Strong Mental

Cognitive Training

Summary

Commercial brain-training reliably makes you better at the brain-training game and almost nothing else — the "far transfer" to general intelligence, memory, work, or school that the apps are sold on is, after two decades of large trials, close to a flat zero — so if your goal is a sharper, more resilient mind, your money and minutes are far better spent on the levers that actually move cognition: aerobic exercise, enough sleep, and effortfully learning a real new skill.

Why Strong

The headline near-transfer-only / far-transfer-null finding — Tier 1 (Strong):
• because: it is supported by a very large independent RCT (Owen 2010, N≈11,430), multiple independent meta-analyses converging across every training modality (Melby-Lervåg 2016; Sala & Gobet 2017-2023), a formal consensus review (Simons 2016), large cross-sectional user data (N≈60,222), and a clean mechanistic explanation for both the near gains and the placebo-driven false positives (Foroughi 2016). For a behavioural-science claim this is about as strong as negative evidence gets.
• NOT weaker because: the conclusion is robust to design quality — indeed it strengthens with better controls (effect size inversely related to design quality), the opposite of a fragile finding.

The speed-of-processing-in-older-adults exception — Tier 2 (Moderate):
• because: a large NIH-funded RCT (ACTIVE, N=2,802) shows durable trained-ability gains, everyday-function (IADL) transfer at 10 years, and a booster-dependent 20-year dementia-incidence signal (HR ≈ 0.75 with boosters) — far more than any commercial product can claim.
• NOT Tier 1 because: everyday-function transfer is partly self-reported in an unblinded no-contact-control trial; the dementia result is a long-term follow-up from a single trial lineage, partly commercially tied, not independently replicated, with the benefit confined to the booster subgroup [VERIFY endpoint pre-specification].
• NOT Tier 3 because: it is a real, large, randomised trial with durable effects and a coherent bottleneck mechanism — well past "plausibility plus anecdote."

The better levers (exercise, sleep, engaged learning) — Tier 1-2: tiered in their own home entries; cited here as the higher-yield alternatives. Exercise and sleep effects on cognition are well-replicated meta-analytically; engaged novel-skill learning (Synapse) is Tier 2 (one strong RCT, plausible mechanism).

Practical takeaway

If your goal is general cognitive sharpness or decline prevention — do NOT buy a brain-training subscription for that purpose. It is the lowest-yield option on the menu. Spend the money and (more importantly) the daily motivation on the levers below.

The high-yield cognitive-health protocol (free, evidence-led):
1. Move aerobically. Moderate-intensity aerobic exercise (brisk walk, cycle, jog, swim), >30 min, >3×/week, ongoing. This is the single best-evidenced cognitive lever you control. (See physical_exercise_mental_health; pair with resistance training for the broader health picture.)
2. Protect sleep. Treat 7-9 hours and consistent timing as a cognitive intervention, because it is one. A week of short sleep costs you more measurable cognitive function than a year of brain-training would ever add. (See sleep_foundations_for_baseline.)
3. Learn something genuinely hard and new. A language, an instrument, a craft, a complex skill — ideally social and effortful, ideally something that keeps presenting new challenge rather than plateauing. This is the real "use it or lose it." A demanding hobby beats a brain-training app on every dimension that matters.
4. Stay socially and intellectually engaged in varied, real contexts (conversation, games with other people, problem-solving that matters to you) rather than isolated drill.

If you simply enjoy the games — that's a legitimate reason, and the honest framing matters:
• Playing Lumosity/Peak/Elevate because you like them is completely fine. They are low-risk entertainment and a pleasant routine. Just hold the right expectation: you are getting better at the games, not at life. Don't pay a premium believing it's cognitive insurance.

For older adults specifically (the one place to consider structured training):
• If you want the one training type with real everyday-function evidence, it is speed-of-processing / useful-field-of-view training (the ACTIVE-lineage protocol; some clinical and commercial versions exist, e.g. BrainHQ's "Double Decision"-type tasks). The evidence suggests it can help maintain instrumental daily function, and completing booster sessions over time is associated with the dementia signal — so dose and persistence matter. Frame it honestly: a specific, modest, narrow tool — not a substitute for exercise, sleep, hearing/vision care, blood-pressure control, and social engagement, which carry the larger dementia-risk evidence (see chronic_disease_risk_mitigation).

What "working" looks like — and what it doesn't. A rising in-app score is not the marker; it only proves practice. Honest markers of cognitive health are functional and slow: holding sustained attention through a demanding task, learning the new skill you're actually trying to learn, recovering mental sharpness when you fix your sleep, sustaining energy and focus through a working day. If a product's only evidence of "working" is its own internal score, treat that as marketing, not measurement.

Evidence detail

Why This Entry Exists

A Realised user wants to "keep their brain sharp," fears age-related decline (often after watching a parent's), or has spent money on Lumosity / Peak / Elevate / CogniFit / BrainHQ and wants to know whether the rising scores mean anything. The marketing is seductive and specific: train fifteen minutes a day, watch a dashboard of "cognitive ability" climb, protect yourself against dementia. The implicit promise is that practising little games of memory, attention, and processing speed will make you a generally smarter, more focused, more decline-proof person.

This entry exists because that promise is, for the overwhelming part of what it claims, not true — and because the honest correction is genuinely useful rather than merely deflating. The science here is unusually clear for a wellness topic, and it points in two directions at once:
• The bad news the industry buries: training a narrow cognitive task improves that task and tasks very close to it ("near transfer"), but it does not generalise to untrained abilities, everyday function, work, or school ("far transfer"). This is one of the better-replicated negative findings in all of cognitive science.
• The good news the industry obscures: the things that do reliably improve cognition and protect it with age are mostly free, already in Realised's wheelhouse, and not games on a phone — moving your body, sleeping enough, and engaging effortfully with real, novel, meaningful skills and people.

What bad advice does this protect against? The belief that a brain-training subscription is a meaningful investment in cognitive health or dementia prevention. The belief that a climbing in-app score reflects a sharper mind (it reflects practice at that score). The displacement error — spending the limited time and motivation you have for "brain health" on the one intervention with near-null transfer, while the high-yield levers (exercise, sleep) go unaddressed. And the opposite over-correction: dismissing all structured cognitive activity as worthless, when learning a demanding new skill genuinely helps and one specific clinical training (speed-of-processing in older adults) has a real, if narrow, evidence base.

Evidence

Read this as three evidence bases of very different character, kept deliberately apart — mixing them is exactly the error the marketing makes.
A. The big finding — far transfer is close to null (Tier 1-strength negative)

The largest single trial — independent, enormous, unambiguous:
Owen, Hampshire, Grahn, et al. (2010). "Putting brain training to the test." Nature 465:775-778. ~11,430 participants completed a 6-week online training programme (52,617 registered), trained on reasoning, memory, attention, visuospatial tasks. Funding: UK Medical Research Council + the BBC (the trial ran as a BBC "Bang Goes the Theory" study) — government/public-broadcaster, no brain-training-industry money.
• Participants got measurably better at the specific tasks they trained on.
• No transfer to untrained cognitive tasks — "even when those tasks were cognitively closely related." The benchmark general-cognition scores of trainees were no better than controls who simply answered trivia questions online for the same period.

The meta-analytic verdict on working-memory training (the flagship "make you smarter" claim):
Melby-Lervåg, Redick & Hulme (2016). "Working Memory Training Does Not Improve Performance on Measures of Intelligence or Other Measures of 'Far Transfer'." Perspectives on Psychological Science 11(4):512-534. 87 studies, 145 comparisons. Independent/academic.
• Reliable near/intermediate transfer (you improve on other working-memory tasks immediately after training).
• No convincing far transfer to nonverbal ability, verbal ability, word decoding, reading comprehension, or arithmetic when compared against a treated (active) control — and the gains shrank or vanished as study design improved.

The cross-modality synthesis — "crystal-clear":
Sala & Gobet, and Gobet & Sala (2017-2023), incl. "Cognitive Training: A Field in Search of a Phenomenon" (Perspectives on Psychological Science, 2023), pooling 14 first-order meta-analyses — 332 samples, 1,555 effect sizes, ~21,968 participants — across working-memory training, action and non-action video games, exergames, music training, and chess training. Independent/academic.
• Verbatim: "the overall effect of far transfer is null, and there is little to no true variability between the types of cognitive training." Against active controls the far-transfer effect size is close to zero, and "effect sizes are inversely related to the quality of the experimental design" — i.e., the better the study, the smaller the effect.

The consensus review the field uses as its reference point:
Simons, Boot, Charness, Gathercole, Chabris, Hambrick & Stine-Morrow (2016). "Do 'Brain-Training' Programs Work?" Psychological Science in the Public Interest 17(3):103-186. Independent/academic.
• Verbatim conclusion: "extensive evidence that brain-training interventions improve performance on the trained tasks, less evidence that such interventions improve performance on closely related tasks, and little evidence that training enhances performance on distantly related tasks or that training improves everyday cognitive performance." And: "no evidence for broad-based improvement in cognition, academic achievement, professional performance, and social competencies that derive from the decontextualized practice of cognitive skills devoid of domain-specific content."

The real-world user check (the cleanest test of the actual claim):
Stojanoski et al. and the Owen-lab cross-sectional work; large internet-cohort analysis (total N ≈ 60,222). Independent/academic.
• People who actually use brain-training products show no cognitive advantage on working-memory, reasoning, or verbal scores over non-users. Notably, people who regularly play ordinary computer, board, and card games did score slightly higher on some measures — a hint that broad, varied, socially-embedded cognitive engagement beats decontextualised drill (consistent with section C below; observational, so confounded by who chooses each activity).

The placebo/expectation confound that explains most of the "positive" studies:
Foroughi et al. (2016). "Placebo effects in cognitive training." PNAS 113(27):7470-7474. Independent/academic.
• Recruiting people with a flyer that advertised the brain-boosting effect, then giving a single 1-hour training session, produced an apparent fluid-intelligence "gain" equivalent to a 5-10 point IQ increase — purely from expectancy, in people primed to expect improvement. Many older positive brain-training studies used passive/no-contact control groups (controls who did nothing), which cannot separate real training effects from expectancy, novelty, attention, and motivation (the Hawthorne effect). When studies use proper active controls, the transfer effects largely evaporate. This is the methodological key to the whole literature.
B. The one genuine narrow exception — speed-of-processing training in older adults (Tier 2)

This is the part the skeptics sometimes over-flatten, and Realised will not. One specific, non-commercial training — useful-field-of-view / speed-of-processing training in healthy older adults — has the strongest positive evidence in the field, and it is real.

ACTIVE trial (Advanced Cognitive Training for Independent and Vital Elderly). N = 2,802 adults aged 65+, randomised to memory, reasoning, or speed-of-processing training (~10 sessions over 5-6 weeks, some with later "booster" sessions) vs no-contact control. Funded by the US NIH/NIA (government/independent). Reports: Willis et al. 2006 (JAMA); Rebok et al. 2014 (10-year, J Am Geriatr Soc); Edwards et al. 2017 (Alz & Dementia: TRCI); Coe et al. 2026 (20-year claims-based dementia, Alz & Dementia: TRCI).
• Transfer to everyday function: at 10 years, all three training groups reported less difficulty with instrumental activities of daily living (memory ES ≈ 0.48; reasoning ≈ 0.38; speed ≈ 0.36) — a genuine far-transfer-to-real-life signal that the commercial literature lacks.
• Durable trained-ability gains: speed-of-processing training held a large effect on its targeted ability at 10 years (ES ≈ 0.66).
• The dementia signal (handle carefully): in later follow-ups, participants in the speed-of-processing arm who completed one or more booster sessions had a lower incidence of diagnosed dementia/ADRD (HR ≈ 0.75, 95% CI 0.59-0.95; ~39.8% vs ~48.7% in controls over ~20 years). Crucially, speed-trained participants with no boosters showed no benefit (HR ≈ 1.01, 95% CI 0.81-1.27). The effect is real but dose-dependent and arm-specific — it did not appear for memory or reasoning training.

Why this stays at Tier 2, not Tier 1: the everyday-function transfer rests substantially on self-reported IADL difficulty (susceptible to expectancy in an unblinded trial with a no-contact control); dementia was a long-term, claims/clinical-diagnosis follow-up rather than the original primary endpoint [VERIFY: confirm pre-specification status and ascertainment method in Edwards 2017 / Coe 2026 full text — abstracts accessed, full text paywalled]; the dementia result is from one trial lineage and has not been independently replicated; and the booster-only pattern, while biologically interpretable (consolidation/dose), also has the shape of a subgroup finding that needs confirmation. This is much more than the commercial apps can claim, and far less than "brain training prevents dementia."
C. The better levers — what actually moves cognition (Tier 1-2)

The honest pivot: the interventions with the largest, best-evidenced effects on cognition are not cognitive games.
• Aerobic exercise (Tier 1-2). Meta-analyses show moderate effects of aerobic exercise on cognition, especially executive function and attention (aerobic SMD ≈ 0.53; executive-function SMDs across analyses ~0.2-0.6; attention ~0.56), with the best returns from moderate-intensity aerobic exercise, >3×/week, >30 min/session, sustained ≥10 weeks. Effects on fluid intelligence are present but small. This is a genuinely transferable cognitive benefit — see exercise_mental_health, physical_exercise_mental_health.
• Sleep (Tier 1). Sleep restriction and deprivation measurably degrade attention, working memory, and processing speed (sustained-attention lapses g ≈ 0.78; memory-formation impairment from 3-6.5h vs 7-11h sleep g ≈ 0.29). The single fastest way to make anyone "cognitively worse" is to underslept them; conversely, restoring sleep restores the function. Sleep is the cognitive lever the brain-training dashboard never mentions — see why_sleep_matters, sleep_foundations_for_baseline.
• Effortful, novel, real-world skill learning (Tier 2). Park et al. (2014), the Synapse Project (Psychological Science, N=173 adults 60-90, independent/NIH): older adults randomised to learn a demanding new skill (digital photography, quilting) for ~15 hr/week over 3 months improved episodic memory and showed more efficient neural function vs receptive/social or placebo-task controls — benefits partly lasting a year. The active ingredient appears to be sustained, effortful new learning in a social context, not decontextualised drill. This is the honest "use it or lose it" — but it must be real learning, not a game that plateaus.

Mechanism

Why near transfer happens and far transfer doesn't. When you practise a task you get better at it through a mix of strategy discovery, task-specific automaticity, and familiarity with the stimuli and the test format. Those gains are bound to the trained task and its close relatives — they are essentially skill acquisition, the same way getting good at Tetris makes you good at Tetris. "Fluid intelligence," general processing capacity, and real-world cognitive function are not a single muscle that any hard mental task strengthens; they are substrate properties that decontextualised drill does not appear to reshape. The brain-training premise — that cognition is a generic resource you can pump up with reps like a bicep — is the part the evidence rejects.

Why the dashboard rises while nothing transfers. The climbing in-app score is near transfer made visible: you are getting better at that game, learning its quirks and your optimal strategy. The score is a faithful measure of practice and a near-useless measure of general cognitive change. This is why the most motivating feature of these products (a number going up) is also the most misleading.

Why exercise and sleep work where games don't. They act on the substrate, not the task. Aerobic exercise drives cerebral perfusion, BDNF and neurotrophic signalling, hippocampal plasticity, and acute arousal/executive-network effects — system-wide changes that show up across many untrained cognitive measures. Sleep consolidates memory, clears metabolic load, and restores attentional/prefrontal capacity; under-sleeping degrades the hardware that every cognitive task runs on. Effortful novel learning recruits broad networks under genuine challenge in a meaningful context — closer to how cognition is actually built and maintained than isolated speed drills. In Realised's terms, these are recovery-and-capacity interventions (restore and protect the system), whereas brain-training is an optimisation pitch aimed at a narrow performance number — and the optimisation pitch is the one that fails.

The honest mechanistic caveat for the speed-of-processing exception. Why might one specific older-adult training transfer to everyday function when generic working-memory training does not? The leading interpretation is that useful-field-of-view / processing speed is a rate-limiting bottleneck for many real-world tasks in ageing (driving, reacting, navigating), so improving it can ripple into instrumental function in a way that abstract working-memory span does not. This is plausible and partly evidenced — but it is a mechanism that explains a narrow transfer in a specific population, not a general endorsement of brain-training.

Risks And Contraindications

Direct physical risk: none. Tapping a screen game is harmless.

The real harms are indirect and worth naming:
• Opportunity cost / displacement. The principal harm. Time, money, and the finite daily motivation people have for "brain health" get spent on the one intervention with near-null transfer, while the high-yield levers (exercise, sleep, real learning) go unaddressed. For an older user worried about dementia, this can mean a false sense of protection that crowds out the genuinely protective behaviours.
• False reassurance about dementia. Believing a subscription is "preventing Alzheimer's" (a claim the FTC specifically sanctioned — see Controversy) can delay attention to the actually-evidenced risk factors: hypertension, hearing loss, physical inactivity, social isolation, sleep, smoking, depression.
• Cost and dark-pattern billing. These are recurring subscriptions; the FTC's Lumosity action specifically required an easy auto-renewal cancellation route after finding deceptive practices. Watch for auto-renew traps.
• Anxiety / over-monitoring. For some health-anxious users, a daily "cognitive score" becomes a source of worry and rumination rather than reassurance — the opposite of recovery.

Where there is genuinely no good evidence to claim benefit: treating brain-training as therapy for ADHD, stroke recovery, traumatic brain injury, or as dementia prevention in the general adult population. (The FTC found the dementia/ADHD/stroke claims unfounded; clinical cognitive rehabilitation is a separate, supervised context and not what consumer apps deliver.) For ADHD specifically, see mental_adhd_comprehensive — the evidenced interventions are not brain-training games.

Controversy

Nature: an inverted Realised controversy. The usual pattern is industry suppressing a cheap, unpatentable intervention. Here a multi-billion-dollar industry over-sells a thin finding, and the corrective (deflation) is the pro-consumer position — with a regulator on record.

Position A — Brain-training industry (Lumosity/Lumos Labs, Posit Science/BrainHQ, CogniFit, Elevate, Peak; commercial: subscriptions, apps, "adaptive engines," healthcare/insurer partnerships).
• Claim: training tasks improves memory, attention, processing speed, work and school performance, and protects against age-related decline and dementia.
• Best supporting evidence: real near-transfer gains; the genuine ACTIVE speed-of-processing everyday-function/dementia signal (which the industry generalises far beyond its narrow base); a 2014 open letter signed by ~133 scientists/practitioners asserting the literature shows benefits.
• Limitation: the broad claims rest heavily on passive-control studies that confound training with expectancy/Hawthorne effects, and they extrapolate one narrow older-adult finding to "everyone gets smarter." The FTC fined Lumos Labs $2 million in January 2016 for deceptive advertising — specifically for unfounded claims that the games improve work/school performance and reduce or delay cognitive impairment from age, Alzheimer's, stroke, TBI, and PTSD — and for undisclosed solicited testimonials. The commercial incentive to maximise breadth is direct and large (market estimates run from ~$3B to >$16B with double-digit projected growth).

Position B — Independent cognitive-science mainstream.
• Claim: brain-training produces near transfer (better at the trained task) but little-to-no far transfer to general cognition or everyday performance; exercise/sleep/engaged learning are the better-evidenced levers; one narrow exception (older-adult speed-of-processing) is real.
• Best supporting evidence: Owen 2010 (N≈11,430, MRC/BBC); Melby-Lervåg 2016; Sala & Gobet 2017-2023 ("null... close to zero"); Simons 2016 consensus review; Foroughi 2016 (placebo); cross-sectional user studies (N≈60,222, no advantage). A 2014 consensus statement from 70+ scientists (Stanford Center on Longevity / Max Planck) stated brain games do not provide a scientifically grounded way to improve cognition or stave off decline.
• Limitation: a strict "it's all worthless" read under-credits the genuine ACTIVE everyday-function/dementia evidence and the value of effortful learning; "no far transfer" partly reflects that decontextualised drill is the wrong vehicle, not that cognition is unimprovable.

The funding/bias dimension (points the opposite way to most Realised entries): the distortion here is commercial over-claim, not suppression. The cleanest, most-cited evidence is the non-conflicted evidence — Owen 2010 (MRC/BBC), the academic meta-analyses, Simons 2016, ACTIVE (NIH) — and it lands in the honest middle: near transfer real, far transfer null, one narrow older-adult exception genuine. The pro-claim case leans disproportionately on commercial actors and on studies with weak (passive) controls. Note one fairness point in the other direction: ACTIVE-lineage speed-of-processing research is partly tied to commercial products (BrainHQ/Posit Science), so its strongest positive evidence is not entirely conflict-free either — which is part of why it sits at Tier 2.

Realised Position: Do not buy brain-training to get smarter or to prevent dementia; the far-transfer promise it is sold on is, after the largest trials in the field, essentially zero. If you enjoy the games, play them as games. The cognitive-health budget — money, time, motivation — belongs with exercise, sleep, and effortful real-world learning, which actually move the substrate. The single honest exception is older-adult speed-of-processing training, which has real (if narrow, modest, and dose-dependent) everyday-function and dementia-signal evidence and may be worth doing alongside, never instead of, the larger levers. See What Would Change Our Mind.

INDUSTRY BIAS ANALYSIS (HIGH relevance — inverted pattern)

The value-add here is naming that the bias runs opposite to the typical Realised entry.
• The incentive is to over-sell, not suppress. Brain-training is delivered via recurring subscriptions, "adaptive engines," and healthcare/insurer partnerships. The commercial incentive favours the broadest possible claim ("smarter, sharper, dementia-protected — for everyone") and a memorable, trackable metric (your daily cognitive score) that users can be coached to improve and re-test. Breadth and a number-that-goes-up sell subscriptions; both outrun the evidence.
• The signature metric does marketing work the data don't support. The rising in-app score measures near transfer (practice at that task). It is presented as if it measures general cognitive change. It does not.
• There is a regulator on record. The FTC's 2016 Lumosity action ($2M, deceptive-advertising) is a rare formal finding that specific brain-training claims (work/school performance; protection against Alzheimer's, stroke, TBI, PTSD, ADHD) lacked adequate scientific support.
• The cleanest evidence is the non-conflicted evidence, and it deflates the broad claim: Owen 2010 (MRC/BBC), the academic meta-analyses, Simons 2016, the large cross-sectional user studies. Industry-aligned evidence leans on weak (passive) controls.
• One honest counter-note (avoid the over-correction): the absence of far transfer is a real, tested negative here — not a funding gap. Unlike most Realised topics, "no evidence" does not mean "underfunded"; the trials have been done at scale and the answer is near-zero. The genuine positive (ACTIVE speed-of-processing) is itself partly commercially tied, so Realised credits it cautiously at Tier 2 rather than amplifying it.

Cross-Pillar Connections

• Physical (exercise_mental_health, physical_exercise_mental_health): the highest-yield cognitive lever. Aerobic exercise produces moderate, transferable gains in executive function and attention — the real version of what brain-training promises. This is the primary redirect for any "keep my brain sharp" user.
• Sleep (why_sleep_matters, sleep_foundations_for_baseline): the cognitive lever the brain-training dashboard never mentions. Sleep loss reliably degrades attention, working memory, and processing speed; restoring sleep restores them. Fix sleep before considering any cognitive product.
• Mental (mental_adhd_comprehensive): brain-training is not an evidenced ADHD intervention; users seeking attention help should be routed to the evidenced options there, not to games.
• Mental (belief_effects_and_honest_framing): the Foroughi placebo finding is a clean case study in expectancy-as-effect — relevant to how Realised frames any intervention whose "evidence" is largely a felt or self-scored improvement.
• Mental (cognitive_bandwidth_as_finite_resource): reinforces the displacement argument — finite attention/motivation spent on near-null brain-training is bandwidth not spent on high-yield recovery behaviours.
• Mental (mental_mindfulness_meditation): a comparison point — like brain-training, meditation is sometimes over-sold for "cognitive enhancement"; its better-supported benefits are in stress/affect regulation, not general intelligence. Hold both to the same transfer standard.
• Cross-pillar (chronic_disease_risk_mitigation): the genuinely evidenced dementia-risk levers (blood pressure, hearing, activity, social engagement, sleep, smoking) live in this neighbourhood — the honest answer to the dementia-fear that drives many users to brain-training.

What would change our mind

Falsifiability: explicit upgrade/downgrade criteria from source

We would UPGRADE the case for commercial brain-training (toward "far transfer exists") if:
• Adequately-powered RCTs with active, expectancy-matched controls showed durable transfer of consumer brain-training to untrained cognition, everyday function, work, or school — reversing the Owen/Melby-Lervåg/Sala-Gobet pattern rather than adding more passive-control studies.
• A registered, pre-specified replication confirmed broad real-world transfer that survives correction for publication bias and placebo recruitment (the Foroughi confound).

We would UPGRADE the speed-of-processing exception (B) from Tier 2 toward Tier 1 if:
• An independent trial lineage (not ACTIVE-derived, not commercially tied) replicated both the everyday-function transfer and the booster-dependent dementia-incidence reduction, ideally with objective (not self-reported) function measures and a pre-specified dementia endpoint with active controls.

We would DOWNGRADE / further deflate if:
• The ACTIVE everyday-function and dementia signals failed to replicate, or were shown to be artefacts of self-report expectancy, differential attrition, or the no-contact control design — which would collapse the one genuine exception toward the general null.

What would NOT change our mind: more rising in-app scores (near transfer, not far), more testimonials, more passive-control studies, or more industry-funded open letters. None of these speak to the actual claim (transfer to untrained, real-world cognition under proper controls).

Sources (21)

Related entries

Belief Effects and Honest FramingStrongCognitive Bandwidth as a Finite ResourceStrongExercise for Mental HealthStrongADHD: Comprehensive Evidence ReviewStrong
HIGH that commercial brain-training does **not** produce broad "smarter" gains; MODERATE on the speed-of-processing exception; HIGH that exercise and sleep are larger, better-evidenced cognitive levers.Last updated June 2026
Open in the Library: search, filter, every entry →

We set no cookies and run no ad trackers. We count visits with Cloudflare's cookieless, privacy-first analytics. The only thing stored on your device is which example you last viewed.