Prototype abstraction and classification of new instances as a function of number of instances defining the prototype.

Abstract

Ease of prototype abstraction and recognition of new instances belonging to the prototype was facilitated by increasing the number of instances sorted together during original learning. The 5s sorted distorted dot patterns into groups of 3, 6, and 9 instances, each group containing distortions generated from a single prototype. Following the sorting task, half of the 5s were tested immediately on their ability to correctly classify old and new patterns as well as the prototype; another group of 5s was tested 4 days later. It was found that (a) correct classification of both the prototype and new instances increased as a function of the number of old instances sorted together in the original learning task; and (b) old instances exhibited some forgetting over the delay, but neither the prototype nor new instances did. It was concluded that the abstraction of a prototype undergoes repeated change as a function of the number of instances which define it, and that the ability to correctly recognize new exemplars of a concept is dependent upon the number of instances as well.A number of recent experiments have attempted to understand how information is abstracted from a set of instances that are classified together (Franks & Bransford, 1971;Posner & Keele, 1968Reed, 1972;Strange, Keeney, Kessel, & Jenkins, 1970). Unlike more typical concept formation tasks (e.g., Bourne, 1966), which require the discovery of verbalizable rules along well-defined dimensions, Posner and Keele (1968, 1970) have investigated the abstraction process with stimuli which are not easily verbalizable and whose dimensions are poorly understood. The basic paradigm requires that an initial "prototype" be determined (by randomly assigning 9 dots to a 30 X 30 matrix), and that instances of that prototype be derived (usually by distortion of the prototype, in which each dot is moved according to some statistical decision rule). A number of prototypes are formed in a like manner, each with its own set of instances. In the first part of the experiment, 5s are shown only the instances of the prototype in a random order, and learning is complete when the instances of each prototype are classified together. In the second part, Ss are tested 1 Requests for reprints should be sent to Donald

Recognition vs. classification of ill-defined category exemplars

“…In addition, the prototypicality of the foils presented in the recognition task was a significant factor in recognition performance: More prototypical foils were more likely to be incorrectly recognized than were less prototypical foils. Finally, it was shown that the learning prooedure produced the same effect of category size on classification that had been obtained previously by Homa et al (1973). Classification of new exemplars was enhanced by increasing the number of exemplars representing a category during learning.…”

“…Since classification of new items and the prototype remained constant across the delay, the authors suggested that their classification was mediated, not by the deteriorating information about the learning exemplars, but by the abstracted prototypical representation, which had been maintained. The findings of Posner and Keele have been expanded by Homa and his colleagues (Homa, Cross, Cornell, Goldman, & Shwartz, 1973), who showed that classification of new exemplars can be influenced by the number of exemplars that represent the category during learning. When subjects were taught to classify three categories represented by three, six, and nine exemplars, transfer to new items increased with the size of the learning set.…”

Differential forgetting of prototypes and old instances: Simulation by an exemplar-based classification model

“…The assumption of all-or-none forgetting of features borrows from Bower's (1967) multicomponent theory of the memory trace, and this suggests that some version of that theory would predict the behavior shown in the present simulations. What we have demonstrated most clearly here is that differential forgetting of prototypes and old exemplars, as shown by Homa et al (1973), Posner and Keele (1970), and Strange et al (1970), cannot be taken as conclusive evidence for the abstraction of a prototype during classification learning, for there is at least one set of assumptions that predicts this result purely from the forgetting of old exemplars, with no ad hoc mechanisms required. Our findings thus provide added support for the notion that classification learning may be based solely on memory for past instances, without involving the representation of a prototype or abstract idea.…”

“…The most compelling evidence for this conclusion has come from studies comparing changes over a retention interval in classification performance on exemplars and on the prototypes from which the exemplars were derived. Specifically, several studies have shown that the ability to classify old exemplars declines substantially over time, whereas the ability to classify prototypes, not seen during origina11earning, declines little and in some cases may even improve (Homa, Cross, Cornell, Goldman, & Schwartz, 1973;Posner & Keele, 1970;Strange, Kenney, Kessel, & Jenkins, 1970). This differential forgetting suggeststhat the prototype has a representation of its own.…”

A parallel rule activation and rule synthesis model for generalization in category learning

“…In well-defined categorizations, transfer performance is nearly perfect and improves slightly with the distance between the to-be-categorized exemplar and the category boundary (see, e.g., Nosofsky, 1991; Vandierendonck, 1988Vandierendonck, , 1991. If the categorization is fuzzy, transfer performance is not quite so good, and it varies as a function ofthe similarity between the new exemplar and the old stimulus patterns (see, e.g., Homa, Cross, Cornell, Goldman, & Schwartz, 1973;Medin & Schaffer, 1978;Nosofsky, 1984Nosofsky, , 1986. In fact, it seems that transfer gradients occur with both kinds of category structure.…”