Студопедия

Главная страница Случайная страница

КАТЕГОРИИ:

АвтомобилиАстрономияБиологияГеографияДом и садДругие языкиДругоеИнформатикаИсторияКультураЛитератураЛогикаМатематикаМедицинаМеталлургияМеханикаОбразованиеОхрана трудаПедагогикаПолитикаПравоПсихологияРелигияРиторикаСоциологияСпортСтроительствоТехнологияТуризмФизикаФилософияФинансыХимияЧерчениеЭкологияЭкономикаЭлектроника






Conditional discrimination and rule learning






 

In a wide sense, conditional discrimination is a discrimination in which the reinforcement of responding to a stimulus depends on or is conditional upon other stimuli. Conditional discrimination is one of relevant methods for empirically determining whether ordered pairs of events or interrelated stimuli make up an equivalence class. Using matching-to-sample (MTS) procedure, relations are established with explicit reinforcement contingencies in stimulus pairs. The function of one stimulus depends on the nature of another stimulus. Thus, the S+ would be S-1 and not S-2 when both are presented with one contextual stimulus S-x, but the S+ would be S-2 and not S-1 when both are presented with another contextual stimulus S-y. This technique involves different sets of tests from simple to very complex ones that have been applied successfully for studying intelligence in animals as well as for diagnostic of normal aging and mental disorders in humans.

The use of complex tests on conditional discrimination in animals demands infinite patience from experimenters. Nissen (1934) managed to teach a chimpanzee for solving the problem as follows: choose the smallest from two squares if they differ by only one characteristic (colour, design, and border) but choose the biggest one if they differ by two characteristics. The experimenter did not assert, however, that the ape defines a problem using the same concepts as humans. However, it is remarkable that Frank, the chimp, reached a level that exceeded 70% of correct answers after 1700 trials. Later Warren (1965) applied matching-to-sample procedure in order to train a chimpanzee to choose a stimulus that differs from a sample by shape if stimuli were presented on white background and by colour if stimuli were presented on black background.

Like many compound tests, tasks of conditional discrimination should involve multiple controls, and results not always reach the optimistic point. Let us consider classical experiments of Lashley (1938) with three rats as subjects. The apparatus was a two- alternative jumping stand. The first task for animals was to choose an upright triangle rather than an inverted triangle when both were presented on a black background. In the second problem, rats learned to choose an inverted triangle rather than an upturned triangle when both were presented on a striped background. In subsequent mixed trials, rats reliably (> 90%) chose the appropriate triangle based on the nature of the background. What are the rats learning in this situation? Should we say they are learning and applying a generalised " if-then" rule? By asking if the rat is learning a generalised " if- then" rule, we are asking whether the rat is learning " if the background is black, then A and not B are correct, and if the background is striped, then B and not A are correct." On the other hand, are they learning to respond to a specific combination of stimulus plus background? Do they learn simply to pick the compound of upright triangle/black background and the compound of inverted triangle/striped background, as independent problems? This question can be answered by training rats in a series of conditional discrimination problems. A pair of stimuli should be presented, first on a black background. One stimulus would be correct and the other incorrect and the speed with which they acquired the discrimination would be measured. Then the same pair of stimuli would be presented on a striped background, now with the formerly incorrect stimulus being correct, and vice versa. If the rat was learning a generalised if-then rule, one would predict that after the rat had learned some number of these sorts of problems: whenever it was presented with a new problem it would learn at a faster rate when the background was striped than when it was black. For example, if the rat was learning an " if-then" rule, one would predict the rat would choose the correct alternative on the very first trial of the second exposure (i.e., with the striped background) with a very high probability. In contrast, if the rat was learning to respond to a specific combination/compound of stimulus plus background, one would predict that problems on the second exposure would be learned at the same rate as those on the first exposure.

In further tests, Lashley trained rats with 4 new problems (large circle versus small circle; cross versus X; circle versus vertical bar; star versus square). One stimulus was correct when both were presented on the same black background. The other stimulus was correct when both were presented on the same background with. The question was how fast the rats learn the second exposure to the problem, when the stimuli are presented on the striped background. The rats learned the discrimination with the striped background at more or less the same rate as they had learned the original discrimination with the black background. There was no evidence of a generalised tendency to reverse and choose the alternate stimulus on the first trial with the striped background. So the implication could be done that rats learn to respond to a specific combination of stimuli (e.g., a compound of upright triangle/black background; or a compound of inverted triangle/striped background), rather than a generalised " if-then" rule.

Development of laboratory studies on higher-order conditional discrimination provides evidence of behavioural flexibility in non-human animals, together with limitations in some species. Members of several species such as rats (Preston et al., 1986), pigeons (Santi, 1978; Edwards et al., 1982), and monkeys (Fujita, 1983) appeared to take multiple-rule strategies. For example, in experiments of Nevin and Liebold (1966), pigeons could solve the multiple discrimination task following four rules: “if red is presented in the lit chamber, then peck red”, “ if green is presented in the lit chamber, then peck green”, “if red is presented in the dark chamber, then peck green”, “if green is presented in the dark chamber, then peck red”.

To further explore the possibility of conditional rule learning in animals, Nakajima (1997, 2001) devised another way of testing. Pigeons were tested in a three-key operant chamber, where a correct response (a left- or right- side key peck) depended on three preceding events. Pecking the left or the right key was followed by a food reward according to the colour (amber or blue) of the centre key, presence or absence of flashing of the three keys with green colour prior to the centre colour presentation, and the house-light illumination condition (dark or light) of the chamber.

The problem posed was: can pigeons learn a hierarchical conditional rule or do they solve the task using other strategies?

With eight pigeons, seven out of eight possible types of trials were trained in a step-by-step fashion, and then the remaining trial type was tested. All birds responded correctly to the training test types, but their poor performance in the untrained test type indicates that they solved the training task by rote learning of the individual trial types. Their test performance enabled Nakajima (2001) to suggest that the birds probably had learned to configure each sequence of events into a unique stimulus to respond properly during the training, and their test performance reflected generalisation from that learning. Another possibility is that the birds had learned seven rules of event sequences (multiple rule strategy), and in testing they resorted to the rule that was most similar to the test type. In sum, pigeons did not derive an adequate response to a novel trial type from the familiar trial types by completing the hierarchical structure. It remains unclear whether this failure reflects the limitation of the birds’ cognitive abilities or an unfavourable procedure to test such abilities.

 

15. CATEGORISATION, ABSTRACTION, AND CONCEPT FORMATION: ARE ANIMALS LOGICAL?

 

In defence of animal reasoning, McGonigle and Chalmers (1992) exclaim in the title of their paper “Monkeys are rational! ” This echoes an earlier paper which, in its title, asks “Are monkeys logical? ” (McGonigle and Chalmers, 1977). Then “…The Paleological Monkey and the Analogical Ape” appeared in the title of Thomson and Oden (2000), who in turn are referring to “paleo-logicans” in the sense enunciated by von Domarus (1944) in his interpretation of reasoning by schizophrenics: Whereas a normal person accepts identity upon the basis of identical subjects (i.e. conceptual equivalence), a paleo-logican accepts identity basing upon identical predicates (i.e. shared features). Of course, Thomson and Oden (2000) do not claim that monkeys are schizophrenics; they do claim, however, that monkeys discriminate categorical equivalence classes on the basis of perceptual identity of features, or some combinations, possibly configurable.

In fact, there is a wide range of ideas concerning concept formation in non-human animals. In terms of operant learning, the formation of discrimination is based on a class of stimuli such that an organism generalises among all stimuli within the class but discriminates them from those in other classes. Such classes play much the same role in analysis of discriminative stimuli as operants do in analysis of response classes. Following a pioneering experiment of Herrnstein and Loveland (1964), much work was concentrated on experiments on discrimination between sets of stimuli. The stimulus sets are usually defined in terms of human concepts, e.g. person vs. non-person, fish vs. non-fish, or artificial concepts defined by specified multiple features. Most discussion is centred on the question of whether animals need to possess concepts in order to perform concept discriminations, and what does it mean for an animal to " possess a concept". It is important to note that in order to dismiss simple stimulus generalisation (see Part II), it must be demonstrated that stimuli to be classified in the same category certainly differ from one another (Vauclair, 2002).

Cognitive ethologists concentrate on what Thomson and Oden (2000) call conceptually mediated behaviour that permits animals to adjust their behaviour to novel objects and events by virtue of a membership in an already familiar class. To infer that an animal has a concept one must provide evidence that it applies the same judgement in the form of an explicit response rule or cognitive operation to objects or events that are perceived to be common members of the same physical or relational class. The measure of conceptual categorisation is based on experimental procedures that match different levels of complexity in animal conceptually mediated behaviour.


Поделиться с друзьями:

mylektsii.su - Мои Лекции - 2015-2024 год. (0.009 сек.)Все материалы представленные на сайте исключительно с целью ознакомления читателями и не преследуют коммерческих целей или нарушение авторских прав Пожаловаться на материал