Background: Pupillometry is commonly used to assess listening effort. For typical hearing (TH) listeners, the magnitude of the pupil response associates with the signal-to-noise ratio (SNR). However, recent studies have shown that the pupil response for CI users is less sensitive to SNR and depends much more on the type of speech material. Most of these studies assessed peak pupil dilation, which reflects the maximal amplitude of the pupil response during or just after sentence offset relative to baseline, but omits the temporal dynamics of the pupil response.

Objective: To assess the time course of the pupil response for CI users and TH listeners at different SNRs and speech stimuli.

Methods: Pupil responses were collected from 18 CI users and compared to 18 age-matched listeners using Matrix and LIST sentences administered at three SNRs: speech recognition threshold, where speech intelligibility (SI) was 50% (SRT); +6 dB re. SRT; and in quiet. Pupil waveforms were analyzed with generalized additive mixed modelling (GAMM).

Results: Preliminary analyses revealed that SNR effects were most prominent for TH listeners, where pupil dilation was significantly larger at more challenging SNRs across all SNR pairs for Matrix and LIST sentences. For CI users, SNR effects were smaller. For the LIST sentences, all SNR pairs differed significantly, but this was not the case for Matrix sentences. For both groups, SNR effects on pupil dilation were most pronounced seconds after sentence offset, but much less prominent during stimulus presentation. CI listeners showed significantly larger pupil dilations than TH listeners during sentence presentation and after sentence offset, in quiet and at +6 dB re. SRT. This effect was also most prominent after sentence offset. At 0 dB re. SRT, where speech intelligibility was identical for both groups, no differences were observed.

Conclusions: SNR effects for both groups were most pronounced after sentence offset, suggesting a more sustained listening effort under more challenging listening conditions. Similarly, differences between CI users and TH listeners were most prominent during the release of effort, which develops long after sentence offset and may not be captured by classical peak-picking outcome measures, notably the peak pupil dilation response.

Funding: ‘INTENSE’ project (grant #17619) within the Crossover Research Program of the Dutch Research Council (NWO), co-funded by Advanced Bionics (Hannover, Germany).

Acknowledgements: We thank the study participants for their dedication, and Paula L. Jansen, Robin van Deurzen, and Nicolas Furnon (Advanced Bionics) for technical support.