Experimental studies on tool use and cognitive abilities in animals

Comprehensive experimental investigation of solving problems using tools by animals may bring knowledge about the level of their apparent “understanding” of the features that give tools their particular functions. Studies of tool use in animals include physical relations involved in animal’s actions, categorisation of different sets of things by their functional properties, taking invisible forces into account by tool users, as well as animals’ ability to anticipate their actions.

Experimental studies of relations between tool use and intelligence in animals have a dramatic history. During long time primates have been in the lead and recently new insights came from birds. Recent achievements of wild tool users are very impressive and may shift our imagination about animal intelligence into acknowledgement of the high level of their understanding of causal relationships. At the same time there is enough room both for enthusiasm and scepticism. For instance, just as Kö hler referred to “good and bad mistakes” related to tool use in his chimpanzees, Visalberghi (2002) referred to a delicate balance between “doing and understanding” in capuchin monkeys. The fact that firstly strokes her interest in problem solving by capuchins was the observation in the zoo of an adult male pounding an unshelled peanut with a boiled potato. Since that time, the researcher has conducted many experiments that demonstrated, on the one hand, how successful capuchins are in solving problems, and, on the other, how relatively little they understand of what they do. So let us try to find together the balance between causality and idle enthusiasm in wild tool users, analysing results of experiments aimed to studying instrumental problem solving.

In fact, the first chimpanzees that appeared in laboratories at the beginning of twentieth century received sticks and faced instrumental problems. More than a century ago Hobhouse (1901) suggested a problem that later would initiate many modifications and thus form a new experimental paradigm. The experimenter presented his chimpanzee Professor with a “tube-problem”, suggesting it to push an item of food out of a short tube. Professor easily solved this problem. Yerks (1916b) suggested a variant of this problem to two apes. The task was complicated by that the tube was too long to be taken in hands (170 sm of the length) so it was placed on the ground. A chimpanzee solved this problem after 12 days of attempts. An animal appeared to understand that to be grasped a food item has to be pushed away. A gorilla mastered this task only after being demonstrated by the experimenter how to do this.

Roginskii (1948) and Ladygina-Kohts (1959) investigated primates’ tool behaviour working mainly with Paris, the most curious and enthusiastic adult male from a group of chimpanzees in Moscow Zoo.

Together with Kö hler’s and Yerks’, these experiments were the first to thoroughly study animals’ problem solving by means of tool use. They became prototypical for many modified versions that would be elaborated, so we consider results of prototypical and modern experiments all together here.

Means-end relationships. Understanding means-end relationships is one of the key elements of cognition in animals and this is also an important step in human cognitive development: a significant transition occurs at around the age of 8 months, when infants move beyond a reliance on what Piaget (1952, 1954) called “circular reactions” (in effect, operant conditioning) to an understanding of means-end relationships (Willatts, 1999).

In order to study the understanding of means-ends relationships, it is necessary to use problem-solving tasks in which a solution to the problem can in principle be perceived directly, without trial and error or previous experience of similar tasks. Means-end understanding is most clearly demonstrated if the subject shows an “insightful” solution to the problem on the first trial, since correct performance at the end of a period of training may well represent the effects of operant conditioning. The usual way of studying means-end understanding in animals, introduced by Kö hler (1925), is by offering a possible physical connection to an out-of-reach object of desire, for example a string that is attached to a piece of food (it is also called the “string-drawing problem”). The food itself is out of reach of the animal, but the near end of the string is accessible. If the animal understands the physical properties of the string it uses it as a means to an end, i.e. pulls the food into reach with the string. It has been

revealed in early studies that chimpanzees and monkeys readily cope with this task (Harlow and Settlage, 1934).

Recently Nissani (2004, 2006) has applied the string-drawing paradigm to Asian elephants. Having a great deal of experience with studying insects’ behaviour (Nissani, 1977), he has compared design of this problem with a natural situation described as early as by Fabre (1879) from the life history of a digger wasp. The wasp seemed to lack mens-end understanding. For instance, according to Fabre’s observations, the wasp could not understand that she can grab her paralyzed prey by a leg instead of an antenna. If the head appendages of her particular prey species suddenly ceased to exist, “her race would perish, for lack of the capacity to solve this trivial problem” (cf. Griffin, 2001; Nissani, 1977). In Nissani’s (2004, 2006) experiments elephants did not exceed insects in causal reasoning. In fact, the understanding of means-end relationships, as many other displays of problem solving that have already been mentioned, give a very mixed picture for comparative analyst. Let us consider different variants of this experimental paradigm.

One variant of problems to be solved by an organism in order to demonstrate understanding means-end relations is a “support problem”.

In this problem a reward is placed on a cloth. The reward itself is outside the subject’s reach but one of the ends of the cloth is within reach. The solution of a problem consists of pulling in the cloth to bring the reward within reach. Roginskii (1948) suggested this problem to the chimpanzee Paris, and Piaget (1952) studied this problem on human infants. One-year old children not only readily pull in the cloth, but, more importantly, they withhold pulling when the reward is not in contact with the cloth. This indicates that children at this age understand that spatial contact is necessary for the tool to act on the reward. The chimpanzee Paris displayed the same level of competence in solving the support problem. Spinozzi and Potì (1989) administrated more detailed support experiments with infants of apes and monkeys. In one condition the reward was placed on the cloth to the side. All primates (1 Japanese macaque, 2 capuchin monkeys, 2 long-tail macaques, and 1 gorilla) responded appropriately by pulling in the cloth when the reward was on the cloth, and withheld pulling when the reward was off the cloth. In a second experiment Spinozzi and Potì (1989) tested the generality of these finding by modifying the conditions of the off-cloth condition by placing the reward near the end of the cloth rather than to the side of it. The authors reasoned that if subjects had simply learned to respond appropriately to a specific configuration of the cloth and the reward rather than more general relations between them, they would respond inappropriately to this novel configuration. Under these circumstances, all subjects pulled in the on-cloth condition but not in the off-cloth condition. Later Spinozzi and Potì (1993) suggested this problem to two infant chimpanzees and one of them succeeded (pessimist would rather say “only one of them” succeeded). This illustrates a great deal of individual differences in problem solving in animals which will be discussed in Part VII.

Hauser and colleagues (1999) turned next to a modified version of this task in which tamarins were suggested to choose one of two pieces of cloth to obtain a food reward. Monkeys focused on changes to the cloths that affected its affordances. For instance, they rejected pulling cloths made of material which did not afford pulling such as pieces of cloth connected with chipped wood, sand or a broken rope. At the same time, they chose cloths of radically different shapes (i.e. triangles, circles, teeth-shaped) that functionally supported the food reward. Further they distinguished between cloths that supported the food reward and cloths that were merely in contact with the food and thus functionally inappropriate. Thus, tamarins distinguished the features that were relevant for the tool’s function from those that were not (Santos et al, 2003).

Some variants of the support problem turned out to be a challenging task for primates. This concerns a “rope problem”, or “string-pulling”: a subject is required to pull a cap containing a food reward with the use of a rope. The simplest variant is that in which the rope is visibly attached to a knob of the cup. This task has been easily solved by apes and monkeys in many studies. Roginskii (1948) conducted plenty of experiments with two species of baboons (Papio porcarius and P. hamadryas), rhesus macaques, pig-tailed macaques, and a mandrill. Monkeys were tested with 38 variants of the rope problem in which some ropes were connected with the reward whereas other were not, or were broken so as not to support the cup. All monkeys displayed good results in choosing relevant ropes, and one pig-tailed macaque championed all tasks with the exception of one variant in which a very long (2.5 m) rope being connected to the cup by many zigzags. The macaque first chose a short straight rope that was not connected with the reward and learned to choose the long crooked rope after many trials (Fig. VI-3).

In other series of Roginskii’s (1948) experiments the chimpanzee Paris was presented with a piece of food placed into a cap out of reach and was expected to bring the cup within reach pulling it by a rope that must be passed through a knob of the cup so as two ends of the rope were put together in a doer’s hand. Paris did it from the first time but it was likely to be by chance because he was wrong for the second time, and learnt to perform this task perfectly and systematically only after 30 additional trials. He made mistakes again when problem specification was slightly changed, and thus was failed to demonstrate full understanding of interrelations between the cup and the rope.

String-pulling behaviour has been studied in many species, and, as it has been already mentioned, the picture is mixed. In particular, the string-pulling paradigm has been tested in a variety of birds. The crucial step in solving the problem for birds is a combination of several behavioural steps that must be completed in the right sequence. This includes reaching down, pulling up the string with the bill, placing the string on the perch, stepping on it with a foot, letting go off the bill, reaching down again, and repeating this cycle for at least 5 times. Most studies showed that the birds either had a hereditary coordination of movements that was adapted to normal feeding habits or they simply would learn the task through trial and error. A great individual variance has been observed in greylag gees (Fritz et al., 2000) and songbirds (Vince, 1956, 1958, 1961; Blagosklonov, 1974). Studies on ravens (Heinrich, 1995, 1999, 2000), and psittacids (Funk, 2002) showed that at least some individuals solved the task without any former training or trial-and-error learning thus implying that the mechanism used was insight. Kea showed an extraordinary success rate in this study (Fig. VI- 4). Except for a fledging, all individuals tested spontaneously pulled the reward in their first trial implying that they understood the mean-end connection and the underlying physical properties of the task. When kea had to choose between two strings in eight different tasks they significantly selected the correct string in all tasks even when strings were crossed. Therefore kea clearly comprehended the functional connection between the string and the reward. Great individual differences both in the performance and the methods used to obtain the reward exclude innate behavioural patterns (Huber, 2002; Werdenich and Huber, 2006).

Different species of old world and new world monkeys and lemurs solved the problem successfully (for a review see: Hauser et al., 2002).

There are long-term controversies concerning whether string-pulling behaviour is “insightful” or is conditioned by trial and error in cats (Adams, 1929), rats (Tolman, 1937), and dogs (Shepherd, 1915; Kö hler, 1925; Scott and Fuller, 1965; for a detailed review see: Osthaus et al., 2005). Recent experiments with the use of combined experimental techniques and a great sample of dogs showed that although dogs can learn to pull on a string to obtain food, they do not spontaneously understand means-end connections involving strings (Osthaus et al., 2005). Instead of grasping regularities in mean-end relationships, dogs rather demonstrate a high level of social intelligence relying on their owner’s experience (Hare et al., 2002). These results are in conformity with Pepperberg’s (2004) data obtained on grey parrots. Pepperberg found that two language-trained parrots demonstrated no means-end understanding, but simply asked their human trainers to give them the treat, whereas parrots that had had no language training solved the problem easily. It appears that the availability of human-aided solutions to problems can sometimes inhibit the expression of animals’ cognitive capacities. It is likely that very modest feats of intelligence displayed by elephants in solving the mean-end task in Nissani’s (2004, 2006) experiments cited above have been caused by the same effect, but this is not, of course, a single possible explanation of these results.

Stick problem. Stick problem consists of using a tool to bring in a reward that is not in direct contact with the tool. This situation entails putting the tool into contact with the reward and then sweeping the reward within reach. Solving this task demonstrates an ability to understand complex causal relations such as that the stick must be of appropriate size and material (long and rigid) and that only certain kinds of contacts (with the certain force and directionality) should be successful.

Roginskii (1948) suggested to the chimp Paris a classic Kö hler’s problem, that is, to bring a food reward placed out of reach within reach pulling it by a stick. Getting used to the problem, Paris then selected appropriate tools and successfully adapted objects that did not satisfy his requirements, for example, he selected a stick of appropriate length, and splintered thick sticks into flimsy ones in order to act effectively. Later primates including baboons, macaques, orang-utans, chimpanzees and tamarins were demonstrated to be capable of solving the stick problems. Some of them used multiple objects in a combinatorial manner, placing one object inside another (see: Tomasello and Call, 1997; Call, 2000, for reviews).

Jones and Kamil (1973) presented the first detailed description of spontaneous solving of the stick problem in laboratory-raised blue jays Cyanocitta cristata that is likely to be based on understanding of causal relations by birds. The pioneering female jay was seen engaged in the following behavioural sequence. She ripped a piece of newspaper from the pages kept beneath its cage, manipulated with the piece of paper, and then proceeded to thrust it back and forth between the wires of its cage, raking in food pellets too distant to be picked up directly with its beak. Experimenters were very surprised because there were no reports about tool using in wild blue jays. It seemed that sophisticated tool manufacturing displayed by that bird had been acquired by food deprivation that was aimed to maintain the bird’s health. After the authors observed the jay manufacturing tools, they isolated her from the other jays in their colony in a cage equipped for filming and presented her with pieces of papers. The bird displayed many times how co-ordinated beak and feet manipulations resulted in paper wisps. She then positioned a wisp on one side of the pellets and by repositioning its grip on the paper made successive sweeping movements of the paper to the opposite side of the pellets, thus sweeping pellets in an arc nearer a point where they could be reached with the beak from between the wires. On different occasions the experimenters presented the jay with a feather, a thistle, a piece of straw grass, a paper clip, and a plastic bag tie. In all cases the bird thrust the object between wires and, except when using the thistle, was successful in raking the pellets. The authors then tested a number of blue jays in their colony. Of eight hand-raised birds tested, five showed definite tool use, two displayed some components of the behaviour, and only one showed no sign of tool use. Even this single jay showed a high level of manipulation with the paper. Jones and Kamil (1973) consider tool making in their subjects as indicative of the particular potential for behavioural adaptations typical for some species with highly generalised feeding behaviour, such as the Northen blue jay.

Tube problem. As it was noted before, the tube problem for more than a century serves as a relevant procedure to study animals’ understanding of the causal relations between the elements of the task. The simplest variant of a laboratory device is a transparent tube with food visible inside. To acquire the food, subjects have to poke it out the distant end with a stick.

Ladygina-Kohts (1959) presented the chimpanzee Paris with many variants of the problem involving different types of tools that required different solutions. Tools were presented in groups of three in order to allow the chimpanzee to make a choice between more and less appropriate tools. He was also required to take some tools for completion by re-shaping or detaching, or- in other cases- to combine several things together (see Fig. VI-5 and VI-6). Three main variants of these tasks (the bundle task, the short-sticks task, and the H-tool task) were later suggested by Visalberghi and Trinca (1989) to capuchin monkeys and became popular tasks for studying possible mental manipulations of the elements of tasks in tool using animals (Limongelli et al., 1995; Povinelli, 2000; Tebbich et al., 2001). In the bundle task, subjects were given a bundle of sticks taped together that as a whole was too wide to fit in the tube, the solution consisted of breaking the sticks apart. In the short-sticks task, subjects were given three short sticks that together added up to the length required; the solution consisted of putting them all in the same end of the tube to displace the food out the other side. In the H-tool task, subjects were given a stick with transverse pieces of either end that prevented its insertion into the tube; the solution consisted of removing the blocking piece from the tool.

As an honour to a galaxy of chimpanzees participating in cognitive experiments, Paris coped with the majority of tasks displaying remarkable quick-wittiness and enthusiasm. He straightened the wire, winded off the rope, took off cross-pieces, picked off widening by his teeth; he effectively re-shaped, detached, or combined tools, compared elements of tools at different stages of manipulations with them, and periodically checked whether the tool was already suitable for inserting in the tube. At that, like all chimpanzees, Paris chronically lacked precision. For instance, he straightened the wire to the extent that only allowed inserting it into the tube with effort. One more turn would allow coping with the task easily but the chimpanzee preferred to puff over this work for a long time trying to insert the under- straighten wire into the tube until a goal was attained.

The tube problem has been suggested to chimpanzees, bonobos, and orang-utans. Together with Ladygina-Koths’ results, it is likely that apes operate with foresight, at least, they obtained rapid success in the basic task, that is, in selecting sticks of the appropriate diameter to fit the tube. It does not mean that apes are foresighted without a limit. They proved less successful in the H-tool task, and some chimpanzees failed this variant of the tube problem.

Visalberghi and co-authors (Visalberghi and Trinka, 1989; Fragaszy et al., 2004) suggested the tube problem to capuchin monkeys. The subjects were presented with three variants of the problem, namely, the bundle task, the short-stick task, and the H-tool task. Although all capuchins eventually solved these variations of the task, they made a number of errors such as attempting to insert the whole bundle and inserting one short stick in one end of the tube, and another short stick in the other end. Moreover, these errors did not decrease significantly over trials, suggesting that capuchins understood little about the causal relations between the elements in the task.

Firsov (1977) suggested a variant of the tube problem that can be named a “ reverse tube problem ” (Fig.VI-7). Firsov conducted plenty of experiments aimed at studying tool use with a group of chimpanzees that were freely housed in a small island in a lake for several summer months during several years. The “reverse tube problem” looked as follows. Four apes (Silva, Gamma, Taras, Boy) were required to extract a small food item from a hole in the ground. The hole was 80 cm deep, that is, about 10 cm more than the length of the chimpanzees’ arm. First reactions were similarly simple in all animals, that is, ineffective attempts to fit the hole with the use of each of four extremities in turn. After a short rest, an ape repeated attempts, now with less enthusiasm, and then started to look farther ahead. From four apes, only Silva manufactured tools to fit the hole. She prepared up to four tools and then selected the most appropriate, that is, long and elastic enough to pull the food item up by clasping it to one side of the hole. Other apes did not make tools from tree branches, instead, they picked sticks up, brought to the hole and detached with the help of their teeth and fingers.

A new problem presented by Visalberghi and Limongelli (1996) allow further examining limits of understanding of causal relations in animals. This problem, that was named tube-trap problem, “punishes” subjects who do not foresee the consequences of their behaviour. The apparatus consists of a horizontal tube with a closed " trap" at its centre. Food is placed in the small metal cup inside the tube, and a stick is provided. The subject has to remove the food from the tube without it falling into the trap. Because the subjects are able to pull objects towards them as well as push them away, they have to either pull the cup when it is between them and the trap, or push the cup if it is on the far side of the trap (see Fig. VI-8). Visalberghi and Limongelli (1996) found that one of their capuchins systematically solved the task pushing the reward away from the trap. Although this female seemed to be planning her moves in advance, the authors noted that in half of the trials she inserted the tool in the wrong side of the tube and upon seeing that the reward was moving into the trap, she withdraw the tool, reinserted in the other end and pushed the reward out. The experimenters probed further her understanding of the relation between the trap and the reward inverting 180⁰ so that the trap was on the top of the tube where it did not “rob” the reward of the user. The monkey, however, persisted on her strategy of pushing the cup away from the trap which indicated that she only learned the simple rules without deep understanding of the problem.

Limongelli et al. (1995) presented the trap-tube task to five chimpanzees who behaved at chance level for the first 70 trials, although two of them learned to avoid the trap during 70 additional trials. The experimenters then varied the location of the trap in the tube in order to learn whether the apes understood relations between the positions of the reward with respect to the trap, or whether they were simply using the simple rule of pushing the reward out of the side to which it is closest, thus avoiding the trap. In some variants of the new set of tasks the trap was located very close to one end with the food just beyond it, so that subjects actually had to push the food out the end from which it was farthest. In other cases, the opposite arrangement was used. The chimpanzees easily solved the problem. It should be noted, however, that the variations used in this experiment could still be solved by the rule “push the food away from the trap”, which could have been learned during the previous trials (Call, 2000). Unfortunately, the authors did not use the variant with the inverted (ineffective) trap. However, Reaux et al. (1999) used the inverted trap condition with one chimpanzee that was the only one from the group who was successful in the regular tube trap task. Despite her mastery on the basic task, she continued to avoid the inverted trap – in the same manner that Visalberghi and Limongelli’s (1996) capuchin.

Although apes and monkeys, or at least some of them, demonstrated high level of selectivity choosing relevant tools, and even manufactured tools for solving some variants of the tube problem; they have not displayed full measure of understanding the relations between the elements of the tube task.

Recent studies portray birds as being not less adept in passing these tests. Kacelnik et al. (2004) conducted series of experiments with New Caledonian Crows applying a number of different techniques, including the " trap tube" experiment used by Limongelli et al. (1995) and Povinelli (2000) to test primates. After about 100 trials with the apparatus, the crow Betty reached criterion (trap avoided on 8/10 trials or more on three consecutive blocks of ten trials). This performance was comparable with that of apes and capuchins. When the trap was inverted during the testing phase, responding did not return to random, instead, Betty continued to avoid the now irrelevant trap. This is just the same result which had been observed in apes and monkeys. However, in similar experiments with woodpecker finches, one finch, Rosa, easily reached high criteria in solving the basic trap tube problem, including re-shaping tools when solving the H-tool problem (Tebbich and Bshary, 2004). To the credit of primates be it said that other five woodpecker finches in the group displayed the same low understanding of the physical problem as apes and monkeys did. Besides, as Tebbich and Bshary (2004) reason, the conditions of the experiment did not confirm whether Rosa understood the function of the trap.

Do animals apply causal reasoning to tool use tasks and, if yes, how often? Although animals often demonstrate incompetence in solving complex instrumental problems, there are many examples of how individuals effectively use their brains to exploit resources by means of tools (see Reznikova, 2006 for a review). One of the most intriguing questions related to tool using in animals, is whether subjects are able to take invisible forces into account to guide their actions with different elements of a problem. Let us start with two controversial examples, both coming from early studies.

The first example concerns Rafael, a chimpanzee who participated in experiments with putting the fire out and was required to get a cup filled with water and move from one raft floated on a lake to another one balancing on a thin bridge with the filled cup in his hands (see Chapter 17). In a special series of trials Schtodin (1947) and Vatzuro (1948) inspected whether the chimpanzee was able to grasp inter-relations between subsidence of water and a hole in the cup. New Archimed did not come in the person of Rafael. When he received the cup having a hole, Rafael tried to fill it with water 43 times in vain. He did not catch sight of the fact that accidental cut-off the hole by his palm retained water in the cup, so he did not use such a simple way to keep water in. Being shown how to stop a gap with the use of a short stick, Rafael took the stick out and continued to fill with water the cup with the hole. At last, Rafael was presented with a small metal ball and was lucky to stop a gap from the first time with the use of the ball. It was completely accidental. Playing with the ball, Rafael took it in his mouth, then took some water in the mouth and spat water together with the ball out, into the cup; the ball hit the hole and thus stopped the gap. So the problem was solved. The chimpanzee fixed the connection between his actions and stopping the leak. But what may be more amazing; he later repeated all sequence of acts without any changes. Contrary to any sense, he always placed the ball in his mouth, and split it out in the cup. Being presented with an intact cup, Rafael nevertheless dropped the ball into it, and when he received two cups, the intact one and the cup having a hole, he preferred to have the broken cup to perform his ritual with the ball and the hole again and again.

Perhaps the word “ritual” is a key word here. Not only chimpanzees but members of many species persist on the behaviours that brought luck at least once. As it was mentioned in Part I, Tony, the dog of the famous psychologist Morgan, refused all ways to lift the latch of a door his owner ever tried to teach him; instead, Tony repeated the behaviour that brought about such a fortuitous outcome. Since that, experimenters have described a great deal of similar cases. As Ladygina-Koths (1935) reasoned, all chimpanzees are enslaved with their past experiences and hardly improve their methods of work. At the same time in the books of Goodall (1986), McGrew (1992, 2004), DeWaal (2001) and others one can find many examples of how chimpanzees brightly cope with new problems in their natural life and, vice versa, how often they come to a standstill with repetitions of previously obtained solution becoming senseless in changeable situations.

The second example concerning chimpanzee’s reasoning about physical forces came from Firsov’s (1977) experiments with his “island colony of chimps” (see above in this section). In this group, a young male Taras distinguished himself by very high frequency of spontaneous “insightful” solutions of vital and artificial problems. For instance, he used a long stick in order to lift a sunken rope up and thus to pull a boat to him without getting his feet wet. The experiment was the following. Chimpanzees were presented with a food reward placed into an experimental apparatus equipped with a mechanical traction. A door of the apparatus was equipped with a latch spring and opened behind a subject’s back when he pulled a handle that hang down from the opposite side of the apparatus. A rod was too long to reach the door with one hand holding it with another hand or with a foot (Fig. VI-9). A problem turned out to be too complex for chimpanzees and gave rise to confusion. One male, Boy, being faced with this task, swallowed in the sorrow. A jar filled with jam was visible through transparent walls of the apparatus and attracted the chimpanzees. Taras, after several unsuccessful efforts, turned round sharply and walked to the border of a forest. He broke off a long dry stick and returned to the apparatus with it. After several manipulations with the stick and the door Taras was lucky to fix the door and reach the reward through the jammed door. The next year, when meeting with the same problem, Taras made no unnecessary actions and obtained the reward from the very first time using a long stick.

In fact, the distinction between performance and competence in wild users of physical world is a critical one for understanding whether they really take invisible forces into account. Some evidences that came from recent studies support the claim that animals are able to touch the ground of solved problem. Among them, experiments of Kacelnik and colleagues (Kacelnik et al., 2004; Chappell, Kacelnik, 2004) with New Caledonian crows were directly inspired with similar experiments with chimpanzees described in Povinelli’s book. The obtained results have enabled the authors to suggest that crows operate with some general principles rather than with specific associations.

In “rigidity” experiment a crow Betty was pre-exposed to two rake-like objects with different levels of rigidity in a non-functional context, and she was tested in a situation where only one of the tools would serve. The idea was to examine whether, when she needed to pick a tool among a set of objects that were familiar to her, but had not been used before as tools, she would choose according to suitability afforded by the object’s properties. The rakes differed in their wide ends. One had a solid end made of wood, whereas the other had a flexible head made of thin plastic. Betty was allowed to freely manipulate with the tools for several days. The rakes were then placed into a box with a transparent lid that was internally divided into two compartments each containing a food-filled cup placed in front of the head of each rake. The cup could be retrieved from the box by pulling the rake with the rigid, but not the flexible head. The results allow suggesting that Betty had learned the properties of the rakes and used this knowledge when choosing the tool. She was 100% accurate on the first trial on each day (although she seemed to loose motivation quickly). Betty’s success on the first trial of each session contrasts with the results of Povinelli’s chimpanzees where six out of seven subjects performed at chance level throughout the experiment, and the only successful subject reverted to chance when a different experimental design was used. These results can also be considered related to the phenomenon of latent learning, whereas the results of the next, ”innovation” experiment, indicate that a crow is able to innovatively shape tools in anticipation of specific needs.

Researchers had an insight into the level of individual creativity of New Caledonian crows through a serendipitous observation that was made during the course of an experiment on tool selectivity. The question was whether the crows would choose a hooked piece of wire over a straight piece, where the task was to lift a bucket containing food (using the handle) from a vertical tube. On one trial Abel, a male crow, took the suitable (hooked) wire away, leaving Betty with an unsuitable straight wire. After attempting unsuccessfully to extract the bucket with the unsuitable straight wire, Betty spontaneously secured the distal end in a crevice and made a hook by pulling perpendicularly on the proximate end. With the hook thus made, she proceeded to retrieve the food. To explore this phenomenon further, experimenters repeated the task, but offered only the straight wire. Now, Betty bent the piece of wire and used it successfully on virtually every opportunity. She used at least two techniques to bend the wire and she often corrected the shape of the tool several times before attempting to use it. Betty did not have any experience with flexible wire or similarly pliant material prior to this episode nor were the technique she used possible with natural materials (see Fig.VI-10).

These results again contrast with achievements of chimpanzees in similar and even simpler tasks (Povinelli, 2000). For instance, apes did not suspect to straighten a hooked wire in order to pull it in a hole and thus to pin a piece of an apple down and fish it out. This difference is amazing, the more so that the chimpanzees had previous experience of manipulation with both straight and hooked wired. These results do not establish a gap between levels of intelligence of clever birds and silly apes; instead, they make clear that our knowledge about animals’ competence in laws of nature is far from absolute. Some recent data are evidence of relatively high level of understandings of tool’s functional features in primates.

Santos et al. (2003) examined whether primates understand which features are causally most relevant to an artefact’s function in the absence of any direct physical experience with that type of artefact. To this end, experimenters used an expectancy violation paradigm. The logic behind the expectancy violation paradigm is that subjects will look longer at events that violate their expectations about the physical world than at events that are consistent. Researchers habituated captive cotton-top tamarins to an event in which a novel object - a purple L-shaped tool pushed a grape down and onto a lower platform. The subjects then were presented with two trials. In one test trial, tamarins saw a tool of a different colour but a similar shape push the grape down the ramp. In the other test trial, subjects saw a tool of the same colour but a different shape (an I-shape tool) pushing the grape down the ramp. The flat base of the new tool was too short to effectively push the grape. Results showed that subjects looked longer at the new shape test trial that the new colour test trial, suggesting that a change in the tool’s shape was more important to its functioning than a change in the tool’s colour. Experimenters extended their work to free ranging rhesus macaques. They conducted the same expectancy violation experiments and obtained similar results.

How then can we reconcile evidences of animals’ competence with many “bad mistakes” (in Kö hler’s sense) introduced by members of different species in different experiments?

Perhaps, an insight would come from studying ontogenetic development of tool use in animals. Recent studies have revealed surprisingly high level of predisposition for manipulation with tools and – what is more important- for understanding of their functional features in members of species that are not natural tool users. For example, in the absence of experience, infant cotton top tamarins recognise that when an object is used as a tool, changing its colour or texture is functionally irrelevant, whereas changing in shape, size and orientation is functionally relevant (Kralik and Hauser, 2002). It is doubly astonishing, first, because the matter concerns infants, and second, because this species do not naturally use tools. As the authors suggest, tamarins possess innate recognition of functionally relevant features.

Experiments with the main technical bird species, woodpecker finches (Tebbich et al., 2001, 2002) and New Caledonian crows (Kenward et al., 2005), revealed a very high degree of innate predisposition for manipulations with tools together with high selectivity of tools which is, like in previous study on tamarins, based on innate recognition of functionally relevant features. In Part VII we will consider in details problems concerning predisposition of some species to develop complex behaviours including tool using and tool manufacture.

19. NUMERICAL COMPETENCE IN ANIMALS