middle of the tank. They can rapidly learn to do this, regardless of where they are initially put into the tank and even though the platform itself is invisible. (The invisibility of the platform is shown by the following: if the platform is moved, the rat will swim straight past it, heading instead toward the position it used to occupy.) Rats in these experiments are not simply approaching a single landmark; they locate their goal by reference to its spatial relationship with a whole series of landmarks, no one of which is necessary. This can be established by using half a dozen arbitrary but easily identified objects as landmarks during maze training. Removal of any one or two of them in no way disrupts the rat's behaviour. If all the landmarks are systematically rotated around the room, the rat will identify a new arm of the maze as correct (the one that has the same relationship to the landmarks as the initially correct arm). If, however, the landmarks are rearranged in such a way as to destroy their original spatial relationship to one another, the rat does not know which arm to choose. The processes involved in this sort of learning are not well understood. Some psychologists have been sufficiently impressed by the rat's flexibility in these experiments to argue that the animal is constructing a map of its environmentnot, obviously, a written map but an internal, maplike representation that encodes a complete set of spatial relationships between major landmarks. The best evidence for such a maplike representation would be if a rat could take an unfamiliar route when its original route to a goal is blocked. Unfortunately, there is little evidence of such performance in rats, except in the not especially critical case where the goal, or a stimulus very close to it, is clearly visible from the choice-point. On the other hand, studies of long-range navigation have shown that some animals can do just this. Types of learning Simple nonassociative learning When experimental psychologists speak of nonassociative learning, they are referring to those instances in which an animal's behaviour toward a stimulus changes in the absence of any apparent associated stimulus or event (such as a reward or punishment). Studies have identified two major forms of simple nonassociative learning, which are to some extent mirror images of one another: habituation and sensitization. Habituation A classic example of habituation is the following observation on the snail Helix albolabris. If the snail is moving along a wooden surface, it will immediately withdraw into its shell if the experimenter taps on the surface. It emerges after a pause, only to withdraw again if the tap is repeated. But continued repetition of the same tapping at regular intervals elicits a briefer and more perfunctory withdrawal response. Eventually, the stimulus, which initially elicited a clear-cut, immediate response, has no detectable effect on the snail's behaviour. Habituation has occurred. Habituation can be defined in behavioral terms as a decline in responding to a repeatedly presented stimulus. As such, it is a very widespread phenomenon, one that can be observed in animals ranging from single-celled protozoans to humans. Most animals behave differently to novel and familiar stimuli: the former sometimes elicit startle responses, sometimes investigatory or exploratory responses; the latter often apparently are ignored. The suggestion that habituation is a simple form of learning, however, implies that it can be distinguished from some even simpler potential causes of this sort of change in behaviour. One reason why an animal might stop responding to a stimulus is that it no longer detects the stimulus; i.e., some form of sensory adaptation might have occurred. Another potential cause is fatigue: perhaps some temporary refractory state is produced by repeated elicitation of the same response, making it impossible to perform that response again. Whether or not one would want to call either of these processes a form of learning is doubtful. But both behavioral and physiological evidence establishes that habituation cannot be explained in these terms. The critical behavioral evidence is that habituation can be disrupted by almost any change in the experimental conditions. If repeated presentation of one stimulus leads to habituation of a response, the same response can still be elicited by a different stimulus. Even if the experimenter presents a novel stimulus that does not itself elicit the response in question, its presentation may restore the response on the next trial in which the originally habituated stimulus is presented. This latter observation, usually referred to as an instance of dishabituation, seems to rule out any simple sensory adaptation; both observations rule out simple effector fatigue. Neurophysiological analysis of habituation in various mollusksfor example, in the sea snail Aplysiahas confirmed that habituation need not depend on changes in the activity of sensory or motor neurons. In the case of Aplysia, researchers have studied the gill withdrawal reflex, a response that rapidly habituates to repeated stimulation of the snail's siphon or mantle shelf. But habituation still occurs even if it is elicited by direct, electrical stimulation of the motor nerve, bypassing the sensory receptors completely; and recording from the sensory nerve during normal habituation reveals no decline in its level of activity. These observations eliminate sensory adaptation as a possible cause of the animal's having ceased to respond to the stimulus. Effector fatigue can be ruled out by showing that direct stimulation of the motor neurons controlling the withdrawal response can still elicit a perfectly normal reaction even after the response has completely habituated. Research shows that habituation in Aplysia depends on changes in the activity of more central neurons. Repeated tactile stimulation of the siphon, leading to habituation of the withdrawal response, causes changes in the activity of the motor neurons innervating the response. Specifically, these motor neurons show a decline in excitatory postsynaptic potential, which is the electrical change that enables the nerve impulse to cross the gap (synaptic cleft) that separates one neuron in the pathway from the next. The decline in excitatory postsynaptic potential short-circuits the response. Moreover, the presentation of a novel stimulus, sufficient to dishabituate the behavioral response, restores the postsynaptic potential. Habituation occurs even in animals without a central nervous systemprobably in single-celled protozoans; certainly in animals such as the coelenterate Hydra, which have a diffuse nerve net and do not appear to be capable of associative learning. Among mammals, habituation of certain reflex responses can be observed even in spinal subjects, that is, those whose spinal cord has been severed from the brain. There can be little doubt, then, that habituation is not only widespread, but that it also can be a relatively simple phenomenon. There is, however, no guarantee that it is the same phenomenon wherever it appears. The waning response to a repeatedly presented stimulus admits of a number of different explanations. In principle, as we have already seen, it might be due to sensory adaptation, effector fatigue, or a more central neural change. These distinctions make rather little sense in the case of a single-celled animal. And one should not necessarily expect the habituation observed in a spinal mammal to involve precisely the same mechanisms as those responsible for comparable behavioral effects in an intact animal. Some psychologists have proposed theories of habituation that appeal to processes of classical conditioning. Such a theory is not likely to apply to the habituation observed in an animal that shows no capacity for classical conditioning. Habituation is usually, as here, classified as an instance of simple, nonassociative learning. It is supposedly nonassociative because all that happens in the course of habituation is that a stimulus is repeatedly presented and the animal's behaviour changes; there is, on the face of it, no other event with which the stimulus can be associated. Habituation must therefore, it appears, be understood by reference to some change in the pathway between stimulus and response, and the work with Aplysia and other mollusks shows how this analysis may proceed at the physiological level. But if habituation is not always the same phenomenon, it is possible that different processes may underlie the habituation of the startle response to a loud noise in an intact mammal. And despite appearances to the contrary, those processes may involve some associative learning. One suggestion is that novel stimuli elicit a biphasic response: an initial increase in startle responses, which include components of emotion or anxiety, followed by a rebound in the opposite direction. Habituation occurs when the latter, rebound response becomes conditioned to the stimulus, occurring sooner and sooner with each repetition of the stimulus and thereby damping down and eventually canceling out the initial reaction. An alternative possibility is that long-term habituation depends on associating the repeatedly presented stimulus with the context in which it occurs, a suggestion that would explain why presentation of the stimulus in a different context sometimes leads to dishabituation. The generality of habituation implies that this behavioral phenomenon has considerable adaptive significance; if true, it would be quite reasonable to expect that a number of different mechanisms might have evolved to produce the behavioral result. The adaptive value of habituation is not difficult to see. A novel stimulus may signify danger, and an animal should react to this stimulus either by withdrawing or at least by orienting toward it to see what will happen next. But if the same stimulus occurs again with no further consequence, it is probably safe: regular repetition of the same stimulus implies that it is part of the background, such as the waving of a branch in the wind or the shadow caused by a piece of seaweed floating with the waves. If the stimulus is not dangerous, time should not be wasted on it. Withdrawal, especially in the case of a snail into its shell, is a time-consuming effort, incompatible with such vital activities as searching for food. If it is important, therefore, for animals to be wary of novel stimuli, it is equally important that they should discriminate the novel and potentially dangerous from the familiar and probably safe. Types of learning Complex problem solving Experimental psychologists who study conditioning are the intellectual heirs of the traditional associationist philosophers. Both believe that the complexity of the human or animal mind is more apparent than realthat complex ideas are built from simple ideas by associating simple elements into apparently more complex wholes. According to this perspective, the only relationship between these ideas is their association, and the determinants of these associations are themselves relatively simple and few in number. Neither conditioning theorists nor associationist philosophers, however, have lacked for critics, who claim that intelligent problem solving cannot be reduced to mere association. Although allowing that the behaviour of invertebrates, and perhaps that of birds and fish, may be understood in terms of instincts and simple forms of nonassociative and associative learning, these critics maintain that the human mind is an altogether more subtle affair, and that the behaviour of animals more closely related to mannotably apes and monkeys, and perhaps other mammals as wellwill share more features in common with human behaviour than with that of earthworms, insects, and mollusks. The idea that animals might differ in intelligence, with those more closely related to humans sharing more of their intellectual abilities, is commonly traced back to Charles Darwin. This is because the acceptance of Darwin's theory of evolution was at the expense of the ideas of the French philosopher Ren Descartes, who held that there is a rigid distinction between man, who has a soul and can think and speak rationally, and all other animals, who are mere automatons. The Cartesian view had, in fact, been challenged long before Darwin's time by those who believed (as seems obvious from even the most casual observations) that some animals are notably more complicated than others, in ways that probably include differences in behaviour and intelligence. It was, however, the publication of Darwin's Descent of Man (1871) that stimulated scientific interest in the question of mental continuity between man and other animals. Darwin's young colleague, George Romanes, compiled a systematic collection of stories and anecdotes about the behaviour of animals, upon which he built an elaborate theory of the evolution of intelligence. It was largely in reaction to this anecdotal tradition, with its uncritical acceptance of tales of astounding feats by pet cats and dogs, that Thorndike undertook his studies of learning under relatively well-controlled laboratory conditions. Thorndike's own conclusions, already noted above, were distinctly Cartesian: animals ranging from chickens to monkeys all learned in essentially the same way, by trial and error or simple instrumental conditioning. Unlike man, none could reason. This controversy actually involves two questions, which are worth keeping apart. The first is whether theories of learning based on the results of, say, simple conditioning experiments are sufficient to explain all forms of learning and problem solving in animals. The second question is whether new and more complex processes operate only in some animals, that is to say, whether some animals are more intelligent than others. The distinction between these questions is not always easy to preserve, for they are clearly related, and an answer to one usually has implications for the other. The remainder of this article is organized around the first question; in cases where the behaviour of an animal does, in fact, seem to indicate that more complex processes are involved, the second question is also considered. Discrimination of relational and abstract stimuli Laboratory studies of habituation and conditioning usually employ very simple stimuli, such as lights, buzzers, and ticking metronomes in Pavlov's experiments. Some of the other examples of learning considered earlier have already suggested that animals can actually respond to additional, more complex stimuli. Even the solution of simple spatial discriminations in the laboratory requires the animal to learn about spatial relationships between different landmarks; migration or navigation over hundreds of miles demands abilities at least as complex as this. Song learning requires the young bird to discriminate between different sequences of subtly varying notes and calls, and the individual recognition involved in imprinting requires response to elaborate configurations of features. Thus, one way in which a problem may become more difficult is if its solution depends on response to more subtle changes in stimuli. Numerous laboratory studies have examined the abilities of a variety of animals to perform such discriminations. The phenomenon of transposition, first studied in chicks by the Gestalt psychologist Wolfgang Khler, suggests that animals may solve even simple discriminations in ways more complex than the experimenter had imagined. Khler trained his chicks to perform simple discriminationssay, to choose a large white circle (five centimetres in diameter) in preference to a small white circle (three centimetres in diameter). He then sought to discover whether the animal was responding to the relationship between the two stimuli or to the absolute characteristics of the stimuli. In other words, had the chick learned to select the larger of the two circles, or had it learned to pick the five-centimetre circle? If the former were the case, Khler reasoned that given the choice between the five-centimetre circle and an even larger one (eight centimetres in diameter), the animal should transpose the relationship and choose the larger circle. This was indeed the result, demonstrating that the animal was responding in terms of the relationship between stimuli rather than, or at least in addition to, their absolute properties. Transposition experiments show that animals can respond to relationships between stimuli varying along a particular continuum of physical characteristics: size, brightness, hue, etc. Another question is whether animals can respond to an abstract property of a stimulus array, independent of the actual physical stimuli making up that array. In experiments on counting, the animal must choose between an array containing, say, five stimuli and one containing three. The actual stimuli in the array vary from trial to trial, in order to rule out the possibility that the animal is responding in terms of other features, such as differences in total area or brightness, between the arrays. Counting experiments have been tried on birds more frequently than on any other class of animal, and several species, notably ravens, rooks, and jackdaws, have solved this type of problem. This success may not be entirely by chance, for there is reason to believe that the stimulus that controls when a female bird stops laying eggs is something to do with the number of eggs already laid and in the nest. Chimpanzees, however, have been trained to label pictures of various objects (e.g., spoons, shoes, padlocks, and balls) with the numeral specifying the number of objects in the picture. Moreover, rats and other standard laboratory animals have solved similarly abstract discriminations, for example, of temporal duration. A rat can learn to perform one response after a stimulus has been turned on for two seconds and a different response after the stimulus has been turned on for five seconds. The nature of the actual stimuli employed can vary without disrupting the rat's discrimination, suggesting that it is the duration of the stimuli to which the rat responds. Concept learning makes up another class of discriminations that may be solved by the abstraction of a particular property or set of properties from a very wide array of individual stimuli. In a typical experiment, a pigeon is shown a large number of colour photographs of natural scenes: half of these contain, somewhere within the scene, all or part of a tree or group of trees; the other half contain no tree (although there might be flowers, a climbing rose, or other plants). Responding to the pictures of trees is rewarded, but responding to the remaining pictures is not. Pigeons rapidly learn the discrimination. In one sense, perhaps this is not surprising: birds that roost in trees, one is inclined to argue, must be able to recognize them. But pigeons can learn other discriminations with almost equal facility; for example, they can be trained to distinguish between underwater scenes containing a fish and similar views with no fish present. In such cases, the class of stimuli in question is one for which their evolutionary history can hardly have prepared pigeons. The question, of course, is how the pigeons solve such problems. Are they, in some sense, abstracting a conceptual rule for categorizing the world into classes of stimuli? Or are they responding to what is no doubt a very large number of particular features that differentiate trees or fish from other objects in the world? Pigeons, in common with most birds, rely more heavily on vision, and certainly have better developed colour vision, than most mammalswith the exception of primates. There is evidence that monkeys can solve the concept discriminations that have been set to pigeons, but there is no evidence that other mammals can. For extensive comparative analysis, therefore, it is necessary to turn to different kinds of tasks. One that has been studied almost to excess is discrimination reversal. In reversal tasks, an animal is first trained on a simple discriminative problem: for example, to choose the left-hand arm of a T-maze, where it is rewarded, rather than the right arm, where it is not. Once the animal has solved the problem, the experimenter reverses the reward assignments, so that the food is now in the right arm rather than the left. Training continues until the animal has learned this reversal, whereupon the assignment of reward is switched back to the left arm. And so on. Rats trained on this series of reversals eventually become extremely adept at the task. Although the initial reversal causes considerable problems, with animals making many more errors than on the original discrimination, after a few more reversals these difficulties vanish. Eventually, rats solve each new reversal in fewer trials than they took to solve the original discrimination, often with no more than a single error. Similarly efficient performance has been observed in a relatively wide range of mammals. More interesting was the early suggestion that the few species of fish (goldfish, African mouthbreeders, and Paradise fish) trained on similar problems showed no evidence of the increase in efficiency displayed by mammals. The fish would learn the first reversal slowly and laboriously, and the 20th reversal equally slowly. Subsequent experiments have established that this was an unfairly pessimistic assessment, for improvements in experimental techniques have been accompanied by a significant improvement in the fish's performance, a finding that highlights the extreme difficulty of assessing the relative efficiency of widely differing animals on supposedly the same task. Nevertheless, it remains doubtful that goldfish are as adept at reversal tasks as rats are. The theoretical question, however, is how rats attain such efficiency. What processes allow them eventually to learn the reversal of a discrimination faster than they originally learned the discrimination itself, and often with only a single error? The most plausible suggestion is that they develop a winstay, loseshift strategy. They learn, in other words, to characterize the alternatives between which they must choose not in terms of their physical features but in terms of whether or not they chose it on the previous trial. They then learn that, if the alternative they chose on the last trial was rewarded, choice of that alternative will be rewarded again on the current trial; while, if it was not, choice of the other alternative will now be rewarded. A variety of other experiments have shown that rats can rapidly learn to use the outcome of one trial to predict the outcome of the next, and hence keep track of regular sequential dependencies in the availability of food or other rewards.