Abstraction in animals

Numerous experiments have demonstrated that animals have the ability to abstract. They are capable of recognising the essential features common to different phenomena and thus by abstracting certain relevant characteristics – arrive at concepts.

In Ré vé sz’s (1924) pioneering study chicks learnt to peck at the smaller one from two geometric stimuli such as a circle, a square, or a triangle. Chicks responded to sizes of stimuli irrespective of their geometric shape. Similar experiments were conducted by Protopopov (1950), Hilchenko (1950) and Markova (1962) with dogs, monkeys, chimpanzees and human infants. For example, in one of these experiments baboons had to choose one of two boxes with figures attached to lids. A reward was always placed into that box to which a smaller square was attached. The boxes were situated in different places in order to avoid using landmarks by a subject to orient. When the animal managed to choose the smaller square, it was tested with figures that were never seen in experiments. Monkeys insisted on choosing the smallest figures irrelevant to their shape and colour. In special series of tests the subjects were shown with volume figures which at the second stage were replaced by their plane projections. In some experiments plane figures were consequently replaced by their outlined and at last dotted lined images. In general, monkeys managed with all tasks but the less figures style resembles the initial one, the worse results were achieved.

Experiments conducted by Rensch and Dü cker (1959) with a civet cat revealed a considerable ability to sift various sensory impressions for certain characteristics essential to the significance of the whole (generalising abstraction). The animal was trained to distinguish between two parallel semicircles (meaning " food") and two straight lines (meaning " punishment"). It was then presented with increasingly complicated patterns in which these two recurred in modified form. The cat showed that it could eventually distinguish between the concepts " bent" and " straight." It also, in similar fashion, formed the twin concepts " equal" and " unequal."

Many modern experiments on examining abstraction and concept formation are derived from the problem used by Neet (1933), Gellerman (1933) and Harlow (1949) to show that human infants, chimpanzees, rhesus monkeys (but probably not rats) were able to form abstracts concepts, such as a concept of triangularity (Fig. V-3). The same experimental paradigm has been repeatedly applied for wide variety of species.

For example, in experiments of Sappington and Goldman (1994) with Arabian horses, the stimuli were grouped into in 6 problems in such a way as to bring animals to shaping a concept of triangularity. Problems 1 through 4 involved simple discrimination learning with only two stimulus patterns: correct and incorrect. Problems 5 and 6 each had three different stimulus patterns that were variations of triangles which were always correct, and three stimulus patterns that were non-triangular shapes, which were always incorrect. On each trial, one card from the set of correct stimuli was randomly paired with one card from the set of incorrect stimuli. These two problems thus involved the concept of triangularity. From them, the Problem 6 was a stricter test of the concept of triangularity, as it involved three triangles never before seen by the subject paired with three new, non-triangular shapes. The results clearly demonstrated that at least some horses were able to form the concept of triangularity. One of four subjects completed both problems 5 and 6 at a criterion level of 85% for two consecutive sessions, while another was performing at more than 80% correct on Problem 5, the first test on the triangle concept. Although both subjects did not reach the point at which they responded correctly to a novel triangle the first time it was presented, which would be the strictest test for concept formation, the fact that they could learn relatively quick that one of several triangles was correct regardless of what it was paired with suggests an ability to respond to a common characteristic of diverse stimuli (Sappington and Goldman, 1994).

In experiments of Delius and Habers (1978) pigeons had to conceptualise symmetry. Birds of two groups were given pairs of stimuli that differed by several minor attributes. For one group only symmetric stimuli were rewarded, irrespective of other characteristics such as shape, ornament and colour, while for the second group the asymmetric stimuli were reinforced. When a level of 80% of positive responses was reached with one pair of stimuli, experimenters suggested to birds another pair of stimuli that consisted of unfamiliar items, again differing by the characteristic of symmetry for one group and asymmetry for another group. When the pigeons grasped regularity within a context of a given problem, they accurately chose only symmetric (or asymmetric in another group) stimuli irrespective of other attributes. In testing experiments baboons were demonstrated to be able to categorise spatial relations such as “above” and “below” categories, as well as “short” and “long” distances (DJpy et al., 1999).

Feat of abstraction was performed by honey bees in experiments of Mazokhin-Porshnyakov (1969, 1989) in which insects conceptualised paired/non- paired objects and solved some other complex tasks. The bees started from simple tasks in which they had to ignore irrelevant experimenter-depended (“conventional”) feature of stimuli such as geometric shape in favour of relevant ones such as sizes. The technique of experiments was as follow. A feeder (rotating table) was situated in several meters from the hive. An experimenter placed a trough (watch crystal) with a drop of syrup to starting foragers at a bee-entrance. When one of interested bees began to eat, it was labelled by paint and then transported to the feeder just on the trough. There was a set of pictures covered by a glass on the feeding table. Rotation prevented the bees from the use of landscape landmarks for determining location of the food. At the training stage of experiments the bee was presented with the trough filled with a drop of syrup placed on the rewarded figure while troughs filled with water were placed on other, non-rewarded figures (see Fig. V-4). In a relatively simple task the same stimuli were used as in previously described experiments with vertebrate animals. Bees had to choose the smallest figure irrespective of its shape and colour. After several visits the bees memorised the rewarded figures. In order to test this, a series of examination was performed during which the bee met all troughs filled with water in order to avoid possible use of sugar smell. The main part of the experiment – like of many procedures described before – was the use of novel stimuli which previously had not been seen by the subject. If the bee was continuing to choose the smallest figures from the new sets of stimuli, the experimenter concluded that the insect had grasped the concept of the task and marked out the main (experimenter-dependent) property (i.e. size) of the stimuli independently of accessory properties (such as shape and colour).

The bees also managed more complex tasks such as conceptualising “novelty of coloration” or “dichromatic/monochromatic”. Here are two examples of the most complex problems solved by the most gifted bee family members. Both problems are concerned with choosing one out of a pair of compound figures, namely, chains of small pictorial elements. The conventional feature of the stimulus in one task was location of a single black element at the end of the chain. The pictorial elements varied in different tasks while their numbers varied in different chains. The second problem was to choose a chain consisting of paired elements versus non-paired. After hundred visits to the feeder the foraging bee learned that the food was placed on the picture with the chain consisting of paired elements while the trough which was placed on the picture with non paired elements was always empty. After several error-free solutions the bee was presented with chains consisting of new elements. Tens of choosing of paired chains consisting of different novel stimuli enabled experimenters to conclude that honey bees can master “paired/non paired” conceptualisation.