henhaobio
Sign in

bioRxivNeurosciencePreprint

Delta rhythm and voice familiarity organize neural entrainment in the infant brain

Isabelle RambossoniD, Damien BenisiD, Francisca Barcos MunoziD, Claire KabdeboniD, Léonardo CeravoloiD, Didier GrandjeaniD, Manuela FilippaiD

PDFDOI
In-depth full-text digest available · Jump to digest ↓

Abstract

The delta frequency band (~0.5-3.5 Hz) represents a phylogenetically conserved tempo of biological communication, and neural entrainment to this timescale would be a key mechanism through which the brain might generate temporal predictions. While entrainment at different levels has been characterized in adults, the mechanisms of auditory temporal prediction remain poorly studied across developmental trajectories, particularly regarding the roles of stimulus rhythmicity, frequency specificity, and voice familiarity. Using high-density EEG, we examined neural entrainment in thirty 6-month-old infants exposed to vocally rhythmic syllables produced by their mother or a stranger at delta (2 Hz) or theta (4 Hz) rates, as well as a non-rhythmic control. Auditory temporal regularity elicited a broad frontotemporal entrainment response. Critically, delta-band entrainment was frequency-specific across frontal and temporal clusters, with right-frontal selectivity progressively strengthening throughout the entrainment period as expected if this region actively refines a temporal model of the input rather than passively resonating to it; no significant theta-band specificity was observed. Entrainment was further shaped by voice familiarity: the left temporal cortex entrained selectively to the mother's voice, whereas the right frontal cluster responded selectively to the stranger's voice. This is consistent with an implementation in which familiarity-based processing within mid-temporal language-related regions interfaces with novelty-sensitive frontal mechanisms, and reflects the ongoing updating of internal predictive models rather than passive sensory registration. The co-occurrence of frequency-specific delta entrainment and experience-dependent hemispheric asymmetries reveals a neural functional network for hierarchical temporal predictions, shaped by early auditory experience and already in place at 6 months.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (13 Sept 2026) · CC BY 4.0

In 30 six-month-olds, 2 Hz syllable sequences selectively raised delta EEG power in right frontal and left temporal clusters; right frontal tuning grew over blocks; voice familiarity altered theta responses.

Most infant studies of prediction use visual paradigms, and it was unresolved whether 6-month-olds use delta-band timing in speech to anticipate input. Rather than measuring cortical tracking of natural speech, this study sets syllable rate against a non-rhythmic control, follows change across blocks and crosses rate with the mother's versus a stranger's voice, which the authors present as among the first tests of familiarity effects on frequency-specific entrainment this early.

Key results

  1. ShownDelta-rate (DEL, 2 Hz) sequences raised delta-band power above both theta-rate (THE, 4 Hz) and non-rhythmic (NR) sequences, but only in the right frontal (RF) and left temporal (LT) regions of interest (ROIs).

    Rhythm × frequency-band interaction χ² = 7.15–27.76 (ps ≤ .028) in RF and LT, with DEL > THE and DEL > NR for delta power (ps < .0001). Left frontal had an interaction (χ² = 30.01, p < .001), but DEL did not differ from NR (p = .4402). The right temporal, right parietal and posterior ROIs showed no tuning (ps ≥ .298). · 2. Neural entrainment is frequency-specific and…

  2. SuggestedIn the RF ROI, delta selectivity grew across successive blocks of the same condition, while the LT pattern stayed stable over the session.

    Rhythm × frequency × block-order interaction in RF χ² = 9.51, p = .009; slope contrasts DEL vs THE and DEL vs NR ps ≤ .0005. Left frontal: χ² = 11.76, p = .003, but only for DEL vs THE (p < .0001). LT: no such interaction (p = .167). Block order was recoded 1–4 within each condition, so slopes reflect within-condition change rather than session-wide habituation. · 2. Neural entrainment is frequency-specific and…; 4. Statistical analysis

  3. ShownVoice familiarity reorganized theta-band responses in opposite directions by region: the mother's voice gave a graded DEL > THE > NR ordering over left temporal sites, and the stranger's voice did so over right frontal sites.

    Voice main effect and rhythm × frequency × voice interaction in RF (χ² = 7.79, p = .005; χ² = 23.71, p < .001) and LT (χ² = 23.32, p < .001; χ² = 32.39, p < .001). LT: mother DEL > THE > NR (ps ≤ .0045); stranger only DEL above THE and NR. RF: stranger gave the full ordering (ps < .0001). Between-voice contrasts showed larger DEL-NR and DEL-THE differences for the stranger in both clusters (ps ≤ .0001). · 3. Voice familiarity shapes theta-band…

  4. ShownBoth rhythmic conditions changed spectral power relative to non-rhythmic timing across a broad set of frontal, temporal, parietal and posterior regions.

    Main effect of rhythmic structure in every ROI (χ² = 11.85–224.04, ps ≤ .003). DEL and THE each differed from NR in the left and right frontal, left and right temporal, right parietal and posterior ROIs (FDR-corrected pairwise ps ≤ .0382). · 1. Rhythmicity drives neural entrainment in the…

  5. ShownNo theta-specific tuning was found: delta-rate input raised theta-band power more than theta-rate input did, which the authors take to suggest that at 6 months delta-rate stimulation spills into the theta range rather than producing band-limited responses.

    The theta superiority criterion (THE > DEL and THE > NR) failed in all three clusters. Delta-rate exceeded theta-rate stimulation in the theta band (ps < .0001); theta-rate exceeded NR only in RF and LT (ps ≤ .0035). · 2. Neural entrainment is frequency-specific and…; 2. Right frontotemporal cluster as an early…

Methods, data and reproducibility

Infants heard 4-syllable sequences recorded from their own mother or from a stranger, at 2 Hz, at 4 Hz or with non-rhythmic timing, over 12 pseudorandomized blocks while 128-channel EEG was recorded. Baseline-corrected multitaper power in the delta (2–3 Hz) and theta (3–5 Hz) bands for each of eight scalp ROIs was modelled with Gaussian generalized linear mixed models and FDR-corrected post hoc contrasts.

  • Term-born infants recorded with high-density EEG while listening to rhythmic syllables (mean age 6.13 months, SD 0.21; 12 female) · human · n = 33 recruited; 30 analysed (3 with partial sessions)
  • Syllable stimuli (/ba/, /ga/, /da/) recorded by each infant's mother on a cell phone, plus a stranger's voice · human
  • Analysis scripts (Python EEG preprocessing pipeline and R statistical analyses) · 10.26037/yareta:kyq65sjt7vecpkjqy6mjxeix3m
Methods & tools
MNE-Python, PsychoPy v2023.2.3, EGI Geodesic Sensor Net, 128 channels (GSN-HydroCel-128 montage), ANT Neuro eego amplifier, Picard ICA, R v4.5.1, glmmTMB, emmeans, performance (R package), simr, G*Power 3.1, PSI, EPDS and IBQ questionnaires
Reagents & environment
PsychoPy v2023.2.3; R v4.5.1 with glmmTMB, emmeans, performance and simr; G*Power 3.1; MNE-Python. Analysis scripts (Python preprocessing pipeline and R statistical analyses) are archived under DOI 10.26037/yareta:kyq65sjt7vecpkjqy6mjxeix3m.
Reproducibility gaps
Raw EEG cannot be deposited under the ethics approval (BASEC No. 2022-02154); de-identified data are available on reasonable request, subject to ethics approval. The MNE-Python version is not given, and the full pipeline parameters are only in the Supporting Information.
Limitations (6)
  • With 30 infants, power is limited, especially for higher-order interactions. Post hoc power in the RF ROI was 0.50 for the rhythm × frequency model (0.88 for the voice model), and simr simulations showed limited power for rhythm-specific contrasts. · stated by the authors
  • Isolated syllables rather than continuous infant-directed speech give strict control of the rate but lower ecological validity, and may partly explain the missing theta-band specificity. · stated by the authors
  • The results rest on spectral power and are compatible with frequency tagging. Without phase coherence with the stimulus envelope or post-stimulus persistence, they do not establish true oscillatory entrainment. · stated by the authors
  • The design is cross-sectional and auditory-only: no developmental trajectory can be inferred, and the proposed domain-general predictive mechanism is untested in other senses. · stated by the authors
  • There is no concurrent behavioural measure of voice preference, so the neural familiarity effects cannot be linked to attention or learning. · stated by the authors
  • Between-voice contrasts favoured the stranger's voice for delta-versus-other differentiation in both the LT and RF clusters, which fits awkwardly with the abstract's framing that the maternal voice specifically engages left temporal cortex. · noted in this summary

Read this if you study infant auditory or speech development with EEG and want a design that crosses syllable rate (2 Hz, 4 Hz, non-rhythmic) with maternal versus stranger voice, analysed with mixed models.

Citations
1

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISRambosson '26Fernández‐M '26Leong '17Mai '19Jessen '25Ives '22
79% semantic matchNeuroscience
Isabelle Rambosson, Damien Benis, Francisca Barcos Munoz+4 more·bioRxiv·Sep 24, 2026