Abstract
PURPOSE: Lexical stress errors are a reported characteristic of childhood apraxia of speech (CAS) in American English-speaking children. However, the ability to reliably identify stress errors may vary given several factors. This study examines the interrater reliability of lexical stress accuracy ratings derived from phonetic transcriptions. METHOD: Fifty-six General American English-speaking participants with CAS aged 9-16 years engaged in picture-naming and multisyllabic word imitation tasks at up to four different time points. Some participants (n = 37) engaged in face-to-face data collection, and others engaged remotely via telehealth (n = 19). Among 23 transcribers, a subset of three raters per session completed narrow transcriptions of words, including lexical stress transcription. Data analysis included descriptions of the interrater reliability of the judged lexical stress accuracy of words, with comparison to the reliability of consonant transcriptions, in both face-to-face and telehealth modalities. Intrarater reliability across nine transcribers was also explored. RESULTS: Intraclass correlation coefficient and Fleiss' kappa were used to describe reliability across three raters per item. Results suggest that transcription of lexical stress for children with CAS is moderately reliable. Lexical stress reliability among transcribers was lower than the reliability observed for percent consonants correct from the same transcribers on the same tasks. Listeners were more reliable in transcribing productions from participants in the telehealth group than in transcribing productions from face-to-face participants. CONCLUSIONS: Lexical stress judgments were found to have good intrarater reliability. The role of lexical stress as a moderately reliable construct in the speech profile of school-aged General American English-speaking children with CAS is discussed in regard to clinical use.