Animals’ natural concepts: Classification at different levels of abstraction

As we have seen in previous chapters, animals of different species have shown some evidence of the acquisition of “natural concept formation”, that is, they can generate concepts which represent real objects and events. The ability to create concepts is closely related to the capacity for abstraction and classification, and demands generalisation of knowledge about the whole category as opposed to memorisation of specific exemplars.

When the Tahitians first saw a horse, introduced to the island by de Bougainville's French sailors, they immediately classified it in terms of the mammal they knew that most resembled it... a pig. It was obviously closer to the horse than the other two mammals of which they had any experience: the dog and the Polynesian rat. Very sensibly, they referred to the horse as a " man-carrying pig". This was a sophisticated attempt to understand the strange beast in light of their experience of similar beasts. There is no reason to suppose that animals do not do likewise (Goldsmith, 1998).

It is intuitively clear that the acquisition of concept formation in animals should be limited. Abstract concepts are defined by logical rules rather than perceptual features. Hunan infants are able to determine category membership at an early age. When they address to a cat, or a caw, or a dog with a word “bow-bow” they apparently mind common features of these animals. Parents will later suggest an idea to a child about that neither a cat nor a cow does not say bow-bow and that the cat differs from the two others. Nevertheless these pedagogical efforts are only aimed to teach a child to select classes of objects more precisely. The tendency to classify objects is innate to human beings. The ability to select character features and group animals, plants, and many other things as classes of objects makes the process of cognition more parsimonious and effectively improves our adaptiveness.

It is of no surprise that tests of the ability to make use of categories are widely applied for diagnosing mental diseases as well as local damages of the brain. Age psychologists also routinely use these tests. For example, psychologists assume that there is a possible developmental shift from similarity-based to theory-based categories in 2-year old children. At this age, children are likely to realise that category membership can be a better basis than perceptual appearances for drawing certain inferences. Children can detect the category membership of some novel pictures on the basis of subtle perceptual cues. Being presented with pictures, two-year olds recognise a lamb-like dog as a dog on the basis of its nose, feet, and tail, despite its overall colour and fur (Anglin, 1977; Gelman, Coley, 1990). An interesting situation has been described by Makarova (1990), a teacher in the art school for children. She presented a 2.5 year-old boy with a box filled with different buttons. The boy began to play with buttons; spontaneously, he began to sort them by sizes and then by colour but very soon he confused and raked all buttons together. To the teacher’s surprise, the boy then quickly found the way to sort buttons: he put buttons with two holes to the one side, and buttons with four holes to the other side.

In the early twenties of the 20- th Century K`hler and Ladygina-Koths revealed that chimpanzees make out things on pictures quite well, and this created the necessary prerequisites for studying nonhuman capacity for classification with the use of pictorial stimuli. Kellogs who reared the chimpanzee Gua together with their little son Donald demonstrated that the chimpanzee could recognise pictures and point at them by a request of her “parents” at more early age than the child could (Kellog and Kellog, 1933). For example, 17-months old Donald and 15-months old Gua were showed with paintings of a cup, a dog, a boot and a house. Donald pointed the dog in response to a request “show bow-bow” but he did not know the names of the rest three things. Gua pointed at the dog, at the cup, and at the shoe. Like Ladygina’s (1935) pupil Ioni, Gua found difficulty in recognising things whose sizes were dramatically changed on pictures in comparison with the originals (such as the pictured house in comparison with the realistic one). It is interesting to note that during his leisure hours Ioni sorted gaily coloured pieces of silk by colour, for example, he grouped yellow pieces and red ones separately.

Recent studies examined concept discrimination learning in animals at various levels of abstraction. As concepts become more abstract, exemplars within the category share fewer features in common. At the concrete level exemplars within a concrete category may share more readily perceivable attributes such as size, shape, and colour. More abstract categories contain members that may be difficult to discriminate on the basis of sensory properties (for example, all members of the category “animals” share the ability to breath and reproduce).

Roberts and Mazmanian (1988) examined concept discrimination learning in pigeons, squirrel monkeys and humans. At the most concrete level, subjects were asked to select photos of king-fisher birds from photographs of other bird species. All king-fisher photos shared readily distinguishable feature such as size, colour, and shape that were not shared by members of the non-reinforced category. All three species tested easily learned this discrimination. At an intermediate level of abstraction, subjects were asked to select photos of birds from various other animal species. The bird members of the reinforced set shared several features but there was more variance between them. At the most abstract level, subjects were asked to select animal photos from those of non-animals. Pigeons and monkeys learned the most abstract category more easily than the intermediate discrimination. There was some evidence that intermediate discrimination was easier to learn if non-animal photos comprised the negative category. It is possible, however, that pigeons and monkeys were not necessarily operating on the basis of conceptual processing but may have been relying on the perceptual features of the stimuli - like, as Huber (2002) suggests, pigeons in Herrnstein and Loveland (1964) study.

Vonk and MacDonald (2002) conducted experiments with a young gorilla Zuri basing on the analogous tasks with that of Roberts and Mazmanian (1988) and intending to represent the same three levels of abstraction. Zuri was presented with photographs on a touch-screen computer. The experimenters were trying to eliminate the possibility that their subject was using irrelevant aspects of the photos to make the discriminations, both by varying as many aspects of the pictures belonging to the same categories as possible, and by detail analysis of Zuri’s errors.

Four discrimination tasks were considered to be concrete discriminations with the use of many pictures: (1) gorillas or orang-utans versus humans; (2) orang-utans versus other primates; (3) gorillas versus other primates; (4) orang-utan colour test. Zuri was asked to select orang-utan photos by attending to a single feature: their reddish colour. In order to illustrate the fullness of classification, the second task could be described here: orang-utans versus other primates: three sets of orang-utans photos versus other primates were used, for a total of 60 photographs. The orang-utan photos included subjects from various stages of development and of both genders (Zuri has never seen an orang-utan before). The photos of other primates included a wide range of primate species including gorillas. Zuri showed a high degree of transfer to novel photos of both gorillas and orang-utans when these photos were contrasted with photos of humans. She was generally not distracted by irrelevant features of the stimuli, such as orientation, size, or gender. In order to avoid the possibility that Zuri discriminated the two sets of exemplars by attending to specific features, rather than by attending to a general concept of “gorilla” or “orang-utan” (for instance, she may have learned to “choose a black face” or “avoid a white face”), the experimenters examined her performance on photos that might be expected to be difficult had she been using simple cues. For instance, the transfer set included a photo of an albino gorilla with a pink face. At this stage of concrete discrimination as well as at two other stages (intermediate discrimination such as primates versus non-primates including reptiles, insects and so on; and abstract discrimination such as animals versus non-animals), the experimenters were trying to distinguish between the use of a perceptual and a conceptual strategy. The young gorilla seemed to have the most difficulty with the intermediate discrimination. The authors acknowledge that the ease with which Zuri learned the most abstract discrimination was partially due to a generic “learning to learn” phenomenon, because this was the last discrimination tested. In sum, acquisition and transfer involving abstract stimulus sets suggest a conceptual basis for gorilla’s categorisation.

Savage-Rumbaugh et al. (1080) demonstrated that chimpanzees are able to sort real objects or pictures into two trays according to their belonging to “food” or “tool” categories. The chimpanzees were able to perform this task with great accuracy, even with objects that had not been used during training. In Part VIII we will see how language trained primates perform categorisation on the basis of these language-based skills.