A study of nearly 1,000 children has identified a specific molecular signal consistently associated with the brain structural organization that underlies mathematical and reading abilities. The finding, published July 4 in Nature Communications, bridges a gap between molecular neurochemistry and the large-scale brain architecture that supports academic learning, and points to a potential target for interventions addressing learning disabilities like dyscalculia and dyslexia.
Researchers at Stanford University, led by corresponding author Vinod Menon, professor of psychiatry and behavioral sciences, analyzed two independent cohorts of children totaling 991 participants. Rather than looking at brain activity or structure alone, the team mapped brain-wide structural phenotypes associated with academic performance onto a comprehensive PET atlas of 19 neurotransmitter receptors and transporters, the molecular machinery that allows neurons to communicate.
The result was unambiguous. Across both domains tested (mathematics and reading) and across both independent cohorts, the distribution of NMDA-type glutamatergic receptors showed the most consistent and replicable association with brain structure linked to academic abilities.
The replication statistics tell the story. For mathematical ability, the Bayesian replication factor exceeded 9 x 10⁴, extremely strong evidence. For reading, the Bayes factor was greater than 4, indicating moderate to strong replication. No other neurotransmitter system came close. Dopaminergic, cholinergic, serotonergic, and GABAergic systems all showed weaker, non-replicable associations.
“It’s the glutamatergic system that consistently maps onto the brain structures supporting these cognitive skills,” said first author Yuan Zhang, a postdoctoral researcher in Menon’s lab at the Wu Tsai Neurosciences Institute. “That wasn’t obvious before this study.”
Shared and domain-specific mechanisms
The researchers found that NMDA receptor density corresponded with multiple functional networks for mathematical abilities, a broad, distributed pattern. For reading, the associations were more spatially focused, concentrated within visual networks. This suggests that while glutamate signaling is a common molecular substrate for both skills, the brain implements each through different large-scale network configurations.
“Math seems to engage a more distributed set of networks through glutamatergic signaling, while reading is more tightly linked to visual network organization,” said co-author Hyesang Chang, also a Stanford researcher. “That makes intuitive sense, reading is fundamentally a visual-to-linguistic transformation, but the molecular specificity is new.”
The NMDA receptor is a glutamate-gated ion channel critical for synaptic plasticity and learning. It has long been studied in the context of long-term potentiation, the cellular mechanism of memory formation. But linking its regional distribution directly to academic performance in children at the whole-brain scale has not been done before.
What this means for learning disabilities
The study’s most direct implication is for children with learning disabilities. Dyscalculia (difficulty with numbers) and dyslexia (difficulty with reading) affect an estimated 5-10% of children worldwide. Current interventions are largely behavioral and educational, intensive tutoring, phonics programs, multisensory learning approaches. While these can be effective, they work at the level of behavior and instruction, not at the level of brain chemistry.
If glutamatergic signaling is the molecular lynchpin connecting brain structure to these skills, it raises the possibility of targeted pharmacological or neuromodulatory interventions. The authors are careful not to overstate this, the study is correlational, establishing a link between receptor distribution and ability, not causation. But the replication across two independent cohorts, combined with the Bayesian analytical approach, makes this one of the strongest molecular-level associations reported in the educational neuroscience literature.
The study also had negative findings worth noting. Dopamine, long suspected to play a role in learning and motivation, did not show replicable associations with brain structure for either domain. Nor did serotonin, acetylcholine, or GABA. This does not mean these systems are irrelevant to learning, they may act through different mechanisms, such as modulating attention or motivation rather than shaping structural organization, but it narrows the search space for molecular targets.
How the study was done
The team used a multi-step analytical pipeline. First, they quantified brain structural phenotypes, cortical thickness, surface area, and gray matter volume, in regions known to support academic skills. They then correlated these structural measures with standardized math and reading scores. Finally, they overlaid a PET-derived atlas of receptor and transporter distribution to ask: which neurotransmitter system best explains the relationship between brain structure and academic ability?
The two cohorts allowed for built-in replication. The first cohort served as the discovery sample; the second was used to test whether the associations held. The use of Bayesian analysis, rather than frequentist null-hypothesis testing, allowed the researchers to quantify evidence for the null hypothesis as well as the alternative, a more informative approach for complex brain-behavior relationships.
Caveats and next steps
The NMDA receptor atlas used in the study comes from adult PET imaging data. Whether the distribution of glutamatergic receptors in children’s brains matches the adult atlas is not yet known, though cross-species and developmental studies suggest receptor distributions are largely conserved after early childhood. The study also cannot distinguish whether the glutamatergic associations reflect genetic predispositions, experience-dependent plasticity, or both, a question for future longitudinal research.
The findings do, however, open a concrete molecular avenue for intervention research. Drugs that modulate NMDA receptor function already exist, including memantine (used in Alzheimer’s disease) and D-cycloserine (studied in anxiety disorders). Whether any existing or novel compounds could safely and selectively support the brain structural organization underlying math and reading in children with learning disabilities is a question the field can now begin to ask with greater precision.
Source: Zhang Y, Chang H, El-Said D, et al. Glutamatergic signaling underlies brain structural organization for mathematical and reading abilities in children. Nature Communications (2026). DOI: 10.1038/s41467-026-75102-9