How do neural oscillations and brain networks develop in children aged one to five?

A wearable MEG helmet for ages one to five

The study asked how two features of brain activity change between the first and fifth birthdays: the power spectrum of neural oscillations, and the functional connectivity between cortical regions. Standard scanning struggles in this age range because young children rarely stay still for a fixation-cross resting state, so the team used optically pumped magnetometer magnetoencephalography (OPM-MEG), a system built around 3D-printed helmets in a range of sizes rather than a fixed scanner.

Children sat in a magnetically shielded room wearing a size-matched helmet fitted with up to 56 OPM sensors, giving whole-head coverage of up to 128 magnetic field channels sampled at 1200 Hz. Head position was tracked continuously with infrared cameras, and each child's head was scanned in 3D so sensor positions could be related to the brain. Instead of resting quietly, children watched five minutes of standardized 'Baby Einstein' video clips, a passive movie paradigm intended to hold attention across this age range.

The analysed sample came from 136 children and 192 sessions; 163 sessions (85%) passed quality control, drawn from 111 unique participants, 72 with a single session and 39 with two or more sessions about a year apart. Ages ranged from 0.84 to 5.80 years (mean 3.14 ± 1.3). Because head motion differed by age, the team matched trial counts across age groups, leaving an average of 108 ± 31 trials; including head motion as a fixed effect did not change the significance of the reported periodic and aperiodic results.

Aperiodic slope and offset rise with age

The power spectrum of a brain region contains a periodic part, the visible oscillations at particular frequencies, and an aperiodic part, a broadband background that declines with frequency. The paper separates them with the specparam toolbox and tracks two aperiodic parameters: the slope, which describes how steeply power falls off with frequency, and the offset, which captures overall broadband power.

Both parameters increased across the 1–5 year range, but not equally everywhere. The aperiodic slope rose across the cortex, with the strongest association in temporal regions (B = 0.1, p-FDR = 8 × 10⁻⁷), and the offset rose most strongly in the left supplementary motor cortex (B = 0.08, p-FDR = 3 × 10⁻⁵). No significant sex effects appeared in either case. The authors describe the slope as linked to the balance between excitation and inhibition, with a steeper slope reflecting a lower excitation-to-inhibition ratio, and the offset as a broader marker that may track synaptic density, cortical thickness or myelination.

That mechanistic reading is the authors' interpretation, not a direct measurement. The paper also places the increase inside a non-linear trajectory: the slope and offset rise over 1–5 years, within a wider pattern that may include a decrease in infancy and the well-established decrease or flattening from mid-childhood to adolescence. The regional differences are the practical point, because they suggest the two aperiodic features index partly separable aspects of cortical maturation rather than one global developmental clock.

Periodic power shifts from theta toward alpha and beta

After removing the aperiodic component, the periodic power changed its frequency composition with age. Theta power (3–6 Hz) decreased across the brain, with the largest age effect in the left superior temporal cortex (B = −0.002, p-FDR = 2 × 10⁻¹⁰). Alpha (6–13 Hz) and beta (13–30 Hz) power increased, with peak age effects in the right motor cortex (B = 0.001, p-FDR = 4 × 10⁻⁵ and B = 0.0001, p-FDR = 5 × 10⁻⁵ respectively).

The distinction between periodic and absolute power matters here. Absolute theta power, measured without removing the aperiodic component, showed no significant age effects, so the apparent theta decline depends on separating the two components. Absolute alpha and beta power did increase with age, peaking in the right mid-motor cortex, while low gamma (30–55 Hz) and high gamma (65–80 Hz) showed no significant regional age effects.

The authors argue that MEG makes this shift more interpretable than scalp EEG, because MEG is minimally affected by the conductivity changes in the skull and scalp that can alter EEG amplitude and topography. On that reading, the frequency shift reflects neurodevelopmental processes rather than a measurement confound, and the paper adds regional detail about where it occurs.

Global connectivity falls as occipital links strengthen

Functional connectivity was estimated between all pairs of 38 cortical regions using amplitude envelope connectivity on source-estimated time series, filtered into a broadband 3–30 Hz range and into age-adjusted frequency bands, with symmetric orthogonalization to account for signal leakage. Each connection was tested against age with a random intercept for participant, and network-level error was controlled with a permutation-based network-based statistics approach using 5000 permutations.

One network increased with age and one decreased. The increasing network linked occipital regions to the posterior cingulate cortex, which the authors describe as task-relevant connections given that children were watching a video; the average connectome likewise showed the strongest relative connectivity at occipital nodes. The decreasing network was a widespread bilateral frontoparietal network, and global connectivity declined across the 1–5 year range.

The same pattern appeared in the age-adjusted bands. Theta, alpha, beta and high gamma each showed a comparable widespread frontoparietal network that decreased with age, with theta and high gamma denser in frontal connections and alpha and beta showing highly connected parietal nodes; high gamma also contained a small frontoparietal network that increased with age. The authors interpret the decline as pruning of diffuse connections in favour of more efficient, targeted pathways, consistent with structural and functional MRI findings of synaptic refinement and network segregation. That interpretation is offered as an explanation, not tested directly here.

Direction of change separates early childhood from later development

The paper's most consequential comparison is with older children. Earlier work cited in the discussion reported overall increases in connectivity from around mid-childhood, roughly 10 years old, whereas this cohort shows decreasing global connectivity between 1 and 5 years. The authors use that contrast to argue that developmental change in these signals is not linear and can reverse direction: the aperiodic slope and offset rise here but are expected to fall or flatten later, while connectivity falls here and rises later.

This is why the age range, not just the sample size, is the study's contribution. The authors describe 1–5 years as disproportionately underrepresented in neuroimaging, and they attribute the gap to poor cooperation, motion and temperament-related variability. A wearable helmet plus a passive movie paradigm addresses compliance, but it also means the measured connectivity mixes intrinsic network dynamics with stimulus-driven effects, since the paradigm cannot separate the two.

The regional results give the direction change a spatial shape. The steepest aperiodic slope increase sat in the right temporo-parietal junction, a region the authors associate with social cognition, and they connect the pattern to maturation of inhibitory signalling and to developing higher-order social and attentional processes. Behavioural measures were not included, so the link between these neural changes and cognitive improvement remains a hypothesis the authors say future work should test.

What the sample and atlas cannot resolve

Several boundaries qualify how far the results transfer. The sample was skewed toward highly educated and high-income families, with 50.4% of participants reporting a postgraduate parental degree and 79.3% reporting household income above $100,000 CAD, largely because of the city-centre recruitment site; the authors call for validation in a more economically diverse sample. Longitudinal sampling was uneven, even though participant ID was modelled as a random effect.

The methods also impose resolution limits. Source estimation used age-matched template MRIs rather than individual scans, and a relatively coarse 38-region atlas, chosen to keep source estimation robust across ages but potentially obscuring finer-grained developmental patterns. Finally, the authors note that aperiodic and oscillatory power metrics underlie multiple biological processes and cannot be attributed entirely to a single mechanism, which is why the excitation-inhibition and pruning readings should be treated as interpretations rather than measured causes.

Read the aperiodic and connectivity results as separate trajectories

For anyone comparing developmental studies, the paper's key instruction is to keep three measurements apart. The aperiodic slope and offset are not the same as band power: theta declined only after the aperiodic component was removed, while absolute theta showed no significant age effect. Global connectivity is not the same as the region-specific networks: the overall decline coexisted with strengthening occipital-to-posterior-cingulate connections. And the direction of change depends on the age window, with this 1–5 year cohort moving opposite to the increases reported from mid-childhood.

The most useful single takeaway is that early childhood is a period of regional reorganisation rather than uniform maturation. The strongest aperiodic changes clustered in temporo-parietal and motor regions, the periodic shift ran from theta toward alpha and beta, and connectivity moved from diffuse frontoparietal patterns toward more targeted occipital involvement. Any claim that these signals simply increase or decrease with age misses the paper's central comparison: the sign of the change depends on which metric, which region and which part of childhood is being measured.

Sources used in this answer