What the Research Says About Strength Training

Resistance Training Boosts Working Memory by 0.4 Standard Deviations in Healthy Older Adults — But Only If You Do It Unstable

Here’s the finding that should make you rethink your gym routine: a double-blinded RCT found that 10 weeks of resistance training performed on unstable surfaces (balance boards, BOSU balls) improved working memory, processing speed, and response inhibition in healthy adults aged 65–79, while the same exercises on stable machines did not. The effect size for working memory was a Cohen’s d of 0.4 — a moderate, clinically meaningful improvement. This isn’t a vague “exercise is good for your brain” claim. It’s a specific, testable intervention with a measurable outcome. For anyone running an n=1 experiment on strength training and cognition, the type of resistance training matters more than most people assume.

What the research actually shows

The evidence base for strength training and cognitive function is surprisingly robust, but you have to look past the headlines. A 2025 meta-analysis of 17 RCTs involving 739 participants aged 60+ found that resistance exercise significantly improved overall cognitive function (Hedges’ g = 0.31, 95% CI: 0.12 to 0.50), working memory (g = 0.42, 95% CI: 0.18 to 0.66), and verbal learning (g = 0.36, 95% CI: 0.07 to 0.65). These are not trivial effects — a 0.4 standard deviation improvement in working memory is roughly equivalent to reversing 3–5 years of age-related decline in that domain.

The strongest evidence comes from RCTs, not observational studies. The Otago Home-Based Strength and Balance Retraining trial (n=74, aged 70+) found that a home-based program combining resistance and balance training reduced falls by 44% to 53% and specifically improved response inhibition — the ability to suppress inappropriate responses — compared to a control group. That’s a concrete, measurable safety benefit alongside the cognitive one.

But the type of training matters. The instability resistance training study (n=68, aged 65–79) directly compared unstable-surface free-weight training to stable machine-based training. Only the unstable group showed improvements in working memory, processing speed, and response inhibition. The stable machine groups showed no significant cognitive improvements. This suggests that the cognitive benefit comes not just from loading muscles, but from the additional demand on balance, coordination, and motor planning.

For clinical populations, the effects are even more pronounced. A 6-month RCT in 86 women aged 70–80 with probable mild cognitive impairment found that twice-weekly resistance training improved conflict resolution and selective attention on the Stroop Test, as well as associative memory, compared to a balance-and-tone control group. The resistance group also showed functional brain plasticity changes on fMRI — actual changes in brain activity patterns during cognitive tasks.

The nuance most people miss

The cognitive benefits of resistance training are not universal. They depend on dose, population, and outcome measured. The meta-analysis of 17 RCTs found no significant effect of resistance training on processing speed (g = 0.16, 95% CI: -0.06 to 0.38) or executive function broadly defined (g = 0.18, 95% CI: -0.02 to 0.38). The benefits cluster in specific domains: working memory, verbal learning, and response inhibition. If you’re measuring the wrong cognitive test, you’ll see nothing.

Age also moderates the effect. The same meta-analysis found that improvements in overall cognition were larger in adults aged 60–70 compared to those over 80. This doesn’t mean it’s useless for older adults — the Otago trial showed benefits in people over 70 — but the effect may diminish with advanced age.

The dose-response relationship is unclear. The studies that worked used frequencies of 2–3 times per week, session durations of 30–60 minutes, and intensities of 60–80% of 1-rep max. But no study has systematically varied these parameters to find the optimal dose. The 4-year EPD study protocol (not yet completed) aims to test this with 48 months of multimodal training, but results aren’t available yet.

A critical confound: most studies combine resistance training with other interventions. The OptiTrain trial in breast cancer patients added HIIT to resistance training. The multimodal cancer trial included relaxation and massage. You can’t isolate the resistance component in these studies. The purest evidence comes from the instability resistance training RCT and the meta-analysis of 17 resistance-only trials.

Practical implications

• Train on unstable surfaces if you want cognitive benefits. The instability resistance training study used free weights (dumbbells, barbells) while standing on balance boards, BOSU balls, or stability cushions. This added a balance and coordination demand that stable machine training didn’t provide. If your gym doesn’t have these, try single-leg exercises or standing on a folded yoga mat.

• Aim for 2 sessions per week, 30–60 minutes each, at 60–80% of your 1-rep max. This is the dose used in the successful RCTs. The 6-month MCI trial used twice-weekly sessions. The instability training trial used 10 weeks, 3 times per week. Consistency matters more than intensity — the 4-week COPD trial showed cognitive improvements in long-term memory and reasoning with combined training, but that’s a short window.

• Measure working memory and response inhibition, not just “brain health.” The meta-analysis shows benefits in specific domains, not global cognition. Use a free online test like the N-back task for working memory or the Stroop test for response inhibition. Track your scores weekly. A positive result is an improvement of 0.3–0.4 standard deviations over 8–12 weeks — roughly a 15–20% improvement in reaction time or accuracy on these tests.

Design your own experiment

What to test: The effect of unstable-surface resistance training (using free weights on a balance board or BOSU ball) on working memory, compared to stable-machine resistance training.

Intervention and dose: 3 sessions per week for 12 weeks. Each session: 30–45 minutes of resistance exercises (squats, lunges, overhead press, rows) performed on an unstable surface. Use a load that allows 8–12 reps per set, 3 sets per exercise. For the first 2 weeks, practice balance without weights to avoid injury.

What to measure: Working memory using a dual N-back task (free apps available). Test yourself 3 times at baseline (same time of day, same conditions) to establish a reliable baseline. Then test once per week during the intervention. Also measure response inhibition using a Stroop test (also free online).

Duration: Minimum 8 weeks to see an effect. The instability training study saw results at 10 weeks. The MCI trial saw results at 6 months. If you see no change by week 8, consider increasing the instability challenge (e.g., move from a BOSU ball to a balance board, or close your eyes during exercises).

Confound to watch for: Learning effects on the cognitive tests. The N-back task improves with practice even without any intervention. To control for this, use a dual N-back (harder to learn) and compare your rate of improvement to published norms. Also track sleep quality, stress, and caffeine intake — all of which affect working memory more than most people realize.

What a positive result looks like: A 15–20% improvement in N-back accuracy (e.g., from 70% to 84% correct) and a 10–15% reduction in Stroop interference time (e.g., from 25 seconds to 21 seconds on the incongruent condition). If you see these changes while your stable-machine control period shows no change, you’ve got a real effect. If both conditions improve equally, it’s probably just practice.