Reading a series of similar texts: Testing a schema-based learning theory

Abstract

Thorndyke and Hayes-Roth (1979) proposed, and presented data to support, a schema model of learning from simple texts. They assumed that repeated exposures to similar texts promoted acquisition of a schema for these texts, which in turn benefited recall for new instances of texts that conform to this schema. They further suggested that with repeated use of the schema, the increased number of details associated with the schema disrupted recall. Their data confirmed these expectations. However, Thorndyke and Hayes-Roth did not use a true 0-exposure control to assess baseline recall of the target text. We assessed baseline recall using both a true 0-exposure control and Thorndyke and Hayes-Roth’s derived 0-exposure measure. The predicted inverted-U-shaped curve for recall as a function of prior training texts was replicated with Thorndyke and Hayes-Roth’s baseline measure but not with a true 0-exposure baseline. With the 0-exposure baseline, recall was a monotonically decreasing function of number of training texts. This result suggests that recall of similar passages reflects nothing more or less prosaic than proactive interference.