Detecting and correcting mispronunciations: A note on methodology

Abstract

Two different techniques were employed to investigate subjects' ability to detect mispronounced words in continuous speech. Twenty two‐syllable words were mispronounced in each of three prose passages, changing either the stress pattern, the voicing of obstruents, or the frontback dimension of vowels. Each passage contained only one mispronunciation type; the mispronounced words were equated for predictability from context and for frequency of occurrence in English. Subjects were tested under two conditions. In the first, subjects indicated mispronounced words on a script while listening to the passages. In the second, subjects were instructed to stop the tape when they detected a mispronounced word and to say what the word was. The subjects' spoken responses were recorded on tape. The first condition provides information regarding subjects' ability to detect mispronunciations under optimal conditions. The second condition provides detection information as well as information about the subjects' ability to correct the mispronounced words and the location of the detection in the stream of speech. The subjects' responses to stress mispronunciations were different from those to the other two mispronunciation types, in both test conditions.