In a Signal Detection experiment, observers attempted to identify a word masked by white noise which occurred in a tape recorded utterance. This was done by using a Yes-No procedure in which the response item sometimes fitted the context of the utterance on the recording, and sometimes did not. When the response item was in context observers were more biased towards reporting that it was the masked word than when the response word was out of context. Also there was a small improvement in sensitivity for words in context which suggests that context not only affects the response system but also has perceptual consequences.Amongst the factors affecting the intelligibility of speech presented under degraded conditions is context. Miller, Heise and Lichten (1951), O'Neill (1957) and Martin (1968) have found that spoken speech masked by white noise is more accurately perceived when presented in the context of a sentence than when presented in isolation. The time for which a word need be exposed before it can be accurately reported is also lessened if it is presented in context (Gross, 1966; Morton, 1964a) and the amount of material which can be correctly reported when a string of letters is briefly shown for a fixed duration, is influenced by contextual changes (Miller, Bruner and Postman, 1954).The effects of context can be likened to that of the Word Frequency Effect in which familiar words are better perceived than unfamiliar ones (Howes, 1954; Howes and Solomon, SolomonandHowes, 1951). In the latter case it is the familaity ~ 1 The authors wish to acknowledge financial assistance from the Australian Research Grants Committee. 'Now at the University of New South Wales.per se which affects the observer's intelligibility score; in the former case a word is rendered more familiar in that its context makes its appearance more predictable.The early experiments on the Word Frequency Effect employed classical psychophysical procedures to determine the threshold for familiar and unfamiliar words and found that more familiar words gave lower recognition thresholds. More recently a number of studies have ascribed the Word Frequency Effect to response rather than stimulus factors. Goldiamond and Hawkins (1958) found a Word Frequency Effect in the absence of any stimulus word, and Broadbent (1967), using procedures based on Luce's (1959) Choice Theory, was able to separate perceptual and response effects, and concluded that the superior perception of common words was due entirely to response bias. This experiment used measures derived from Signal Detection Theory to determine whether the superior recognition scores for words masked by white noise and presented in context are due to response bias or enhanced discriminability.
305