Academic journal article Canadian Journal of Experimental Psychology

Abstract Analogies and Positive Transfer in Artificial Grammar Learning

Academic journal article Canadian Journal of Experimental Psychology

Abstract Analogies and Positive Transfer in Artificial Grammar Learning

Article excerpt

Abstract

Following Brooks and Vokey (1991), we show that positive transfer to new items generated from an artificial grammar in which the vocabulary has been changed from training to test can be based on "abstract analogy" to specific training items (specific similarity) rather than abstraction of a grammar and symbol remapping rules, even with remapping unique to each test item. The results confirm that transcendence over symbols provides no support for the implicit learning of abstract structure. Ironically, they also show that the effect of specific similarity does not depend on surface characteristics of the items, but the residual effect of grammaticality does.

Brooks (1978) introduced the then-radical idea that positive transfer in structural learning situations can be accomplished without experimental participants' directly representing any aspect of that structure. Partly in response to Reber's (1967, 1969, 1976) earlier claims for abstraction of structure in artificial grammar learning, Brooks proposed instead that participants could be responding at test by analogy to specific, remembered training exemplars. Because such a distributed knowledge base preserves many of the statistical properties of the domain, responding on the basis of similarity to prior instances enables relatively accurate classification of novel items because within-domain items generally will be more similar to individual studied items from the domain than will items from other domains. This property has been exploited by all subsequent instance and distributed memory models (e.g., Estes, 1980; Hintzman, 1986, 1988; Logan, 1988; Medin & Schaffer, 1978; Nosofsky, 1986), although for these computational models, the memory mechanism has been one of similarity to all instances in memory (a "chorus" of instances), rather than analogy to specific instances. With the elaboration that it is the memory for specific processing episodes (e.g., Jacoby & Brooks, 1984), this view has since become a dominant explanation of performance in structural learning situations, and particularly in artificial grammar learning (e.g., Brooks & Vokey, 1991; Higham, 1997a, b; Higham & Brooks, 1997; Vokey & Brooks, 1992, 1994; Whittlesea, 1987; Whittlesea & Dorken, 1993; Whittlesea & Wright, 1997).

Consider dissociations that arc obtained in artificial grammar learning experiments if the grammaticality of test items is manipulated orthogonally to their specific similarity (i.e., the similarity of a given test item to a ' specific training instance). Vokey and Brooks (1992) showed that the magnitudes of the effects of both grammaticality and specific similarity varied significantly but differentially as a function of the various encoding manipulations and test tasks. As another example, Higham (1997b) produced clear dissociations of grammaticality and specific similarity as a function of requiring participants to exclude cither "old" or "similar" items at test, and by dividing attention at test. These dissociations suggest that the effects of specific similarity and grammaticality are the result of different aspects of the encoding and retrieval of the stimuli, at least in part. One explanation for such dissociations might be that the effects of grammaticality and specific similarity are derived from functionally different memory systems (cf. Reber & Allen, 1978): Grammaticality is based on an implicit memory system that abstracts the rules of the grammar, and that is relatively impervious to limited attentional resources at test, whereas the effect of specific similarity is based on an explicit memory system (i.e., explicit retrieval of specific training items) whose operation is impaired by a secondary task.

However, beginning with Brooks and Vokey (1991), Brooks and his students have advocated a more parsimonious explanation: Both the grammaticality and specific similarity effects derive from similarity to prior instances stored in an episodic database, but application of this single episodic memory system can have multiple, dissociable effects (e. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.