Relationships Between Subjective and Objective Measures of Listening Accuracy and Effort in an Online Speech-in-Noise Study

Objectives: Speech-in-noise performance is of paramount importance to daily function, and there exists a bewildering array of outcome measures to capture the many dimensions of this concept. The aim of the present study was to provide insight into how different speech-in-noise outcome measures relate to one another, how they behave under different test conditions, and how researchers or practitioners might go about selecting an outcome measure (or measures) depending on the context and focus of their enquiry. Design: An online speech-in-noise study was conducted using the Labvanced experimental platform. A total of 67 participants (42 who reported having normal hearing, 25 who said they had some degree of hearing loss) completed the Effort Assessment Scale (a self-reported measure of daily-life listening effort), followed by a sentence recognition task in which BKB sentences were presented in speech-shaped noise at signal to noise ratios (SNRs) of −8, −4, 0, +4, +8, and +20 dB. Participants were instructed to listen to each sentence and then repeat aloud what they heard. Responses were recorded through participants’ webcams and later independently scored by 2 research assistants. Several outcome measures were used to tap into both accuracy and listening effort. Specifically, we examined: (1) objective intelligibility (percentage of keywords correctly repeated); (2) subjective intelligibility; (3) subjective listening effort; (4) subjective tendency to give up listening; and (5) verbal response time (VRT) extracted from the audio recordings. Data were analyzed using Bayesian statistical methods. Results: Hearing loss and age were associated with speech-in-noise outcomes. Specifically, we observed lower intelligibility (objective and subjective), higher subjective listening effort, and longer VRT (time to verbal response onset) in hearing-impaired compared with normal-hearing listeners, and reduced objective intelligibility and longer VRT in older compared with younger listeners. When moving from highly favorable to more adverse listening conditions, subjective listening effort was the first measure to show sensitivity to worsening SNR, followed by subjective intelligibility, objective intelligibility, subjective tendency to give up listening, and, finally, VRT. Participants, especially those with normal hearing, consistently underestimated their own performance. Conclusions: The present findings offer useful insight into how different subjective and objective measures of listening accuracy and effort respond to variation in hearing status, age, and SNR. Although speech intelligibility remains a measure of primary importance, it is a sensitive measure only under adverse listening conditions, which may not be representative of everyday listening. Under more ecologically relevant listening conditions (generally speaking, at moderate, positive SNRs), listening effort becomes a crucial factor to consider to adequately describe the listening experience. VRT may provide a useful objective marker of listening effort, but caution is required to deal with measurement variability, differences in definition, and the potentially confounding effect of age.