Martha M. Shiell1, Sergi Rotger-Griful1, Martin Skoglund1,2, Johannes Zaar1,3, Gitte Keidser1
Eriksholm Research Centre, Oticon A/S, DK-3070 Snekkersten, Denmark
2Department of Electrical Engineering, Linköping University, Linköping, Sweden
3Department of Health Technology, Technical University of Denmark, Kgs.
Lyngby, Denmark

Although hearing aids can be effective at restoring sensory input, some users still struggle to benefit from their devices in real-world communication situations. To better understand and address the needs of these users, we need test paradigms that can capture more ecologically-valid outcome measures. In this presentation, we will describe a range of recent efforts from Eriksholm Research Centre where we explored methods to develop such paradigms, and share recommendations for continued work in this direction. Our efforts spanned seven independent experiments with various groups of hearing-impaired and normal-hearing participants. Over these experiments, we advanced traditional speech intelligibility tests with two additions: (1) stimuli that reflected some real-world complexity, and (2) an accuracy measure aimed capturing the listener’s understanding of the speech material (i.e., comprehension). The stimuli were audiovisual recordings of three unscripted talkers, two of whom engaged in a Diapix task and a third that improvised a monologue. This high level of realism produced the expected challenges to experimental control, but also produced high participant engagement – to the extent that engagement may even have distracted somewhat from the experimental tasks. The accuracy measure was calculated from responses to questions on the speech content. Three styles of questions and their iterations were implemented, in which we targeted the listener’s comprehension while avoiding formulations that rewarded word-recognition strategies. We speculate that, as a side effect of this strategy, our question-response systems evoked increased cognitive abilities for reading, semantic abstraction, and working-memory. Overall, our experiences emphasize the necessity of developing stimuli that are catered to the desired task. Furthermore, they highlight the challenges that are associated with measuring comprehension via extensions of traditional speech intelligibility tests. As such, we suggest that new innovations in behavioural testing may be required for more ecologically-valid outcome measurements.

Acknowledgements: This work was financially supported by the Swedish Research Council (Vetenskapsrådet, VR 2017-06092 Mekanismer och behandling vid åldersrelaterad hörselnedsättninggrant).