A Speech Perception Test in Simulated Reverberation Conditions.

Abstract

This study developed new test materials by applying various reverberation treatments to sentences having high and low contextual redundancy. The Speech Perception in Noise-Revised (SPIN-R; Bilger, Nuetzel, Rabinowitz, & Rzeczkowski, 1984; Kalikow, Stevens, & Elliott, 1977) sentences were modified (SPIN-Reverb) with reverberation times (RT60) from simulated environments: unprocessed, RevCond 1 (RT60 = 600 ms), RevCond 2 (RT60 = 1200 ms), and RevCond 3 (RT60 = 3600 ms). Phase 1 investigated list equivalency among 75 listeners with normal hearing; Phase 2 examined the utility of SPIN-Reverb for 15 cochlear implant (CI) recipients. Equivalent lists within each reverberation condition (unprocessed, RevCond 1, 2, and 3) were identified using nonparametric bootstrapping. Analysis of variance (Phase 1) demonstrated significant differences across conditions for high predictability and total scores. Listening performance decreased for both high and low predictability as RT60 increased for listeners with normal hearing and CI recipients. Unprocessed, RevCond 1, RevCond 2, and RevCond 3 conditions were significantly different from each other. Within RevCond conditions, high- and low-predictability sentences were significantly different from each other. RevCond 1 and RevCond 2 may be useful supplements to the current CI battery. The SPIN-Reverb has potential as a set of clinically feasible materials that are graded in difficulty and representative of real-life acoustic challenges for the evaluation of sensory devices.