Does L-Theanine Add Anything to Caffeine? Part One: The Best-Measured Trial, and What It Cannot Answer

Abstract

The claim that L-theanine and caffeine work better together than caffeine alone is the most commercially valuable sentence in the theanine literature. It is also the sentence our own formulation rests on, which is why it deserves auditing rather than repeating.

This series examines the three randomised trials published between 2023 and 2025 that tested the combination directly. This part covers the best-measured of them: a crossover trial in sleep-deprived adults that recorded brain electrical activity while participants responded to traffic hazards.

Its results are clear and its design is careful. It also cannot answer the question in the title of this series, and the reason it cannot is worth understanding before reading the two parts that follow.

1.The Trial

Gayani Nawarathna, Dewasmika Ariyasinghe and Tharaka Dassanayake at the University of Peradeniya in Sri Lanka published a double-blind, placebo-controlled, counterbalanced two-way crossover trial in the British Journal of Nutrition in 2025.

Thirty-seven healthy adults aged 22 to 30, twenty-one of them men, completed it. Each received either 200 mg of L-theanine with 160 mg of caffeine, or 360 mg of starch, on two occasions separated by three to seven days. The trial was registered in advance with the Sri Lanka Clinical Trial Registry.

The distinguishing feature is the state the participants were in. For seven days beforehand they slept at least six hours a night; they gave up alcohol for 48 hours and caffeine for 24; then, after a two-hour afternoon nap, they stayed awake for seventeen hours and were tested at eight in the morning.

That choice is the point. As the authors put it, people consume L-theanine and caffeine when they feel tired or sleep deprived, while most of the existing literature tests healthy volunteers under optimal conditions. This trial tested them in the condition the compounds are actually taken in.

The task was a visual oddball paradigm built from traffic scenes photographed as a driver would see them. Twenty per cent showed an imminent accident and required a button press; eighty per cent were safe scenes to be ignored. Four hundred and fifty images, fifteen minutes, with a thirty-two-channel electroencephalogram recording throughout. Participants were tested before dosing and again fifty minutes after.

2.What Changed in Behaviour

Accuracy improved on the combination and did not improve on placebo. The probability of correctly catching an accident scene rose significantly after the combination and did not move after placebo. The same held for A′, a measure of how well targets were being told apart from distractors.

Compared against each other, the combination beat placebo on hit probability and on discriminability. False alarm rates did not change under either condition, which rules out the dullest explanation — that participants simply pressed the button more often.

Reaction times improved under both conditions, and that detail is worth pausing on. Placebo produced a 13.97 ms improvement from pre-dose to post-dose; doing a task twice makes you faster at it. The combination produced 52.08 ms. The difference between the two, which is the number that means anything, was 38.1 ms with a confidence interval from 13.48 to 62.73.

A trial that measured only the post-dose session would have credited the practice effect to the capsule. The pre-post crossover structure is what prevents that, and it is the reason this design is stronger than most in the field.

3.What Changed in the Brain

The electroencephalogram was recorded to extract the P3b — a positive deflection peaking around 300 ms after a stimulus, largest over centro-parietal scalp sites, and one of the better-characterised markers of attentional processing.

Two properties were measured. Amplitude indexes how much neural resource is allocated to processing the target. Latency indexes how quickly that allocation happens.

Latency fell by around 30 ms at every one of the six scalp sites examined, significantly at all of them, while placebo changed latency at none. Amplitude rose at four of the six sites — at the vertex, for instance, by 1.92 microvolts with an interval from 0.57 to 3.26.

So the behavioural improvement has a physiological correlate recorded at the same moment, in the same participants, in the direction the behaviour would predict. That is less common than it sounds, and it is what lifts this above a reaction-time study.

One methodological note the authors make themselves: they did not adjust their p-values for multiple comparisons, on the grounds that both the comparison and the scalp sites were specified in advance from the study hypothesis. That is a defensible position and it is not the only defensible position. A reader should know it was taken.

4.The Arm That Is Missing

The trial had two conditions: the combination, and starch. There was no caffeine-alone condition and no theanine-alone condition.

This means the trial establishes that 200 mg of theanine with 160 mg of caffeine outperforms an inert capsule in sleep-deprived adults. It cannot establish that the theanine contributed anything, because the comparison that would show it was never run.

Caffeine on its own is a well-established stimulant with documented dose-dependent effects on selective attention. Everything reported above is compatible with the caffeine having done all of the work.

The authors do argue for synergy, and their argument is worth stating precisely because of how it is built. Their own earlier trial gave 200 mg of theanine alone to sleep-deprived adults and found a reaction-time improvement of about 20 ms; here the combination produced about 40. That is a comparison across two studies in two different samples, not a comparison within one.

Between-study comparisons of that kind are an argument. They are not a measurement, and they are exactly what a four-arm trial exists to replace.

This gap is not our invention. The 2025 meta-analysis of theanine as a single agent, which we covered in our theanine series, closes its discussion by raising the same question: whether theanine affects cognition only in combination with caffeine, and whether that combination is more effective than caffeine alone. The question was open when this trial was published and this trial does not close it.

5.What It Does Establish

Stated at its correct width: in acutely sleep-deprived young adults, a single dose of 200 mg L-theanine with 160 mg caffeine improved both the accuracy and the speed of responses to hazardous traffic images, and concurrent electroencephalography showed faster and larger allocation of attentional resources.

The authors list their own limits. A laboratory hazard task is not driving, which involves operational, tactical and strategic processes their paradigm does not touch. Chronic sleep deprivation may behave differently from one engineered night. And habitual users of theanine and caffeine may develop tolerance, so the findings do not transfer to people who take these compounds daily.

To which we would add: one dose, one population, one occasion, and no active comparator.

Editorial Comment

Boost contains 250 mg of L-theanine and 100 mg of caffeine. This trial used 200 mg with 160 mg — roughly 1.25 parts theanine to one part caffeine, against our 2.5 to one. It is not a test of our product, and the ratio difference is not a detail, because the whole premise of the combination is that the ratio does something.

We make no claim about driving, and we want to be unambiguous about the one inference this article must not license. The paper's own closing sentence suggests the combination may help sleep-deprived drivers respond faster to potential accidents. Someone too tired to drive safely should not drive. A capsule that shortens reaction times to photographs in a laboratory does not make an exhausted driver safe, and treating it as permission to set off would be the most dangerous possible misreading of this literature.

Part Two turns to the trial that has the arm this one lacks — a four-condition design that could, in principle, settle whether theanine adds anything at all.