['Animal Thought' © Stephen Walker 1983]
gif pdf file of this chapter

9   Knowing and meaning in monkeys and apes




Although there are many uncertainties about the course of vertebrate evolution, it is pretty obvious that monkeys and apes are physically more like us than other species are. They look a bit like us, and such details as teeth, or the biochemistry of the blood, suggest that we are rather close relatives of the great apes—the chimpanzee, the orang-utan, and the gorilla (e.g. Goodman, M., 1974; King and Wilson, 1975; Zihlman et al., 1978). Most importantly for the theme I am pursuing, the brains of monkeys and apes are much more like the human brain than are those of other mammals, not to mention the brains of birds and lower vertebrates. If size of brain alone was considered, then we might expect that the functions of a brain a third or a tenth the size of ours (in a chimpanzee or a rhesus monkey) should bear more relation to the functions of the human brain than the things done by brains that are a hundred, a thousand, or ten thousand times smaller. An equally compelling reason for predicting that there should be similarities between human and primate mental characteristics is that the details of the anatomical arrangements of the human brain reflect those in other primates. Much work has been done since Huxley emphasised 100 years ago that ‘every principal gyrus and sulcus of a chimpanzee brain is clearly represented in that of a man’, but there is nothing which contradicts his conclusion that the differences between the human and chimpanzee brains are remarkably minor by evolutionary standards—as compared with, for instance, the difference between the brain of a chimpanzee and that of a lemur or marmoset. Other physical features which might have a bearing on psychological development, such as sensory and motor apparatus (in particular the structure of the eyes and the hands), would lead us to expect similarities between primate and human perception and action. Social


and emotional parallels between man and the primates are perhaps more debatable (see Goodall, 1965, 1979). But emotional, social and cognitive development during the life- span of individuals is certainly one of the areas where the affinities between man and other primates are clearly greater than those between man and mammals generally. Occasionally very strange things are said about the speed of human growth, and the length of human infancy, so it is worth pointing out that differences between man and the anthropoid apes in age- related physical changes are not terribly dramatic. Pregnancy is 8 months long in the chimpanzee, and gorilla gestation is a day less than the human term, in so far as one accepts available figures (Napier and Napier, 1967). All such figures are rather approximate, especially if derived from field observations, but it seems that civilised human lifespan, at 70—75 years, is about twice as long as that of the apes, at 35 years plus. If this is so, then the rate of progress through the life-span, in proportion to the duration of the span, is roughly the same in man and other primates. If infancy is measured as the period of close dependency on the mother, which means in apes that the mother continues to breast-feed and carry the infant, then great-ape infancy lasts about 3 years. The comparable human figure is clearly influenced by culture and is rather arbitrary, but is often given as 6 years (Napier and Napier, 1967; Jolly, 1972). Even more arbitrary is the dividing line between sub-adult and adult life. However, it can be said fairly reliably that the apes reach sexual maturity when about 10 or 11 years old, which fits well enough with the convention that human adulthood begins at 21. Obviously one’s doubts about this comparison would be because the years of human sexual activity, if not those of discretion, may begin rather earlier than at 21. The point is that the apes remain immature for about a quarter of their life span, and there is no case for saying that the proportion of the human life-span spent as a juvenile has been radically extended. The rhesus monkey, the commonest subject of psychological investigations of primate behaviour, lives to be 28 or 30 and is juvenile for the first 7 or 8 of these years—again a juvenile phase of up to a quarter of the life- span.

Cognitive development in primate infants

The time intervals involved in physical development are not in themselves particularly informative about the processes of psychological


development which may or may not be occurring. But much of interest has been found in the establishment of mother- infant affectional ties, the gradual acquisition of social skills, and the progress of emotional adjustment, in the rhesus monkey and other non-human primates (Harlow, 1958; Harlow and Mears, 1979; Hinde, 1974). There is less work on the separate assessment of cognitive and intellectual development in monkeys and apes. One study may be used as an example of the sort of results which may be obtainable (Redshaw, 1978). Differentiation between linguistic and non- linguistic knowledge, as far as that is possible, is of course necessary if comparisons between human and primate intellectual development are to be made. It happens, perhaps quite fortuitously, that the theories and methods of Piaget, which provide the major body of work available on human cognitive development in childhood, concentrate on changes which are supposedly independent of verbal fluency, or the receipt of propositional knowledge expressed in language. This allows for clearer comparisons between human and non-human cognition. than would be possible if one were to emphasise intellectual accomplishments known through speech or writing. I do not mean to discount the importance of human speech for human cognition—it is because human thought is so largely based on language that psychological comparisons with other species are so difficult. But this fact should make us seize all the more eagerly on opportunities to assess the non- linguistic capacities of animals in any way which allows some form of comparison with human abilities.

Perhaps the only time that this can be done with any confidence is in the period before human children make use of language, i.e. Piaget’s ‘sensory-motor period’ during the first one-and-a-half years of the life of human infants. Redshaw examined 4 gorilla infants (hand-reared in a zoo) and 2 human infants, over a period of 18 months, using tests designed and standardised to assess changes in the cognition and perception of human babies at this stage of life (Uzigiris and Hunt, 1974). In these, there are about forty individual achievements, such as ‘Finds an object hidden behind one of two screens’; ‘Grasps toy when both toy and hand are in view’; ‘Uses one object as a container for another’; ‘Attempts to wind up a mechanical toy following a demonstration’. Each achievement is measured in terms of how many weeks old a particular infant is when it first does it. On the basis of Piagetian theory, the individual tests are rather arbitrarily grouped into four scales: ‘Visual pursuit and object permanence’; ‘The


development of means for achieving desired ends’; ‘The construction of object relations in space’; and ‘The development of operational causality’. The examples given above represent one from each scale. There is a fairly reliable and predictable standard order in which human babies progress through the tasks: at 10 weeks their eyes follow moving objects, including hands, and by 50 weeks they can find objects which have been moved about behind screens, pull things up on the end of strings, detour around obstructions to retrieve toys, and wind up clockwork toys.

The findings for the gorilla babies are pretty straightforward: generally they do all the things which human babies do, and progress through the scales in the same sequence, but are a few weeks more advanced than the human babies in everything except playing with clockwork toys, where they are a few weeks behind. There are just four things none of the gorilla infants were observed to do in Redshaw’s study: they did not use a stick to retrieve an out-of-reach toy; they did not make towers by putting one block on top of another; they did not point to familiar persons; and they did not give their toys back to the human adult who had handed them out. The last two tests may not be entirely fair for gorillas. Young gorillas presumably can identify familiar gorilla adults, notably their mothers, although they may not be naturally inclined to point or to hand objects back and forth. The other failures indicate that infant apes may already be falling behind on the conceptualisation of space and/or causality, although, as we shall see shortly, adult chimpanzees, if not infant gorillas, can under some conditions use sticks to retrieve objects, and pile up two or three boxes to form towers. The main point, which would not necessarily have been predictable from the normal lifestyle of adult animals, is that the infant gorillas appear to go through the same preliminary stages as human babies: following falling objects, pulling in a cloth if a toy is resting on it, but not when the toy is held just above it, and so on. This provides empirical support for Piaget’s theoretical contention that the beginnings of human knowledge of the external world arise from rather basic forms of object perception (see Chapter 3, pp. 95—7).

It is likely that, even in the first year of life, the experience of interacting with reality, by playing with toys, crawling about, being fed, and so on, contributes to the internal organisation of human perception. This makes it all the more notable that the zoo-reared apes got as far as they did, lacking the degree of practice and imitation


available to a human baby in a human home. They did have the benefits of the care of a human residential parent- substitute, but this is not quite the same thing.

Most of the research which I shall now go on to discuss involves environmental enrichment of one kind or another for apes at various ages. This is certainly artificial, but it is arguable that all human environments from the stone ages on are artificial too. The question of what types of mental ability are available to apes in their natural conditions of life is an important one, but if they can be persuaded to exhibit greater cognitive capacities by certain kinds of experimental intervention, this is in some senses an even better basis for comparison with the human abilities which we may regard as natural, but which we always observe in individuals who have been surrounded by artifacts from the moment of birth.

The Mentality of Apes

Material pertinent to the cognitive development of apes after the first year of life is sparse. The Hayeses (Hayes, 1951; Hayes and Nissen, 1971) and the Kelloggs (Kellogg, 1968) reported on the gradual development of problem-solving and discriminatory abilities in young chimpanzees which they reared in their homes with the hope of observing the appearance of speech. As only two or three imitative vocalisations emerged, their efforts are usually regarded as wasted, although the non-verbal abilities of the animals should have some interest. (On the basis of rigorous comparisons with human toddlers, the Hayeses concluded that their chimpanzees had at least the rudimentary stages of several human mental abilities.) The concentration on the development of gesture, or other specialised efforts at communication in more recent intensive training of chimpanzees, which I will discuss later, has meant that other indications of intelligence have tended to be ignored. We must therefore move on from the relationship between chronological age and intellectual abilities, and look simply at object perception, and object manipulation, achieved at any age. The classic study of this type is Kohler’s The Mentality of Apes (1925). The Prussian Academy of Sciences maintained a research station on the island of Tenerife between 1912 and 1920, in which chimpanzees were kept, in fairly restricted conditions. Kohler stayed there from 1913 to 1917 observing and


testing these animals, although most of the data reported in his books were collected in the comparatively short space of six months during 1914. His book is remembered for the evidence of the mental solution to problems, by ‘insight’, prior to the execution of problem-solving actions, but is based on phenomena which have been frequently replicated by subsequent investigators (e.g. see Menzel, 1978). The main findings were that chimpanzees would drag food towards them, through the bars of a cage, by the use of a stick; would plug one piece of cane into another to make a longer stick, when this was necessary; and would construct towers, from wooden crates, in order to reach bananas hung far out of reach above them. Considerable effort has been spent on showing that chimpanzees will not do these things in the absence of prior experience of appropriate kinds (e.g. Birch, 1945; Davenport and Rogers, 1970). This is certainly the case: a chimpanzee will not drag in out-of-reach fruit with a stick on the very first occasion it sees a stick, and preliminary play and practice in the manipulation of objects is necessary if subsequent problems are to be rapidly solved. But this does not conflict with the details of Kohler’s observations, in which it is made clear that the animals needed to ‘familiarise themselves’ with the use of the objects concerned. It is true, however, that Kohler liked to describe his animals as ‘striking on solutions’ after having first scratched their heads in doubt. This did not go down well with sceptical and behaviourist readers, and perhaps Kohler went a little too far in the direction of imaginative interpretation. But stripped of the more enthusiastic Prussian anthropomorphisms (Kohler has it that some of his apes marched around in circles, one behind the other, with accentuated rhythmic stamping of the feet, feeling stateliness and pride and heightened bodily consciousness—pp. 96—7), The Mentality of Apes still contains much of value.

A task which was one of Kohler’s first tests, and which seemed to elicit a rapid perceptual solution, involved a swinging basket. The basket, containing fruit and some heavy ballast, was out of the reach of the animals when hanging still at the end of a rope, but if given a big enough push (by Kohler), would swing like a pendulum, getting close enough to scaffolding to be caught. This is an ‘extrapolation’ problem. Some of the chimps, on first seeing the basket swinging like this, unsuccessfully leapt up towards it, but others, in less than a minute sprang up the scaffolding and, so Kohler says, waited with arms outstretched for the basket to come to them. Solving the other problems was never so immediate, but Kohler’s theory was apparently


that the chimps eventually ‘saw’ a solution in the same way that some of them were able to extract from the movements of the swinging basket sufficient information to direct their movements up the scaffolding. I think, from reading his own accounts, that he was probably wrong in this, and that the later problems involved a greater element of habit, and the remembering of previous experience, but this is largely a matter of opinion.

Kohler himself certainly acknowledges the element of gradually acquired skills, in the case of the use of a ‘jumping pole’. The problem of getting at bananas suspended high up in the air was one solved in a number of different ways. Kohler’s animals had a succession of problems of this type, which meant that they had general experience of manoeuvres aimed at suspended objects, and if there are any inherited mental abilities in chimpanzees, then these should be exhibited in getting at bananas, if in nothing else. Sultan, a fully grown male with a history of using smaller sticks, is credited with the invention of jumping with the aid of a pole or board, but most of the animals did this at some time or another, and a young female, Chica, became particularly skilful. According to Kohler, the jumping was primarily a matter of play, which could be put to practical uses if the opportunity arose. The technique was not like human pole vaulting—one end of the pole was planted in the ground without any preliminary run, and the chimp then gained upward momentum by climbing straight up the stick and pushing off, before it fell over. Chica, after showing promise with short sticks and boards, and a pole two metres long, was given a bamboo pole of four metres, with which she was able to make five-metre vaults, being one metre tall herself. It is this instance that Kohler uses as an example of acquired skills of muscular co-ordination and balance. Jumping for Chica was evidently an end in itself, and the bamboo pole a valued possession. Deprived of food and the pole, and then returned to the enclosure with access to the pole, and to food placed on the ground, she would alternate between eating and jumping.

Sticks of one sort of another were used in a variety of ways by Kohler’s animals, but usually to get food. The only common use of this kind of tool observed in natural conditions is the collections of termites on small twigs (Goodall, 1965). Kohler saw something very similar—the pushing of sticks and straws through wire netting to collect ants on the other side. Sticks were also used for general poking, and in particular for surprisingly vicious treatment of chickens kept on the other side of a wire-netting fence. Some of the chimps appeared to


Kohler to be charitable, throwing pieces of bread (which the chimps themselves did not like very much) over the netting, and watching the chickens eat. Others were less kind, since they habitually held pieces of bread against the wire netting until a hen came to peck at it, and then pulled it away and ate it themselves. A perverse variant of this was to attract the chickens with bread, and then repel them by poking with sticks or wires. Stronger sticks were used in digging for roots. Digging with the hands is seen in the wild in chimpanzees and gorillas (Schaller, 1963), but not the use of tools, which apparently requires human encouragement. One specific form of encouragement used by Kohler was to bury fruit underground, within sight of the animals. This can be a test of memory, but clearly provides a reason for digging that is not present in nature.

It is in general quite impossible to tell, from Kohler’s account, how far the chimpanzees were being trained in certain particular techniques, as opposed to arriving at them by their own insights. This is not always crucial, but in the ‘building’ experiments, in which the chimpanzees made a tower of two or three boxes, it is obvious that the final behaviour depended on the accumulation of previous experience at simpler versions of the task, and not the spontaneous idea that a tower of three boxes is needed to climb to a particular suspended banana. The animals first had to be trained to pull one box underneath suspended fruit—by leaving them in a room with a box which was just the right size. After this, they were left with the fruit hanging much higher, and two available boxes. They always first tried bringing just one box under the fruit, and only after some time brought the second one. After an attempt of this kind, Kohler records that he himself placed the second box firmly on top of the first one, and held it in position while the chimpanzee involved, Grande, climbed up to fetch the reward (Kohler, 1925, p. 141). This animal ‘progressed with time’ eventually building towers from three or four boxes. Such an ability to manipulate objects in space demonstrates many capacities including patience and a certain amount of spatial imagination needed for positioning the boxes on top of one another. But the building technique as such was clearly more a matter of memory than of inspiration.

Menzel (1972, 1973, 1978) observed the apparently spontaneous acquisition of the technique of using large stripped branches as ladders or bridges in a group of chimpanzees living in a one-acre field, but argues that such behaviours arise from ‘specific motivational factors


and perceptual-motor coordinations’ which are heavily influenced by social facilitation and early social experience (Menzel, 1978, p. 386). The one-acre field was enclosed from surrounding woodland by concrete and barbed wire fences, and trees inside the enclosure had live electric wires wrapped round the base of their trunks to prevent climbing. Photographic evidence is supplied to support the observation that the animals leant tree branches against the walls and against the wired-up trees in such a way that they were able to escape over the walls, and bypass the wired bases of the trees. It is also claimed that on some occasions one animal held the bottom of the ‘ladder’ while others climbed up, and that when the branches were confiscated, the chimps pulled posts out of the ground and used them in the same way. It would seem appropriate to deduce that these individuals possessed comparatively rich mental representations of the objects they manipulated and mental organisation by which the objects were systematically related to future goals. Although it took several years for the animals to develop the ability to use branches and poles in the way that they did, and social interactions during these years were necessary, this does not make the cognitive achievements negligible.

It is difficult to say what these cognitive achievements are. Clearly if chimpanzees use poles or piles of boxes as means to attain desired ends, this implies a recognition of poles and boxes as things which may be used for getting to otherwise inaccessible places, but if the manipulations become accustomed routines it can be claimed that the object perceptions and associated motor skills are not held together by any abstract inner knowledge. Pavlov repeated Kohler’s experiments with apes kept in his own laboratory and managed to get his chimpanzee, Raphael, to pile up six boxes to reach suspended fruit. Raphael was also trained to turn on a water tap which was positioned over flames which guarded a box with fruit in it, ‘and when the tap ran dry he took a bottle with water in and poured it on to the flame’ (Pavlov, 1955, p. 594). But Pavlov firmly interpreted this as a form of association, not different in principle from what he studied by the conditioned reflex method. Since he then went on to say that human thinking is also just a form of association, his objection rather loses its force. It seems to me sensible to assume that internal associations and connections of whatever kind, that govern the learned use of tools by chimpanzees, are very much more like human, thinking than are simple anticipatory reflexes which may be observed in decerebrate goldfish. Clearly this is rather vague, and better tests to distinguish higher forms of cognition


in apes from simpler discriminations in other animals would be helpful. But Kohler’s observations should stand as evidence for sophisticated spatial memory and object manipulation in the chimpanzee.

Estimation of quantities

In the Piagetian theory of the development of human cognition in childhood, the use of tools would be a rather primitive aspect of ‘sensory-motor’ intelligence, and a more advanced type of mental operation would be one involved with the perceived stability of quantity over sequences of qualitative change. This may or may not be a worthwhile distinction to make for the purposes of comparing species: it happens to be reliably correlated with age changes in human children. A child of 3 or 4 tends to be misled by superficial changes in the appearance of given quantities of liquids or solids since if orange juice in a short wide glass is poured into a tall thin one, the 3-year-old will give every appearance of believing that there is more juice in the thin glass than there was in the wide one. A 7- or 8-year-old, on the other hand, should know that there is something important about orange juice which does not change when it is poured from one glass to another—this is known as ‘conservation of volume’. Similarly a round ball of clay rolled into a sausage shape does not lose or gain any volume, as far as an adult is concerned, but a very young child is likely to be more impressed by the change in shape than the constancy of substance, and is very reluctant to admit that the sausage is the same quantity as the ball (Piaget and Inhelder, 1969).

Although this is a reliable observation about the behaviour of young and old children respectively, it is not obvious that it ought to have very much theoretical importance. However, belief in the immutability of clay, or orange juice, is an abstract concept, which requires immediately perceived dimensions, such as length or height, to be ignored, in favour of memory. It is only because we remember that the sausage of clay has been formed from the ball it was before, without anything being added or subtracted, that we judge the sausage to be the same as the ball. If non-human animals can demonstrate anything like this judgment about the constancy of material over time, it may similarly indicate a triumph of recent memory over immediate perception.

Children are given conservation tests by being asked questions —‘Is


the orange juice the same?’ or ‘Is there more orange juice?’—we might ask after pouring it from one glass to another. To some extent the answers may reflect what a child thinks words refer to, and such niceties may. not apply to tests designed for animals. It is possible, however, to give choices to monkeys which require non-verbal judgments of constancy and change. Pasnak (1979) conducted an elaborate investigation along these lines. A monkey was shown two balls of clay of equal size. One of these balls then had a lump detached from it. The other ball was then rolled into a sausage shape, both operations being performed in the animal’s sight. The monkey then had to choose the clay which had been changed in volume rather than shape (getting a piece of apple if it chose correctly). The problem was difficult because sometimes a lump was added, rather than subtracted from the ball being changed in mass, and sometimes the changed ball, rather than the unchanged one, was rolled into a sausage. (And at other times both balls were squeezed into sausages, after one had been changed in size.) The obvious way for the monkey to solve the problem was, therefore, for it to remember which ball had been changed in size, and to use this information whether or not the objects changed in shape. Apparently the monkey (a rhesus macaque as usual) was able to do this, since it chose correctly with this example, and also with 24 others, in which clay balls were squeezed into various shapes after size alterations, or in which stacks of sponge cubes or piles of drinking straws (or various other collections of objects) underwent analogous transformations, first in size and then in shape. A second rhesus macaque went through the same series of tests, always rewarded for choosing the object or pile which had not been changed in quantity, whether or not it was changed in shape. A number of careful modifications to the procedures make it almost certain that the animals were not responding on the basis of the static appearance of the objects present at the moment of choice, but on the basis of whether a given object, as it appeared, had previously undergone addition or subtraction of some of its material. This is not precisely the same thing as a child giving the right answers on a conservation test, but Pasnak was entitled to claim that his animals had demonstrated cognitive components necessary for conservation, namely the differentiation between the adding and subtracting of material on the one hand, and changes in superficial appearance of objects on the other.

A somewhat closer approximation to the test for remembered sameness of quantities as it is administered to children was used by


Woodruff et al. (1978) with a 14-year-old chimpanzee, Sarah, who had had ten years of experience of interacting with human experimenters by the manipulation of bits of plastic (see below, pp. 357—64). Sarah was shown two objects, and was not required to choose between them, but rather to make a judgment about their equivalence. Food titbits were given at the end of a long testing session, whether or not the judgments were correct, so there was no direct reinforcement of particular responses.

One series of problems involved two lumps of modelling clay, but I shall quote the details of the other series, in which glass jars of three different sizes were filled with various amounts of blue water. At the beginning of a trial, two glass jars of the same size and shape were placed in front of the chimpanzee. Either they contained the same volume of blue liquid (reaching the same height in the two jars), or they contained different amounts (the heights reached in the two jars differing quite clearly—by 5cm or more). The experimenter then took one of the original jars, and poured its contents into a glass container of different proportions. The jars were selected so that if the original jars contained water to the same height, after the pouring the new jar had a different height of water than the other; but if the original jars held different amounts of water, after the pouring the water in the new jar came up to the same height as the water in the old one. The experimenter then handed the chimpanzee a covered dish containing two pieces of plastic, one of which she was accustomed to use to indicate ‘same’, the other of which she had used for many years to indicate ‘different’. The experimenter left the room and the chimpanzee picked one of the pieces of plastic, and put it between the two jars. Roughly 8 times out of 10, Sarah selected ‘same’ when the two original jars had held the same amounts, and ‘different’ when they had held different amounts. By itself, this might not mean very much. However, preliminary tests had shown that ‘same’ and ‘different’ were normally used appropriately for two jars of the same shape containing equal or unequal amounts of water. Also, as a check that the state of the original two jars on each trial determined the final judgment, the procedure was repeated, with the chimpanzee being allowed to see only the final stage of the sequence, that is two jars of different shapes containing either equal or unequal amounts. Judgments in this case were not accurate: there was a tendency for ‘different’ to be selected, whatever the amounts of liquid present, but this was not quite statistically significant. This suggests that when the chimpanzee was allowed to


observe the initial state of the two jars, followed by the pouring episode, it selected the ‘same’ and ‘different’ responses on the basis of the initial state, and ignored their final appearance. This would only require remembering the initial state for a minute or two, well within the chimpanzee’s range, but it means that the animal must be interested enough in the heights of liquids to remember whether the two jars were initially the same or different, and be able to discount the heights of the liquids immediately present when the judgment is given.

This may not be quite the same thing as ‘conservation of liquid quantity’ as it occurs in a 7-year-old child, but it shows flexibility of memory, and an independence from current vivid stimuli. Sarah was certainly capable of watching the pouring of liquids very closely; at the final stage of this experiment (Woodruff et al., 1978) the two jars always started out with the same amount of water in them, but when the experimenter did the pouring from one of the original pair into the jar of a different shape, he sometimes poured a little of the water into a non-transparent cup, or used water already in this cup to add to the new jar along with all the original water. Sarah was again about 80 per cent accurate on this procedure, giving ‘same’ when pouring took place normally and ‘different’ when the cup was used for adding or subtracting. The very least this demonstrates is that the chimpanzee noticed when the experimenter was fiddling with the cup, and used this as a cue for the subsequent selection of one from two bits of plastic. Since the rhesus monkeys used by Pasnak (1979) were not fooled by the experimenter patting clay with a spatula, when adding or subtracting with the spatula was important, it seems sensible to assume that something about the liquid was being perceived. A further experiment with the chimpanzee Sarah suggests that she readily ‘sees’ quite abstract relationships in visual displays (Gillan et al., 1981). Presented with a cut apple with a knife on one side, and a piece of paper and scissors on the other, she selected the ‘same’ symbol to put between them, whereas given the Cut apple and knife paired with the paper and a bowl of water, she gave the symbol ‘different’. On the basis of a number of tests like this the experimenters concluded that the chimpanzee was bringing to bear a process of analogical reasoning’ in order to deduce similarities which reside in relationships between objects. ‘Reasoning’ is a contentious term to use, but clearly something was going on in the chimpanzee’s head during these tests beyond the plain comparison of visual patterns.

What exactly was going on in the two experiments on ‘conservation’


of quantity by non-human primates (Pasnak, 1979; Woodruff et al., 1978) is uncertain, and whatever it was probably depended on the exact conditions of training in each case. But it is not unreasonable to suggest that monkeys and apes can come to know things about objects which progress beyond the simplest sensory-motor levels of the hierarchies of cognition observable in man. In the natural environment such cognitions are presumably put to the service of eating and social adjustment. One can well imagine that it is useful to know that a large banana is large, whether it is perceived in a vertical or horizontal alignment, and to some extent this must be regarded as a form of conservation. It cannot be denied, however, that detailed comparisons between such forms of cognition in primates and corresponding human knowledge must always be qualified by the absence of verbal expression of ideas in primates other than man. It is only when a child becomes able to give a verbal explanation of why orange juice poured from one glass to another is ‘the same’ that we are likely to accept that his knowledge approximates that of an adult (Piaget and Inhelder, 1969). The child is eventually able to argue that a given volume of liquid must stay the same in different containers: either by explaining that sameness depends on nothing being added or subtracted; or by pointing out variations on this theme, such as the possibility of restoring the liquid to its original appearance in the original container. Few would suppose that the chimpanzee’s knowledge of liquids and solids could ever be given this sort of expression. But on the basis of data such as Kohler’s, which apparently demonstrates comprehension of ends and means, many have felt that these animals if no others, should be able to give verbal expression to their more concrete inner ideas, if given sufficient encouragement.

Communication by sign and symbol in the great apes

Unfortunately, captive apes display a notable lack of interest in imitating human speech. Of the great apes, the orang-utan, being largely solitary, has little use for frequent vocal signals, while gorilla social groups are fairly quiet (Schaller, 1963) and the noise produced by chest-beating in aggressive male displays is probably as important as vocally produced sounds. The chimpanzee is more vocal and more actively social than the other great apes, but gesture, posture and touch seem to have a wider application in social exchanges than vocalisation.


The natural social vocabulary of even the chimpanzee is almost certainly less extensive than that of Old World monkeys (e.g. Green, 1975; Seyfarth, et al., 1980) or of the lesser apes, the gibbons (Tenaza and Marler, 1977; Tenaza and Tilson, 1977). It is often suggested that the sound- producing apparatus of apes—the resonating cavities of the throat and mouth and the adjusting mechanism of lips and tongue—is physically incapable of producing human speech sounds (e.g. Lieberman, 1975). To the naked eye physical specialisations for speech in man seem remarkably minor, but whether the difference lies in the anatomy of the tongue and throat, or in psychological inclinations to utilise the anatomy to make interesting noises, there is no doubt that apes intransigently resist any temptation to cross the Rubicon of human speech.

The evidence is mostly negative: apes kept in zoos, or in human homes, are highly imitative, form strong social attachments to people, and are prepared to adopt human habits of personal hygiene and social decorum, but have never been known to imitate speech. It is the more strongly negative, in that deliberate attempts to foster speech, by bringing up infant chimpanzees in human homes, and maximising the social and the more tangible incentives for vocalisation, have had very little success. One of the most domesticated apes was probably the gorilla John Daniel (Cunningham, 1921; quoted in Tilney, 1928) who behaved himself well in company, but died when only 6. (‘His table manners were rather exceptionally good. He always sat at the table, and whenever the meal was ready would pull up his own chair to his place. . . He always took afternoon tea of which he was very fond, and then would eat a thin slice of bread with plenty of jam’: Tilney, 1928, p. 635.)

Two psychologists, W. N. and L. A. Kellogg, brought up an infant chimpanzee in their own home in the 1920s (see Kellogg, 1968) and were able to draw comparisons between its social development and that of children, but saw no signs of speech- like vocalisation. This enterprise was repeated by K. J. and C. Hayes two decades later (see Hayes, 1951; Hayes and Hayes, 1951, 1952; Hayes and Nissen, 1971). They were able by dint of explicit training, to get their animal, Vicki, to make noises that sounded vaguely like human words, but only four of them (Mama, Poppa, cup and up). Vicki also made other sounds in specific contexts which were not recognisable as English words, but which were not natural chimpanzee vocalisations either. This was a considerable achievement, if the anatomical barriers to speech


production in chimps are as great as is now generally supposed, but the four words themselves do not expand very much either the chimpanzee’s cognition or our interpretation of it. If a chimpanzee could say ‘cup’ when a cup is presented at various angles and distances, this would demonstrate the perceptual capacity of object recognition, but the same inference could be drawn from observations of a thirsty chimp reaching for cups of various shapes and sizes.

Training chimpanzees to make gestures

It may be that the verdict on the numerous efforts to teach chimpanzees sign-language will be rather similar—just as the attempt to make a chimpanzee talk produced only the most minimal approximation to human speech, attempts to make chimpanzees communicate by a gestural system usable by the human deaf produce only a pale reflection of human performance (Terrace, et al., 1979). There is a myth that we could converse with animals as easily as with other people, if only we did something rather special and magical, such as drinking dragon’s blood. One can detect in some of the discussions of chimpanzee gesturing (Gardner and Gardner, 1975, 1978; Linden, 1976) an air of the false hope that sign-language might be the modern equivalent of dragon’s blood, allowing us to communicate with apes on an equal footing. It is obvious that this dream is not going to come true. But it is equally obvious that chimpanzees’ cognitions are not, as some still claim (Terrace et al., 1979), as far removed from human thinking as those of a pigeon or rat. Training apes to make a large number of distinct hand movements, which correspond to objects and perceptual categories, reveals a good deal about perception and memory in the chimpanzee, even if it is a very poor substitute for dragon’s blood.

Before discussing the evidence which is obtained from chimpanzees trained to make artificial gestures, I want to make a clear distinction between claims that these animals do things that are strictly comparable to the human use of language (e.g. Gardner and Gardner, 1975) and more limited inferences about perception and memory. Because claims for the conversation-like attributes of chimpanzee gesturing have been exaggerated, it is sometimes suggested that, on the contrary, their gestures have no meaning at all. It is perfectly possible, however, that something useful may be learned from experiments which fall


short of demonstrating human levels of linguistic competence in apes, and I believe this is true of the projects designed to teach chimpanzees sign-language.

The first of these was begun by R. A. and B. T. Gardner, with the chimpanzee Washoe (see Gardner and Gardner, 1969, 1971), The idea was that, as chimpanzees never imitate sounds but frequently imitate actions, they should have less difficulty in learning the arm and hand movements of American Sign Language, widely used by the deaf in North America, than they do in learning speech. It is customary to describe the individual signs in this system by giving translations: it is begging the question to translate gestures made by animals directly into ordinary words, but there seems to be no realistic alternative. I shall follow the usual convention, but the reader is warned not to take translations of chimpanzee gestures as evidence that the mental states accompanying the gestures were the same as those which may accompany the verbal translations. Examples of gestures eventually performed by Washoe, and their translations, are as follows. An extended arm moved upward, with or without the index finger also extended, is interpreted as ‘up’. A beckoning motion with the whole hand or the ends of the fingers is ‘Come-gimme’ ; a palm patting the top of the head is ‘hat’; slapping the thigh is ‘dog’; an index finger drawn over the back of the other hand is ‘tickle’ (Gardner and Gardner, 1969). Some of these, such as ‘up’, are iconic or intuitive, while others such as ‘dog’ are more arbitrary.

The training programme was first to make gestures of this kind a ubiquitous part of the only social life the animal experienced. Washoe was wild-caught, but after entering the Gardners’ laboratory at the age of approximately 12 months, she lived in an environment of human artifacts and human companions. One or more persons were present during all Washoe’s waking hours, making ASL (American Sign Language) signs directed at the chimpanzee, and between themselves. Imitation was therefore possible, but direct guidance, by putting the animal’s hands in the desired position, was used as well, and deliberate reward and punishment was pervasive (for instance, tickling only when Washoe made the signs for ‘tickle’ or ‘more tickle’).

As a result of these efforts, Washoe exhibited a number of gestures which corresponded quite closely to standard ASL signs: 30 when she was about 2 years old, and 132 after two further years of training. It seems safe to conclude that a vocabulary of gestures is more readily acquired than a vocabulary of spoken words, but it is prodigiously


difficult to determine how far the gestures are performed at random, or in immediate imitation of human companions, in the ‘free-living’ context of play and persuasion which is filmed or photographed to provide the most appealing record of the results of this type of experiment (see Patterson, 1978; Terrace, 1980). Close examination of some of these photographic records suggests that imitation of recent signs produced by the trainer is one of the things that happens (Terrace et al., 1979; Seidenberg and Petitto, 1979). Claims by trainers that Washoe, or other apes subjected to similar procedures, ‘learned sign language’ thus need to be treated with some caution. Fortunately, there is sufficient evidence from carefully controlled tests to make it clear that individual gestures are reliably connected with particular contexts or referents. The disagreements are not so much about whether chimpanzees learn to associate particular gestures with particular objects, and particular actions, but about whether they achieve the higher levels of mental organisation that would be indicated by the understanding of relationships between words.

Evidence that individual gestures have individual referents (or individual meanings, to use that term loosely) can be obtained simply by recording (using films or videotapes and independent observers) that an ape uses a sign—say ‘cat’ when it sees a cat—without any help from human companions. This was done with Nim Chimpsky, a male chimpanzee reared under more or less the same conditions as Washoe by another team of people (Terrace et al., 1979). Alternatively the chimpanzee can be shown a random series of objects, or of pictures of objects, with an observer, who cannot see these objects, noting the gestures made (Gardner and Gardner, 1971, 1978). With both these techniques, it is apparent that particular object categories elicit appropriate signs. In the reverse direction, if the chimpanzee sees a human trainer make a gesture, it is able to identify appropriate objects. Some of the evidence for this is rather anecdotal, it is true. Terrace (1980) reports that signed instructions of the type ‘give me brush’ were responded to by Nim selecting whatever object was asked for from an array of possibilities, and that this animal would also leaf through the pages of a picture book to find and point to illustrations of objects referred to by gesture of the human trainer. A more systematic and elaborate test of the received meanings of sign-language gestures is reported by Fouts et al. (1976). The subject was a chimpanzee born in captivity, and reared in a human home hearing spoken English as well as seeing ASL gestures. The design of the experiment involved


establishing to begin with reliable indications that this chimpanzee could identify 10 objects, which it did not know by gesture, by hearing spoken words (the 10 were spoon, foot, curtain, water, banana, shocker, raisin, nut, leaf and pillow). If the experimenter said ‘spoon’ the animal fetched one, but if the word spoken was ‘pillow’, it fetched a pillow. At the next stage the human trainer spoke the word and at the same time made the corresponding gesture, in the absence of the appropriate object. This was done with individual words, one at a time. The test was that, after sound-gesture experience, the chimpanzee was assessed by another experimenter, who was not aware of which gestures the chimpanzee was supposed to know. The assessment was to pick up each object in turn (the experiment was split up into two groups of five objects) and see if the animal could make the appropriate gesture. The result was that the 10 individual gestures were acquired by the animal, in the order in which it received instruction via speech and gesture conjunctions, in the absence of the objects referred to.

It would be unwise to place a great deal of weight on any single anecdotal or experimental report, but unless there has been widespread fraud or delusion, it seems necessary to accept that under the conditions described, chimpanzees form mental associations between perceptual schemata for manual gestures and others for object categories. This is not to say, in Romanes’s phrase, that they can mean propositions, informs such as ‘all chimpanzees like bananas’, although it is arguable that individual animals make gestures that bear some relation to the proposition ‘I would very much like a banana now’. The relation is more interesting than that between ‘I would like a banana’ in human speech, and tugging at the sleeve of someone who is holding a banana, but, since it has not been convincingly demonstrated that one chimpanzee gesture modifies another, or that there is any approximation to syntax and grammar in the comprehension or expression of artificial gestures, the similarity between the use of individual signs by apes, and the use of words by people, is definitely limited.

Objects as symbols for other things

Premack (1970, 1971, 1976) developed a training system for use with chimpanzees which, according to his interpretation, allows them to comprehend and express propositions of the form ‘all the biscuits are


round’ and ‘these two objects are the same’. This interpretation is extremely suspect, since it depends on the translation of fairly simple actions into verbal terms, but Premack’s evidence deserves serious consideration. Most of it was obtained from one wild-born chimpanzee, Sarah, who was kept in a laboratory cage from the time she was acquired, when she was less than a year old. Not surprisingly, her social attachments to people were not as marked as those of chimpanzees reared in homes, or with constant human companionship. All of the initial training with the Premack system took place in a two-year period between the sixth and eighth year of Sarah’s life, the human experimenters entering the cage during training sessions amounting to an hour or so per day, for the first sixteen months, but remaining outside the cage, because of the chimpanzee’s aggressiveness, for the last eight months.

Given the limitations of this sort of laboratory training, it is remarkable that so much data was obtained. But Sarah is the only one of four chimpanzees used by Premack (see Premack, 1976) to get very far, while the method of immersion in gestures has been successful with dozens of animals, trained by different teams of people. Vocalisation as used in human speech has a number of advantages as a means of communication. It requires very little physical energy, and self produced signals can be monitored in the same way as sounds received from other sources. If necessary vocalisation can take place under almost all circumstances of bodily activity. Gesturing with the arm is somewhat more energetic, and can only be done if the arms are not being used for other purposes, but it does not require any artificial equipment. Premack’s communication device required both artificial equipment and use of the arms, since objects and concepts were signified by the placement of metal- backed coloured plastic shapes on a magnetised board in the chimpanzee’s cage. As with vocalisation, self-produced messages corresponded closely to externally presented patterns. The great advantage of the plastic symbols, if the animal pays attention to them, is that they circumvent the problem of memory. If a sequence of instructions is given to the animal, in the form of plastic symbols placed in a particular serial order, the internal retention of the sequence is not crucial, as it would be with transient gestures or sounds, since the animal can continually refer to the external record. In a sense, then, the plastic symbol system used by Premack avoids the requirement of memory for signs or sequences of


informative signals, and thus makes things considerably easier for the chimpanzee.

In the early stages of training, however, the transactions between man and animal were depressingly familiar. After having a particular plastic shape always placed alongside an apple, the chimpanzee was required to place this shape on the magnetised board before she was given the apple, then she had to place in sequence several shapes, translated as ‘Mary give apple Sarah’ (Mary being the human trainer). But with preliminary experience of this kind, more complicated interchanges took place. The bulk of these can be construed as questions, put to the animal by the experimenter, who placed an array of plastic shapes on the metal board, or on a shelf, the chimpanzee answering by making alterations to this array. It seems probable from the nature of these questions and answers that the plastic shapes signalled something other than themselves, as we shall see, but the most direct evidence that they carried messages, or conveyed meaning, arose when sequences of the plastic tokens implied an instruction.

The simplest instruction was the reversal of the chimpanzee’s requests. Instead of the animal writing out ‘Mary give apple Sarah’, the trainer wrote out ‘Sarah give apple Mary’ or ‘Sarah give apple Gussie’ (another chimp). Occasionally these instructions elicited refusals, in the form of tantrums or re-ordering of the symbols, but by rewarding compliance the experimenters usually obtained obedient reactions. When Sarah had acquired a larger vocabulary, via various kinds of association between symbols and events, other actions could be elicited by the use of sequences of’ tokens—for instance, ‘Sarah apple cut’ and ‘Sarah apple wash’ could be used as instructions for the chimpanzee to cut or wash an apple (Premack, 1976, p. 76). These messages would be used individually, but the most elaborate instruction which Sarah followed was a compound of two instances of the same command, ‘insert’. After considerable experience of single instances, such as ‘Sarah banana pail insert’ and ‘Sarah apple dish insert’ when the two objects and the two containers were available, she was able to put the correct item in the container indicated. She was then given instructions in the form ‘Sarah banana dish apple pail insert’, so that two actions were required in the same command. Once the banana had been put in the dish in this example, a desire for regularity might have prompted the assignment of the remaining object to the unfilled container. But various complications of the instruction were given with two containers and


several objects of’ various categories available. Sarah seemed to have little difficulty in complying with the command to put chocolate in both the dish and the pail, or the command to put apple and a cracker in the dish and just a cracker in the pail (Sarah apple cracker dish cracker pail insert: Premack 1976, pp. 325—30).

The final variation was to use two dishes of different colours, and to identify them with the colour ‘words’ that Sarah already knew (which were particular plastic shapes, not coloured themselves). This gave instructions like ‘Sarah cracker candy yellow dish cracker blue dish insert’, which the chimpanzee complied with correctly about 8 times out of 10. This sort of performance is not as complicated as it looks, because the general practice of putting either one or two kinds of food into each container was well established. But the details of exactly which food has to be put in which dish could only be extracted by noticing the positions of symbols in the sequence (which was a vertical line of the plastic tokens, in which top-to-bottom corresponded to left-to- right:

the chimpanzee appeared at first to prefer the vertical arrangement). It may be misleading to call this syntax, since it is a single case, and rather a simple one. However, the order of the symbols, and not just their presence or absence, was clearly an important component of this kind of instruction. Human patients with certain kinds of brain damage sometimes can read individual nouns, but have difficulties in understanding or expressing phrases whose meaning is determined by the order of words, and are thus diagnosed as ‘agrammatic’ (Goodglass and Geschwind, 1976; Goodglass, 1976). Although it would be foolish to claim that Sarah the chimpanzee demonstrated human levels of syntax or grammatical competence, it would appear that in some limited sense she was not completely ‘agrammatic’.

A particular deficit in human ‘agrammatic’ patients is the inability to read function words, such as prepositions like ‘on’ or ‘under’, which includes some instances of failure to follow instructions involving simple interpretations of prepositions. It is interesting, therefore, that Sarah was able, albeit in the same limited conditions, to respond appropriately with a plastic token designated as ‘on’.

When she was already familiar with the series of achromatic shapes which corresponded to particular colours, she was introduced to on by manipulations of pairs of coloured cards. The trainer wrote Red on Green, gave Sarah a red card, and guided her hand so that she placed it overlapping a green card already lying on the work table in front of her. After very little practice of this kind, the trainer put up


instructions, with three cards lying on the table, in the same form (‘Red on Green’, ‘Green on Yellow’, etc.) which the animal followed by picking up the first mentioned card and placing it over the second. The chimpanzee was also able to describe pairs of overlapping cards put in position by the trainer: the trainer, for instance, might put the yellow card on top of the red card, giving Sarah the three colour tokens and the ‘on’ token, with the result that the chimpanzee stuck the tokens of the board in the order ‘Yellow on Red’. A second chimpanzee, Elizabeth, who was the only one of the others tested to reach Sarah’s standards, soon followed ‘on’ instructions when they were introduced with objects, instead of cards. Knowing all the object names, she was first instructed in ‘keys on clay’, and then successfully tested with ‘chow on shoe’ or ‘shoe on chow’ (Premack, 1976, pp. 107—11).

As the sequences of tokens were written vertically from top to bottom, it may be that both ‘on’ and ‘insert’ succeeded as instructions because of the similarity between the format of the tokens and the required positions of the objects. Since the top-down sequence was used in many other instructions, such as give, take, push, cut and wash, I suspect Sarah would have eventually managed to respond to ‘Yellow under Red’ by putting yellow and red cards in the opposite relation to their corresponding tokens on the board, but it is a great pity that this was not, apparently, attempted.


Questions and answers

The vertical sequence of tokens was abandoned for many of the tests, since these involved questions about real objects, which could not be conveniently attached to the metal board. Objects and tokens were laid out on a shelf or table instead. The simplest questions were, one presumes, perceptually very easy, since they required only that the chimpanzee detect whether two real objects were identical or not. The perceptual judgment itself was tested very directly by giving Sarah two spoons and a cup, and encouraging her to place the two spoons together and the cup to one side. The transfer to this perceptual judgment into the manipulation of tokens was accomplished by designating particular plastic shapes as same, different, yes, no and question, and establishing their usage in certain contexts. Let me first make explicit my convention here of identifying real objects by words, and plastic tokens with putative meanings by italicised words. Using this convention, the layout for a question might be like this: ‘cork question cork’ laid in line from left to right, with same and different placed


in line an inch or so nearer to the animal. The answer in this case would be to remove the question token, and replace it with same. Clearly if the array was ‘cork question scissors/same, different’, the answer is to replace question with different. The solution to these problems seemed to be clear enough to the chimpanzee, even if the problem arrays took more complicated forms, such as: ‘key different question/key, rubber band’; ‘clothes-peg same clothes-peg question/yes, no’: (Premack, 1971; 1976, pp. 147—52).


Linguistic functions of tokens

Reaching the final arrays of same between two similar objects and different between two different objects could be construed as merely the assignment of particular tokens to perceptions of’ similarity and dissimilarity which could be assessed just as well by other methods. But the correct addition of yes and no to triplets containing same and different seems to require the animal to make judgments about the meanings of the tokens rather than just about the real objects. Decisions about the relations between tokens perhaps come rather closer to the kind of abstraction achievable with speech than decisions about the relations between objects as objects. This may seem slightly obscure, but there are some fairly straightforward examples.

For instance, a particular bit of plastic was designated ‘name’, and (to begin with, later they were separate) no was glued to name to make ‘noname’ (‘not the name of’). Sarah was then given the arrays ‘apple name apple question/yes, no’, and ‘banana noname apple/yes, no’ and modifications of these such as ‘apple question banana/name, noname’ . This is very much like the same-different judgments, but the similarity is between object and the token for the object, rather than between two objects. One would expect the main source of confusion to be the double negative necessary in ‘apple noname apple’ which requires the addition of no. It is hardly surprising that Sarah made more mistakes when negatives were involved, in the course of achieving the standard level of 8o per cent correct answers on a series of problems about the correct names for apricots, raisins, dishes and pails (Premack, 1976, pp. 162—4).

In a similar way, tokens about tokens, rather than tokens about objects, were introduced in the case of colour, shape and size (Premack, 1976, pp. 189—98). Of course, before this could be done, the chimpanzee had to assign tokens for individual colours, shapes and sizes to appropriate objects, for instance, yellow to a banana but red to an apple, round to a button but square to a box, and small to a grommet but large to


a sponge. Then she could be given problems in the form ‘ Question colour banana/red, yellow’; ‘Question shape ball/round, square’; ‘Question size sponge/large, small’. Such problems were answered correctly more than 8 times out of 10, except that performance on the size problems was slightly worse since Sarah tended to put large when the experimenters wanted small.

The most abstract of all the problems solved by Sarah, it seems to me, were those in which she had to identify the tokens indicating classes of properties on the basis of instances. For instance, she was given ‘banana question banana’, ‘yellow question banana’, ‘brown question chocolate’, ‘chocolate question chocolate’, with the alternatives name and colour to choose from. On one of the rare occasions when there were three answers to choose from rather than two, a series of arrays like these were given with name, shape and colour as possible choices. Thus something like ‘round question apple’ was answered by replacing question with shape, while ‘red question apple’ resulted in question being replaced by colour. Premack gives only a brief’ account of this, but if anything it appears that Sarah found it easier to choose classes than to choose individual properties since fewer errors than usual were made, and in some series such as the choices between colour and shape, no errors were made at all. Premack (1976, p. 198) suggests that the improved performance came about because by this time Sarah had had plenty of practice with tokens involved, but in any case it seems that identifying yellow as needing to be accompanied by colour rather than shape or name, that is, remembering associations between tokens rather than associations between objects and tokens, presented no particular difficulties.

The tokens indicating classes of property were also used to identify new plastic labels with new stimulus qualities and object. For instance, Sarah was given the array ‘question colour chocolate’ provided only with the token brown, with which she replaced question, subsequently demonstrating the meaning of brown by selecting a brown card from among four others on the instruction ‘Sarah take brown’ (Premack, 1976, pp. 202—3). After being introduced to the tokens for figs, peaches and other appealing food items by seeing propositional arrays in the form ‘fig name fig’ she used the appropriate tokens to request whichever item she was shown (Premack, 1976, pp. 196—266). This implies that associations between particular tokens and the object classes or stimulus properties which they signified could be formed very quickly. In fact, towards the end of Sarah’s training, new tokens as labels for


objects could be introduced simply by the trainer holding up the token and an object together.

All these results might be taken to suggest that the plastic tokens were capable of signifying other things to the chimpanzee, and that seeing a token resulted in inner representations of qualities or objects not immediately present. Colloquially, seeing a token made the chimpanzee think of something else. Great emphasis is placed by Premack on descriptive choices elicited by tokens. If shown a real apple between the tokens red and green or between the tokens round and square, the chimpanzee promptly chose red and round as properties which went with the apple. And if shown the token for apple (which was in fact a blue triangle) in similar circumstances there was no hesitation in the similar choices of red and round, even though these qualities were not being observed at the time. The same results were obtained if cards that were actually red and green, and square and round, instead of the tokens for these qualities, had to be selected from.

On the basis of these results, the plastic tokens were very informative to the chimpanzee, and the relations between token and token, and token and object, may imply a theoretically vital degree of mental organisation possible in this species. But it is misleading, in my view, to say that the tokens had the same function as words in human speech and writing. In practical terms the tokens were only used in a very restricted set of circumstances, and in theoretical terms there is no evidence that the involvement of tokens in such things as grammar and semantic categorisation approaches the way that words are involved with these things in man. The importance of Premack’s results, taken together with other work on apes, is not that the apes can be trained to do things that are equivalent to human language. Rather, apes can be trained to do things which might plausibly be equivalent to a preliminary stage in the evolution of human language, or which indicate that apes have a level of cognitive organisation which one can imagine might be sufficient to make the beginnings of speech useful

Keyboard experiments, and exchange of information between apes

There are many who suspect that both Premack’s work with plastic tokens, and the Gardners’ method of interpreting learned gestures, may be seriously at fault because of’ artifacts in the procedure, statistical unreliability, and so on (Terrace, 1979; Seidenberg and


Petitto, 1979). It is just as well that a third body of work on chimpanzee communication is available, which gets round some of the problems of human participation by training the animals to operate a computer-controlled console like a simplified typewriter. Initially this system was used so that a single animal could exchange information with the computer, eliminating the possibility of the human trainer (by subtle changes in posture, tone of voice, or facial expression), providing external cues which would allow the chimpanzee to exhibit correct responses without working them out for itself. (This is known as the ‘Clever Hans’ phenomenon, after a German horse of that name which stamped out answers to arithmetical problems written on a blackboard, but only because it sensed from the trainer’s reactions when it should stop raising and lowering its hoof.)

But the most important development with this keyboard system has been the exchange of information between two chimpanzees. Chimpanzees in the wild respond to each others’ gestures, postures, facial expressions and vocalisations, but it is usually argued that this is a case of fixed emotional reactions to social signals, rather than the communication of ideas. Anecdotal reports of chimpanzees communicating to each other in any way at all using the sign-language gestures which they have been taught by human intervention are extremely fragmentary (Linden, 1976). The artificial nature of the keyboard system makes early training rather difficult (Savage- Rumbaugh et al., 1978a), but has the advantage that, once animals are familiar with visual patterns used as symbols to denote objects, relatively clear evidence of reference to absent objects can be obtained, in communications made between two animals.

The keyboard is an array of small rectangular screens. Particular visual symbols made up of lines, circles and dots against coloured backgrounds can be projected on any one of these keys, so that the visual pattern itself, rather than a given position in the array (as in a• typewriter) must be attended to (Rumbaugh et al., 1973). When a pattern in the array is touched, it is put up in a line of projectors above the keyboards, giving a record of the sequence of symbols pressed, going from left to right. Certain patterns are assigned by the experimenters to certain items of food and drink, which can be delivered to the animal automatically, and other patterns are required as the first and last symbols in an acceptable sequence, and as connection symbols which are given verbal translation, such as ‘machine’ and ‘give’. Not surprisingly, chimpanzees have little natural


tendency to connect two-dimensional patterns of lines and squiggles with other events, and it is probable that early training with these visual cues leaves them with a rather limited symbolic significance, by comparison with the manipulable plastic chips used by Premack, or the imitable gestures of the Gardners’ sign-language method (Savage and Rumbaugh, 1978). Lengthy training is needed, in which the chimpanzees press keys in the sequence ‘Start give banana slop’ to get a banana, ‘Start pour coke stop’ to activate an automatic drink dispenser, and so on. The original chimpanzee which learned this system, Lana, could make a wide variety of requests via the keyboard, including ‘Start trainer tickle Lana stop’, if the trainer was in the room, and ‘Start machine make window open stop’ (Rumbaugh and von Glaserfeld, 1973; Rumbaugh, 1977). Social relationships with human beings, and interaction with people via pointing and gesturing on their part, and some imitation by the animals, appears to be necessary with this method of training, as it is with the others, but after training has taken place, the trainers can absent themselves, leaving the animal alone with the computer, to provide the crucial tests of the animal’s abilities (Savage-Rumbaugh et al., 1978b).

Intensive individual training of the female Lana established that a chimpanzee could use the keyboard system to make a wide range of requests for desired objects and events, and that the chimpanzee could be induced to select patterns designated as the names of things such as sticks and keys if these were present (Rumbaugh, 1977). More specialised and limited training given to two young males, Austin and Sherman, resulted in their gradually acquiring the theoretically valuable ability to communicate with each other. In the first instance of this the identity of a hidden item of food was transmitted from one animal, who had seen the food hidden, to the other, who had not (Savage-Rumbaugh et al., 1978b) and in the second, one chimpanzee requested and received from the other an object which could be used as a tool to obtain food (Savage-Rumbaugh et al., 1978a). A preliminary step was to test the ability of each animal to press the appropriate key when a particular kind of food was visible. Eleven types of comestible were used altogether: beancake, banana, chow, milk, orange drink, juice, cola, pieces of orange, sweet potato, bread and candy. The chimpanzees were also given practice at repeating, by pressing a key, the symbol for one of the foods included in a series of symbols displayed on the projector line above the keyboard. In tasks like this some practice is needed on the symbols which correspond to foods the


animals do not actually Like to eat such as beancake and the commercially produced chow. The first version of the communication experiment went as follows. One of the animals (either Austin or Sherman—they alternated as sender and receiver) was led out of the room with the keyboard to another room,’ where it watched the experimenter place a single item from the eleven possibilities in a container, which was then sealed. This ‘sender’ animal was then taken back, with the container, to the keyboard, and given signals on the projector line which served as an instruction to name the food present in the container, and elicited the pressing of a key. The receiving chimpanzee then had to repeat the same interaction with the keyboard. If both the ‘sender’ (who had seen the food hidden) and the receiver (who had not) selected the appropriate symbol, they were allowed to share the food in the container. If either of them was wrong, they were both allowed to look at what had been hidden, but not to eat

In fact, they were very rarely wrong, since both responded correctly on more than 90 per cent of the trials in the initial procedure, and throughout a number of variations designed to eliminate unwanted explanations of the results. In these it was established that a ‘Clever Hans’ effect, of utilizing the knowledge of the human experimenter, was not occurring, since the experimenters present at the keyboard did not know the answer, not having seen which of the eleven foods had been hidden. The second animal was not simply copying the movements of the first, since he could do just as well by only being allowed to see the results of the first one’s movements in the symbols on the projector line. And the second animal was not just copying the visual symbol, although that would have been sufficient for success, since after seeing the symbol he could select a photograph of the correct food on that trial, Out of a choice of three alternatives (Savage- Rumbaugh et al., 1978b). Apparently the arbitrary visual patterns had come to elicit a mental image, or inner representation, of the foods they signified.

A further variation was to put Austin and Sherman in adjacent compartments, with a window between them, each with his own keyboard which the other could see. This arrangement was used for the transmission of the symbol for the food in the sealed container experiment, and also for a different test of communication. In this, one chimpanzee was given a tray containing a variety of foods, but the other was given none. Without any prompting from the experimenters,


the have-not animal used his keyboard to request food. With encouragement from the experimenters, the animal with food passed what was asked for through a gap in the window. After some alternation of roles between giving and receiving, compliance with requests from the other animal became fairly routine, especially for non-preferred items.

The passing of things through the window, by one animal to the other, was examined more systematically when the things which had to be passed were tools rather than items of food (Savage-Rumbaugh et al., 1978a). In this case the chimpanzees were first taught to use six objects to gain access to food, by imitating demonstrations made by the experimenter. These tools were: (1) a key, used to unlock padlocks on boxes and doors; (2) a wrench, used to unscrew bolted doors; (3) washers, used like coins to operate vending devices; (4) a stick, used either to dip into a yoghurt or puddings in containers not accessible in other ways, or to push a piece of fruit out of a horizontal tube; (5) a long straw, used to obtain liquids when this could be done by pushing the straw through a small hole in the tops of containers; (6) a sponge on a string, used by dropping it down a tall vertical container, as well as to soak up liquids from fiat surfaces. Investigation of tool-using as such was not the purpose of this experiment, and it became highly practised, but it is worth noting that each chimpanzee had to master a variety of unnatural manipulative skills. The important point was that the chimpanzees should associate a particular tool with a particular visual pattern used on the keyboard. This took a certain amount of training, but both animals were eventually able to select the appropriate key with 100 per cent accuracy if the experimenter held up one of the tools, and they were almost as good at picking out the correct object in response to instructions in the form of symbols appearing on the projector line above the keyboard, and in punching the appropriate key if they saw that a particular tool was needed—for instance if they watched the human experimenter put some fruit in a box and padlock the box they were accustomed to press the symbol for ‘key’ rather than any of the others.

The test for co-operation between the two chimpanzees came when they were placed in adjacent rooms, one being shown food or drink that was obtainable only with a particular tool, and the other being given the complete set of six tools. The room in which food was’ concealed contained a complete set of vending devices, horizontal and vertical tubes, padlocked boxes and so on, and the window between the two rooms was covered while the tool-requesting animal (in turn,


either Austin or Sherman) was shown one of these places being baited. When the window was opened, neither chimpanzee could see the hidden food, and thus only the tool- requester knew which tool was needed. On the first day of this testing, the requester animal readily used the keyboard to select the symbols give straw (or whatever tool was necessary) but directed his attention to the experimenter, who was present, and who had previously functioned as a donator of tools. The experimenter encouraged co-operation between the animals by pointing and gesturing to draw their attention to each other’s presence. After this the requesting animal began pointing out the symbols he had produced on the projectors to the other chimpanzee with the tools. By the end of the second day co-operative behaviour was well established —the chimpanzee with the tools quickly looked to see what the other chimpanzee had produced on the projector display, and handed over a tool, and the tool-using animal (again after encouragement from the experimenter) occasionally handed back through the window some of the food he had obtained with the aid of the tool. Errors were made quite often at this stage, although the level of accuracy (correct tool both requested and provided) was well above chance at 76 per cent. By the fifth day, however, without any intervention from the experimenter, 55 Out of 6o (92 per cent) correct tool transfers were recorded (Savage-Rumbaugh et al., 1978a). The day after this, the requester’s keyboard was turned off to see if the animals could get anywhere by gesturing, but the donor chimpanzee under these circumstances offered the same wrong implement repeatedly, or offered all the tools in turn.

There can be little doubt, in the case of this experiment, that the visual patterns used in the keyboard system had mental associations with objects, and that the chimpanzee who punched a particular key did this in the expectation that the other animal would hand him a particular tool. It depends, of course, on what one means by ‘intention’, but this seems to be a reasonable approximation to intentional communication. It may be that the intentionality of this use of symbols by the chimpanzees Austin and Sherman contributed to their later development of capacities for comprehension of the meaning of symbols which go beyond those demonstrated in any other individual animals. When they were trained with arbitrary symbols assigned to the two object categories ‘foods’ and ‘tools’ Austin and Sherman successfully selected the appropriate category, when shown arbitrary symbols which were the names for particular foods or tools (Savage-Rumbaugh et al., 1980). That is, they were able to label


labels, rather than merely label objects: for instance if shown the arbitrary pattern indicating ‘banana’ they responded by pressing the key meaning ‘food’, but if shown the symbol for ‘wrench’ they pressed the ‘tool’ key.

Intentional communication and deception by pointing

Before assessing the implications of the various more elaborate artificial methods of communication used by specially trained chimpanzees, it is worth looking at a study of one of the simplest means of intentional exchange of information—pointing by limb extension and gaze direction. Woodruff and Premack (1979) devised yet another variation of the ‘hidden food’ problem, in which a chimpanzee indicated the location of hidden food to a human partner, or vice versa. A small room contained a cage area partitioned off with wire mesh. In the first phase of the investigation, which lasted for 5 months, a chimpanzee sitting in the cage was shown that a piece of food was concealed in one of two containers (a box and a cup) placed in the room outside. After some to-ing and fro-ing, a human partner, who did not know where the food was, came into the room, and the chimpanzee was returned to the cage (the only access to the room was apparently through the cage and so the chimpanzee had to be removed while the human partner went through it). The task of the human participant was to guess, from the posture of the chimpanzee, where the food was. A peculiarity of the experiment was that the human participants were of two types. ‘Co-operative’ participants wore standard green laboratory clothing and were instructed to behave in a generally friendly and soothing manner. ‘Competitive’ participants wore black boots, a white coat and hat, and dark sunglasses, and were instructed to speak gruffly. Whether these features of dress and mannerism as such made any difference to chimpanzees is uncertain, but the two types ought to have been readily distinguishable. What presumably did make a difference was that the ‘co-operative participants’, if they found food at their first attempt, gave some of it to their animal partners, whereas the competitive participants kept it. But if the competitive partner failed to find food, the chimpanzee was released to retrieve it for itself. Four chimpanzees were used, with a number of human partners, and the animals differed considerably in their behaviour. However, it is claimed that all the chimpanzees conveyed more information to the


co-operative human participants, and that two of the chimpanzees deliberately misled the competitive humans by pointing to the object that did not contain the food. One of the chimpanzees, Sadie, quite unambiguously stuck a leg under the wire caging in the direction of one of the containers, and stared at it—pointing to the location of food in the case of a co-operative participant, but to the wrong container in the case of competitive participants.

In later stages of the experiment, lasting more than a year and starting 10 months after the first stage, the human participant remained inside the cage, knowing the location of food, while the chimpanzees had to guess where the food was, with the aid of the postures of the humans. The co-operative human participants attempted to mimic what had been observed in the chimpanzees, by sitting on the floor of the cage facing the correct container, extending an arm or a leg in that direction. The competitive humans did the same thing, but pointed towards the wrong container. With the co-operative humans, three of the chimpanzees were able to choose the correct container from the start, and another one gradually learned to do so. With the competitive human partner, all the animals were initially misled, and chose the empty container pointed at. One of them continued to do this, but the other three gradually learned to avoid picking up the container pointed at by a competitive partner, and chose the alternative one instead, being allowed to keep the food they thus acquired.

In many ways this is a rather messy experiment, but it provides support for the contention that postures and gestures may be given varying interpretations by chimpanzees depending on the human individual in whom they are observed, which is perhaps not surprising. More important, it is arguable that the chimpanzee’s own postures in this experiment were modified by their communicative intent. If they hoped that their human partner would succeed in finding food (because they would get some themselves) their posture was helpful, but if it was to the advantage of the chimpanzee that their human partner fail, it was not.

The relation between human speech and the use of artificial systems of communication by apes

It has often been argued that a chimpanzee (Gardner and Gardner, 1969) or a gorilla (Patterson, 3978) that makes a certain number of


gestures which approximate those of American Sign Language can be said to have ‘acquired language’, thus bridging an important gap between ape and man. Premack (1976) has made similar claims on the basis of chimpanzees’ manipulation of plastic tokens and, although Savage-Rumbaugh et al. (1978a, 1980) have been more cautious in interpreting the results they obtained with the keyboard system, they argue that the two animals in their experiment were ‘able to comprehend the symbolic and communicative function of the symbols’.

I should like first to emphasise that analogies between the demonstrated abilities of chimpanzees and the human use of speech are extremely shaky. There has of course been no shortage of sceptics, ready to point out the limitations of communicative abilities demonstrated by apes (see the collection of articles collected by Sebeok and Umriker-Sebeok, 1980: especially Chomsky, 1979; Bronowski and Bellugi, 1970; Mounin, 1976; Terrace, 1979; also Petitto and Seidenberg, 1979; Seidenberg and Petitto, 1979; Terrace et al., 1979). The most serious criticism of all is the assertion that the results obtained in communication experiments with apes are bogus, and not what they scent, because of surreptitious sensing by the apes of cues other than the ostensible causes of their communicative acts—the ‘Clever Hans’ problem. I shall argue that this criticism could apply to some, but not all, of the data. What remains, as achievements of the animals rather than interpretations by their trainers, can easily be distinguished from human speech on a number of grounds, among them modality of expression, syntactic complexity, disengagement from context, separation from emotions and goals, and internalisation. The problem is not so much to show that human speech is different from anything done by other species, but to say exactly how and why it is different.


Surreptitious use of human directions

Osker Pfungst (1911) was able to debunk the performances of Clever Hans, the horse that stamped out answers to sums chalked on a blackboard, by presenting the horse with questions which the horse’s trainer could not see. Evidently the horse started and stopped stamping its hoof according to cues of very slight changes in posture and facial expression given—unwittingly—by the human intermediary.

Umiker-Sebeok and Sebeok (1980) have put forward a blanket


condemnation of all work with apes of the type discussed in this chapter, on the grounds that similar social cues could enable the animals to make the observed responses even in ‘double-blind’ tests, in which the person accompanying the animal supposedly does not himself know enough to be of any help. In my view a certain amount of doubt of this kind must be attached to most of the work with the plastic token system discussed by Premack (1976), since many of the responses of the chimpanzees required the selection of one token from a choice of two—something easily influenced by the trainer glancing at the correct one. When a trainer was used who did not himself know the correct answers, the chimpanzee was considerably less accurate, although she performed well enough to rule out the possibility that dependence on cues from the trainer was total (Premack, 1976, pp. 32—6). In some of the later experiments, the chimpanzee became accustomed to making choices after the trainer had left the room, which provides much stronger evidence (Woodrufl et al., 1978; Gillan et al., 1981).

When two chimpanzees exchanged information between themselves, using the computer-controlled keyboard system, with experimenters not in the same room (Savage-Rumbaugh et al., 1978b), the evidence seems relatively robust. In this case, the animals had to choose one from six possibilities (and in the first test of communication between the same two animals, one from eleven different symbols for food objects: Savage-Rumbaugh et al., 1978b), and this sort of choice could not so easily be influenced by extraneous cues. All in all, with the computer- controlled keyboards, and with the carefully controlled tests of naming with the gesture-sign method (Fouts, 1973; Gardner and Gardner, 1978) the explanations in terms of social cuing appear tome to be much more far-fetched than the assumption that the animals pay attention to the ostensible objects and signals.

The most important type of unwitting human direction of behaviour which has been interpreted as the product of the mental organisation of the apes themselves is in the ‘prompting’ of sequences of gestures in animals trained with the American Sign Language method. Terrace et al. (1979), by detailed examination of film and videotape records, have shown that sequences of gestures such as me hug cat are usually interspersed with gestures from a human companion, such as you, just before me and who just before cat. Since, in this instance, the trainer was holding a cat (in one arm, while signing with the other), at the beginning of the sequences, and the chimpanzee was allowed to


hold the cat after gesturing for it, one might reasonably assume that the separate gesture-sign, cat, was elicited by the sight of the cat. But the order in which the signs were given was very clearly open to the influence of the trainer. As practically all instances of sequences or combinations of gestures by chimpanzees or gorillas are made in the context of interactions with a human companion, there is virtually no evidence of this kind which is not vulnerable to the charge that the human contact determined the sequence of combinations observed.

Syntactic complexity

It is especially unfortunate that one must discard evidence concerning regularities in the serial combination of signs of chimpanzees and gorillas, because these would bear on the question of whether any grammatical rules whatsoever can be observed within this kind of use of symbols. If it was the case that internal organisation, rather than external cues, led to such regularities, the use of more before rather than after signs for such things as banana, drink and tickle (e.g. Patterson, 1978) might be the starting point for arguments about whether this should count as a primitive form of syntax. However, even with the benefit of human prompting, the serial combination of gesture-signs by apes is limited in the extreme. Nim Chimsky, the chimpanzee studied by Terrace et al. (1979) recorded ‘banana Nim eat’ occasionally, but ‘banana eat Nim’ occurred almost as often. What are listed as ‘Four-sign combinations’ turn out to be such things as ‘eat drink eat drink’ and ‘banana Nim banana Nim’. It may be that banana has a rather high priority, for obvious reasons, and tends to come first if it is going to be signed at all, but this would be only a marginal form of syntax. Thus, even accepting such gesture sequences as have been reported as genuine products of ordering by the animals, and taking a gesture as equivalent to a spoken word, variety and regularities in gestures that appear to be strung together by the trained apes bear comparison only with the speech of children who use two- and three-word utterances for a few months after they have begun to talk, and this comparison itself is very superficial.

Although the order of words in a sentence or phrase is not as significant a feature of syntax in other languages as it is in English, it is often emphasised that hierarchical recombination of elements is a pervasive characteristic of human language. In speech, phonetic


elements such as the positions of the tongue and lips, voicing and aspiration are combined into phonemes (e.g., roughly, the English vowels and consonants); these are combined into units of meaning (morphemes) such as words and parts of words in English; and morphemes are combined according to grammatical rules to make phrases. The artificial systems of communication used with apes to all intents and purposes circumvent the need for hierarchical permutations and combinations of elements by providing the animals with a ready- made fixed correspondence between a set of referents (mostly objects and simple actions) and a set of symbols which are all individually recognisable. Does this mean that apes can label individual objects with individual gestures, or other symbols, but lack the ability to associate qualities and features of objects and perceived events with detachable components of gestures or symbols? Is the facility for syntax and grammar, which allows rules for the recombination of linguistic elements, the crucial achievement of human intelligence, entirely absent in other primates (e.g. Chomsky, 1979)? In effect, this is what the data suggest. However, there are a number of other limitations in the communicative performance of apes. Take for example the lack of disengagement of communication from context (Bronowski and Belugi, 1970; see below). Is the lack of disengagement from context due to the absence of syntactical rules, or could the difficulties with syntax arise from the lack of disengagement from context? ,

Disengagement from context and separation from goals: ends and means

Children may use the early stages of speech for the expression of wants and needs, and more generally, as a means to the end of social attention. But words typically become ends in themselves if they form questions and answers between child and parent, and exchanges of narrative, even if only in the utilitarian social contexts of the child saying ‘Daddy gone’, or the parent telling a bedtime story. Comment on the external world, rather than a change in the world, seems to become a primary motivation for speech very early on (Clark and Clark, 1977). By comparison, the use of artificial systems of symbols by apes appears to be very closely tied to the achievement of immediate tangible goals. It is not true that the animals have to get a banana, or be tickled, every time they gesture or manipulate a symbol, but the


frequency of external rewards, and the frequency of deliberate social instigation of symbols used by human experimenters, is very high. It seems necessary to conclude that communicative acts by apes remain very closely tied to specific external goals, and the exchange of information does not become an end in itself (Mounin, 1976). Proposing an innate tendency for social communication confined to the human species is rather unwieldy, but for whatever reason, the actual quantity of communication induced in apes is small, and it is possible to characterise it as goal-directed problem solving (Terrace, 1979) as opposed to an independent means of information exchange.

Internalisation and inner speech: communication and thought

In English, grammatical irregularities, especially of plural nouns and the past tense, make it apparent that 3- and 4-year-old children do not merely mimic everything that is said to them, they deduce rules, and are thus led into errors such as ‘mans’ and ‘mouses’ and ‘falled’ and ‘corned’. For this and more obscure reasons, children are said to create their own internal system of grammar. Whether this results from an inherited linguistic blueprint, or a more general tendency to extend perceived regularities, is still a moot point. Of course it would be expected that apes should lack anything as specific as innate devices to deal with linguistic problems per se, whereas they ought to possess a good measure of any tendency to generalise perceived regularities. But if the interest that apes have in the symbolic communication systems presented to them is largely confined to the achievement of particular tangible goals, this in itself might prevent the emergence of internal grammatical regularities. In any event, the incorporation of external communication skills, such as they are, into an internal syntactical system is not a reliable consequence of the training schemes so far devised. This is one aspect of internal mental organisation which is bound up with the human use of language and which apes appear to resist.

A more intuitively obvious question about the internalisation of communication concerns the use of language for inner speech—self. instruction, imaginary conversations, verbal thought, and the organisation of conscious experience. In the child, the use of speech for self-instruction can sometimes be observed because the child’s instructions to himself are spoken out loud, and overheard conversations with dolls


or reports of discussions held with imaginary friends leave little doubt that the internal use of speech goes far beyond the more overt behavioural manifestations. The most coherent and widely accepted theory of the relation between speech and thought in childhood (Vygotsky, 1962) suggests that verbal and non-verbal mental processes are initially independent, and gradually intertwine. But adults, and especially academic adults, are often reluctant to accept that their own conscious thought is possible without the internal use of language. Some still argue that visual imagery is theoretically indistinguishable from verbal propositions (e.g. Anderson, 1978, 1979).

Although chimpanzees such as the Gardners’ Washoe, who have been taught sign-language gestures, use signs to label objects seen in picture books, and in association with actions in play, the communicative abilities of these and other apes provide little evidence that the associations of symbols and objects give rise to anything analogous to inner speech. In most cases, the assumption should be that the animals think in terms of perceptual schemata (of bananas, sticks, being tickled and so on) and apply labels to their inner perceptions in order to achieve a current goal, or as a result of prior training. The main exceptions are those when the chimpanzee Sarah (Premack, 1976; see pp. 357—64 above) apparently identified symbols in terms of other symbols: as when a plastic token designated name was placed beside tokens that designated objects but a token called colour was placed next to those which referred to particular hues. If it is genuine, this result is extremely significant since it implies a form of mental organisation in which artificial labels are related to each other rather than only directly related to internal representations of external events. But the test is suspect (a) on the grounds of social cueing from the trainer; and (b) the correct token was selected from only two alternatives—this kind of ‘forced-choice’ is not particularly convincing. It may very well be possible for apes to make some distinctions between categories of symbols which they have been taught, but even so, this would be a far cry from the propositions which are characteristic of human speech and thought.

Labels imply mental representations

Apes trained to employ artificial systems of symbolic communication ought not, therefore, to be said to have acquired a language, in the


sense that people acquire a language. Human language is unique to humans, and although some of the distinctive features of human speech, such as the mimicking of sounds, may be observed in other species, the resemblance between, for instance, the trained gesturing of a chimpanzee and communication via sign-language among the human deaf is in some senses no greater than the resemblance between the speech of a parrot and that of its owner. This being clear, what may be deduced from the study of the mental capacities of non- human primates? I suggest that the experiments on symbolic communication support work such as that of Kohler (1925) On problem-solving and that of Menzel (1978) on spatial memory: we should conclude that apes are capable of organised mental knowledge about objects and actions.

Experimental evidence of problem-solving and memory might by itself be sufficient to make this case, but doubts about whether even chimpanzees can categorise objects are surely resolved if symbolic labels can be attached to their perceptions. I have suggested that perceptual categorisation and the retention of inner descriptions of objects are intrinsic characteristics of brain function in many other animals apart from the anthropoid apes. The additional capacity which appears to be demonstrated by training apes with artificial methods of communication is the generation of mental associations between inner descriptions and representations for one set of perceptual categories which deal with a world of food items, physical space and social relationships, and another set of categories within the system of gestures or visual patterns imposed by specialised training. In so far as it can be demonstrated that the apes establish a collection of associations between signs and objects, then the results of their training extend further than any previously observed form of animal learning, but it is not clear that they need a substantially different kind of ability to make these associations from that which may be used by other mammals to respond to smaller sets of signals. A dog which usually hears a buzzer before being given a piece of liver may be said to form an association between the arbitrary sign of the buzzer and the object category of ‘liver’: is this any different from the mental linkage between a banana and the human gesture which precedes bananas, established in the training of chimpanzees? One difference is in the sheer quantity of associations—over 100 distinct signals are often claimed to be understood by a chimpanzee thoroughly exposed to the gesture-sign method, whereas experimental techniques used with


other species rarely provide evidence that an animal can give distinct responses to more than one or two arbitrary signals. This may be partly a restriction in the method rather than in the animals—sheepdogs, working horses and working elephants may respond to more than just one or two commands, although it seems likely that the number would be far less than 100.

The distinctive behavioural fact about the chimpanzee experiments is of course that the animals themselves produce or select signs rather than only reacting to external commands. This in itself may be helpful in establishing connections between signs and references but it is difficult to say whether the productive use of signs is a result of a superior kind of mental organisation or of the greater physical opportunities provided. It is theoretically important that the production of signs is accompanied by communicative intent. This is not something which can be quantitatively assessed, but there is general agreement that a social context to the production of any form of signing is essential. Even when a computer-controlled keyboard is used, so that tests can be made in the absence of a human presence, social interactions between human trainers and the animal being trained are apparently necessary if the animal is to show any interest in using the keyboard (Rumbaugh, 1977; Savage-Rumbaugh et al., 1985).

Names and labels and words and things

One way of describing the outcome of the experiments in which apes learn gesture-signs or other symbols is to say that these animals can be taught to use names, ‘in a way which strongly resembles the child’s early learning of words’ (Bronowski and Bellugi, 1970). Apes do not get very much further, at least with the training methods now used. Therefore it may be misleading to talk about the ‘ability to name’, since the use of names by human adults is certainly not the same as the use of names by young children (Clark, 1973), and may reflect syntactic and other internal complexities which the apes lack. However, as long as naming is opposed to other kinds of linguistic competence it may be useful to emphasise that the use of symbols by apes is more like naming than anything else. In so far as they string gestures together we might call this ‘serial naming’, a phrase used by Luria (1970) to describe agrammatic speech in human aphasia due to left-hemisphere brain damage. But a word used as a name in normal


human speech may carry with it many more implications than a symbolic label apparently serving a roughly similar purpose in the rudimentary signing of an ape. We may assume that a person who says or reads the word ‘canary’ has access to a network of knowledge expressible in propositions such as ‘a canary is a bird’ and ‘a canary has wings’ (Collins and Quillian, 1969; Collins and Loftus, 1975). We cannot assume that an ape which makes a gesture designated as the name of its human companion has access to similar propositions by which that individual is identified as a mammal, with a list of mammalian characteristics therefore attached. Names in human speech, even if used singly, are potentially insertable into purely verbal propositions, whereas inmost cases signs used as names by apes are not.

Any theory of why this should be so must be largely a matter of speculation, but in order to defend the contention that there are significant continuities between the functions of the human brain, and those of other vertebrates, some account needs to be taken of the evident discontinuities. The first step must surely be to acknowledge that the utilisation of speech as we know it amplifies the cognitive capacities of the primate brain almost, but not quite, beyond recognition. Some try to avoid this step by exaggerating the similarities between natural human speech and the linguistic competence of apes (Premack, 1976; Gardner and Gardner, 1978) ; others, once they have taken the step, simply assert that the human species is biologically different from any other, to such an extent that any comparison with other species is futile (Chomsky, 1976, 1980; Lenneberg, 1967).

In between these positions, theories of how human language may have evolved are legion (Hewes, 1975). What I would like to suggest is that, whatever the sources and stages in the evolution of human language, there remain sufficient similarities between human and animal brain function to allow for comparisons. The similarities, which I have discussed in previous chapters, depend on the assumption that animal brains are devices for selecting and organising perceived information, and that the neural systems which accomplish perception and memory exhibit evolutionary continuity. Behavioural data from tests of perception and memory support the contention that mental organisation exists in animals, in particular in non-human primates, as a necessary part of the perception of objects and their localisation and interrelationships in space and time. It does not seem unreasonable to suppose, therefore, that the development of the distinctively human capacity for language took place via the deployment of the same


forms of mental organisation for new purposes. Gregory (1970), for instance, suggested that human language capacity makes use of the existence of a ‘grammar’ of rules which the brain uses to interpret patterns of light impinging on the retina in terms of objects. O’Keefe and Nadel (1978) imagine that the brain organisation necessary for cognitive maps in strictly spatial co-ordinates becomes used in man for verbal relationships. We have Chomsky’s imprimatur to continue such speculation, provided proper caution is observed, (Chomsky, 1979; p. 132).

It does not seem necessary to add further speculative details, except perhaps to observe that hearing seems to have been neglected and that linking the identity and location of objects to sounds ought to be one of the prerequisites for human speech. But generally, work I have reviewed in previous chapters supports the assumption that animals perceive and remember things in terms of inner descriptions and cognitive schemata. We can say then that the human and non-human species may in the first place perceive and remember, without the assistance of language, in a roughly similar way. In addition we humans may perceive, remember and reproduce words, instead of things. It is important that words can serve as labels for things, but perhaps even more important that words become alternatives to things, so that we perceive and remember relationships between words, as if the words were themselves things. According to this view the human brain is able to accommodate mental organisation in the form of relationships between words because the brains from which it evolved were already accommodating mental organisation in the form of relations between and within perceptual schemata. Language can be internalised in the human brain because the vertebrate brain in general and the mammalian brain in particular serves to construct internal representations as a means of adapting to external realities.

Although this is vague, the assumption that advanced vertebrate brains construct models, images or schemata with which to interpret the outside world is not unusual (e.g. Piaget, 1971; Craik, 1943). It would certainly be very mysterious if adapting this system to treat vocalisations as objects allowed for human language and all that flows from it, but it would be even more mysterious if the functions of the human brain bore no relation whatever to the functions of brains which are physically rather similar.