Auditory Training With Synthesized Voice Anchors: Effects on Rater Agreement.

Purpose: To investigate the effects of auditory training with synthesized voices on intra- and inter-rater agreement of the auditory-perceptual voice analysis of roughness and breathiness.

Methods: This was an experimental study consisting of four auditory training sessions. The sample consisted of twenty raters, students from a Speech-Language Pathology course, who had previous experience with auditory-perceptual assessment. The raters participated in the four training sessions with a seven-day break in between sessions. Each training consisted of three tasks: 1) Pre-training activity: Participants were asked to rate 20 natural voices, normal and dysphonic, from zero to three, according to the parameters of roughness and breathiness; 2) Training activity: Synthesized voice anchor stimuli were presented, and participants were asked to rate them from zero to three. Four stimuli were related to roughness and four to breathiness. Participants heard 20 voice stimuli and were instructed to pair the natural voice with the synthesized anchor stimulus that most resembled it; and 3) Post-training activity: the 20 voices from the pre-training activity were randomized and participants rated the same voices, without prior knowledge that these were repeated. Statistical analysis of data was performed using the AC test, to assess the extent of agreement between raters, and the Friedman test to compare the training sessions. A 5% significance level was considered.

Results: For the auditory-perceptual voice analysis of roughness, intra-rater agreement results ranged from 79% to 86% between the first and fourth auditory training session, with improvement in intra-rater agreement from the fourth session forward (P = 0.005). For the analysis of breathiness, results ranged from 88% to 92% between the first and fourth auditory training sessions, with improvement from the fourth session forward (P = 0.036). Inter-rater agreement results for the auditory-perceptual analysis of roughness ranged from 23% to 34%, and from 48% to 61% for breathiness, with no differences regarding training (P = 0.855).

Conclusion: The auditory-perceptual breathiness parameter had a higher AC indicator compared to the roughness parameter, suggesting better agreement. The intra-rater agreement showed improvement starting from the fourth auditory training session for the assessment of roughness and breathiness. The auditory training program did not show a positive inter-rater agreement impact.