Does L-Theanine Add Anything to Caffeine? Part Two: The Design That Could Answer It

Abstract

Part One ended on a missing condition. The best-measured trial of the L-theanine–caffeine combination compared it only against an inert capsule, so it could not show that the theanine contributed anything.

This part covers a trial that has the missing condition. Twenty-two elite curling athletes were tested under four conditions — caffeine alone, theanine alone, both together, and placebo — with a Stroop task and scored shots on ice.

Its stated conclusion is that the combination beat everything. Its reported statistics do not support that conclusion, and the problem is arithmetic rather than interpretation.

1.The Design Is the Right One

Umut Yilmaz, Yusuf Buzdagli and colleagues at Erzurum Technical University published this double-blind randomised crossover trial in the Journal of the International Society of Sports Nutrition in 2023.

Twenty-two athletes from the Turkish national curling team attended four sessions three days apart, at the same time each morning. In each they received one of: 6 mg/kg caffeine, 6 mg/kg L-theanine, both at 6 mg/kg, or 400 mg of maltodextrin as placebo. Testing began sixty minutes later.

For a 70 kg athlete, 6 mg/kg is 420 mg of each compound. That is a large caffeine dose — above the 100 to 400 mg range in which the observational findings in our caffeine series cluster, and above what any consumer product supplies.

Each session opened with a computerised Stroop task measuring reaction time and error rate across neutral, congruent and incongruent trials. Athletes then took eight shots in each of three curling styles — guard, draw and take-out — scored one to four points by which zone the stone came to rest in.

Curling is an unusually good test bed for this question. It demands precision under sustained concentration, and a scored shot is an outcome that exists independently of the laboratory. The design is what Part One was missing: four arms permit the comparison that two arms cannot.

2.What Was Reported

On all three shot styles, the paper reports that the combination significantly outperformed placebo. On the Stroop task, it reports that the combination produced faster reaction times and lower error rates than caffeine alone, theanine alone and placebo on the neutral trials, and better performance than placebo on the congruent and incongruent trials.

The conclusion, as written: the best shooting scores and cognitive performance were achieved with the combined intake of caffeine and theanine.

Taken at face value, that is a direct answer to the question this series asks — the first one available, from a design capable of producing it.

3.The Numbers Do Not Reconcile

For each outcome the paper reports three quantities: an F statistic, a p-value, and a partial eta squared effect size. Given the degrees of freedom, any two of those determine the third. With twenty-two participants and four conditions, a repeated-measures analysis of variance has 3 and 63 degrees of freedom.

Take the draw shot. The reported F is 1.647, and the reported p is below .001. An F of 1.647 on 3 and 63 degrees of freedom corresponds to a p of roughly 0.19 — an unremarkable null result. There is no degrees-of-freedom combination this design could plausibly have under which an F of 1.647 yields p < .001.

The same outcome reports a partial eta squared of .485. To produce that effect size at these degrees of freedom, F would need to be close to 19.8. The reported value is 1.647.

The guard shot behaves the same way. Its F of 3.452 corresponds to p ≈ 0.02 and a partial eta squared near 0.14, against a reported p below .001 and a reported effect size of .842.

We are not making any allegation about the underlying data, which we have not seen and which may be perfectly sound. We are reporting something narrower and checkable: the three numbers printed for each outcome are mutually incompatible, and a reader has no way to determine which of them to believe.

4.The Conclusion Contradicts Itself

The paper's closing section states, of the individual compounds: "While it was effective to use these supplements separately, it was not more effective than taking them in combination."

Two paragraphs earlier, in the same section, it states: "Although using these supplements single was effective, using them together was not more effective."

The first sentence says the combination wins. The second says it does not. They cannot both be describing the same result, and the paper does not adjudicate between them.

A similar collision appears in the discussion of theanine alone, where a sentence reporting that theanine improved shooting scores, reaction times and accuracy is immediately followed by one stating that this predicts it does not have a positive effect.

More generally, long stretches of the discussion appear twice in near-identical wording, as though a paraphrase were left in beside the original. That is not itself a scientific fault, but a manuscript that was not read closely before publication is a manuscript whose numbers were probably not checked closely either.

5.What Survives

Within this trial, the combination did best on most measures, and the Stroop data show it beating caffeine alone on neutral reaction time and neutral error rate — the specific comparison this series is about. That is worth recording even when the statistics surrounding it are not.

But consistency of direction is a weak form of evidence, and it is not what this trial was supposed to provide. A study whose reported statistics do not reconcile cannot settle the question its design was built to settle.

The loss is specific rather than general. This is the trial with the arm Part One lacked, in a design well suited to the question, on an outcome that exists outside a laboratory. It is the study that should have answered this, and it does not.

Editorial Comment

The doses here are far above ours. Six milligrams per kilogram of each compound is roughly 420 mg apiece for a 70 kg adult, against the 250 mg of theanine and 100 mg of caffeine in Boost. Nothing in this trial describes our product, and we would not cite it as though it did.

We include it because the honest answer to whether theanine adds anything to caffeine depends on trials of exactly this shape, and because reporting only the studies that read well would misrepresent where the evidence actually stands.

The arithmetic we applied here is the arithmetic we would apply to a paper that favoured us. A finding is not more credible for pointing in the direction of our catalogue, and a reader who watched us wave this one through would be right to discount everything else in this library.

Part Three covers the third trial, which makes the most attractive claim of the three: that theanine removes what people dislike about caffeine.