Reliability and clinical relevance of segmental analysis based on intelligibility assessment.

OBJECTIVE: In persons with a speech pathology, improved intelligibility is the fundamental goal of treatment. Identification of specific phonetic features that affect intelligibility, contributes to selecting intervention targets. The main purpose of this study was to investigate whether a segmental analysis based on a perceptual phoneme intelligibility assessment is reliable and can be used for target selection. Secondly, the hypothesis that the degree of agreement for phoneme identification may depend on the intelligibility of the speaker is addressed. PATIENTS AND METHODS: 30 patients with pathological speech (11 dysarthric, 10 hearing impaired, 9 laryngectomees) were assessed by means of a standardized phoneme intelligibility assessment. The samples were judged twice by 9 experienced listeners. The inter- and intra-rater reliability for each test item were examined. RESULTS: The inter-rater reliability per patient for phoneme identification varied between fair and almost perfect (k: 0.24-0.89). The higher the patient's intelligibility, the higher the level of agreement (p < 0.001). The intra-rater reliability for phoneme identification across all patients was good (k: 0.603-0.787). CONCLUSION: A segmental analysis based on a phoneme intelligibility assessment is only reliable and clinically relevant in patients with a slightly to moderately decreased overall intelligibility level.