Effort allocation during speech perception: Insights from pupillometry
Patrycja Alina Lebiecka-Johansen · 2023
Untreated hearing impairment affects daily-life communication, increases likelihood of social isolation and risk of cognitive decline in later life. Effortful listening is one of the major complaints of adults with hearing loss. Pupillometry (measurement of pupil size) has been identified as a suitable method to measure and quantify objectively listening effort allocation. Previous studies showed that the pupil size is sensitive to both acoustic and memory task demands during listening as well as to hearing status. In daily life listening, however, more factors are playing a role, and these may also influence the pupil size. One example is luminance. This factors still hinders use of pupillometry outside the controlled laboratory setting. Therefore, there is a need to identify and develop pupil measures that can disentangle the effects of task demand during speech perception from the effects of illumination level on the pupil response. This thesis had two main aims. Part I aimed to identify pupil measures sensitive to changes in the acoustic task demands (signal-to-noise ratio, type of background noise) that would reveal alterations to effort allocation during speech perception due to the interference from a non-target speech, change in luminance, or in presence of additional task demands (i.e., due to an additional recall task). Part II aimed to apply the most promising pupil measures to predict likelihood of a successful recognition or recall task performance and investigate elements of the hearing-related alterations to effort allocation during speech perception. This thesis provides for a comprehensive and up-to-date overview of pupil measures potentially informative of effort allocation during speech perception. Findings from the Part I can be used to guide the selection of analysis tool for the further research on (listening) effort allocation. Both single-value pupil measures (PPD, MPD, BPS) as well as time series analysis techniques (cluster-based permutation, GCA, PCA) were found to be sensitive to substantial changes in the acoustic task demands (e.g., speech intelligibility at 50% vs. 100%). However, time series analysis techniques were further found to be sensitive to subtle changes in SNR. Besides, they revealed alterations to effort allocation within a single sentence perception due to an additional recall task, presence of an interfering talker or change in luminance. Findings from Part II indicate that likelihood of successful task performance in a word recall task can be predicted from a single-value pupil measure (PPD) extracted from a pupil response during word presentation. Although, a combination of single-value and temporal measures (PPD, BPS, PCA) would be necessary to predict performance in a more complex task such as sentence recognition or differences in hearing-related alterations to effort allocation.