Children's detection of mispronounced words in fluent speech

Abstract

Children's ability to recognize mispronounced words in fluent speech was measured using an error-detection task in which the children rang a bell when they heard a mispronounced word in a story. Mispronunciations of the same word occurred in contexts in which they were semantically predictable and semantically unpredictable. The mispronunciations occurred word-initially on stop consonants and consisted of changes in voicing or place of articulation. Data were analyzed for the effects of semantic context, feature change, age, and sex on the percentage of errors detected and on the latency of response. Results indicated that four- to six-year-olds were very accurate (nonsignificant age differences) at detecting mispronunciations (⩾78%). Detection of mispronunciations in semantically predictable contexts were significantly better (and latencies were significantly shorter) than in unpredictable contexts, for all age groups. There were no significant differences in the overall detection of mispronunciations involving a change in voicing versus place; however, significant differences for direction of change within feature occurred for both voicing and place. Females were significantly more accurate than males at detecting mispronunciations.