Does brain training work? What the big reviews and consensus statements say

Does brain training work? It depends on what you want to change. Learning to play a game better is one outcome. Improving a different task is another. The major reviews give these outcomes different levels of support. Daniel Simons and colleagues found extensive evidence for gains on trained tasks, less evidence for closely related tasks, and little evidence for everyday benefits in their 2016 review.

Before accepting a claim, finish the sentence: “This training improved…” The words after that matter more than the phrase “brain training.” A score increase answers a narrower question than a promise about your life.

Three outcomes to keep separate

Researchers call improvement beyond the practised task transfer. Near transfer means a gain on a closely related task. Far transfer means a gain on a substantially different task or ability. These distinctions guide the Simons review.

Here is an illustrative way to read a claim. The examples describe possible measurements, not results from freegama.

Outcome	What someone might measure	What the result would support
Practised task	Remembering longer lists in the game you trained on	Improvement in that game
Near transfer	Recalling unfamiliar lists in a separate, similar task	Improvement on that related task
Far transfer	Solving a different reasoning task after memory practice	A broader claim requiring separate evidence

Keep the outcome's name in your description. If a study measured list recall, write “list recall.” Do not quietly replace it with “memory in everyday life.”

What the 2014 statements said

The Stanford Center on Longevity and the Max Planck Institute for Human Development issued a 2014 consensus statement signed by scientists. It challenged exaggerated commercial claims. It accepted that practised skills improve, though it described the gains as small and narrow. It found little evidence that brain games improve broad abilities or everyday functioning.

The statement also explained why a score can rise without a general ability changing. Players can discover strategies suited to one task. That learning is real, but its scope matters.

The same year, more than 100 other scientists published an open letter in reply. They agreed with parts of the statement. They argued that some training programmes can improve cognitive function, including in ways that carry into daily life, and they worried readers would conclude that no training works. The experts disagreed, which is why the later reviews below matter.

What the big reviews added

Simons and colleagues published Do Brain-Training Programs Work? in Psychological Science in the Public Interest in 2016. They examined the evidence cited by brain-training companies and assessed its methods. Many studies had design or analysis weaknesses. Their review helps explain why a list of papers is insufficient: the papers must establish the advertised outcome.

Melby-Lervåg, Redick and Hulme's 2016 meta-analysis examined 87 publications and 145 experimental comparisons of working-memory training. It found immediate gains on related verbal and visuospatial memory measures. Comparisons with groups receiving another activity gave no convincing support for the broader outcomes examined, including reasoning, language, reading and maths.

Sala and colleagues, 2019, combined results from meta-analyses. They found near transfer from working-memory training, while far-transfer estimates were small or null. After accounting for comparison activities and publication bias, their overall far-transfer estimate was zero. That conclusion concerns the programmes, populations and measures included. It does not prove that every future intervention must fail.

Why the size of a benefit needs context

An effect size describes the magnitude of a measured difference. Take one concrete example. Sala and Gobet's 2020 analysis covered 41 studies of working-memory training with 2,375 typically developing children. It found small to medium gains on memory tasks. For intelligence and school measures, the estimate with active controls was 0.001, effectively none. These are standardised differences, not percentage improvements or points on your game score.

An active control gives the comparison group another activity. It helps examine whether improvement exceeds effects of taking part in a programme. Publication bias arises when positive findings are easier to find in the published record. Melby-Lervåg and colleagues examined this problem, so read a positive average alongside the bias analysis.

Before trusting a headline, use this checklist:

• What exact task improved?
• What did the comparison group do?
• How large and uncertain was the difference?
• Was it measured again after training ended?
• Does the claim match the participants and programme studied?

What freegama can honestly say

freegama offers a finite Daily 4, optional practice and records of game scores. New choices include Number Grid, Cord Garden, Bell Cove and Track the Few. Their rules describe particular activities, and none establishes transfer. We describe the tasks you practise and the results you achieve. We do not promise higher IQ, better grades, protection against decline, treatment or diagnosis. A technical skill label names a task family. It does not establish a wider benefit.

For example, Middle Arrow asks you to tap the direction of the framed middle arrow, with rules that change what an arrow means. Its score runs to 1,000, with different weights for faster and slower correct answers and penalties for mistakes. That score describes how you played under those rules. It cannot be converted into a measure of general attention.

These reviews cover specific evidence available when they were written. They do not validate freegama or establish lasting everyday benefits from our games. For the separate questions about later life and newer research, read what brain games can and cannot do for older adults.