Abstract
Background: The differential diagnosis between apraxia of speech (AOS) and aphasia with phonemic paraphasia is challenging because the disorders share similar speech characteristics. Quantification of the speech profile of AOS is needed.
Aims: To characterise the relative frequency of sound distortion errors in speakers with aphasia with and without AOS and to evaluate inter-observer reliability for this metric among independent listeners.
Methods & Procedures: Fourteen adults with left hemisphere lesions were recruited from a consecutive clinical sample. They were assigned to groups of probable apraxia of speech (P-AOS) or probable aphasia with phonemic paraphasia (P-APP), based on the multisyllabic word production rate and perceptually evident sound errors. Four listeners generated independent phonetic transcriptions from an audio-recorded motor speech evaluation. The percentage of distorted segments and inter-observer reliability were calculated.
Outcomes & Results: Inter-observer reliability was satisfactory for frequency of the overall distortion errors (ICC = 0.73, p < .001). Distortion frequencies were greater in the P-AOS group than in the P-APP group, but varied along on a continuum. Severity effects could not be ruled out, as the overall sound error frequencies were higher in the P-AOS group.
Conclusions: With narrow phonetic transcription, independent clinicians can expect to code relative distortion frequency reliably, even though they may differ in the type of distortions they favour and in the absolute rates they observe. Distortion errors are present in speech produced by speakers with aphasia whose speech is characterised by sound error production, and rates are generally higher in stroke survivors who meet prosodic criteria for AOS. The relationship between phonemically salient sound errors and sound distortion frequency should be addressed in future research.
Acknowledgements
We thank Angel Talavera, Shakia Forbes, Tiara Huff, Kate Winterbottom, and Leigh Wallmeyer for their assistance with data collection.