The Intelligibility-Based Repeat-Recall Test: I. Bayesian-Guided Estimation of Multiple Speech Reception Thresholds.
TL;DR
The ezSRT test is capable of producing reliable SRT estimates in 6 min that are sensitive to different experimental conditions and listener groups and that result in different SiN performance levels when later tested in a fixed SNR configuration, and using the ezSRT test as part of the new i-RRT protocol to determine SNRs targeting specific intelligibility levels.
Abstract
Objectives
The intelligibility-based Repeat-Recall Test (i-RRT) seeks to compare secondary speech-in-noise (SiN) test outcomes, such as listening effort, against normative references measured at equal levels of SiN intelligibility. The present study evaluates the reliability, sensitivity, and accuracy of the easy speech reception threshold ("ezSRT") test, the first part of the i-RRT, a Bayesian-guided adaptive SiN test for quickly estimating a listeners' complete performance-intensity functions and deriving estimated SRTs.
Design
Single-blind mixed design. Twenty normal hearing and 19 hearing-impaired older adults participated in the study over 2 visits. In the first visit, listeners performed the ezSRT test with target speech sentences presented at a fixed level from a single loudspeaker located directly in front. Colocated background maskers, either speech-shaped noise or two-talker babble noise, were presented at varying levels to produce signal to noise ratios (SNRs) determined by the ezSRT test software. Listeners' SRTs were derived from performance-intensity functions estimated by ezSRT based on performance measured over 30 sentences (~6 min) drawn from a refined RRT speech corpus. Sentences were either high or low in their availability of semantic context. Listener SiN performance was scored based on correct repetition of three to four preidentified target words per test sentence. The ezSRT was administered twice to assess within-session test-retest reliability. In the second visit, the validity of SRT estimates was evaluated by testing listeners with 30 sentences at fixed SNRs corresponding to their individualized SRTs for 50%, 75%, and 90% performance at each combination of masker type and context availability. The ezSRT test was also readministered in visit 2 to assess between-session test-retest reliability.
Results
Both within- and between-session test-retest reliabilities of SRTs estimated by the ezSRT test wa "good" to "excellent" across conditions, with intraclass correlation coefficients (ICCs) ranging from 0.84 to 0.95. Estimated SRTs also showed significant sensitivity to expected effects of semantic context availability, masker type, and listener group. Validation of SRT estimates via fixed SNR testing showed improved SiN performance at greater SRT performance levels, as expected, for both sentence contexts, listener groups, and masker types. However, listeners tended to outperform expected intelligibility performance criteria, particularly at SNRs corresponding to 50% speech recognition (SRT50).
Conclusions
The ezSRT test is capable of producing reliable SRT estimates in 6 min that are sensitive to different experimental conditions and listener groups and that result in different SiN performance levels when later tested in a fixed SNR configuration. These results support using the ezSRT test as part of the new i-RRT protocol to determine SNRs targeting specific intelligibility levels.
