Abstract
The present study used pupillometry to assess listening effort in normal-hearing adults during a speech perception task. Early research on speech perception relied on self-report or subjective measures to assess effortful listening, while more recent studies employ objective measures to better quantify listening effort in various listening conditions. Further, audiological evaluations reveal that cochlear implant (CI) recipients experience greater difficulty understanding speech due to the spectrally-degraded quality of speech input. To test this, Defenderfer et al. (2017) measured neural activation to investigate the effect of speech quality in normal-hearing adults using speech-in-quiet, simulated CI-speech (vocoded), and speech in background noise (SIN). The SIN condition was associated with significant increases in neural activation, yet no significant differences were observed between speech-in-quiet and vocoded conditions, despite the degraded quality of vocoded speech. Expanding on these findings, the present study used pupillometry to elucidate the impact of vocoded speech on listening effort. Changes in pupil size have been shown to reflect the amount of cognitive resources required to carry out a task; hence, the pupil response may be more sensitive than neural measures at detecting effort. Using the same conditions as in Defenderfer et al. (2017), we recorded pupil size while normal-hearing adults performed a speech perception task. First, mean pupil sizes were significantly larger when participants were listening to the sentences compared to the pre-sentence baseline (-1000ms) for all three conditions. This effect increased as the quality of speech became more difficult, suggesting greater effort. Specifically, changes in pupil size were greatest for the SIN condition (M=248 a.u./16% change) followed by the vocoded (M=160 a.u./11% change) and then by speech-in-quiet (M=45 a.u./3% change). These results indicate that vocoded speech demands more cognitive resources than speech-in-quiet, suggesting that pupillometry is a more sensitive measure of effort than neural measures during listening tasks.