What Cognitive Training Does and Does Not Improve: The Evidence on Transfer

An AI-assisted, human-verified evidence synthesis of the published scientific and regulatory literature on what computerized cognitive training ('brain training') does and does not improve, organized around the load-bearing distinction between near transfer (gains on the trained task and close variants) and far transfer (gains in general intelligence, everyday cognition, or real-world functioning). The largest single test — an online six-week randomized study of tens of thousands of adults (Owen et al., Nature 2010; 11,430 analyzed) — found large improvements on trained tasks (effect sizes ~0.72–1.63) but no transfer to untrained cognition, even for closely related tasks. The most authoritative consensus review (Simons et al., Psychological Science in the Public Interest 2016) found extensive evidence for trained-task gains, less evidence for closely related tasks, and little evidence that training improves everyday cognitive performance. Two working-memory meta-analyses (Melby-Lervåg and colleagues, 2013 and 2016; the 2016 review covering 87 publications and 145 comparisons) found reliable short-term working-memory gains (g ≈ 0.28–0.31) but no convincing far transfer to intelligence, reading, or arithmetic against active/treated controls (far-transfer g ≈ 0.05–0.06). In 2016 the U.S. Federal Trade Commission settled deceptive-advertising charges against Lumos Labs (Lumosity) for exactly this overreach. Genuinely supportive evidence exists but is bounded: the ACTIVE trial (N = 2,832 older adults; Ball et al., JAMA 2002; Rebok et al., JAGS 2014) shows durable, domain-specific gains in its reasoning and speed-of-processing arms — not broad transfer — and its downstream dementia-risk signal (Edwards et al. 2017, hazard ratio 0.71, 95% CI 0.50–0.998, p = .049) is statistically fragile, vendor-conflicted, and contested. A measurement caveat governs all of it: simply retaking a cognitive test raises the score (retest effect d ≈ 0.26; Hausknecht et al. 2007, 107 samples, 134,436 participants), so an uncontrolled before/after gain is not by itself evidence of a cognitive improvement. The defensible synthesis: cognitive training works for what you train, in the population and on the ability trained, and does not reliably generalize beyond it. This deposit is the full evidence review (the citable version of this work); a summary brief is published alongside it. Ungated and fully cited; not peer-reviewed and not medical advice. Published by Holistic Quality LLC at https://holisticquality.io/research/cognitive-training-efficacy-full .Competing interests: the author, Levi Robey, is the founder and principal of Holistic Quality LLC, the commercial publisher of this report, which develops regulator-facing safety-data and compliance products; a sibling property under the same parent entity, the Institute for Cognitive Sovereignty, and a planned cognitive-training surface may cite this report. Because this synthesis defines the evidentiary boundary for a product line the author has an interest in, the competing interest is direct and material. To mitigate it, every quantitative claim and citation was independently verified against its published source (fetch-no-memory) by verifiers blind to any product framing, the limits of the evidence are stated throughout, the genuinely-supportive evidence is represented with its boundary conditions rather than dismissed and the unsupported far-transfer claims are represented as unsupported rather than softened, and the author retained full and sole editorial control — no commercial customer or third party reviewed or influenced the content. Funding: none; produced internally by Holistic Quality LLC with no external, grant, or third-party funding.