19/05/2026
運動幫助大腦回春，特別是50+！
After 50, your hippocampus shrinks by 1 to 2% per year. A 12-month aerobic exercise trial in older adults reversed that. The exercise group gained 2% in hippocampal volume. The stretching controls lost 1.4%. That 3.4-percentage-point gap reverses roughly one to two years of typical age-related decline. Spatial memory improved, and the gains correlated with hippocampal volume changes.
That trial, run by Kirk Erickson's team at Pittsburgh and published in PNAS in 2011, remains one of the strongest single pieces of cognitive-aging intervention evidence we have in healthy older adults. It hasn't been overturned. It's been extended and validated in larger meta-analyses since.
A paper published this month in the Journals of Gerontology adds the mechanism we can't directly measure in humans. The team took 18-month-old mice (roughly equivalent to a person in their 50s or 60s, depending on the conversion table), induced neuroinflammation, and gave them eight weeks of moderate aerobic exercise. Hippocampal mitochondrial homeostasis recovered. Inflammation, oxidative stress, and apoptosis all fell. Memory impairment reversed.
Mouse to human translation isn't automatic. We can't measure mitochondrial dynamics in living human hippocampus, because no one is biopsying living human brain. The mechanism is inferred in humans, not proven. But the human imaging outcomes are real, replicated, and unambiguous.
This is what gets lost in cognitive aging marketing.
Brain health products line shelves. Nootropics, mushroom blends, lion's mane, NAD precursors, methylene blue. Every one is sold with mechanism claims (often plausible) and weak or absent cognitive outcome data. The single intervention with replicated structural brain outcomes in healthy older adults is exercise. It's free, requires no bottle, and takes 150 minutes a week.
Sleep is the other real competitor. The cognitive evidence for sleep quality is solid, and the structural brain effects of chronic poor sleep are well documented. Treat sleep and exercise as the two foundations, not as alternatives.
What the data actually supports:
150 minutes of moderate aerobic exercise per week, split into 3 to 4 sessions of 30 to 45 minutes. Moderate means you can hold a conversation but not sing. Heart rate around 60 to 70 percent of your maximum. Brisk walking on an incline, light jogging, cycling at a sustainable pace, swimming, rowing. The Erickson protocol used progressive walking, ramping up duration over the first weeks until participants were sustaining roughly 40-minute sessions at moderate intensity.
Vigorous exercise produces similar benefits in roughly half the time, though the evidence base is smaller in older adults specifically.
Resistance training has growing evidence for cognitive aging. The SMART trial in Sydney showed it slowed brain atrophy in adults with mild cognitive impairment. It's worth doing for several reasons, including independent mobility, bone density, and fall prevention. But the cognitive-specific evidence stack is thinner than for aerobic exercise.
What the data does not support:
That compounds or cognitive training apps substitute for exercise. Nothing in those categories has reached the structural brain outcomes Erickson demonstrated.
That casual walking counts as the trial dose. The intervention used elevated heart rate, sustained for 30 or more minutes, multiple sessions per week. Walking your dog around the block is good for you. It's not what was tested.
The bottom line:
If you're over 50 and not currently doing 150 minutes a week of moderate aerobic exercise, this is the single highest-confidence intervention with replicated structural brain outcomes for the second half of life. Sleep is the other foundation. Everything else, every compound and app and stack, sits below those two on the evidence ladder.
Start with 10 minutes of brisk walking three times this week. Build to 30 minutes. Get to 150 minutes a week by month two. That's the kind of progression that produced 2% hippocampal growth in 12 months.
Li et al., J Gerontol A, 2026
Erickson et al., PNAS, 2011
Colcombe et al., J Gerontol A, 2003
Northey et al., Br J Sports Med, 2018
Suo et al., Mol Psychiatry, 2016
