Word Identification as a Function of Contextual Information
Gary A. Klein, Helen Altman Klein · The American Journal of Psychology · 1973
Subjects were asked to identify word units within unbroken strings of letters that varied in contextual structure: in Experiment I the strings were first-, third-, and sixth-order approximations to English, and a prose passage; in Experiment II they were ungrammatical, anomalous, and grammatical passages. The subjects in both experiments showed highly significant increments in performance with increasing contextual structure, results that are discussed in terms of word-recognition models postulating distinctive-feature tests. High-speed sequential identification of words in auditory or visual passages is a complex skill that is only beginning to be understood. Neisser (1967) has argued convincingly that these phenomena cannot be explained simply in terms of the successive deciphering of stimulus units. Some type of predictive, hypothesis-generating behavior seems to be involved, in which prior context is used as a basis for inferring properties of material to come. The present research was designed to investigate the use of contextual information for facilitating word prediction in a visual task. There have been several approaches toward defining the ways that contextual cues contribute to word identification. Perhaps the simplest is the probabilistic nature of the material. Miller, Heise, and Lichton (1951) varied the number of alternatives from which an auditory stimulus could be chosen, and found that the larger this set, the less accurate the recognition. Miller and Selfridge (1950) generated a set of passages of varied approximations to English. Their technique involved placing increasing constraints on word selection to force greater conformity to coherent passages. They showed that higher approximations resulted in superior learning. One difficulty of this approach is that higher approximations also show increased semantic and syntactic characteristics, which confounds the interpretation of the results (Tejirian, 1968). Semantic and syntactic features of material have been shown to facilitate word identification in an auditory task using masked presentation. Miller and Isard (1963) developed three types of material: gram