What is Insight?

Let us start this Chapter from a quotation from Wolfgang K`hler’s Address of the President at the sixty-seventh Annual Convention of the American Psychological Association (1959):

What is insight? In its strict sense, the term refers to the fact that, when we are aware of a relation, of any relation, this relation is not experienced as a fact by itself, but rather as something that follows from the characteristics of the objects under consideration. Now, when primates try to solve a problem, their behavior often shows that they are aware of a certain important relation. But when they now make use of this " insight, " and thus solve their problem, should this achievement be called a solution by insight? No - it is by no means clear that it was also insight which made that particular relation emerge. In a given situation, we or a monkey may become aware of a great many relations. If, at a certain moment, we or a monkey attend to the right one, this may happen for several reasons, some entirely unrelated to insight. Consequently, it is misleading to call the whole process a " solution by insight."

We see then that even the “father” of the theory of insight in animals, in his late publications, considered a notion of insight rather fuzzy. As it was described in the Part I, K`hler (1918, 1925) first revealed experimental evidences of insight learning in chimpanzees basing on ideas of Gestalt psychology. His series of experiments showed chimpanzees as using planning and foresight, that is, cognitive reasoning to solve a problem. K`hler devised an arrangement in which all of the elements necessary for the solution of the problem were in a full view of the animal. In 1910-th K`hler conducted thousands of experiments with chimpanzees and summarised them in “Mentality of Apes” (K`hler, 1925). His observations have been the subject of controversy ever since. The period of quiescence that sometimes preceded the solution by animals, its sudden onset, and its smooth, continuous emergence were proffered as evidence that (1) contrary to suggestions of learning theorists of the day, problem solving was not necessarily a trial-and-error process, and (2) constructs such as “insight” were necessary for an adequate account.

One of the most referred examples of problem solving by insight concerns Sultan, a chimpanzee, whom K`hler regarded as the brightest of a number of apes he worked with (and with whom we have already met in Part I). Sultan sat in his cage, in which there was also a short stick. Outside the cage there was a longer stick, which was beyond Sultan’s reach, and even further away was a reward of fruit. Sultan first tried to reach the fruit with the smaller of the sticks. Not succeeding, he tried a piece of wire that projected from the netting in his cage, but that, too, was in vain. Then he gazed about him and after a long pause suddenly picked up the short stick once more, came to the bars directly opposite to the long stick, scratched it towards him with the auxiliary, seized it and went with it to the point opposite the objective which he secured. From the moment that his eyes fell upon the long stick, his procedure formed one consecutive whole.

Modern researchers in the field of animal learning and cognition find it difficult to interpret the majority of K`hler’s experiments because his apes may be simultaneously engaged in the same problem so that it is virtually impossible to understand the problem-solving abilities of any individual. Furthermore, all of his subjects had played with boxes and sticks prior to the experimental trials in which they then reached bananas that were hung from the ceiling of their enclosure by pushing a box under the bananas and them climbing onto the box, or by using a stick to shake the banana down. The absence of trial- and- error responding may thus have been due to the prior experience of the animals. As we have seen from the quotation above, K`hler (1959) himself was frustrated about understanding the nature of subject’s awareness about relations between things that he called insight. In particular, K`hler distinguished two kinds of mistakes related to problem solving in chimpanzees, that is, “good mistakes” and “bad mistakes”. If the chimpanzee tried to affix a box to a wall in order to reach a banana from the top of the wall, it is a “good mistake” that can serve as evidence of the animal’s understanding about dimensions; the chimpanzee is simply unaware of box’s properties. But chimpanzees in K`hler’s experiments also demonstrated a lot of “bad mistakes”. For instance, in one setup, chimpanzees could only get bananas by removing a box. Here was something, Kö hler expected, that even his awkward chimpanzees could “do at once.” And yet, to his astonishment, the chimpanzees had difficulties in solving such problems; they often drew into the situation the strangest and most distant tools, and adopted the most peculiar methods, rather than removing a simple obstacle which could be displaced with perfect ease.

Later insightful behaviour has been defined as “the sudden production of a new adaptive response not arrived at by trial behaviour” and “the solution of a problem by the sudden adaptive reorganisation of experience” (Thorpe, 1963). There are two key words here, “experience” and “sudden”.

In problem solving, the experience need not be directly associated with the problem at hand. Sometimes, when presented with a new problem, an animal will solve the problem on the first attempt because of experience in dealing with the component parts of the problem, although they were never met together in just such a way before.

The important role of past experience in problem solving is illustrated by the experiments concerning insight in pigeons (Epstain et al., 1984). In some sense, this can be called “pseudo-insight”. Researchers replicated with pigeons a classic “hanging banana problem” which K`hler suggested to chimpanzees, and copied the design of K`hler’s experiment. Pigeons that had acquired relevant skills solved the problem in a remarkably chimpanzee-like fashion. They were trained with the use of instrumental conditional technique to perform two behaviours: to push a cardboard box toward a target (a green spot placed at various locations on the floor) and to climb onto a box and peck at a miniature model of a banana which was suspended overhead. Banana pecking was reinforced with the food. These two behaviours were always trained in separate training sessions. Once they had learned, the experimenters confronted the pigeon with a situation it had not encountered during training. The box was placed in one part of the chamber and the banana was in another. At first, the bird appeared to be confused. It stretched toward the banana, turned back and forth from the banana to the box, and so on. Then, rather suddenly, it began to push the box toward the banana, sighting the banana and readjusting the path of the box as it pushed. Finally, it stopped pushing when the box was near the banana, climbed onto the box, and pecked the banana. In the subsequent experiments, Epstain (1987) taught a pigeon to (a) peck at a model of a banana, (b) climb onto a box, (c) open a door, and (d) push a box toward a target. During testing, the pigeon was confronted with a banana hanging over its head and a box behind a door. The bird successfully combined all four behaviours to solve a problem: it opened the door, pushed the box out and moved it under the banana, climbed the box and pecked at the banana.

The possible contributions of different experiences were determined by varying the training histories of different birds. The experimenters concluded that successful performance was dependent upon the animals learning the basic skills in separate contexts that must go together to solve the problem. For example, it was demonstrated that only if the birds have learned to push a box toward a goal will they move it under a model banana hanging from the ceiling of the cage. Epstain argued that it is not enough to teach a bird to push a box; it must have learned to push it toward a goal. Presumably, the chimpanzees’ experience with moving boxes around and climbing on them was necessary for them to solve the hanging banana problem.

Recently students of animal intelligence argue that describing such behaviour as “insightful” does not help us to understand it. If we do not know what behaviours an animal has already learned, a novel and complex sequence of behaviours that solves a problem seems to come from nowhere (Skinner, 1985). Rather than attributing the successful pigeon’s performance to insight, Epstain offered a moment-to-moment explanation of their actions, based on principles of behaviourism (see Parts I and II). It is of no surprise because Epstain had started as a graduate student of Skinner. As Skinner himself noted (1987), “…We collaborated on a variety of research, including a 3-year project which we eventually called " Columban [Pigeon] Simulation. Through careful construction of complex contingencies of reinforcement, we were able to get pigeons to exhibit behavior said to show " symbolic communication, " " spontaneous use of memoranda, " self-concept, " " insight, " and other so-called cognitive or creative processes”.

Pigeon Simulation demonstrated that animals can effectively combine previously learned skills in order to reach a goal even if these skills were never trained together. But this project had little in common with K`hler’s main ideas that were based on principles of Gestalt psychology. K`hler argued that “sudden” (the second key word in a definition of insight) solution of a problem by animals is based on their understanding of a whole situation. He believed that chimpanzees grasp general principles or relationships and “saw” the solution before carrying out an action. Perceptual reorganization of the elements that constitute a problem situation allows animals (and humans) to see at one moment all the parts in relation to each other, forming a meaningful whole. Unlike Epstain’s simulation, K`hler’s scenario often demonstrated animals as being able to find unexpected techniques to reach a goal with the use of unexpected accessory items.

For example, Sultan learned to use such improvised means as people in order to solve hanging banana problem. He seized a trainer by the hand and pushed him in order to jump on his shoulders and reach the prize. Sultan was very angry at the trainer when he did not grant an ape’s wish. Once Sultan climbed on the trainer in order to use a man as a step stool but the trainer stooped so that the chimpanzee was not able to grasp the banana. Then Sultan jumped off, caught hold of trainer’s belt and tried to lift the man up, with groaning. It soon became a habit with chimpanzees to use each other as step ladders so they were crowded under bananas suspended over chimpanzees’ heads. Chica, the most athletic chimpanzee in the group, preferred to stand the long bamboo stick upright and quickly climb up the 7 meters and grabbed the bananas before the stick fell over.

One task for the apes was to reach the hanging banana in a situation when only one box was available filled with four heavy stones. Among nine chimps only Sultan was clever enough to pull out all stones, one by one. He came to this solution after many attempts. First, he pulls out only one stone and tried to push the box with all his might but failed. Two stones were enough to pull out in order to push the box with an effort. After four repetitions, Sultan came to the most effective way of solving the problem and since that he always pulled all stones out of the box. Other apes did not grasp the relationship between stones placed into the box and difficulties they met when trying to move it forward. For instance, Grande took one stone out but not for making the box movable but to use the stone as a step stool. She then estimated a situation correctly and did not try to climb onto the stone because it was too low. Nevertheless, like other apes, she did not guess how to act with stones in order to make the box light.

Indeed, it is reasonable criticism that many of “sudden” solutions of the chimpanzees in these experiments were achieved rather slowly in reality. Critics argue that only after years of practice and repetition did the apes learn the behaviour needed to solve the problem. Another criticism is that chimps have been observed to commonly swing sticks at nothing. They have also been seen enjoying themselves by climbing up on boxes and stacking them without a prize in view, so that the prize didn't trigger the behaviour. Moreover, many factors were not estimated properly. In particular, some unaccounted factors could fix sequences of behavioural acts of animals in unknown way thus made them “ritualised”. Perhaps this was just the case in the experiments of K`hler’s opponent, Pavlov, who replicated K`hler’s experiments in order to proof that chimpanzees’ behaviour is governed by contingencies of sequences of stimuli and reactions rather than by “insightful” behaviour. At the same time, results of experiments of Pavlov and his students can set K`hler’s results off and thus help us to understand insightful behaviour more clear.

Pavlov purchased two chimpanzees, Rafael and Rose. He performed, from 1933 to 1936, a number of experiments, including a replication of K`hler's building experiment. Confirming K`hler's findings, Pavlov explained the problem-solving process in terms of unconditional reflexes and the establishment, by Pavlovian conditioning and the Thorndikian method of trial and error, of temporary neural connections identical, on the psychological level, to associations. According to Pavlov, insight is achieved progressively - as the result of the organism's problem-solving behaviour. This statement contradicts K`hler's theory of a sudden subjective reorganisation of the environmental situation.

The most complex scenario was performed by Pavlov’s student (Vatzuro, 1941). Rafael, the chimpanzee, was forced to tie together many complex behavioural patterns which he had already learned in separate training sessions. The chimpanzee was requested to (1) stack several boxes on top of one another and get a hung stick; (2) to open a box with the use of the obtained stick and thus to get a rope from inside the box; (3) to open the second box with the use of the obtained rope as a mechanical traction and take a cup there; (4) to fill the cup with water from a large mess-dish; (5) to put on a fire out in an apparatus and get the prize. This scenario was implemented on two rafts floating in a lake. In order to fill the cup Rafael had to get over the second raft with the use of a thin long stick as a bridge and thus to reach a cistern with water there. Then he returned to the first raft balancing on the thin bridge with the cup filled with water. After putting flame out, the happy chimpanzee could take his prize from the apparatus. Why wasn’t Rafael clever enough to bend over the raft surface and scoop some water with a cup? It is likely that he learned the complex sequence of feats as a ritual. It is also possible that Rafael learned to configure a sequence of events into a unique stimulus, like pigeons in the experiment of Nakajima (2001) which was described in Chapter 14.

Later Voronin and Firsov (1967) obtained rather different results in their experiments with Lada, a chimpanzee. Confronting with the same situation, Lada easily came to a natural solution to scoop water from the lake. The fact is that experimenters did not dictate Lana an obligate sequence of behaviours connected with the cup and the cistern. During the training sessions, Lada was free to get water from different containers such as a jar, an aquarium, a bucket, and a pan. She thus acted with water itself rather than with a concrete container. Like K`hler’s apes, Lada was likely to achieve a goal through understanding the relations between things in a context of a situation.

So what is insight in animals? Basing on recent data on animals’ capacities for analogical reasoning, we can attribute insightful behaviour to animals’ ability for fast integration of behaviours gained from their past experience and effectively applying this experience for solving a problem as a meaningful whole, in a context of a situation. There is a distinction between insight and insight learning. Insight is a part of the mental process that relates to understanding relationships; insight learning is demonstrated in the behavioural solution of the problem (Klopfer and Hailman, 1967). Insight could be considered a part of rule extraction. To “attain” insight, another part of rule extraction is necessary to be obtained by a subject, namely, the ability to “learn to learn”.