Within-speaker speech intelligibility in dysarthria: Variation across a reading passage

Abstract

Understanding factors underlying intelligibility deficits in dysarthria is important for clinical and theoretical reasons. Correlation/regression analyses between intelligibility measures and various speech production measures (e.g., acoustic or phonetic) are often reported in the literature. However, the analyses rarely control for the effect of a third variable (severity of speech disorder, in this case) likely to be correlated with the primary correlated variables. The current report controls for this effect by using a within-speaker analysis approach. Factors that were hypothesized to underlie the intelligibility variations in multiple breath groups within a connected discourse included structural elements (e.g., number of total words) as well as acoustic measures (e.g., F2 variation). Results showed that speech intelligibility in dysarthric speakers with two forms of neurological disease (Parkinson and ALS) does, in fact, vary across breath groups extracted from a connected discourse, and that these variations are related in some cases to a per breath estimate of F2 variation. [Work supported by NIDCD Award No. R01 DC03723.]