Cognition governs whether neural alpha indexes speech-in-noise outcomes

EEG alpha power (8-12Hz) has been proposed as a neural marker of individual effort and attentional engagement, for example, when processing speech in natural listening conditions (e.g., background noise). Yet alpha’s behavioural relevance and reported effect directionality varies widely across studies limiting interpretability and translational potential. Drawing on resource-based models of listening, we report the first-to-date test on whether the literature’s variability primarily emerges from interactions between acoustic conditions and individual differences in cognitive capacity. Using electroencephalography, we recorded neural activity in 53 older adults with mild-to-moderate hearing loss during continuous speech comprehension under no noise, low noise (+ 8 dB SNR), and high noise (+ 3 dB SNR). First, alpha power showed non-monotonic modulation across noise levels, decreasing from no noise to low noise and rebounding under high noise. Second, alpha–behaviour coupling depended on cognitive capacity and acoustic context: under low noise, higher selective attention was associated with more positive alpha–accuracy coupling, whereas higher working memory was associated with stronger inverse alpha–effort coupling. Under high noise, these relationships were attenuated, consistent with performance becoming increasingly constrained by sensory input. Pure-tone average exerted robust main effects on accuracy and effort but did not reliably moderate alpha–behaviour coupling in post-hoc analyses. Together, these findings reconcile prior inconsistencies in the directionality of alpha effects and position alpha modulation as a capacity-dependent neural index of control deployment during effortful listening – paving the way toward its use in personalized technology and precision medicine.