Both very quiet speech and speech in noise present challenges for the listener. Understanding quiet speech is demanding due to the intrinsic weakness of the signal, which, at very low levels, can be further interfered with by internal noise. In contrast, speech in noise involves energetic masking at all signal-to-noise ratios (SNRs) and might require additional cognitive resources to discriminate the target speech from background noise. Therefore, in this study, we asked whether understanding speech in noise is more effortful than understanding quiet speech.

In the current study, participants were asked to repeat back sentences played with or without a speech-shaped noise masker (Noise vs. Quiet) while having their task-evoked pupil response (TEPR) tracked as a measure of listening effort. Each participant first went through a staircase procedure to determine either the SNRs (Noise condition) or the dB SPL thresholds (Quiet condition) corresponding to 90% and 50% accuracy. Participants then completed the main listening task, in which they were again asked to repeat back sentences presented at these individually tailored intensity levels. The design, therefore, involved Listening Condition (Noise vs. Quiet) and Difficulty Level (90% vs. 50%). Subjective ratings of tiredness, effort, and accuracy were also collected for each of these four conditions.

The current findings are based on preliminary data from 14 participants (target N = 64). As expected from the staircase procedure, average performance was at approximately 90% and 50% accuracy in the respective Difficulty Levels across the two Listening Conditions, with mean SNRs at -2.68 dB and -6.18 dB for the Noise condition, and mean SPLs at 28.88 dB and 23.44 dB for the Quiet condition (90% vs. 50% condition, respectively). TEPRs were, overall, larger in the 50% than 90% condition. Listening Condition did not appear to significantly influence TEPRs. However, this 50%-vs-90% TEPR difference was substantially larger in the Noise than Quiet condition. Subjective measures correlated more strongly with performance than with the TEPR patterns.

These preliminary results suggest that understanding speech in noise is not more effortful than understanding quiet speech. The reduced 50%-vs-90% TEPR difference in the Quiet condition indicates that perceiving quiet speech might involve consistent cognitive demands regardless of intelligibility and performance levels. Quiet speech might therefore induce a steady state of attentional vigilance largely independent of signal quality and possibly linked to a constant level of internal noise. In contrast, the TEPR pattern in the Noise condition was more closely linked to signal availability (SNR) and performance, as observed in previous studies.