Form discrimination in young children and the concept of similarity

A brief account is given of differentiation theory as it relates to form discrimination studies, and relevant experimental work is reviewed. Some preliminary comments on the concept of similarity, arising out of this and pertaining to the extent to which children can be said to possess an adequate grasp of what is implied by "the same as," are made as an introduction to the first experiment. This examines the performance of a group of children twice, at mean ages 3-8 and 4-9, in matching-from-sample discrimination tasks with stimulus material varying in orientation only, and in form and orient¬ ation. Considerable improvement in terms of increased number of correct responses is shown. A sequence of three stages is described. In the first, characterised by a small number of correct responses and a small number of multiple responses (MR), that is responses where more than one comparison figure are matched to the standard, performance is affected as much by extraneous factors such as position of a figure in the comparison array as by stimulus features. The second stage shows an increase in both MR and number of correct matches; here global features of similarity appear to be being detected and used. The final stage, with most responses being correct and few MRs being given, represents the most competent level. It is indicated that failure to detect stimulus features, rather than the presence of a deviant notion of the meaning of "same, "/"same," is responsible for the error patterns shown. The concept of similarity is then examined in greater detail. A number of distinctions are drawn, in particular, between the specification of similarity relations between members of a stimulus set in terms of its attribute structure and the perceived similarity of the same set as expressed by subjects' judgements. The importance of providing a normative model as a baseline for the assessment of perfomnance is emphasised. A number of ways of specifying stimuli are described, together with methods of analysing similarities data. The use of a model to link perceived and physical structure, and thus to give some indication of the processes underlying discrimination performance, is considered. A set of experiments is then described which embody these ideas, and it is shown that even in four-year-old children errors in a matching-from-sample task systematically reflect features of the v/hole stimulus set. It is also shown that this does not hold with pair comparison presentation. The capacity of children for redefining the attribute structure of stimulus sets is brought out. It is concluded that much more attention should be given to stimulus speci¬ fication in formal terms if the processes involved in form discrimination are to be elucidated, as opposed to the discriminability of a particular set of stimuli under particular circumstances