WORD IDENTIFICATION IN NOISE

Speech intelligibility has traditionally been measured by presenting w ords mixed in noise to listeners for identification at several differe nt signal-to-noise ratios. The words are produced in isolation or in s entence contexts where the predictability of specific items can be var ied. Psychometric functions are typically obtained relating signal-to- noise ratio to percent correct recognition. Error analyses are often c arried out by examining response confusions to construct similarity sp aces for words which reflect their perceptual organisation and acousti c-phonetic similarity. When using these techniques to measure speech d iscrimination or speech intelligibility in an open-set format, the rec ognition score obtained reflects the combined influence of both the se nsory information encoded in the speech signal as well as the listener 's decision process and response biases. Despite this limitation, the procedure has strong face validity as a measure of word recognition pe rformance in normal-hearing listeners as well as other clinical popula tions which routinely use speech audiometry techniques to diagnose and assess both peripheral and central hearing impairments. All of the ma jor findings and phenomena in the spoken word recognition literature c an be demonstrated and explored with this experimental method. This te chnique continues to provide extremely valuable information about the organisation of words in the mental lexicon and how these sound patter ns are accessed from acoustic-phonetic information in the speech signa l.