Academic journal article Memory & Cognition

Prior Knowledge and Exemplar Frequency

Academic journal article Memory & Cognition

Prior Knowledge and Exemplar Frequency

Article excerpt

New concepts can be learned by statistical associations, as well as by relevant existing knowledge. We examined the interaction of these two processes by manipulating exemplar frequency and thematic knowledge and considering their interaction through computational modeling. Exemplar frequency affects category learning, with high-frequency items learned more quickly than low-frequency items, and prior knowledge usually speeds category learning. In two experiments in which both of these factors were manipulated, we found that the effects of frequency are greatly reduced when stimulus features are linked by thematic prior knowledge and that frequency effects on single stimulus features can actually be reversed by knowledge. We account for these results with the knowledge resonance model of category learning (Rehder & Murphy, 2003) and conclude that prior knowledge may change representations so that empirical effects, such as those caused by frequency manipulations, are modulated.

Frequency has long been known to be an important property of category structure. Rosch and Mervis (1975) argued that the frequency of properties in a category determines how typical category members are. Those that have properties frequently found in the category are more typical than those possessing less frequent properties, and category members possessing properties frequently found in other categories are less typical than those possessing less frequent properties. Although the frequency or familiarity of an object does not itself seem very strongly related to its typicality in natural concepts (Barsalou, 1985; Mervis, Catlin, & Rosch, 1976; Novick, 2003), when item frequency has been experimentally manipulated independently of other variables (such as similarity to category prototypes), it does influence category structure. For example, Nosofsky (1988) showed that repeating one item five times in each block of category learning made the item not only easier to learn, but also more typical after learning. Furthermore, the effect spread beyond the frequent item itself, in that similar items in the same category also were rated as more typical.

Theories of concepts can explain such frequency effects easily (Barsalou, Huttenlocher, & Lamberts, 1998). If exemplar theories assume that each presentation of a stimulus is a stored instance, frequent exemplars will have more stored instances, increasing the typicality of items similar to them. Likewise, if prototype theories assume that the category prototype is based on generalizing from instances, the more an item is repeated, the more influence it will have on that generalization. That is, frequent items will pull the category prototype in their direction. Thus, the effect of frequency seems to be a straightforward example of how category structure influences learning and use of concepts.

One might expect basic variables, such as frequency, to have consistent effects across materials and tasks. However, there are a number of examples in which the effects of category structure are altered when concepts make contact with other knowledge. For example, the standard learning advantage of conjunctive (and) over disjunctive (or) concepts can be overruled when the disjunctive concept is related to prior knowledge (Pazzani, 1991). Wattenmaker, Dewey, Murphy, and Medin (1986) examined the effects of prior knowledge on learning linearly separable and nonlinearly separable categories. Linear separability is a structural variable that refers to whether correct categorizations can be made by independently weighting the category's properties. Wattenmaker et al. showed that linearly separable categories could be made easier or harder to learn than nonlinearly separable categories by varying the categories' content (see also Murphy & Kaplan, 2000). They argued that some conceptual domains encourage summing of evidence, suitable for linearly separable categories, and that other domains encourage configural processing, suitable for nonlinearly separable categories. …

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.