Intelligibility of speech produced with an electronic artificial larynx

Abstract

Five normal-speaking adult males were trained to use an electrolarynx. A speech pathologist trained the talkers ever a six-week period. At the end of the training period each talker read two lists from the Modified Rhyme Test (MRT): a closed-response intelligibility test. Eight listeners provided responses to the MRT recordings. The overall intelligibility for the five talkers averaged 90%. Word-final consonants were more intelligible than word-initial consonants. Liquids, glides, and affricates were more intelligible than stops, fricatives, and nasals. The recorded materials were also transcribed phonetically by seven listeners. This technique revealed more errors than were obtained in the closed-response format. Overall intelligibility was 76%, with word-final consonants being more intelligible than word-initial consonants. Regardless of the method by which intelligibility was assessed, voicing errors constituted the major source of perceptual confusions. Voiceless consonants were often perceived as their counterparts. In contrast, the data revealed few place and manner errors. [Research supported by NINCDS and a grant from the Samuel W. Traylor Foundation.]