Effect of sentence complexity, amplification, and hearing loss on the psychometric function obtained from adaptive speech-in-noise tests

Hearing aid users expect their hearing devices to provide improved understanding in speech-in-noise situations. This requirement can be validated with adaptive speech tests using sentences as test material. Available speech corpora show differences in terms of complexity and predictability which impact intelligibility and processing demands. Speech material can be classified in two main categories: a) matrix sentences with a fixed grammatical structure which provides a uniform test material once the training phase is done, and b) everyday sentences that introduce more variability in terms of structure and word predictability. Both tests determine the speech reception threshold (SRT) with an adaptive procedure to find the signal-to-noise ratio (SNR) at which 50% of the words are recognized.

The current study investigates the effect of sentence complexity on amplification benefit as a function of the tested SNR. SRTs were measured with a matrix sentence test (OLSA) and an everyday sentence test (GOESA) in three listening conditions: unaided, aided with a RITE device, and aided with an ITE device. Test results, from 16 experienced hearing aid users with a mild to moderately-severe hearing loss, were fitted with a logistic function to extract the SRTs and the slope (s50) of the psychometric function at 50 %. The difference between aided and unaided fitted psychometric curves was used to visualize the differences between listening conditions and sentence complexity over a large test SNR interval. Test material, unaided SRTs, unaided s50, age, and subjective satisfaction measured with the APHAB were selected as fixed effects in a mixed effect regression to predict the amplification benefit.

Results indicate an overall SRT shift towards more positive SNRs with the GOESA as compared to the OLSA test material. This shift suggests that sentences with a more complex structure and unpredictable vocabulary increase the test difficulty. The different SNR range achieved with each test could further impact the aided scores due to the fact that SNR dependent features of hearing aids (like directionality) will provide more benefit at lower SNRs. The results from this test provide indications about factors that might explain variations in the performance of different hearing aids or signal processing algorithms when measured with different speech tests.