Differences in spatiotemporal brain network dynamics of Montessori and traditionally schooled students

Education is a pervasive early-life experience with potential long-term effects on cognition and adaptation via experience-dependent neural plasticity. Traditional pedagogy typically features teacher-led instruction, age-homogeneous classrooms, and evaluation focused on memory and recall, whereas Montessori pedagogy emphasizes mixed-age, student-led, trial-and-error learning with self-corrective materials and minimal external evaluative feedback. Prior behavioral studies suggest Montessori students often show advantages in academic outcomes, creativity, self-regulation, and social interaction. However, the extent to which pedagogical experience shapes underlying brain network dynamics across development remains unclear. Recent advances in network neuroscience, including spatiotemporal connectome approaches that integrate structural and time-resolved functional information, allow characterization of integration (system diversity, SD) and temporal stability (spatiotemporal diversity, STD) of brain activity. Building on developmental work showing increasing integration and stability with age, the authors asked whether schooling experience modulates these dynamics. They hypothesized that Montessori students would show higher global SD (greater integration) and lower global STD (greater neural stability) than traditionally schooled peers, particularly in control and learning-related systems including frontoparietal, somatomotor, dorsal attention, ventral attention, and cerebellar networks.

Comparative research between Montessori and traditional pedagogies has found differences in academic, creative, and socio-emotional outcomes, with Montessori students often outperforming peers and showing better self-regulation and social behaviors. Neuroimaging work indicates pedagogy-linked differences: during an fMRI math task, Montessori students showed increased activity in right parietal and frontal areas and greater ACC–frontal connectivity after errors, whereas traditional students showed greater ACC–hippocampal connectivity after correct trials, suggesting distinct learning strategies. In creativity studies, Montessori students scored higher and spent less time in an intra-connected DMN state; results suggested a modulatory role of pedagogy on the salience network linked to cognitive flexibility, potentially reflecting different balances among SN, ECN, and DMN. Developmental connectomics literature reports that from childhood to adulthood, network architecture shifts toward decreased segregation and increased integration, with maturing brains showing increased global SD and decreased global STD, consistent with experience-dependent plasticity. Spatiotemporal connectome methods provide a structural prior to functional dynamics, enabling quantification of Connected Components and derived measures SD (cross-system integration) and STD (temporal stability).

Design: Cross-sectional comparison of students with Montessori versus traditional schooling backgrounds. Participants: 100 healthy participants (ages 4–18) recruited from >20 schools (2018–2021); inclusion required consistent pedagogy since age 4 or for ≥3 years; exclusions included learning/behavioral disabilities and sensory impairments. After excluding 13 for MRI artifacts, 87 participants remained (50 females; age range 4.60–18.00 years; mean 10.02 ± 2.88; 42 Montessori). Group homogeneity measures: parental SES (questionnaire), parental pedagogy interest (questionnaire), and fluid intelligence via PM-47 Raven’s Progressive Matrices (36-item, 15 minutes). No significant between-group differences were found in age, gender ratio, SES, pedagogy interest, fluid intelligence, age distribution, or years in pedagogy; age correlated with years in schooling system. MRI acquisition: 3T Siemens Prisma-Fit with 64-channel coil. Sequences: 3D T1w MP-RAGE (1 mm isotropic), diffusion spectrum imaging (EPI, SMS=6, voxel 1.6 mm isotropic, b=0/700/1000/2000/3000 s/mm²; 10 b0 + 137 DW directions), and 6-minute resting-state fMRI (TR=500 ms, TE=33 ms, 48 slices, 2.4 mm isotropic, 720 volumes). Motion minimized with foam pads; participants fixated a cross and rested. Structural connectivity template: Anatomical and diffusion images processed with Connectome Mapper 3. Cortical surfaces and multi-scale cortical parcellation (symmetric Cammoun/Desikan), combined with hippocampal subfields, brainstem, and thalamic nuclei to create parcellations; analyses used scale with 506 GM regions. dMRI processed with constrained spherical deconvolution and deterministic tractography (SDstream) to compute individual structural connectivity matrices (506×506), averaged across subjects to form a mean structural connectome. fMRI preprocessing: fMRIPrep 20.2.1 (Nipype 1.5.1), including INU correction, motion correction, skull stripping, detrending of non-steady-state, spatial smoothing removal, and ICA-AROMA for motion artifact removal. Confounds extracted (FD, DVARS, global signals); participants exhibited acceptable motion (mean FD 0.27 ± 0.26; DVARS 1.11 ± 0.07). Mean fMRI time series extracted for each of 506 parcels. Spatiotemporal connectome: A multilayer graph captured dynamics constrained by structural connectivity. Point process analysis applied to z-scored fMRI time series using a 2 SD threshold to define regional activation events. Nodes represented parcel–time points; edges linked pairs of anatomically connected regions coactive within the same or consecutive time frames. Weakly connected components (CCs) represented transient spatiotemporal networks. CC metrics: number (amount of activity), length (temporal duration), height (spatial extent). SD and STD: Using CC spatial activation vectors labeled by functional systems (Yeo 7 systems plus cerebellar), SD quantified cross-system integration via entropy, while STD quantified temporal stability via average cosine similarity across CCs. Computation scale: SD and STD computed at global (all CCs) and functional system levels (CCs with ≥20% regions from a given system) on concatenated CCs per group; due to per-subject CC sparsity, SD/STD were computed at group level. Statistics: For CC metrics, ANCOVAs tested effects of pedagogy (fixed factor), age (covariate), and their interaction on number, length, and height. For SD/STD, significance assessed via two-tailed permutation tests with 1000 subject-label permutations; system-level results FDR-corrected; significance p<0.05.

Sample comparability: No significant between-group differences in age (U(87)=906, p=0.75), gender ratio (Chi2=0.86, p=0.35), SES (U(87)=850, p=0.78), parental pedagogy interest (U(87)=744, p=0.13), fluid intelligence (U(87)=837, p=0.45), age distribution (F(1,85)=0.147, p=0.702), or years in pedagogy (U=845, p=0.393). Years in schooling correlated with age (r(85)=0.859, p<0.001). CC metrics: Number and length of CCs showed no significant effects of age, pedagogy, or their interaction (all ps>0.211). CC height (spatial extent) decreased with age (F(1,83)=5.99, p<0.016), independent of pedagogy. Global spatiotemporal dynamics: Montessori students showed higher global SD (greater cross-system integration; p=0.017) and lower global STD (greater temporal stability; p=0.013) than traditional students. Functional systems: - SD: Significant group difference only in the cerebellar (CBL) system, with higher SD in Montessori (PFDR<0.001); no differences in VIS, SM, DA, VA, LIM, FP, DM (PFDR>0.496). - STD: Montessori students exhibited lower STD (greater stability) in dorsal attention (DA; PFDR=0.020), ventral attention (VA; PFDR<0.001), somatomotor (SM; PFDR<0.001), frontoparietal (FP; PFDR=0.020), and cerebellar (CBL; PFDR=0.020) systems; no differences in visual, limbic, or default mode systems (PFDR>0.110). Overall, Montessori schooling was associated with more integrated and temporally stable network dynamics, particularly in attention, control, sensorimotor, and cerebellar networks.

Findings indicate that schooling experience relates to differences in brain spatiotemporal network dynamics beyond demographic and cognitive factors, with Montessori students showing a more adult-like pattern characterized by greater cross-system integration (SD) and higher temporal stability (lower STD). Stability differences localized to key learning-related systems: dorsal and ventral attention networks (supporting flexible attentional control and top-down selection), somatomotor network (sensorimotor processing for action), frontoparietal control network (executive control, working memory, coordination of whole-brain interactions), and cerebellar network (motor–cognitive integration). These differences align with Montessori practices emphasizing self-directed, hands-on, uninterrupted work and trial-and-error learning, which may train attentional control, sensorimotor engagement, and executive regulation. The cerebellar findings support literature linking fine motor and cognitive development and suggest that active, movement-based learning may accelerate maturation in this network. The presence of broader system-level differences in STD than SD may reflect a developmental sequence where network stability emerges before extensive cross-system integration. Together, results suggest that educational environments can shape the maturation of dynamic functional connectivity, with implications for learning and cognitive flexibility.

This study provides preliminary evidence that pedagogy is associated with differences in resting-state brain network dynamics across development. Relative to traditionally schooled peers, Montessori students showed higher global integration (SD) and greater temporal stability (lower STD), with system-specific effects in cerebellar (integration) and attention, somatomotor, frontoparietal, and cerebellar (stability) networks. These findings contribute to understanding how long-term educational experiences may influence the maturation of dynamic functional connectivity in learning-relevant networks. Future research should employ longitudinal designs to assess causality, replicate across diverse socioeconomic and geographic contexts, and examine specific Montessori features (e.g., multisensory, movement-based learning, self-direction) and teacher characteristics that may drive these neural differences. Developing subject-level metrics enabling brain–behavior correlations would further clarify educational effects on neurodevelopment.

• Non-randomized, cross-sectional design limits causal inference and generalizability beyond the sampled population. - Potential unmeasured differences between Montessori and traditional-school families (e.g., cultural factors, openness to experience, motivation) could confound results despite matching on SES and other variables. - Participants were drawn from an upper-class-salary region; results may not generalize to middle or lower SES populations or other regions. - Teacher-related variables (training, experience) were not assessed and may influence outcomes. - SD and STD were computed at the group level (insufficient per-subject CCs), preventing individual-level analyses and correlations with behavioral measures.