A Sequential Sentence Paradigm Using Revised PRESTO Sentence Lists
06 August 2020 (online)
Background: Listening in challenging situations requires explicit cognitive resources to decode and process speech. Traditional speech recognition tests are limited in documenting this cognitive effort, which may differ greatly between individuals or listening conditions despite similar scores. A sequential sentence paradigm was designed to be more sensitive to individual differences in demands on verbal processing during speech recognition.
Purpose: The purpose of this study was to establish the feasibility, validity, and equivalency of test materials in the sequential sentence paradigm as well as to evaluate the effects of masker type, signal-to-noise ratio (SNR), and working memory (WM) capacity on performance in the task.
Research Design: Listeners heard a pair of sentences and repeated aloud the second sentence (immediate recall) and then wrote down the first sentence (delayed recall). Sentence lists were from the Perceptually Robust English Sentence Test Open-set (PRESTO) test. In experiment I, listeners completed a traditional speech recognition task. In experiment II, listeners completed the sequential sentence task at one SNR. In experiment III, the masker type (steady noise versus multitalker babble) and SNR were varied to demonstrate the effects of WM as the speech material increased in difficulty.
Study Sample: Young, normal-hearing adults (total n = 53) from the Purdue University community completed one of the three experiments.
Data Collection and Analysis: Keyword scoring of the PRESTO lists was completed for both the immediate- and delayed-recall sentences. The Verbal Letter Monitoring task, a test of WM, was used to separate listeners into a low-WM or high-WM group.
Results: Experiment I indicated that mean recognition on the single-sentence task was highly variable between the original PRESTO lists. Modest rearrangement of the sentences yielded 18 statistically equivalent lists (mean recognition = 65.0%, range = 64.4–65.7%), which were used in the sequential sentence task in experiment II. In the new test paradigm, recognition of the immediate-recall sentences was not statistically different from the single-sentence task, indicating that there were no cognitive load effects from the delayed-recall sentences. Finally, experiment III indicated that multitalker babble was equally detrimental compared to steady-state noise for immediate recall of sentences for both low- and high-WM groups. On the other hand, delayed recall of sentences in multitalker babble was disproportionately more difficult for the low-WM group compared with the high-WM group.
Conclusions: The sequential sentence paradigm is a feasible test format with mostly equivalent lists. Future studies using this paradigm may need to consider individual differences in WM to see the full range of effects across different conditions. Possible applications include testing the efficacy of various signal-processing techniques in clinical populations.