The five stages of human brain development

Until recently, it was thought that the typical human brain undergoes changes in the number of neurons and their connections (synapses) that are nearly complete by about 25 years of age. From birth to about 5 years of age, about 90% of a child’s brain is developed, with rapid growth in the first few years. Typically, infants are born with approximately 100 billion neurons. However, they have relatively few, weak synapses (connections).

In the first two to three years, rapid brain development occurs. Over one million new connections form per second, often exceeding adult levels. Later, they are refined through synaptic pruning to improve efficiency. By age 2, a child's brain often has 50% more synapses than it does as an adult. The brain reaches nearly its full physical size (around 90-95% of adult volume) by this age, forming over a million new neural connections every second, highlighting the critical importance of early experiences, nutrition, and nurturing. It was thought that by about 25 years of age, brain development stabilized, with a slow degradation in old age.

However, recent research from the University of Cambridge identified five distinct, non-linear structural phases over a lifetime. These phases are defined by key turning points at ages 9, 32, 66, and 83.^1-2 They compared the brains of 3,802 people between zero and ninety years old using datasets of MRI (magnetic resonance imaging) diffusion scans, which map neural connections by tracking how water molecules move through brain tissue. A typical brain’s neural connections transition into adult mode beginning in the early thirties. This is the longest era, which lasts over three decades. This suggests that the brain's adolescence lasts much longer than previously thought, with structural maturity not fully reached until around age 32. A third turning point around age 66 marks the start of an early ageing phase of brain architecture. Finally, the late ageing brain takes shape at around 83 years old.

From infancy through childhood, the neural network in most people’s brains is gradually consolidated. Excess synapses are removed, with the more active ones surviving. Across the whole brain, connections are made from birth until about nine years old. Meanwhile, the volume of grey and white matter increases. Also, cortical thickness – the distance between outer grey matter and inner white matter – reaches a maximum. Cortical folding, the characteristic ridges on the outer brain, stabilizes.

By the nine years of age, the typical brain undergoes a change in cognitive capacity, as well as an increased risk of mental health disorders. During adolescence, the volume of white matter continues to increase. The organization of the brain’s communications networks (connectome) is increasingly refined. Connections become more efficient both within specific regions as well as rapid communication right across the whole brain. This is related to enhanced cognitive performance. On average, these developments peak in the early thirties, which is the most marked topological turning point in a typical person’s lifespan.

At about age 32, the era of adulthood begins. Brain architecture stabilizes compared to previous phases, with no major turning points for thirty years. There is also more segregation, as regions become more compartmentalized. The turning point at age 66 is far milder, and not defined by any major structural shifts, although researchers still found meaningful changes to the pattern of brain networks on average at around this age. Brain networks are reorganized up through the mid-sixties. The connectivity decreases as white matter starts to degenerate. The last turning point comes around age 83. While data is limited for this era, the defining feature is a shift from global to local, as whole-brain connectivity declines even further, with increased reliance on certain regions.

Now that researchers know more about how the brain connectome changes over time, they can learn more about its network.³ Previously, generative network models (GNMs) studied the organization of connectomes by creating synthetic networks according to simple computational rules. These models captured the connectome’s topology (the overall distributions of network modularity, small-worldness, and rich-club structure). However, they typically ignored the different rates of connectivity formation. By omitting this temporal program, GNMs often misplace topological features in physical space.

The Cambridge University researchers added a time-dependent growth term to GNMs. They also used a new model fitness function that weighed topology and topography equally. Topography is the spatial embedding of the network and the actual anatomical positions of tracts. With these advances, they were able to generate synthetic networks that more accurately reproduce the spatial layout of diffusion-MRI connectomes from two independent adult cohorts. Their simulations accurately located cortical hubs and modules and converged on a single caudal-to-rostral gradient of brain maturation.

Understanding the structural organization of the brain helps us understand how distributed neuronal activity supports complex cognitive processes. By mapping the connectome (the ensemble of white-matter tracts connecting distinct brain regions), they described the physical scaffolding that supports information flow across specialized neural circuits. In humans, this organization is typically studied by non-invasive diffusion-weighted MRI. It reveals topological features such as small-world architecture, hierarchical modularity, and rich club organization.

Complex networks—notably brain connectomes—are organized through a combination of small-world architecture (high clustering, short paths), hierarchical modularity (submodules within modules), and rich-club organization (densely connected high-degree hubs). This structure optimizes the balance between local specialization and global integration, ensuring high efficiency, robust communication, and low wiring costs. Small-world architecture by high local clustering (nodes tend to connect with neighbors) and low average path lengths between any two nodes, often created by a few long-distance hub links. It enables both efficient local processing and rapid global communication across the network. Brain networks maintain high topological clustering (similar to a regular lattice) yet possess short path lengths (similar to a random graph).

These networks are organized into modules (densely connected groups of nodes), which are further subdivided into submodules. This promotes robustness and allows for specialized functionality, where local modules handle specific tasks while interacting sparsely with other modules. This modularity supports both specialization and faster evolution, as changes in one module do not necessarily break the entire system.

In a rich-club organization, hubs are more densely connected than expected by chance, forming a central rich club. This provides a backbone for global integration, handling a large proportion of long-distance communication. For example, about 89% of short paths in the human brain pass through rich-club nodes. Rich-club hubs act as major communication hubs, but their central role also makes them critical points of vulnerability, where failure can disrupt system-wide communication.

These features are thought to foster efficient communication among distributed brain areas, enabling the emergence of complex cognitive functions. These organizational principles are often scale-invariant. Small-world properties emerge from a balance of high local connectivity (low cost) and fewer, long-range links. The rich club connects different modules within a hierarchical structure, allowing for efficient communication between specialized zones.

So, scientists integrated biologically grounded processes into GNMs. They closed the gap between developmental mechanisms and computational models of network architecture. They used mathematical equations to define spatially ordered developmental gradients over time. This improves our ability to simulate brain structure and function in health, aging, and disease. Since the brain in the human skull is just part of the neuroendocrine immune system, it also helps us maintain a healthy and balanced immune response while improving our understanding of ourselves.