Neural attentional-filter mechanisms of listening success in middle-aged and older individuals

Abstract

Successful listening crucially depends on intact attentional filters that separate relevant from irrelevant information. Research into their neurobiological implementation has focused on two potential auditory filter strategies: the lateralization of alpha power and selective neural speech tracking. However, the functional interplay of the two neural filter strategies and their potency to index listening success in an ageing population remains unclear. Using electroencephalography and a dual-talker task in a representative sample of listeners (N = 155; age=39-80 years), we here demonstrate an often-missed link from single-trial behavioural outcomes back to trial-by-trial changes in neural attentional filtering. First, we observe preserved attentional-cue-driven modulation of both neural filters across chronological age and hearing levels. Second, neural filter states vary independently of one another, demonstrating complementary neurobiological solutions of spatial selective attention. Stronger neural speech tracking but not alpha lateralization boosts trial-to-trial behavioural performance. Our results highlight the translational potential of neural speech tracking as an individualized neural marker of adaptive listening behaviour.

Fig. 1. Schematic illustration of addressed research questions.

The dichotic listening task manipulated the attentional focus and semantic predictability of upcoming input using two separate visual cues. We investigated whether informative cues would enhance behavioural performance (Q1). In line with (Q2), we also examined the degree to which a spatial (and semantic) cue modulated the two auditory neural measures of interest: neural speech tracking and lateralization of auditory alpha power. Finally, we assessed (Q3) the co-variation of neural measures, and (Q4) their potency in explaining behavioural performance. Furthermore, we investigated the impact of age, hearing loss, and probed ear on listening success and its underlying neural strategies.

Fig. 2. Experimental design and behavioural benefit from informative cues.

a Visualization of used 2 × 2 design. Levels of spatial and semantic cues differed on a trial-by-trial basis. Note that the effects of the semantic cue were of secondary importance to the current analyses. Top row shows the informative [+] cue levels, bottom row the uninformative [–] cue levels. b Schematic representation of the trial structure. Successive display of the two visual cues precedes the dichotic presentation of two sentences spoken by the same female talker. After sentence presentation, participants had to select the final word from four alternative words. c Left: accuracy per cue–cue combination. Coloured dots are individual (N = 155 participants) trial averages, black dots and vertical lines show group means with bootstrapped 95% confidence intervals (CI). Right: Individual cue benefits displayed separately for the two cues (top: spatial cue, bottom: semantic cue). Black dots indicate individual (N = 155) mean accuracy with bootstrapped 95 % CI. Histograms show the distribution of the difference in accuracy for informative vs. uninformative levels. OR: odds ratio parameter estimate from generalized linear mixed-effects models; two-sided Wald test (FDR-corrected); spatial cue: P = 1.36 × 10^–24; semantic cue: P = 0.68. d Left: Response speed per cue–cue combination. Coloured dots show individual (N = 155 participants) mean speed, black dots and vertical lines show group means with bootstrapped 95% CI. Right: Individual cue benefits displayed separately for the two cues (top: spatial cue, bottom: semantic cue). Black dots indicate individual (N = 155) mean speed with bootstrapped 95% CI. Box plots in (c) and (d) show median centre line, 25th to 75th percentile hinges, whiskers indicate minimum and maximum within 1.5 × interquartile range. β: slope parameter estimate from general linear mixed-effects models; two-sided Wald test (FDR-corrected); spatial cue: P = 4.49 × 10^–48; semantic cue: P = 2.49 × 10^–9. Source data are provided as a Source Data file.

Fig. 3. Informative spatial cue elicits increased alpha-power lateralization before and during speech presentation.

a Grand-average (N = 155 participants) whole-trial attentional modulation of 8–12 Hz auditory alpha power. Purple traces show the grand-average alpha lateralization index (ALI) for the informative (solid dark purple line) and uninformative spatial cue (dashed light purple line), each collapsed across semantic cue levels. Error bands indicate ±1 SEM. Positive values indicate relatively higher alpha power in the hemisphere ipsilateral to the attended/probed sentence compared to the contralateral hemisphere. The shaded grey area shows the time window of sentence presentation. Brain models show the auditory region of interest (red). b ALI during sentence presentation (3.5–6.5 s) shown separately per spatial-cue condition and probed ear (left plot) for N = 155 participants. Coloured dots show trial-averaged individual results, black dots and error bars indicate the grand-average and bootstrapped 95% confidence intervals. Box plots show median centre line, 25th to 75th percentile hinges; whiskers show minimum and maximum within 1.5 × interquartile range. For probed-right trials, there was a significant difference in ALI between selective- and divided-attention trials (right plot). Black dots represent individual mean ALI values with bootstrapped 95% CI error bars. Histogram shows the distribution of differences in ALI for informative vs. uninformative spatial-cue levels. β: slope parameter estimate from the corresponding general linear mixed-effects model; two-sided Wald test (FDR-corrected, ***P = 2.65 × 10⁻¹⁷). Source data are provided as a Source Data file.

Fig. 4. Neural speech tracking of attended and ignored sentences.

a Schematic representation of linear backward model approach. Linear backward models estimated on selective-attention trials. Onset envelopes are reconstructed via convolution of auditory EEG responses with estimated backward models and compared actual envelopes to assess neural tracking strength and decoding accuracy (see Supplementary Methods). b Left: grand-average (N = 155 participants, 95% confidence interval (CI) error bands) forward-transformed temporal response functions (TRFs) for attended (green) and ignored (yellow) speech in the left and right auditory ROI. Right: single-subject (N = 155 participants; 95% CI error bars) mean Pearson correlation of reconstructed and presented envelopes shown separately for attended (top, green) and ignored speech (bottom, yellow). c Top: grand-average (N = 155 participants) peri-stimulus time course of neural tracking index shown separately for selective (solid dark green curve) and divided attention (dashed light-green curve) ±1 SEM error band. Histograms show sentence and final-word onsets. The shaded area indicates the final-word presentation interval used for statistical analysis. Bottom: Single-subject (N = 155 participants) mean attended and ignored neural speech tracking during final-word presentation for selective and divided attention, respectively. d Left: neural tracking index shows per spatial-attention condition and for trials in which cued/probed sentences started ahead of (‘probed first’) or after the distractor (‘probed second’). Coloured dots represent the single-subject average (N = 155 participants), black dots and error bars indicate grand-average and bootstrapped 95% CI. Box plots show median centre line, 25th to 75th percentile hinges, whiskers indicate minimum and maximum within 1.5 × interquartile range. Right: significant difference in neural tracking between selective- and divided-attention trials in probed second trials (top plot), and stronger neural tracking in probed-left trials. Black dots represent the individual mean neural tracking index with bootstrapped 95% CI error bars for N = 155 participants. Histogram shows the distribution of differences in neural tracking in informative vs. uninformative spatial-cue trials, and probed-left vs. probed-right trials, respectively. β: slope parameter estimate from the corresponding general linear mixed-effects model; ***P = 1.22 × 10⁻⁹, *P = 0.0233 (two-sided Wald test, FDR-corrected). Source data are provided as a Source Data file.

Fig. 5. Relationship of alpha lateralization and neural speech tracking.

a Hypothesized relationship of alpha power and neural speech tracking within the auditory region of interest. Changes in alpha lateralization are assumed to drive changes in neural tracking. Schematic representation for an attend-left trial. b Independence of neural speech tracking and alpha lateralization during final-word presentation as shown by the predictions from the same linear mixed-effect model. Plots show the predicted, non-significant effect of within- and between-subject variations in alpha lateralization on selective neural tracking, respectively. Blue lines indicate the respective predicted fixed effects with 95% confidence interval, grey thin lines in the left plot show N = 155 subject-specific random slopes (included for illustrative purposes only), and grey dots show average predictions per subject. β: slope parameter estimate from the corresponding general linear mixed-effects model, two-sided Wald test (FDR-corrected). c Grand-average time courses of alpha lateralization and neural speech tracking during sentence presentation mapped to the same peri-stimulus time axis. Shown separately for selective attention (darker, solid curves) and divided-attention trials (lighter, dashed curves). Error bands reflect ±1 SEM. Note how the peak in neural speech tracking under selective attention precedes the peak in alpha lateralization. d Mean normalized cross-correlation of trial-averaged neural speech tracking and alpha lateralization time courses. The upper and lower bound of the shaded areas reflect the 97.5th and 2.5th percentile of surrogate data derived from 5000 independently permuted time courses of alpha power and neural speech tracking. Source data are provided as a Source Data file.

Fig. 6. Neural speech tracking predicts listening behaviour.

a Model predictions for the effect of neural tracking on behaviour for N = 155 participants. Left panel shows the predicted group-level fixed effect (green line ± 95% CI) of trial-to-trial variation in neural tracking on accuracy. Grey thin lines indicate estimated subject-specific slopes. Right panel shows the predicted group-level fixed effect of neural tracking at the between-subject level on response speed (green line ± 95% CI). Grey dots indicate subject-level model predictions. OR: odds ratio, β:  slope parameter estimate from the corresponding general linear mixed-effects model, two-sided Wald test (FDR-corrected). b Summary of results. Black arrows highlight statistically significant effects from (generalized) single-trial linear mixed-effects modelling. Grey arrow shows the effect of additionally modelled influences. Notably, changes in age and hearing loss did not modulate the fidelity of the two key neural measures. Source data are provided as a Source Data file.