What the Research Says
What the Exercise Research Actually Shows
The exercise literature is one of the most robust in behavioral science. Here's what consistently replicates — and where the noise drowns out the signal.
Why Exercise Research Is Unusually Reliable
Exercise science has an advantage most behavioral research doesn't: easy objective measurement. You can measure VO2 max, muscle cross-sectional area, hormone levels, and cognitive test scores before and after an intervention. This makes the exercise literature more reliable than most — though not immune to the usual problems.
What Replicates Strongly
Exercise is the most robustly supported intervention for mood. Meta-analyses covering hundreds of RCTs consistently show effect sizes for depression comparable to antidepressant medication, with fewer side effects. The effect holds across aerobic exercise, resistance training, and mixed protocols. Minimum effective dose appears to be around 150 minutes per week of moderate intensity, but even single sessions show acute mood effects lasting hours.
Resistance training produces cognitive benefits distinct from cardio. Multiple RCTs show strength training improves executive function, memory, and processing speed in adults across the age range. The mechanism appears to involve BDNF, IGF-1, and direct cerebrovascular effects. These effects are partially independent of the cardiovascular benefits of aerobic exercise — both matter.
Zone 2 cardio (conversational pace) drives metabolic adaptation. The "polarized training" literature — high volume at low intensity, with a small proportion of high-intensity work — consistently outperforms moderate-intensity training for endurance athletes. For health rather than performance, the main finding is that sustained low-intensity aerobic work drives mitochondrial biogenesis in a way that moderate-intensity does not.
Exercise timing affects sleep. Morning and afternoon exercise consistently show neutral or positive effects on sleep. Late-evening vigorous exercise raises core temperature and cortisol, delaying sleep onset — though individual response varies and some people adapt.
What the Research Is More Ambiguous On
Optimal protocols depend heavily on goal. The research is clear that something beats nothing by a large margin. But "best" varies by outcome — metabolic health, mood, cognition, longevity risk, and muscle hypertrophy each have different optimal parameters.
Supplements mostly fail to replicate in well-controlled trials. Creatine monohydrate is the notable exception — the evidence for strength and hypertrophy benefits is unusually consistent.
The Individual Variation Problem
Responsiveness to exercise varies more than most people realize. "Non-responders" to aerobic training (those who show no VO2 max improvement despite compliance) appear in most RCTs at rates of 10–30%, depending on the protocol. The same pattern appears in strength training. This means population averages tell you less about your individual response than a personal experiment would.