Comparing roughness in sustained phonations and connected speech using a matching task

Experiments on perception of dysphonic voice quality have typically relied on sustained vowel phonations. The strong preference for vowels could be explained by their ease of production, their time invariance, and by the absence of confounding articulatory (e.g., consonant, dialect) and prosodic (e.g., stress, rate) changes. However, the magnitude of roughness perceived from sustained vowels may not reflect the roughness in connected speech produced by the same speaker. We examined how stimulus type impacts the perception of roughness. Ten naïve listeners judged roughness for both vowel /a/ and sentences selected from ten dysphonic speakers using a single-variable matching task. Stimuli were selected to ensure a continuum of vocal roughness for vowels. The intra- and inter-listener reliability estimated using intra-class correlation (ICC) as well as matching values and variability will be compared across each stimulus type. This work is essential for extending models of dysphonic voice quality perception from vowels to connected speech, which are likely to correspond more closely with perceived handicap and better represent relevant treatment targets for patients with dysphonia.