Concept learning:
reinforcement for
abstract behaviour



Can speech, thought and imagination be reinforced and learned like other behaviours? The behaviourist tradition says yes, up to a point (Ch. 1). Sechenov's 100-year-old slogan, `all acts of conscious or unconscious life are reflexes', is in fact more deterministic than modern behaviourist analyses which include such concepts as 'self-generated stimulus control' and `commitment'. But one can still defend Darwin's belief that even the faculties of speech and imagination, which seem to set humans apart, are elaborations of abilities already present in lower species. It is reasonable to take the view that what is special about human language and thought is not that they are immune from general processes of learning and reinforcement, but that they are uniquely sensitive to these influences. Despite the reasonableness of this view, the reader will be aware that it is a matter of controversy. The tendency for human beings to talk to each other in a highly systematic and relatively meaningful fashion probably depends on innate propensities as much as social pressures and cultural conventions. Linguists since Chomsky (1959) have become convinced that no-one would have either the inclination or capacity to speak grammatically if it were not for innate propensities, while behaviourists have always stressed the importance of upbringing, training and individual experience in verbal as in all behaviour (Skinner, 1957). A nature-nurture debate has raged for the last fifteen years, and seems to be dying down without any satisfactory compromise, but both sides are now well aware of the


contradictions inherent in either extreme view.

Although the language problem in its own right has been the major focus of argument, it also represents to some extent the wider issue of the uniqueness of man. Apart from language, there is an almost endless list of psychological attribute,, which distinguish man from humbler beasts : moral sense, tool-using; foresight; self-consciousness; sense of beauty - these were the kind of characteristics which Darwin felt he had to deal with (Ch. 1). Neither Darwin nor contemporary behavioural psychologists require that we ignore or forget these human traits: rather, they urge us to remember that human behaviour has evolved both biologically and culturally. Therefore, psychological principles which are not uniquely human may apply tc uniquely human behaviours.

Three kinds of activity which are especially rather than uniquely human, where a behavioural analysis can be attempted. are the discrimination of stimulus concepts, learning by imitation, and learning to communicate with signs and symbols.

Discrimination of stimulus concepts

The term `concept' can be applied to classes of objects, such as `chairs', `blue ribbons', `tall trees'; to complicated physical dimensions like `space', `time' and `infinity'; and to abstraci ideas, as in the `concept of democracy'. Used so broadly, ii covers almost all intelligent activity, especially in the context of Piaget's theories (C2). There is a case for distinguishing the several kinds, and I want to concentrate on an elementary sort which I have termed stimulus concepts, to emphasize that discriminations on the grounds of stimulus attributes alone is the criterion. Stimulus concepts in this sense can be defined as discriminations which work for very large numbers of individual stimuli (Ch. 8). If a monkey learns to pick out blue objects o1 various sizes from a wide choice of colours, there are grounds for saying that the monkey has `acquired the concept of blue objects', although it is a lot simpler to say that he is picking out blue objects.

The theoretical question is whether the monkey is responding merely on the basis of perceptual similarity of the objects, or because he has attained an underlying `abstract' idea of blueness. In the case of colour it is difficult to see what difference it makes. However there are some stimulus concepts that are clearly about stimuli, but are abstract in so far as they are not


about particular stimuli. Examples of these are the concepts of oddity, similarity and regularity of stimulus objects.

Oddity. If three or more objects are presented at the same time, one can be different from the others in shape, colour, size or in several ways at once. If a new set of things is shown at each occasion for choice, the only way of always choosing the `odd' one is by the abstract principle of oddness, rather than by perceptual similarity per se. Monkeys and apes can solve oddity problems quite well and rats and pigeons can occasionally, with some difficulty, achieve similar solutions, if the right kind of cues are used. This suggests that we should consider `abstractness' as a matter of degree, and not an all-or-nothing condition. Rats may learn to choose the odd stimulus, whatever it is, in the domain of smells, without being able to transfer the rule to visual stimuli. Monkeys and men have similar capacities for generalizing the `oddity' principle within several visual dimensions (Bernstein, 1961) but monkeys probably could not apply the rule across modalities - from visual displays to sounds for instance - as well as people.

Similarity. In theory, if an animal can point to the `odd' one of a collection of objects it can also indicate the `similar' ones. A technique for getting at this perception of similarity is the 'matching-to-sample' procedure (p. 103). In pioneering work in the 1920s, a chimpanzee was simply given a sample and required to pick a similar object from a tray. A common procedure nowadays is to use three displays, with a middle one containing the sample, so that the animal is rewarded if it correctly indicates which of the two outer displays is more similar to the middle one. Pigeons and monkeys can match up colours or shapes in this procedure, though here again pigeons appear to acquire a lesser degree of abstractness than monkeys. Novel colours are sometimes not matched properly by pigeons, and they match some shapes better than others. Rhesus monkeys, on the other hand, can learn a`rule' of matching for similarity which generalizes to entirely new objects, even after damage to the frontal lobes of the brain (Mishkin et al., 1962). Pigeons can do very well at judging new colours in terms of their similarity to familiar standards (Wright and Cumming, 1969), but this might be construed as more perceptual than


abstract as the judgements relate to the limited domain of the familiar standards.

Regularity. It now becomes necessary to make a very obscure distinction between `perceptual abstraction' and 'non-perceptual abstraction'. This is because some theorists, such as Sutherland (1968), believe that perception itself requires forming and storing fairly abstract rules about sensory features. Sutherland's abstract rules work just to accomplish pattern recognition for one set of stimuli, and do not necessarily produce generalization of concepts to entirely new sets of stimuli, or to new modalities. But these perceptual abstractions add further weight to the notion of continuity between discrimination and stimulus concepts.

The perception of regularities and repeated motifs demonstrates existence of perceptual rules. Rats trained to distinguish between a regular checkerboard pattern and one with a particular `mistake' such as an extra black square can generalize the concept of `regular v. irregular' to many other pairs of patterns differing widely from the originals (Sutherland and Williams, 1969). Even goldfish tend to generalize from a single square with a bump on the top to one with a notch in place of the bump, again presumably on the basis of the `abstract' coding of a discontinuity or irregularity. It is very difficult therefore to draw a hard and fast line between generalization on the basis of perceptual similarity and generalization on the basis of abstract principles, because so much of perception requires a degree of abstraction.

There is an obvious need to differentiate the stimulus concepts possessed by goldfish from those we use ourselves, but it may be that verbal labels we apply to concepts, and the thoughts we have about them, are a lot more important than the process of perceptual analysis in giving our concepts their `higher' properties. An example to think about is the efficiency of the stimulus concept of `people' used by pigeons. When shown a large collection of holiday slides, some showing people and some not, pigeons have no difficulty in distinguishing scenes which contain people from those which do not, although the images of people have little in common as the holiday slides have shown people of various ages, at different distances and angles, wearing clothes and not wearing clothes and so on (Herrnstein and Loveland, 1964). This is a reminder that most


animals have to be capable of categorizing information in a fairly abstract way into concepts such as `dog' `perching place' and `close to home' in order to operate outside the laboratory. As a counterpart to this, we might note that we become capable of using quite difficult concepts perceptually `at a glance'.

Learning by imitation

One of the many unresolved theoretical problems in the analysis of learning is whether or not learning comes in distinct types. Gagne (1965) following Tolman (1932; see Ch. 1) lists eight separate types, starting with classical and operant conditioning and ending with concept learning and problem solving. It is safe to say that we really know very little about the brain processes that bring about one sort of learning, let alone eight different sorts, but it is difficult to avoid some speculation based on the existence of different procedures for learning different sorts of things. Hence the questions of whether operant conditioning differs from classical conditioning, or whether concept learning differs from discrimination learning. Although these particular enigmas are rather esoteric there is a similar but more obvious question of whether it matters if you learn by watching somebody else as opposed to doing it yourself. On the face of it, learning to ride a bicycle is based on a certain amount of trial and error on the part of the rider, but watching someone else first probably helps and hints about which way to turn the handle-bars might cut down on some of the errors in the `trial and error'. Using the most up-to-date techniques, we could also facilitate_ learning by showing the trainee bike rider a videotape - if necessary in slow-motion - of his own mistakes and successes. The problem for learning theory is that there are an infinite number of kinds of situation in which learning can take place, and almost as many `methods of instruction' as there are teachers. It isn't very satisfactory to say that there is a different `type' of learning produced in each case. Gagne's solution is probably as good as any, for the time being, with differences between being told what to do, taking notes from lectures, watching videotapes and so on, coming under the heading of different `media for instruction'. But we might look for major divisions between learning by doing, learning by watching, and learning via language.

Of course if we include listening, smelling etc, along with watching, we have a very general category of learning by


stimulus input that would include habituation and classical conditioning (which can take place perfectly well under paralysis). But the main thing about watching and listening is that they include the possibility of learning by imitation or, as it is termed by Bandura, modelling. Bandura (1973) has been particularly interested in the imitation of violent and aggressive behaviour, but imitation of non-violence is equally important. Imitation of non-fearful or non-neurotic behaviour has also been studied intensively, and is a valuable facet of some forms of behaviour modification (Bandura, 1969, and Ch. 2). Children who are afraid of dogs may be reassured if another child `shows how' a dog can be patted without harm; students who are mildly phobic about snakes can become less afraid of them after watching films of other students handling snakes, as part of a desensitization procedure (see Ch. 3). In all forms of `learning by imitation' interpretation is complicated by . the presence of further 'sub-types' of influence on behaviour. 'Parrot-fashion' mimicry, or copying of response skills, is often combined with either short-term or long-term emotional links between the observer and the observed person, or model. In the short-term it may be a matter of social facilitation of reinforced behaviours. (It is well known that when the main character in a film lights up a cigarette, smokers in the audience reach for their own pack.) This is usually regarded as an elementary process since chickens, whose intellect is not of the highest order, are extremely susceptible and will eat merely because others are eating. A similar effect has been demonstrated in rats, who are also highly aroused by observing the sexual activities of other rats. Probably most social animals are sensitive to the actions and emotional states of others in their group, through a variety of mechanisms.

In the case of human personality development, there is a tendency to `identify' emotionally with parents or heroes so that `being like X' becomes a reinforcer. In this case the status of X has considerable significance. Status may be acquired by being a`nurturant' figure, that is by giving comfort and other necessities to the observer. The parent-child relationship is the obvious one, but imitation of authority figures like concentration camp guards has also been put in this category. Alternatively a model acquires status by receiving or possessing strong reinforcers - a hero (or heroine) with conspicuous riches, fame or talent may be imitated because his hairstyle or attitudes are


associated with such powerful rewards. Bandura's work has shown that there is much to be gained by separate consideration of the effects of reinforcers obtained by the model, and direct reinforcers given to the observer for imitating a model.

Imitation as a response class can be encouraged in children by giving reinforcers such as approval or sweets after particular instances of imitation (Baier and Sherman, 1964), and this is an effective tool in remedial education. Copying of sounds or movements often appears to occur spontaneously, however, and may function as an intrinsically reinforcing activity at some stages of human development. Experiments in which animals learn the standard operant response of bar-pressing either by watching another animal or by conventional methods (Ch. 1) suggest that some kind of copying of movements is readily obtainable in rats, and is a superior training method for cats, at least if the observer and model cats are friends (John et al., 1968). The chimpanzee Washoe (see p. 127) learned some gestures by imitation, although it was not the optimal training method.

Imitation can be regarded therefore as a mode of learning, and an influence on the performance of already learned behaviours, which can be affected by experience and reinforcement in various ways. There is marked discrepancy between the importance ascribed to imitation in the development of personality and social behaviours and the low esteem in which imitation is held by many psychologists who study the growth of thought and language in children. This may reflect the ease with which personal mannerisms and emotional reactions can be acquired, compared with the more demanding tasks required in cognitive development. It appears, however, that imitation can assist the learning of cognitive skills by children in experimental settings. A review by Zimmerman and Rosenthal (1974) deals with language skills, which I shall mention in the next section, and also concept attainment. Stimulus concepts needed to pick out cards with certain combinations of dots and squiggles on them can be learned by watching someone else sort the cards. More important for educational practice is the finding that the more abstract `Piagetian' type of concept can also be learned by observation. The best-known example is the `conservation' problem, in which a four- or five-year-old has to learn that orange juice, say, poured from a short fat glass into a tall thin glass stays the same instead


of becoming `more' because it comes up higher in the thin glass. Learning to imitate friends who say it's `the same' may be another of the many factors which affect children's answers (and `cognitive processes') in this situation. Watching another child respond correctly to one conservation task not only helps children get that particular one right, but also improves their judgement of other conservation problems.

Learning with words

There are two aspects of learning with words that have particularly attracted the attention of psychologists. The first is learning that takes place when adults read, and the second that which occurs when babies learn to talk. A great deal of specialized work has been done in both areas and it is covered in A6, A7 and C2 of this series. I shall confine myself to a brief foray into the linguist/behaviourist battle over the nature of language and words.

For brevity and point the contemporary behaviourist view of learning to talk espoused in 1940 by Bertrand Russell cannot be bettered :

In learning to speak, there are two elements, first, the muscular dexterity, and second, the habit of using a word on appropriate occasions. We may ignore the muscular dexterity, which can be acquired by parrots. Children make many articulate sounds spontaneously, and have an impulse to imitate the sounds made by adults. When they make a sound which the adults consider appropriate to the environment, they find the results pleasant. Thus, by the usual pleasure-pain mechanism which is employed in training performing animals, children learn, in time, to utter noises appropriate to objects that are sensibly present, and then, almost immediately, they learn to use the same noises when they desire the objects... (Russell, 1940, p. 69)

Skinner's view of language. Skinner's analysis of verbal behaviour (1957) is essentially an elaboration and extension of earlier behaviourist treatments. Russell's interpretation was not behaviourist enough for Skinner, as Russell adds a certain amount of introspection and mental imagery. Skinner's approach is more radical because it excludes any non-behavioural factors like mental images, thoughts or cognitions as explanations for what is said or written. This is not a denial that people do or feel the things normally referred to as `thoughts' or `images', but a hypothesis that these can be taken as a special


category of behaviour itself rather than a`higher' or `deeper' source. Although this is a difficult hypothesis to support, since it flies in the face of common belief, not all of what Skinner says is at variance with facts accepted by others. The division of speech into utterances for naming, commenting and requesting seems inevitable, and Skinner found it convenient to term the first two tacts (for `makes contact with') and the third mands (for `demand'). It has been confirmed that the early speech of infants consists largely of simple comments and requests, although it has proved extremely hard to determine to what extent these are influenced by the support of those caring for the child, as opposed to the intrinsic satisfactions of making comments and requests (Brown, 1973).

Skinner deals extensively with literary devices and figures of speech, and includes a behavioural interpretation of parts of Finnegans Wake. Conversation, the recall of memorized verbal material and the skills of translation are subsumed under the heading of intraverbal responses, while certain organizing relations between words, including the rules of grammar, are singled out as autoclitic processes. The central theme is not a linguistic analysis of language, but the accumulation of support for the proposition that what we say and write is determined by the environment rather than spontaneous inspiration or intuition. The environment, however, includes our past and present experience, education, culture, purpose and audience.

Chomsky's view of language. Chomsky's opinion is precisely the opposite of Skinner's, as he feels that the environment, however construed, has an almost negligible influence on the nature of language, or on knowledge generally. Liebniz's remarkable suggestion that even arithmetic and geometry are all in the mind to start with is quoted with approval by Chomsky, who takes the position that language learning is a matter of `drawing out what is innate in the mind' (Chomsky,' 1965, pp. 50-51). Even when language itself has been drawn out we cannot actually give information about verbal ambiguities to someone by telling him of them `but simply arrange matters in such a way that his linguistic intuition, previously obscured, becomes evident to him' (Chomsky, 1965, p. 22). This seems as much a contradiction of common sense as Skinner's radical hypothesis but is not an essential feature of the more moderate Chomskyan view, that the character of language as a gram


matical form is a product of innate predispositions rather than cultural invention or individual experience.

The distinction between the `ideal knowledge' or competence for a language, and the realities of actual speech or performance, make it difficult to apply Chomsky's theories to practical situations. Since actually speaking or writing is not really part of the theory, except as a pale reflection of ideal linguistic knowledge, there is not much to say about any linguistic performance, in particular that of the fairly large numbers of individuals who have speech impediments or difficulties in learning to read and write. Chomsky (1965, p. 58) is concerned with something that is independent of `intelligence, motivation, and emotional state' and comes entirely from inside the individual. Skinner's hypothesis at least gives those on the outside something to do to help, and the evidence suggests that it works (Yule et al., 1974, and see p. 60). Part of Chomsky's argument against a theory of language based on learned discriminations is that a sentence such as `The W Xed the Y in the Z' can be recognized as grammatical in principle even though it bears little similarity, as a superficial stimulus, to any real or meaningful sentence. Chomsky offers a theory of grammar which is independent of particular meanings or words, and can therefore explain why we can understand and say completely novel sentences, and judge whether they are grammatical or not. He has pointed out, however, (1965, p. 148) that a number of specialized rules have to be added to this basic theory of grammar to explain why some grammatical sentences make more sense than others. In areas like this a more behaviourist theory has some advantages. Consider for example these two sentences in the form just mentioned: `The government overpaid the universities in the crisis'; `The opposite appeared the safer in the sky'. The first is nonsensical from bitter experience rather than grammatical criteria. The second needs some grammatical analysis to identify why `appeared' here is not an action, while `overpaid' in the first sentence is. But a straightforward conditioning theory ought to be able to explain why `circumstances' is a more acceptable word than `sky'.

Obviously I do not share Chomsky's conviction that ordinary behavioural principles have nothing to do with language. Although Chomsky's theorizing has dominated most psychological discussion of language for some time, few psychologists have been prepared to abandon behavioural investigation altogether,


and there is a vast spectrum of ideas about language between the polar extremes of Chomsky and Skinner. Apart from the usefulness of imitation and reinforcement in remedial training of language skills in people, evidence that learned discriminations may contribute to methods of communication in animals has some bearing on the issue.

Language skills in animals

The ability of some birds to mimic human speech remains fascinating, but is being overshadowed by the competence of chimpanzees in some forms of artificial communication. However, a monograph by Marler (1970) did suggest some points of similarity between vocal learning in birds and language acquisition in children. In some species it is necessary for the young first to hear the appropriate song-pattern, and then to take time to develop their own vocal ability to match the remembered sounds. This is an example where early `models' of sound production are important, even if imitation does not take place immediately. Since children talk in a different manner from surrounding adults the value of imitation in language learning has been discounted (C2). But by analogy with Marler's birds it may be the case that children's ability to reproduce speech lags behind their capacity to perceive it. The finding that imitation may serve in the acquisition of rule-governed speaking and understanding has enhanced current interest in this kind of possibility (Zimmerman and Rosenthal, 1974).

Washoe's gestures. Several attempts to teach chimpanzees to speak have got no further than two or three recognizable grunts. Human speech requires specialized vocal apparatus which chimps do not have. It is a distinct possibility that the vocal apparatus necessary for speech sounds appeared at the very last stage of human evolution, when homo sapiens replaced neanderthal man (Lieberman, 1974). But we have since then invented numerous forms of communication that do not depend on speech - pictographic writing and bookmakers' tictac among them. It now looks as though chimpanzees have the intellectual capacity to make intelligent use of some of these non-speech systems. Several chimps have successfully been trained to use signs in the American Sign Language (ASL) in which signs represent words rather than letters. Chimpanzees in the wild are accustomed to using gestures in social


interaction, but Washoe and other chimps who learn ASL are taught the names of objects like toothbrush, cat, doll. New signs are taught by imitation, prompting and reinforcement (Ch. 4), and teachers and human companions often use ASL between themselves. Washoe was the first chimp trained by the Gardners (1969) who are now repeating their success with some newborn chimps. (Washoe was six-months old before language training was started.) Her accomplishments include using about 100 separate signs such as drink, up, baby, help, hurry, and several hundred two- and three-sign combinations, from more tickle, to hurry open door, many being of her own invention. One recent story is that she used the sign dirty taught to her during toilet training, to describe an unfriendly rhesus monkey - perhaps chimpanzees can acquire not only language, but bad language (Fleming, 1975).

Sarah's thoughts. Washoe's gesturing was very like human language in the sense of social speech, but did not contain much evidence of academic ability. It was play-talk rather than school-talk, although Washoe was given tests to make sure she could `name' objects and pictures of objects correctly. Sarah was a six-year-old `adolescent' before she began two years of intensive training designed by Premack (1970) to teach a range of intellectual skills. Communication is achieved by physical tokens, rather than by gestures or speech, which are used in a one hour per day training session, rather than being a permanent part of life. Different words such as banana, bucket, insert, same, different, yes, no are represented by distinctive pieces of plastic, varying in shape size and colour. These plastic `words' are metal-backed, so that they can be `written' by experimenter or chimp, by laying them out on a magnetized board. Step-by-step discrimination_ training, using mainly the positive reinforcers of attractive foods, enabled Sarah to demonstrate several skills normally associated with language. First, she followed instructions of the type Sarah insert banana pail apple dish, spelled out in the plastic words. Second, she answered questions about objects by `writing' yes or no or filling in blanks. For instance, real objects like an apple and a car key were shown along side a word which meant same as, with a word for question mark, and Sarah answered by replacing the question mark with no. Thirdly, this technique was used for more abstract concepts, including the concept of name. Given the


words red is question apple Sarah could answer colour of. If she was given apple is question with a real apple, Sarah selected the answer name of. The function of the `word' as a symbol for the real object is demonstrated by using the blue triangle which means apple to ask questions like question colour of apple, getting the answer red. Even though the `word' itself was a blue triangle, Sarah said that it was round and red.

Premack concludes: `The procedures that train animals will also produce words'. Obviously the plastic words are not the `same as' spoken or printed words, as Sarah would undoubtedly tell us if we asked her. The very difference probably accounts for one of the practical applications of Premack's language system, which is to provide a means of communication for individuals who cannot use normal language, either because they have never learned, or because they have had an accident which has damaged speech centres in the brain. Symbols exactly like Sarah's have been learned by children unable to talk normally. The plastic words are probably a lot easier to learn than normal reading and writing, and may be easier than speech for some children. A system that depends on vision is obviously useful for anyone with problems of vocal production or hearing. A more mysterious possibility is that the artificial communication systems bring into play different parts of the brain from natural human speech - a patient who could not comprehend or produce written or spoken words properly after removal of `speech areas' on the left side of the brain has been trained to a higher level than Sarah with a system of pictures drawn on index cards (Baker et al., 1975).

Lana's sentences. The day-to-day living quality of Washoe's communicating and the academic correctness of Sarah's grammar have been combined in the system taught to another girl chimp, Lana (Rumbaugh et al., 1973). She has a console like a large typewriter keyboard on which she can type requests to a computer twenty-four hours a day, as well as use for exchanging messages with her trainer during his visits. Signs on the keys are in the artificial language 'Yerkish', invented for this purpose, which uses combinations of nine stimulus elements to make a rich variety of visual patterns or `words'. Her keyboard is easier than a typewriter in that it has one `word' per key, but more difficult in that the position of each word is varied randomly by the computer. The symbols on the keys which Lana


pushes are `written out' electronically on a display unit above the keyboard.

Training was by the usual step-by-step method with a variety of rewards. The computer could deliver bits of apple or banana and orange juice, play music and open the window. Tickling is a very powerful reinforcer for young chimpanzees and there was a symbol for tickling, although it had to be delivered by the human trainer rather than the machine. The first stage of training was to give apple, music and the rest whenever the appropriate keys were pressed. After this it did not take long for Lana to learn to string symbols together, progressing through the forms please banana, please banana period to please machine give piece of banana period and please machine make window open period. Pressing the period or full-stop key was a special response, because that was the signal for the computer to deliver the goods for a correct sentence or erase the message if words had been typed in an incorrect order. Sometimes Lana would press the period key halfway through a sentence to erase it and start again. This `start again' function was used in a systematic test of whether Lana could tell the difference between `grammatical' and `ungrammatical' sentences. In testing sessions, some `correct' sentences were half-written in for her on the display, e.g. please machine give ..., and Lana always completed the sentence to make up a particular request. But if some words were written up in a jumbled order, such as please banana make .... Lana would reject these wrong words by pressing the period button and starting a new sentence.

It looks as though Lana has begun to recognize the order of words, in a first approximation to 'grammatical competence. This does not mean that she is ready for higher education, but it does add emphasis to the continuity between animal and human cognition. Lana's performance represents two other breakthroughs. The testing by computor avoids any possibility of `Clever Hans' cueing from a human trainer (p. 104). Also the keyboard system seems to facilitate the asking of questions by the chimp. A question can be defined as a mand, or request, for information rather than objects. When Washoe asks time eat?, or a child makes a similar query, the request and the question boil down to the same thing. But the question becomes further removed from demand for an object when Lana asks Tim give Lana name of this? and Tim (her trainer) punches in the answer that it is called a box, before she asks if she can have the box


(Fleming, 1975). If reports like this are supported, and Premack's new work with dogs bears fruit (Premack, 1972), we may have to conclude that Darwin was right after all, and it is possible to communicate ideas to terriers (p. 11).

Summary and conclusions

Chomsky's position, that language can only be understood in terms of inborn human intuitions, and Skinner's, that a complete account of language and thought may be had from contingencies of reinforcement, are both extremes. A compromise will probably be necessary, but it cannot be said that any questions about language or concept learning have been firmly answered. However, the capacity for concept learning and communication in animals is proving to be substantial. Procedures for children which combine reinforcement techniques and imitation also suggest that the acquisition of rule-governed or abstract skills can be brought under the umbrella of a behaviourist approach to learning.

Thus the originally scandalous doctrines of behaviourism (Ch. 1), characterized by the intent to simplify complexities, can be defended on two counts. They continue to provide a down-to-earth base of methods and theory against which the wilder flights of psychological fancy can be tested. And, perhaps less expectedly, behaviourism in its modern forms has some claim to be a fertile source of recommendations for applied fields of human psychology.