The Intelligibility-Based Repeat-Recall Test: II. Performance of Listeners With Normal Hearing and Hearing Impairment...

To gather normal-hearing (NH) reference data for an intelligibility-based Repeat-Recall Test (i-RRT) against which to compare the performance of hearing-impaired (HI) listeners.

A single-blind, within-subjects design involving 20 NH and 19 HI adult listeners who performed the i-RRT protocol in two stages. First, listeners were tested with an adaptive "easy" Speech Reception Threshold test to estimate their performance-intensity functions for speech-in-noise (SiN). Second, listeners were tested at fixed signal to noise ratios (SNRs) corresponding to individualized speech reception thresholds (SRTs) for different performance criteria. NH listeners were tested separately for high context (HC) and low context (LC) materials drawn from the Repeat-Recall Test (RRT) speech corpus, presented from the front at a fixed level of 68 dBA, in two-talker babble-noise from the front (2TBN-0°) or speech-shaped noise back (SSN-180°) masking configurations. NH listeners then performed the i-RRT at SNRs corresponding to individualized estimates of speech reception threshold at a criterion of 25% correct, SRT50, SRT75, and SRT90. The i-RRT involved repeating lists of six RRT sentences, then recalling the content of those sentences, and providing ratings of listening effort (LE) and willingness to stay in noise (i.e., tolerable time [TT]). Each combination of SRT criterion, noise configuration, and sentence context was tested twice to assess test-retest reliability. As a demonstration, HI listeners were tested in the aided mode using only HC speech materials in 2TBN-0° and SSN-180° noise configurations, and the i-RRT was administered at SNRs corresponding to SRT50 and SRT90.

The "easy" Speech Reception Threshold performance of NH listeners was significantly better for HC than LC sentences and for SSN-180° than 2TBN-0° noise configurations. All subsequent outcome measures of the i-RRT (repeat, recall, LE, and TT) were significantly affected by SRT criteria. However, even at equal SRT criteria, the i-RRT outcome measures of recall, LE, and TT were still significantly affected by sentence context; recall and TT outcomes were also significantly affected by noise configuration. Specifically, NH listeners exhibited poorer recall, higher LE, and lower TT for LC than HC sentences and poorer recall and lower TT for 2TBN-0° than SSN-180° conditions. Test-retest reliability analyses showed the lowest reliability for LC recall in both noise configurations and for repeat scores in 2TBN-0°; other measures showed moderate to excellent reliability. HI listeners' recall and LE ratings were typically in the lower 25 to 30 percentile of NH performance.

Conditions of the i-RRT, including sentence context and noise configuration, in addition to listeners' hearing status, not only impacted the SNR required for different listeners to achieve certain SRT criteria, but also affected secondary measures of recall, LE, and TT ratings assessed at matched SRT criteria. These results suggest that certain SiN test factors demand cognitive resource allocation from listeners that cannot be accommodated by SNR adjustment. While still a research tool, the i-RRT may be useful in the future to clinically profile listeners' SiN difficulties and their allocation of listening-related cognitive resources as captured by recall, LE, and TT.