Cold water works, depending on what you measure

A cold plunge produces an unusually vivid sensation of having done something. That is not evidence, but it is a measurement of a sort, and it is the one most people are taking. The trial literature is large and fairly consistent. What it shows is that the answer turns on which outcome you count — and the outcomes cold water reliably improves are the ones you cannot blind.

What improves

The largest synthesis is a 2022 systematic review of 52 trials comparing cold-water immersion with passive recovery in physically active people [1]. After high-intensity exercise, the perceptual outcomes move.

Soreness at the same timepoint moved by SMD −0.89 (95% CI −1.48 to −0.29), also graded moderate. The 2012 Cochrane review of 17 trials and 366 participants found the same direction: soreness lower at 24 hours (SMD −0.55, 95% CI −0.84 to −0.27, 10 trials) and still lower at 96 (−0.58, −1.00 to −0.16, 5 trials) [2]. Its authors called study quality low and the results heterogeneous, but the sign has held across two decades.

What does not

Now the objective outcomes, from the same 52 trials.

The review reports no significant influence on strength recovery after either eccentric or high-intensity exercise. Muscular power does move — SMD 0.34 (0.06 to 0.62) 24 hours after eccentric exercise — but 48 hours after high-intensity exercise the estimate is −0.02 (−0.12 to 0.07), and that null is the only entry graded high certainty. The most confident finding here is a null.

The blinding problem

There is a specific reason to distrust the perceptual side. You cannot blind a person to being in cold water. The review's authors say so themselves:

A 2025 trial randomised 40 national-level male footballers aged 15 to 19 to cold-water immersion at 10 °C, hot-water immersion at 42 °C, or a sham laser on the thigh, all three presented as proven effective [3]. Nothing separated them on 20 m sprint (p = 0.90), countermovement jump (p = 0.66), maximal voluntary contraction (p = 0.13) or time to exhaustion (p = 0.18). Nineteen continued for 15 weeks, and strength rose in every group with no group effect.

A smaller crossover in 12 semi-professional footballers found jump performance held up equally under 15 minutes at 11.3 °C and under a flavoured drink presented as tart cherry juice, while creatine kinase rose the same in both [4]. Its authors read the benefit as placebo rather than reduced muscle damage.

The adaptation question

Separately, there is the worry that habitual cold exposure blunts what training is for. Roberts and colleagues had 21 active men strength-train for 12 weeks, taking either 10 minutes of cold-water immersion or active recovery after every session [5]. Lean mass rose 309 ± 73 g with active recovery and 103 ± 71 g with cold water. Leg press strength rose 201 ± 65 kg against 133 ± 43 kg (d = 1.5, p = 0.033). Type II fibre cross-sectional area rose 17.1 ± 5.1% with active recovery (p = 0.009) and did not reach significance with cold water (p = 0.10).

That is one trial of 21 people, and the meta-analysis pooling it with seven others is more equivocal than its title suggests [6]. Training alone produced hypertrophy of SMD0.5 = 0.36 (95% CrI 0.10 to 0.61); with cold water added, 0.14 (−0.08 to 0.36). The controlled comparison is smaller and less certain than either.

It crosses zero. A paper titled Throwing cold water on muscle growth reports an estimate consistent with a small attenuation and equally consistent with none. Its own wording goes no further than may attenuate, which is the correct reading and not the one the title gets used for.

What this does not establish

Both placebo-controlled trials are small — 40 and 12 participants — and entirely male, one cohort adolescent. That is not incidental: of the 52 trials pooled in 2022, 44 used male-only cohorts against 4 female-only. Whatever these numbers describe, it is mostly young men.

None of it speaks to what cold exposure is now mostly sold for — mood, stress, metabolic effects. Those are separate literatures we have not read. Nor is there a dose answer: the review's meta-regression favoured shorter, colder exposures, but that is a pattern across trials rather than a protocol anyone tested head to head.

The hypertrophy figures were read from the open-access preprint, the published version being paywalled, and we will not quote numbers from an abstract snippet. The Cochrane review is also fourteen years old, and its authors noted that most included trials did not actively survey for adverse events — the safety side here is thinner than the efficacy side, not better established.

Why this is a bad n-of-1 problem

We would normally argue that a population null is not your answer, and that a cheap reversible intervention is what a personal trial is for. Cold exposure is where that argument works against us. Track how recovered you feel and you will find it works — the perceptual outcome is the one no design protects, and unlike a capsule, a plunge has no placebo you can give yourself without knowing.

The trials at least had a sham laser and a control group; one person has neither. The outcomes that survive blinding — power output, a timed sprint, load moved — are the ones the evidence says barely move, so detecting a real effect there alone would take more sessions than the effect is worth. That is a limit on what a platform like ours can tell you, better stated than designed around.