Hearing Loss Is Reflected in Conversational Pupil Responses.

The experimental setup was open: talkers' gaze was not restricted, so accommodation and local luminance differences could partially drive the pupil responses alongside cognitive effort. This is a secondary analysis of data collected for a different purpose (evaluating a noise-reduction algorithm), and the three noise-reduction settings were collapsed across, introducing potential asymmetry in difficulty between the two hearing groups. Turn detection was based solely on voice-activity timing, not on what was said; brief back-channels and longer utterances were treated the same. The task was a structured spot-the-difference conversation, not free-form dialogue, so generalisability to everyday conversation is uncertain. All participants were older Danish speakers; findings may not generalise to younger populations or other language groups. The hearing-loss group spanned mild to severe loss, and the authors note that roughly 38% had moderate or severe loss, so the group is heterogeneous. About 17% of pupil recordings were excluded due to excessive missing data, which could introduce bias if dropout was related to difficulty.

Funded by the William Demant Foundation (grant 21-2520). William Demant is a major hearing-aid manufacturer. The authors declared no other conflicts of interest. The data were originally collected to evaluate a proprietary noise-reduction algorithm, and the current analysis is a secondary use of that dataset.

Do not read the pupil-size differences as definitive proof of a specific cognitive process. The authors themselves state that gaze shifts, accommodation, and other behavioural factors could partially drive the responses, and the hearing-status effect on mean pupil size was only borderline (p = 0.05). The clearly significant finding is the pupil drift (fatigue) difference, not the absolute pupil size difference.