Discrimination and reversal shift

To train animals on discrimination shift, experimenters first suggest them a simple discrimination between pairs of stimuli. For example, in experiments of Dias et al. (1997) marmosets were trained on a simple visual discrimination consisting of either a pair of blue-filled shapes (" shape" group) or a pair of black lines (" line" group) presented randomly and simultaneously on two test boxes positioned on the far right and left of the test apparatus. A response to either “correct” shape or “correct” line resulted in removal of the transparent screen, allowing the monkey access to a food item hidden within the test box, whereas an incorrect response resulted in the replacement of the opaque screen and no reward. Once the correct reaction was attained on the simple visual discrimination, on the following session the reward contingencies were reversed such that the stimulus that had been negatively correlated with reward was now positively correlated with reward and vice versa.

Mackintosh and Little (1969) examined more complex paradigm involving what is called intradimensional (IDS, for short) and extradimensional (EDS) shifts. They started by training animals to make a discrimination on multidimensional stimuli. From two groups of rats, Group 1 had two problems to solve: discriminating between a blue rectangle and a yellow circle, and discriminating between a blue circle and a yellow rectangle. For the Group 1 discrimination was based on colour: Blue was correct, and yellow was not. In Group 2, the correct discrimination was shape: These animals were rewarded for choosing the rectangle over the circle, regardless of what colour each was. The experimenters then took both groups and moved them to the new set of problems. These involved different shapes than earlier and different colours. There were red “plus” and green triangle in one test (“plus” is rewarded) and green plus and red triangle in another test (again, plus serves as S+). The dimension or relation of shape is still relevant: The animal needs to respond to the “plus” signs and avoid the triangles. For the animals placed on this set of problems, whether something is red or green is irrelevant. For Group 2, the transfer problem is an IDS: The dimension stays the same, though new stimuli are presented. But for Group 1, the transfer problem is an EDS: They need to move from responding to the dimension of colour to responding to the dimension of shape. Reversal shifts are a type of IDS, and non- reversal shifts may be regarded as a type of EDS. The ability to solve IDS problems supports a claim of learning about abstract dimensions such as colour or shape, rather than physical values such as yellow or triangular.

When only one stimulus from many stimuli is associated with reinforcement, then concurrent discrimination is examined. For example, collaborating with the young chimpanzee Ioni, Ladygina-Koths (1935) decided to skip the procedure of training to match pairing stimuli and proceeded with testing the concurrent discrimination. Ioni was presented with groups of 8 figures among which he had to select one that matched the sample. Ioni was able to discriminate between 13 plane geometric figures and 10 volumetric figures (see Fig. II-3).

Giebel (1958) reported that a horse learned to discriminate the correct stimulus in 20 pairs of visual patterns concurrently, which ranks it high among other animals that have been tested. Thomas (1986) complied data on concurrent discrimination learning for various fish, reptiles, birds and mammals, including mice, rats, zebras, donkeys, horses and elephants. Of these species tested, only the elephant was able to successfully complete as much concurrent discrimination as the horse.

Chimpanzees championed concurrent discrimination in a study of Farrah (1976) on what he called “Picture memory” in apes. In these matching-to-sample experiments each discrete trial began with one of 24 possible visual samples, varying in form and colour and being presented on a black-lit key. Immediately after the sample was terminated one stimulus was presented on each of four black-lit keys arrange side-by-side in a line horizontally below the sample key. The chimpanzees were rewarded if they pushed the key on which the displayed stimulus was the same as what had appeared on the sample key. Importantly although the four stimuli varied across the 24 possible matching problems, the same set of three incorrect stimuli appeared with each correct “match” on any given problem. The terminal performance levels of all three chimpanzees exceeded 90% of correct matching responses suggesting perhaps that they had learned a generalized matching concept. It is interesting to note that basing on this experimental paradigm, experimenters later have revealed extraordinary abilities of chimpanzees to remember faces of conspecifics and to judge about their kinship (see details in Part VIII).