Some aspects of generalized imitation in subnormal children

Abstract

Explicitly rewarding the imitation of actions demonstrated by a model provides a method for training new behaviour in mentally retarded children. After such training even initially nonimitative children will often copy further novel actions even though such imitations are not rewarded (generalized imitation). A review of the literature on this topic suggested that many aspects of this phenomenon had not been adequately investigated. The research for this thesis therefore studied a number of practical aspects of imitation training in a total of fourteen initially nonimitative subnormal children. Five experiments were performed using a discrete trial paradigm. Experiment 1 compared two methods of training generalized imitation in initially nonimitative children. The first method involved varying the actions demonstrated for imitation from trial to trial (a 'Cumulative' method). In the second method, imitation of each action was trained to criterion performance in isolation (a 'Serial' method). Both methods successfully trained imitation and generalized imitation, but the results suggested that the 'Cumulative' method was the more efficient. Experiment 2 investigated the maintenance of imitation by intermittent reinforcement. The 'imitations' of one group of subjects were reinforced on a variable-ratio schedule and those of the other on a continuous reinforcement schedule. Both reinforcement conditions maintained 'imitations' and 'generalized imitations' at high, stable levels, but the group maintained under variable-ratio reinforcement showed greater resistance to subsequent extinction (the Partial Reinforcement Effect). In Experiment 3, subjects who had been trained to reproduce the actions of a particular model in one setting were tested in different locations and with different models. Changes in both variables resulted in decrements in 'imitative' and 'generalized imitative' response performances. In Experiment 4, a discrimination was established with three subjects by training imitation in the presence of a large ball and non imitation in the presence of a small ball. Imitation was then tested for various other ball sizes. Levels of imitation decreased as the test stimuli increasingly differed in size from the large ball. "Generalized imitations" occurred at about the same level as 'imitations' for each test stimulus. In Experiment 5, all previously trained subjects were tested after an interval of three months with no formal imitation training. Some children then demonstrated decrements in imitative responding but rapidly recovered former levels of performance upon brief refresher retraining. The results suggested that, for clinical purposes, the "imitations" and "generalized imitations" of retarded children may be expected to show some characteristics of a single functional response class. However, some parts of the present results as well as other published data indicate such an account does not completely explain all aspects of the phenomenon.