Neuroplasticity

Elena, a violinist, has been practising the same difficult passage for six years. On her first attempt her fingers stumble, the bow screeches, and the notes blur. Three years later she plays that passage at full tempo without hesitation. The difference is not magic. Repeated use has changed the circuits that carry the skill: synapses have been strengthened, frequently used axons have been insulated, and the coordination among motor cortex, cerebellum, and basal ganglia has become more reliable. The same neurons are still there. The connections between them are not the same.

This is neuroplasticity in the literal sense that laboratory research has documented for decades: the brain's capacity to reorganise itself in response to experience. Knowing that fact and understanding what it implies for mental training are different things.

This chapter examines the biological processes behind that change. Not marketing slogans, and not the claims made for brain-training apps. What the evidence says, what it does not say, and what that means if mental training is going to be used as a tool for sustained growth.

The Mechanisms of Change

For much of the twentieth century the adult brain was treated as largely fixed after childhood. That view is wrong. The brain remains plastic throughout life, though the mechanisms and the scale of change differ from the rapid developmental plasticity of early years.

Three processes mediate adult plasticity, and they work together. Hebbian learning strengthens synapses that are repeatedly activated together and weakens those that are not. Experience-dependent myelination wraps frequently used axons in fatty sheaths that speed signal transmission, which is one reason practised skills become more automatic. Adult neurogenesis, the birth of new neurons, is well established in the rodent hippocampal dentate gyrus and is increased by exercise, enrichment, sleep, and lower stress. Whether new neurons continue to be added to the adult human hippocampus is still contested, and that dispute is taken up below.

These mechanisms shift the focus from quick fixes to sustained engagement. The brain does not reorganise because an app was downloaded or a chapter was read. It changes when deliberate, repeated practice continues over weeks and months. The neuroscience supports that conclusion. The commercial brain-training industry often does not.

Learning by Repetition: How Synapses Strengthen

The phrase "neurons that fire together, wire together" comes from Donald Hebb's 1949 book The Organization of Behavior. When two neurons are repeatedly activated together, the connection between them strengthens. The cellular process is long-term potentiation, one of the best-documented phenomena in neuroscience.

Long-term potentiation occurs at the synapse. Repeated, intense firing releases glutamate into the synaptic cleft. Glutamate binds to NMDA and AMPA receptors on the receiving neuron and makes that neuron more responsive to the same input later. The connection has been strengthened. This has been observed at the microscopic level across decades and multiple species.

The opposite process, long-term depression, weakens connections that are rarely used. Without it the network would accumulate irrelevant links and become inefficient. Potentiation and depression are both required. One reinforces what is used. The other prunes what is not.

Hebbian plasticity is activity-dependent and input-specific. It strengthens only the connections that are actively used. Left unchecked, that process would produce runaway excitation: every strengthened connection would make the next strengthening easier, until the network fired uncontrollably. Homeostatic plasticity is the counterbalance.

Homeostatic changes operate on a different timescale and a different principle. Where Hebbian changes are input-specific, homeostatic changes are largely cell-wide. Synaptic scaling adjusts the strength of a neuron's synapses to keep overall activity stable. Intrinsic excitability changes alter how easily a neuron fires. A related idea, the Bienenstock-Cooper-Munro model from 1982, describes how the threshold for potentiation or depression can itself shift so that the circuit remains adaptable without becoming unstable. Turrigiano and Nelson (2004) reviewed the evidence that these homeostatic mechanisms prevent Hebbian learning from destabilising developing circuits. Together, the two forms of plasticity allow the brain to learn from experience while remaining stable. For mental training, that means practice can strengthen specific connections without collapsing the network. It also means that vague or inconsistent practice may produce little change. The brain needs clear, repeated signals.

A language learner meeting the Spanish word biblioteca for the first time has a thin link between that word and the concept of a library. Each review strengthens the link. Weeks without review allow depression to weaken it. Spaced repetition works because reviews are timed near the point where the connection would otherwise fade. That is Hebbian learning at the scale of ordinary study.

Myelination: Faster Pathways for Practised Skills

If Hebbian learning strengthens connections, myelination is one way the brain makes frequently used connections faster.

Myelin is produced by oligodendrocytes in the central nervous system and by Schwann cells in the peripheral nervous system. It wraps axons, the projections that carry electrical signals between neurons. An unmyelinated axon leaks current. Myelin insulates it, so the action potential is regenerated only at the nodes of Ranvier and the impulse effectively jumps from node to node. Unmyelinated axons often conduct at about 0.5 to 10 metres per second. Large myelinated axons can reach roughly 100 to 150 metres per second. The hundred-fold figure sometimes quoted is an upper-bound comparison, not a constant. The difference is still large enough to matter for timed motor skills.

Myelination is experience-dependent. The brain does not pre-build myelin for skills it will never use. Repeated activation prompts oligodendrocytes to wrap frequently used axons. That is part of how a skill becomes automatic. A new driver checks mirrors, changes gear, and steers with deliberate attention, recruiting motor cortex, visual cortex, and prefrontal cortex together. After months of driving, the same actions need less conscious supervision. Myelination is one contributor. So are synaptic strengthening and a shift of control toward cerebellar and basal ganglia circuits. Attributing the whole change to myelin alone is too simple.

Myelin formation is slow. A single long practice session will not wrap an axon. Weeks and months of regular, focused practice will. Prolonged disuse works in the other direction: without the signals that maintain the sheath, conduction slows again. Consistency matters more than occasional intensity for this reason.

The gap between a novice guitarist and a practised player is not a mystery, but it is also not myelin by itself. The novice places each finger deliberately. The practised player has thousands of hours of corrected repetition behind the movement. Synapses, maps, and myelin have all changed. Conscious supervision is no longer required for every note.

BDNF, Exercise, and New Brain Cells

The claim that the adult brain can generate new neurons overturned a long-standing assumption. In rodents, neurogenesis continues in the hippocampal dentate gyrus, a region important for certain forms of memory, at a much lower rate than in early development. Aerobic exercise is the most reliable behavioural driver of that process in animals, largely through brain-derived neurotrophic factor (BDNF), a protein that supports the survival and integration of new neurons (Cotman, Berchtold, & Christie, 2007).

In humans the picture is less settled. Some studies report immature neurons in the adult dentate gyrus. Sorrells and colleagues (2018) examined hippocampal tissue across the lifespan and found that neurogenesis dropped sharply after early childhood, with little evidence in adults. Other groups, using different markers and tissue handling, report persistence into old age. The disagreement is methodological as well as biological. A chapter on mental training should not treat lifelong human neurogenesis as established fact. What is more secure is that exercise raises BDNF, supports hippocampal volume, and improves several cognitive outcomes in older adults. Those effects do not require the stronger claim that new neurons are being added.

Erickson and colleagues (2011) randomised 120 older adults to a year of aerobic walking or to stretching and toning. Hippocampal volume rose by about 2 percent in the walking group (2.12 percent on the left, 1.97 percent on the right) and fell by about 1.4 percent in the stretching group. Serum BDNF rose with the volume increase. Both groups improved on a spatial memory task, and the group difference in memory was not statistically significant. A commentary in the same journal pointed this out: the title and abstract say exercise improved memory, but the data show a volume effect and a within-group association, not a clear between-group memory advantage. The volume result is the solid finding. The memory claim in the original abstract overreaches.

A 2024 paper by Mohamed Hesham Khalil proposes a "BDNF-interactive model" in which physical activity, cognitive stimulation, and mindfulness regulate growth factors together, and argues that the built environment can support those behaviours. It is a theoretical proposal, not a confirmed extension of the Erickson trial. The broader idea is still useful. Exercise, sleep, and lower chronic stress make cognitive practice more likely to stick. Treating movement as a distraction from mental training cuts against the evidence.

What the Imaging Studies Show

A meta-analysis by Li and colleagues (2025) pooled 24 neuroimaging studies of structured cognitive training in older adults. Training produced a moderate improvement in cognitive function (Hedges' g = 0.38) and increased task-related activation in the left inferior frontal gyrus, a region involved in executive control, and in the bilateral precuneus. Only precuneus activation correlated directly with the cognitive gains. Age moderated the effect. Training type and baseline cognitive status did not.

Reviews of mindfulness and combined cognitive interventions report both functional and structural changes. Functional changes, meaning altered activation during a task, tend to appear sooner than structural changes such as cortical thickness or grey-matter volume. A practitioner may notice better focus within weeks. Structural changes visible on MRI usually take months.

That timeline should set expectations. Studies that document meaningful change follow people for months, not days. The changes are real. They are gradual.

The Brain-Training App Problem

If plasticity is robust, the controversy around brain-training apps needs an explanation. The gap is between what the studies show and what the marketing claims.

Melby-Lervåg, Redick, and Hulme (2016) meta-analysed working-memory training studies with control groups. Training improved performance on tasks close to the ones practised. When the comparison was an active control, far transfer to intelligence, reading, or arithmetic was not supported, and improvement on the trained measure did not predict far-transfer effects. Simons and colleagues (2016) reviewed the studies cited by commercial brain-training companies and reached the same split: strong evidence for improvement on the trained tasks, weaker evidence for closely related tasks, and little evidence for distant tasks or everyday cognitive performance. Many of the company-cited studies failed basic design standards.

Near transfer that survives an active control group is small. Far transfer to fluid intelligence, academic achievement, or general mental ability is not supported. Practice effects, small samples, and publication bias inflate the positive literature. People get better at the games they play. That improvement does not generalise in the way the advertising suggests.

Lumosity and Elevate have both been criticised on these grounds. The pattern in the independent reviews is consistent: users improve on the trained games, and untrained cognitive tasks do not move with them. Targeted practice in a real domain can still produce real gains. What the evidence does not support is the jump from a narrow game to "brain fitness" in general.

Deliberate Practice and the Architecture of Expertise

Ericsson, Krampe, and Tesch-Römer (1993) studied violinists at a Berlin music academy. By age 20, the highest-rated group had accumulated about 10,000 hours of deliberate practice, a middle group about 7,500, and a group headed for teaching rather than performance about 5,000. The study argued that expert performance tracked structured practice more than a simple talent account.

Deliberate practice, in their definition, has clear goals, immediate feedback, tasks set just beyond current ability, full attention, and repetition with correction. It is not the same as time spent playing.

Macnamara, Hambrick, and Oswald (2014) meta-analysed the deliberate-practice literature. Practice explained about 12 percent of the variance in performance overall (meta-analytic r ≈ 0.35). The figure was 26 percent for games, 21 percent for music, 18 percent for sports, 4 percent for education, and under 1 percent for professions. Practice is the most reliable contributor that has been measured. It is not most of the variance. Working memory, motivation, opportunity, instruction, and other individual differences account for the rest. The popular reading of Ericsson, that 10,000 hours is a sufficient rule, does not survive this analysis.

For mental training the practical point is narrower. Thirty minutes of focused, corrected practice will change more than three hours of casual engagement. The hours still matter. The quality of those hours matters more, and even high-quality hours leave most of the difference between people unexplained.

Practical Engagements

1. Practice log with a quality rating (2 to 4 weeks)
For each session, record what was practised, how long, focus from 1 to 10, effort from 1 to 10, and energy before the session from 1 to 10. Note any change in performance, however small. After a few weeks, look for patterns. Longer sessions do not always predict better outcomes. Focus and pre-session energy often do.

2. Pre- and post-session check (about 5 minutes)
Before the session, note whether you are alert, fatigued, anxious, or settled. Afterwards, rate focus, effort, perceived improvement, and how you feel compared with the start. Individual days are noisy. Trends over weeks are the useful signal.

3. Two-domain comparison (2 to 4 weeks)
Choose two kinds of practice, such as a memory task and an attention drill, and run both for a few weeks. Then compare that period with a stretch of single-domain practice. This is not a controlled experiment. It is a way to notice whether one form of practice makes the other easier to sustain, or whether the combination simply keeps engagement higher.

4. Skill-tracking over weeks (4 to 8 weeks)
Pick one skill: a passage, a vocabulary set, a calculation method. Attempt it, practise the weakest part for about 10 minutes, and attempt it again. Repeat across weeks and record the attempts. A narrowing gap between the first and second attempt is a performance change. It is not evidence of myelination. Synaptic change, strategy, and attention can produce the same pattern. The value of the log is empirical awareness of practice, not a window onto myelin.

Limitations

Plasticity is real and it is limited. Most studies that find meaningful change follow people for months. Response varies with age, sleep, stress, baseline ability, and genetic differences. A rested 25-year-old and a sleep-deprived 65-year-old are not the same experimental system.

Most imaging studies show co-occurrence, not a demonstrated mechanism. A region that is more active after training has not, by that fact alone, been shown to cause the behavioural gain. Sample sizes are often small, and publication bias pushes published effects upward. Meta-analyses such as Li et al. (2025) reduce that problem. They do not remove it.

Brain-training apps have been oversold. Far transfer remains weak. Domain-specific practice improves similar tasks. It does not raise general intelligence. Gains also regress without continued use. There is no permanent installation.

Human adult neurogenesis should be treated as an open question. Exercise, BDNF, and hippocampal volume should not be collapsed into a single story about new neurons.

Connections to Other Chapters

An earlier chapter introduced plasticity as the claim that the adult brain remains responsive to experience. This chapter supplies the mechanisms, and the limits on those mechanisms: Hebbian change balanced by homeostasis, myelination as one contributor to automaticity, and exercise as a support for hippocampal health rather than a proven source of new human neurons.

A later chapter on personal regimens is where the practical design belongs: how to structure sessions, how to balance kinds of practice, and how to set expectations that match the timelines above. Exercise, sleep, and stress management are not competing priorities. They are conditions under which cognitive practice is more likely to produce durable change.