Development, reliability, and validity of PRESTO: a new high-variability sentence recognition test

Area of Science:

• Audiology and Speech Sciences
• Neuroscience
• Cognitive Psychology

Background:

• Existing speech recognition tests lack sensitivity to individual differences and core neurocognitive processes.
• The Perceptually Robust English Sentence Test Open-set (PRESTO) was developed using exemplar-based learning theories with high-variability sentences.
• PRESTO incorporates talker and dialect variability from the TIMIT corpus, balanced for keywords, frequency, and familiarity.

Purpose of the Study:

• To evaluate the performance, test-retest reliability, and validity of the PRESTO speech recognition test.
• To compare PRESTO performance with the Hearing In Noise Test (HINT) across various signal-to-noise ratios (SNRs).

Main Methods:

• Phase I: 121 young, normal-hearing adults completed PRESTO at four SNRs.
• Phase II: 40 participants from Phase I were retested on PRESTO and also completed the HINT.
• Data analysis involved performance distribution, test-retest correlations, and comparisons between PRESTO and HINT using ANOVA.

Main Results:

• PRESTO demonstrated a wide range of performance (40.26% to 76.18% keyword accuracy) across SNRs.
• Excellent test-retest reliability was found for PRESTO (r = 0.92, p < 0.001).
• PRESTO showed moderate correlation with HINT (r = 0.52, p < 0.001), with distinct performance patterns at different SNRs.

Conclusions:

• PRESTO exhibits strong test-retest reliability and is sensitive to individual differences in speech perception.
• The distinct performance patterns suggest PRESTO and HINT engage different processing strategies, highlighting the value of high-variability materials.
• Findings underscore the importance of high-variability speech stimuli for accurately assessing individual differences in auditory perception.