Auditory Rhythm Encoding during the Last Trimester of Human Gestation: From Tracking the Basic Beat to Tracking Hierarchical Nested Temporal Structures

Abstract

Rhythm perception and synchronization to periodicity hold fundamental neurodevelopmental importance for language acquisition, musical behavior, and social communication. Rhythm is omnipresent in the fetal auditory world and newborns demonstrate sensitivity to auditory rhythmic cues. During the last trimester of gestation, the brain begins to respond to auditory stimulation and to code the auditory environment. When and how during this period do the neural capacities for rhythm processing develop? We conducted a cross-sectional study in 46 neonates (24 male) born between 27 and 35 weeks gestational age (wGA), measuring their neural responses to auditory rhythms with high-density electroencephalography during sleep. We developed measures to evaluate neural synchronization to nested rhythmic periodicities, including the fast isochronous beat and slower metrical (beat grouping) structures. We show that neural synchronization to beat and meter becomes stronger with increasing GA, converging on small phase differences between stimulus and neural responses near term, similar to those observed in adults. Dividing the cohort into subpopulations born before and after 33 wGA revealed that both younger and older groups showed neural synchronization to the fast periodicity related to the isochronous beat, whereas only the older group showed significant neural synchronization to the slower meter frequencies related to beat groupings, suggesting that encoding of nested periodicities arrives during late gestation. Together, our results shed light on the rapid evolution of neural coding of external hierarchical auditory rhythms during the third trimester of gestation, starting from the age when the thalamocortical axons establish the first synapses with the cortical plate.

Keywords: electroencephalography; music; neural synchronization; premature human brain.

Figure 1.

Development of neural synchronization to the rhythmic hierarchy. A, The auditory sequence. The rhythmic pattern is composed of 300 ms duration tones and rests (top). The dashed lines superimposed on the figure show the beat and metrical levels. The frequency spectra of the stimulus sound envelope are shown below the rhythmic pattern. The topographical scalp distributions of the grand average SI absolute values for both beat- (B) and meter-related (C) frequencies. The black crosses overlaid on the topographical distributions represent clusters where the coupling strength was significant: a frontocentral cluster for the beat frequency (B, p = 0.0014), a frontocentral cluster for the duple meter frequency (C, left p = 0.032) with a relatively more focalized topographical distribution, and a frontocentral cluster for the triple meter frequency (C, right, p = 0.036). D, E, Relation between gestational age and neural response to the rhythmic hierarchy. Scatter plots demonstrating significant positive correlation in the full cohort average z-scored SI values with gestational age for the beat (D, Spearman correlation, ρ = 0.37, p = 0.01), duple meter (E, left, Spearman correlation, ρ = 0.31, p = 0.03), and triple meter (E, right, Spearman correlation ρ = 0.30, p = 0.03) frequency, averaged across the detected clusters presented in (B, C). The black solid line shows a linear fit to the data.

Figure 2.

Comparing neural synchronization to beat and meter periodicities. The topographical scalp distributions of the average SI absolute values for both beat- (A) and meter-related (B) frequencies, separately averaged across the younger and the older age groups. The black crosses overlaid on the topographical distributions represent clusters where the coupling strength was significant: frontocentral clusters for the beat (A, p < 0.001), the duple meter (B, left, p = 0.043), and the triple meter (B, right, p = 0.041) frequencies for the older group (A, B, top row) and a significant frontal cluster (smaller in size compared with that of the older group) only for the beat frequency (A, p = 0.042) for the younger group (bottom row). C, D, Comparing the observed brain–stimulus SI absolute values to chance level for a single electrode (shown over the head map). The black distribution represents 1,000 surrogate SI values, with a significance threshold set at p = 0.05. For the beat frequency, coupling strength is on average above chance level for both age groups (C), while for meter-related frequencies, the older group showed above chance level responses, whereas the responses of the younger group fell below the chance level (D). Extended data on more individual electrodes are presented in Extended Data Figure 2-1.

Figure 3.

Neural oscillations coupling phase with the beat periodicity. A, Circular–linear scatter plot of the evolution of the phase of coupling at the beat frequency with gestational age on one sample electrode. Circular–linear correlation analysis between the phase of synchronization and gestational age was significant (R_c = 0.47, p = 0.007). To further visualize nonlinear circular–linear relationship, a quadratic fit is shown as a black solid curve. B, Comparing the phase distribution between the older (top) and younger (bottom) age groups. The black line illustrates the average phase across subjects within each age group. Rayleigh test verified as significant for the older group (z = 9.84, p < 0.001), while it was not significant for the younger group (p = 0.72). Extended data on more individual electrodes are presented in Extended Data Figure 3-1.