The creatine-makes-you-smarter study everyone cites found an 18-point IQ jump. Nobody replicated it

The number everyone quotes, and where it came from

The claim that creatine — a supplement mostly known for building muscle — also boosts cognition traces back to a specific 2003 study: Rae et al., which tested 45 vegetarian and vegan university students (a population chosen because their diets contain essentially no dietary creatine, unlike meat-eaters). The results were dramatic: backward digit span — a working-memory test — improved by 1.19 standard deviations, roughly equivalent to an 18-point IQ gain, and performance on Raven's matrices, a nonverbal reasoning test, jumped 1.76 standard deviations, roughly 26 IQ points — both significant at p < 0.0001.

Two decades later, that single small study is still the number that shows up in supplement marketing and casual conversation whenever creatine's cognitive effects come up.

What happened when someone actually replicated it

A 2023 study by Sandkühler et al., with 123 participants — nearly three times Rae's sample — and better methodology, found the effect had all but evaporated: 0.17 standard deviations on digit span (about 2.5 IQ-equivalent points, not statistically significant) and 0.09 standard deviations on Raven's matrices (about 1 point). A 2024 meta-analysis by Xu et al. tried to synthesize the broader literature and reported modest positive effect sizes — +0.34 overall cognition, +0.32 executive function — but the article's author flags serious methodological problems with how that meta-analysis pooled its studies. Europe's food safety regulator, EFSA, reviewed ten human intervention studies in 2024 and landed on a blunter verdict

: they "do not show a consistent effect."

The pattern is a familiar one in supplement research — a small, striking early result generates years of citations and marketing before a properly powered replication quietly deflates it. Creatine's muscle-building effect is well-established and safe; its cognitive effect, on the evidence assembled here, looks like it was mostly a 45-person fluke.

"A null result isn't the same as no effect"

The thread's most substantive disagreement wasn't about creatine — it was about how to talk about null results at all. One commenter pushed back on the piece's own conclusion ("I don't know. Maybe a little."), arguing that framing understates the situation: failing to reject the null hypothesis isn't evidence of "maybe a small effect," it's evidence the study couldn't detect an effect one way or the other. Others countered that as an informal, blog-post-level summary rather than a formal statistical claim, the hedge was fair enough.

A more interesting distinction came from a reply noting creatine sits in a different category from most supplements precisely because we know, mechanistically, that supplemented creatine does reach the brain and gets used there — "which is definitely not a claim all supplements can make." That doesn't rescue the 18-point figure, but it explains why researchers keep testing this particular compound rather than dismissing it outright: there's a real biological pathway, even if the cognitive payoff on the current evidence is small enough to be statistically invisible in a well-powered study. One reply, dryly: creatine's cognitive research is starting to resemble the XKCD jellybean cartoon — test enough variables and something will clear p<0.05 eventually, whether or not it's real.