Multimodal Physiological Responses During Sentence Recognition at Individualized Speech-in-Noise Thresholds

By individualizing speech-in-noise level, one can aim to compare physiological responses near a comparable recognition level across listeners with different hearing abilities. We examined whether the aim was realized and whether pupil diameter, skin conductance, and photoplethysmography-derived heartrate provided usable responses during sentence recognition. Thirty-three young adults with normal or near-normal audiograms repeated sentences in multitalker babble at individually calibrated levels. Code audit identified the calibration estimate as a model-based whole-sentence SNR50, rather than the intended SNR75. Actual whole-sentence accuracy averaged 43.6% (ranging from 20% to 70%). Relative to a 500 ms pre-speech reference, mean pupil change was 0.17 mm during encoding-retention (95% CI [0.13, 0.21]; n = 19) and 0.14 mm during verbal response ([0.08, 0.20]; n = 13). Corresponding skin-conductance changes were 0.02 μS ([0.01, 0.04]; n = 30) and 0.07 μS ([0.04, 0.09]; n = 30). Skin conductance was greater during verbal response, an interval also containing speech production, respiration, and movement; this contrast cannot isolate listening-related activity. No cross-modal or physiology-behavior association survived false-discovery-rate correction. The short reference did not support heart-rate change estimation. Individual calibration provided an approximate performance anchor but did not verify equivalent observed performance or effort. The data describe stage-related pupil and conductance reactivity but cannot rank modalities by sensitivity to changing listening demand.