Academic journal article Memory & Cognition

Procedural Interference in Perceptual Classification: Implicit Learning or Cognitive Complexity?

Academic journal article Memory & Cognition

Procedural Interference in Perceptual Classification: Implicit Learning or Cognitive Complexity?

Article excerpt

Researchers have argued that an implicit procedural-learning system underlies performance for information integration category structures, whereas a separate explicit system underlies performance for rule-based categories. One source of evidence is a dissociation in which procedural interference harms performance in information integration structures, but not in rule-based ones. The present research provides evidence that some form of overall difficulty or category complexity lies at the root of the dissociation. The authors report studies in which procedural interference is observed for even simple rule-based structures under more sensitive testing conditions. Furthermore, the magnitude of the interference is large when the nature of the rule is made more complex. By contrast, the magnitude of interference is greatly reduced for an information integration structure that is cognitively simple. These results challenge the view that a procedural-learning system mediates performance on information integration categories, but not on rule-based ones.

A modern debate in the perceptual classification literature concerns the issue of whether or not multiple representational systems underlie categorization performance (e.g., Ashby, Alfonso-Reese, Turken, & Waldron, 1998; Erickson & Kruschke, 1998; Nosofsky & Johansen, 2000; Nosofsky & Zaki, 1998). Furthermore, assuming that there are multiple systems, what is their nature, and how do the systems interact?

One highly influential proposal stems from the COVIS (competition between a verbal and an implicit system) model of Ashby etal. (1998). According to the COVIS model, there are at least two separate cognitive systems that underlie perceptual classification. One system learns categories by using explicit reasoning that is closely related to the formation of verbal rules. A second cognitive system relies on a form of implicit learning. Other multiple-system models have also been proposed that posit combinations of explicit and implicit representational systems. However, a distinctive aspect of the COVIS model involves its assumption that the implicit-learning system in categorization is procedurally based. By contrast, in other multiple-system models, the implicit system might involve the storage of specific exemplars or category prototypes (Erickson & Kruschke, 1998; Pickering, 1997; Reber, Stark, & Squire, 1998).

In recent work, Ashby, Ell, and Waldron (2003) sought to provide distinctive evidence that favors the procedurallearning assumption of the COVIS model. The key idea that motivated their experiments is that procedural learning is closely associated with motor performance and the learning of specific motor responses (Willingham, Nissen, & Bullemer, 1989). By contrast, there is no reason to expect close ties, for example, between exemplar-based learning and motor responses.

To pursue this idea, Ashby et al. (2003) conducted experiments in which forms of response-motor interference were introduced into the learning of different types of categories. The researchers distinguished between two fundamentally different types of category structures. Rule-based category-learning tasks are ones in which the category structure can be learned by an explicit reasoning process that generally involves the formation of a verbal rule. From a formal point of view, in rule-based tasks, the observer makes a separate decision along each individual dimension about the category regions where a percept falls. These separate decisions are then combined to determine the overall categorization response (Ashby & Gott, 1988; Nosofsky, Clark, & Shin, 1989). A simple example is the unidimensional category structure that is illustrated in the top panel of Figure 1. In this example, the stimuli vary along two continuous dimensions, but only one dimension is relevant for performing the classification. Stimuli with large values on Dimension 1 are assigned to Category A, whereas stimuli with small values are assigned to Category B. …

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.