Listening Over Time: Single-Trial Tonic and Phasic Oscillatory Alpha-and Theta-Band Indicators of Listening-Related Fatigue

Abstract

Objectives: Listening effort engages cognitive resources to support speech understanding in adverse listening conditions, and leads to fatigue over the longer term for people with hearing loss. Direct, neural measures of listening-related fatigue have not been developed. Here, event-related or phasic changes in alpha and theta oscillatory power during listening were used as measures of listening effort, and longer-term or tonic changes over the course of the listening task were assessed as measures of listening-related fatigue. In addition, influences of self-reported fatigue and degree of hearing loss on tonic changes in oscillatory power were examined.

Design: Participants were middle-aged adults (age 37-65 years; n = 12) with age-appropriate hearing. Sentences were presented in a background of multi-talker babble at a range of signal-to-noise ratios (SNRs) varying around the 80 percent threshold of individual listeners. Single-trial oscillatory power during both sentence and baseline intervals was analyzed with linear mixed-effect models that included as predictors trial number, SNR, subjective fatigue, and hearing loss.

Results: Alpha and theta power in both sentence presentation and baseline intervals increased as a function of trial, indicating listening-related fatigue. Further, tonic power increases across trials were affected by hearing loss and/or subjective fatigue, particularly in the alpha-band. Phasic changes in alpha and theta power generally tracked with SNR, with decreased alpha power and increased theta power at less favorable SNRs. However, for the alpha-band, the linear effect of SNR emerged only at later trials.

Conclusion: Tonic increases in oscillatory power in alpha- and theta-bands over the course of a listening task may be biomarkers for the development of listening-related fatigue. In addition, alpha-band power as an index of listening-related fatigue may be sensitive to individual differences attributable to level of hearing loss and the subjective experience of listening-related fatigue. Finally, phasic effects of SNR on alpha power emerged only after a period of listening, suggesting that this measure of listening effort could depend on the development of listening-related fatigue.

Keywords: EEG; alpha; fatigue; hearing; listening effort; oscillatory power; speech perception; theta.

FIGURE 1

Hearing levels of participants. Black line shows the mean, gray lines show individual participants.

FIGURE 2

Behavioral results. Error bars show ±1 SE, where SE is scaled to represent within-subjects variance for the repeated-measures design (Cousineau, 2005). Gray lines show individual participants. SRTn80, the signal-to-noise ratio that approximates the 80 percent speech recognition threshold for each participant.

FIGURE 3

Grand-average event-related spectral power difference from baseline. (Top left) Representative sentence waveform; (Top middle, Top right) scalp topography in the time range of 0 s before and 2.5 s after sentence onset for the alpha-band (Middle) and theta-band (Right). (Bottom) Mean event-related spectral power difference from baseline across a trial. Color scale shows power relative to baseline in decibels [10*log₁₀(average sentence interval power/average baseline interval power)] (Makeig, 1993). Time range shown on the x-axis is 0.5 s before and 3 s after sentence onset. Sentence onset is marked as time zero. White dotted and dashed lines delineate the frequency and time range analyzed for the alpha and theta bands, respectively. The mean event-related spectral power difference from baseline is collapsed across level of signal-to-noise ratio, participants, and electrodes.

FIGURE 4

Overall oscillatory power as a function of trial. Shown is the overall oscillatory power as a function of trial in the alpha-band (Top) and theta-band (Bottom) for both sentence processing (red lines) and baseline (blue lines) intervals.

FIGURE 5

Fitted oscillatory power as a function of SNR and time bin. Shown is fitted oscillatory power with error bars showing ±1 SE. Time bins cover the first to fourth quartiles of trial numbers. SRTn80, the signal-to-noise ratio that approximates the 80 percent speech recognition threshold for each participant.