Glottal Stop Production in Controls and Patients with Unilateral Vocal Fold Paresis/Paralysis Using Contextual Speech

(1) To determine whether acoustic glottal stop production (GSP) measures from two sentence-embedded glottal stops (GS) differed between patients with unilateral vocal fold paresis/paralysis (pUVFP) and normal controls (NCs), and if so, their diagnostic accuracy; (2) to examine context specificity by correlating within-speaker GSP measures from sentence-embedded GS with repeated [i] and [isi] tokens; and (3) to test whether GSP measures differed depending on speaker perception as pUVFP or NC.

Sixty-two pUVFP and eight NCs produced "We eat eggs every Easter," containing two potential GS: GS1 (we eat) and GS2 (every Easter). Averages were calculated for four GSP measures (offset and onset intensity differences and slopes) obtained by 2 assessors using a customized Praat script. Group comparisons were analyzed using Mann-Whitney U tests, and diagnostic accuracy by receiver operating characteristics and Youden's Index for combined and separate contexts. Pearson product-moment coefficients were used to assess within-speaker correlations across contexts. Kruskal-Wallis tests assessed differences between GSP measures based on perceptual classifications of the sentence.

Intensity differences were significantly smaller and slopes shallower (P < 0.05) between pUVFP and NCs for both environments combined and for GS2 individually. Diagnostic accuracy was perfect for GS2 measures, but poor for GS1. Significant within-speaker correlations (P < 0.05) were found for GSP measures from most contexts. Significantly greater intensity differences and steeper slopes (P < 0.05) were found from sentences that both listeners perceived as produced by an NC compared to those produced by a pUVFP.

Acoustic GSP measures, regardless of phonetic context, show generalizability in distinguishing between pUVFP and NCs. However, repeated [i] may be more valid because acoustic GSP measures are significantly correlated with measures from other contexts; better diagnostic accuracy, and include one more variable, the number of voicing cessations, already shown to be important in recovery.