Towards more realistic speech tests for hearing aid research

Virginia Best, Jörg M. Buchholz, Gitte Keidser

Research output: Contribution to conferenceAbstract


Sentence tests are routinely conducted in the laboratory and the clinic to assess the ability of a particular listener to understand speech. However these tests are often poor predictors of the performance of that listener in real world listening situations, where a high level of individual variability is encountered. Moreover, current laboratory tests often cannot accurately predict the real world benefit of rehabilitation devices/features. As technology advances, better predictions would be extremely useful for demonstrating a competitive advantage of one device/feature over another. There are several ecologically relevant variables that are missing from current speech tests that may prove to be important, including dynamic variations in spatial and level characteristics of the acoustic environment, the presence of reverberation, and co-ordinated visual information. In addition, verbatim sentence recall does not capture features of real speech communication such as the extraction of meaning and the ongoing engagement of cognitive processes. Our overall aim is to create new speech tests that provide more accurate predictions of real-world performance. Our hope is that by carefully adding realism to speech tests, we will engage the auditory and cognitive processes involved in real communication tasks, and generate more meaningful performance measures. An experiment is described that aims to under-stand the psychophysical effects of introducing realistic variations to speech and noise variables, and preliminary results are presented. Specifically, we examine (a) the effects of conducting a conventional sentence test in a simulated cafeteria that is dynamic and reverberant, and (b) the effects of moving from a speech recall task to a speech comprehension task. The reference task in both cases is a conventional sentence test conducted in anechoic space using spatially diffuse speech babble. Critically, the same set of listeners is used so that correlations across the different tasks can be examined. In addition, psychometric functions are collected to compare sensitivity of the different tests to changes in signal-to-noise ratio. These analyses will establish whether the more realistic tests provide new information over the conventional test, and thus have the potential to better capture individual real-world experiences.
Original languageEnglish
Number of pages2
Publication statusPublished - 2012
EventInternational Hearing Aid Research Conference - Lake Tahoe, California
Duration: 8 Aug 201212 Aug 2012


ConferenceInternational Hearing Aid Research Conference
CityLake Tahoe, California

Fingerprint Dive into the research topics of 'Towards more realistic speech tests for hearing aid research'. Together they form a unique fingerprint.

Cite this