L-Theanine

• Where an effect appears it is on attention: attention switching, digit vigilance, visual information processing, choice reaction time. Nothing in this literature concerns memory, recall, learning or reasoning.
• The effects are acute. Almost every measurement was taken in the first or second hour after a single dose, matching the pharmacokinetics — plasma peaks between 30 minutes and two hours.
• The magnitude is fifteen to twenty-seven milliseconds on reaction time and one to two percentage points on accuracy.
• Effects appear at supplement doses, not tea doses. The lowest dose in any pooled analysis was 50 mg; no included study delivered the intervention as an actual tea beverage.

• Whether an effect of this size has any consequence outside a laboratory. The authors of the meta-analysis explicitly decline to claim it does.
• What daily intake does over weeks. That literature barely exists — one four-week crossover, with a manufacturer among the authors.
• Why simple reaction time deteriorated at the highest dose tested. Mátyus and colleagues state there is currently insufficient literature to explain it.
• Whether the effect exists at tea-equivalent doses in free-living people, which the authors identify as the field's most useful next step.
• How three dose-graded attention trials published between 2022 and 2025 in Nutritional Neuroscience bear on this claim. They test precisely the dose-response question at issue and none is available in full text; we do not write from abstracts, so they are recorded and left unappraised. This is the single largest gap in this assessment.

• Mátyus and colleagues (2025) conducted what they describe as the first meta-analysis of L-theanine as monotherapy for cognition: 2,301 records screened, five randomised placebo-controlled trials in 148 healthy adults, doses 50–500 mg, protocol registered in PROSPERO before the work began with no amendments. The significant pooled result was on rapid visual information processing and recognition visual reaction time: −15.20 ms (CI −28.99 to −1.41), and only at higher doses.
• Payne and colleagues (2025) reviewed 50 randomised trials and pooled 15. For L-theanine alone, choice reaction time in the first hour reached significance: SMD −0.35 (CI −0.61 to −0.10), about 18 ms.
• Hidese and colleagues (2019) ran a randomised double-blind placebo-controlled crossover of 200 mg daily for four weeks in 30 healthy adults, reporting improvement in verbal fluency — letter fluency in particular — and executive function.
• Baba and colleagues (2021) randomised Japanese adults aged 50–69 with self-assessed cognitive decline to 100.6 mg of L-theanine daily or placebo, double-blind, for 12 weeks with an additional single-dose assessment. This is the lowest dose in this claim's evidence and the only trial here in an older population.

• Mátyus found nothing on simple reaction time (−0.46 ms, CI −15.65 to 14.73), nothing on the Stroop test for congruent stimuli (−37.38 ms, CI −86.39 to 11.62), and a non-significant change in the wrong direction for incongruent stimuli (109.28 ms, CI −8.72 to 227.27).
• At the highest dose tested, 400 mg, simple reaction time was non-significantly worse than placebo — a mean difference of 5.91 ms.
• Mátyus's own summary is that the superiority of L-theanine over placebo in enhancing cognition is not proven, with GRADE certainty rated low or very low across all four outcomes, citing small samples, inconsistency and imprecision.
• Inside Sohail's 2021 review, participants given L-theanine alone in Haskell's trial reported more headache and completed fewer correct serial subtractions than on placebo — a negative result for the isolated compound sitting inside a review whose conclusion is positive about the pairing. Source verification of this trial's own record is outstanding.
• In the ADHD study within the same review, L-theanine alone made inhibitory control worse: stop-signal reaction times lengthened, p = 0.053.
• In Baba's trial one of the two pre-specified primary endpoints, MMSE-J, showed no significant difference between groups either before or after the intervention. The reported benefits are specific subtests within the Cognitrax battery, which raises multiplicity: MHES §15 requires that a study whose primary endpoint fails while selected secondary measures succeed is not represented as unambiguously positive.

• L-theanine crosses the blood–brain barrier within about 30 minutes and peaks in plasma between 30 minutes and two hours, which is why the trials measure acutely and why the acute framing is appropriate rather than convenient.
• MHES §20 applies: pharmacokinetic plausibility establishes biological activity and does not establish improved attention. It is listed here as context, not as support.

Poor, and stated by the people best placed to know it. In Payne's risk-of-bias assessment, zero studies were judged at low risk, 12 raised some concerns and 25 were judged at high risk. Payne's own funding is disclosed: the lead author's doctoral studentship is funded by Lipton Teas and Infusions, one co-author is a Lipton employee and another an employee of Unilever. Hidese's trial has two authors employed by the Nutrition Division of Taiyo Kagaku, which manufactures the branded L-theanine used across this literature. Formal per-study RoB 2 domain assessment has not been performed by us and is outstanding.

The one significant pooled effect is 15.20 ms with a confidence interval reaching −1.41 — that is, an upper bound close to no effect at all. Payne's authors state the position directly: the confidence intervals frequently highlighted the uncertainty surrounding the direction and magnitude of these differences. A second illustration of fragility comes from the same paper: an earlier meta-analysis had found a significant alertness effect only after excluding one study as an outlier; Payne's team found no clinical or methodological grounds for that exclusion, kept the study in, and the effect stopped being conclusive.

Significant results cluster at 200 mg and above. In Mátyus's analysis the pooled effect at 100 mg was smaller and its interval crossed zero; the theanine-alone analyses reaching significance in Payne's review were those including the 400 mg arm, with a median dose of 200 mg in theanine-only interventions. The 400 mg finding on simple reaction time points the other way and is unexplained.

Healthy adults tested in a laboratory. MHES §22 applies: nothing here supports extrapolation to clinical populations, to older adults specifically, or to free-living conditions. No study delivered the compound as tea, so the evidence does not describe drinking tea.

EU authorised health claim: None identified.

L-theanine holds no authorised health claim under Regulation (EC) No 1924/2006. A scientific grade and the absence of an authorised claim are separate statements and both are true here — MHES §31. When the European Food Safety Authority evaluated a claim for black tea and attention it based the assessment on caffeine content.

• An adequately powered independent replication of the visual information processing result at 200 mg, pre-registered, in a group with no commercial relationship to the ingredient.
• Any trial demonstrating an effect on a task with practical rather than laboratory relevance.
• A dose–response trial that explains the deterioration observed at 400 mg.
• Trials using tea or tea-equivalent doses in free-living participants, which the field's own authors name as the most useful next step.

• Objective measures did not move. Bulman's pooled analysis found no significant difference for objective sleep onset latency, objective sleep efficiency, objective sleep duration or objective sleep disturbance.
• Subjective measures did. Sleep onset latency SMD 0.15 (CI 0.01 to 0.29), daytime dysfunction 0.33 (CI 0.16 to 0.49), overall sleep quality 0.43 (CI 0.04 to 0.83).
• Neither review found convincing evidence that L-theanine increases total sleep time; Cotter states it directly.
• Tolerability is good across the pooled set, including doses up to 900 mg daily for eight weeks.

• Whether the subjective–objective gap reflects placebo response, mismatched instruments, or a real effect on the experience of rest that current objective measures are not designed to detect. The trials cannot currently distinguish between these.
• The effective dose. Bulman's subgroup analysis pointed below 200 mg; Cotter's conclusion names 200 to 450 mg. The same literature, opposite recommendations, and no dose–response trial exists anywhere in the field.
• Whether the compound's sleep-relevant action is inducing sleep at all, rather than offsetting caffeine. One nine-person EEG crossover found 50 mg of L-theanine returned caffeine-disrupted wake-after-sleep-onset to placebo levels. The sample is tiny and the hypothesis is untested.

• Bulman and colleagues (2025) searched five databases and one register, registered in PROSPERO, and identified 19 randomised trials covering 897 participants, 18 entering the meta-analysis. Doses 50–1,000 mg daily, durations from a single night to eight weeks.
• Cotter and colleagues (2026) took the opposite methodological decision on the same question — excluding every trial combining L-theanine with another potentially active ingredient — leaving 13 trials in 550 participants, of which 11 were RCTs and two were single-arm open-label studies.
• Payne and colleagues (2025) examined sleep as one of three outcome domains and found the evidence inconclusive for tea, theanine alone and theanine with caffeine.

• Every objective outcome in Bulman's meta-analysis was null. Objective sleep duration: mean difference 5.91 minutes, CI −8.08 to 19.91.
• The positive subjective results are fragile under sensitivity analysis. For subjective sleep onset latency, removal of five of the ten contributing studies eliminated significance. For overall sleep quality, heterogeneity was 82.9 per cent traceable to a single study with an effect size of 2.58 against a pooled 0.43; removing it dropped the pooled effect to 0.21, and removing a different study eliminated significance altogether.
• One result runs against the compound: for objective sleep onset latency, removing the study responsible for the heterogeneity left a pooled result significant in favour of the control condition.
• In 11 of Bulman's 19 included studies L-theanine was one ingredient among others — GABA, valerian, lemon balm, saffron, magnesium, lactic acid bacteria, tart cherry, lavender. The authors state plainly that the observed outcomes cannot be solely attributed to L-theanine alone. MHES §17 applies.
• Cotter's positive tally counts trend-level results as supporting findings; p-values of 0.070 and 0.073 appear in the evidence table as such. The authors justify this by underpowering and consistent direction, which is not unreasonable and is also the reasoning that turns a set of near-misses into a positive review.

• No mechanistic account specific to sleep is established in this evidence set. The route both reviews lean towards is indirect — via stress — and that route is assessed separately in MH-EVD-001-C3, where the evidence for objective markers is itself unpersuasive.

The two reviews differ in a way that is itself evidence. The independent team, applying wider inclusion criteria and pooling, reached a hedged conclusion. The team reporting the confident positive conclusion declares that its authors are employees of The Water Street Collective or British American Tobacco, that The Water Street Collective sells products containing L-theanine, and that it is a wholly owned subsidiary of British American Tobacco. That review is transparent about its methods, names the four trials that failed to show benefit, rates most included studies as merely fair, and acknowledges it was not prospectively registered — but its judgement calls, to include open-label trials, to count trend-level results and to decline to pool, all point the same way.

The subjective effects are small, heterogeneous and fragile: the largest, on sleep quality, has a confidence interval reaching 0.04 and loses significance when either of two individual studies is removed. Heterogeneity of 82.9 per cent traceable to one outlying trial means the pooled figure describes that trial more than it describes the literature.

Unsettled and contested. Bulman's subgroup analysis pointed to under 200 mg as the more effective range; Cotter's conclusion names 200 to 450 mg. Neither is established by a dose–response trial, because none exists.

The pooled set spans any age and health status by design, which widens applicability and weakens directness simultaneously. MHES §22 applies: results from mixed clinical and healthy populations should not be read as evidence for healthy adults specifically.

EU authorised health claim: None identified.

No authorised health claim exists for L-theanine and sleep under Regulation (EC) No 1924/2006. MindHeaven makes no sleep product and has no commercial position in this question, which is why this claim is assessed at length rather than briefly.

• A well-designed dose–response trial in pure L-theanine, which both review teams independently name as what the field needs.
• Any objective sleep outcome reaching significance in a pooled analysis of standalone-compound trials.
• A trial designed to test the caffeine-offset hypothesis directly, rather than the sleep-aid framing, in an adequately powered sample.
• Replication of the subjective findings in a trial where the pooled result does not depend on one or two individual studies.

• Moulin's pre-specified primary outcome, change in salivary cortisol at days 14 and 28, was null. There were no significant differences between groups.
• McAllister's acute stress challenge worked — salivary alpha-amylase, secretory IgA, heart rate and state anxiety all rose significantly — and the supplements changed none of it. There was no treatment-by-time interaction on any of the four measures.
• On self-reported symptoms over four weeks there is one reasonably designed positive trial: Hidese's crossover found stress-related symptom scores fell relative to placebo.
• The frequently cited 200–400 mg effective range for stress and anxiety, which subsequent reviews treat as settled background, has not been successfully re-tested in the recent trial literature.

• Whether the positive self-report findings reflect a real effect on experienced stress or a placebo response that objective markers correctly fail to register.
• Whether any effect exists at doses or durations not yet tested in isolation.
• Why eight of nine adverse events in Moulin's trial occurred in the theanine arm, in a compound otherwise notable for tolerability.
• Whether the cortisol reduction reported for a theanine-containing combination drink in 2016 would survive isolation of the compound. No trial has tested that, and the four-ingredient formulation includes two compounds marketed on their own cognitive claims.

• Hidese and colleagues (2019): randomised, double-blind, placebo-controlled crossover, 200 mg daily for four weeks in 30 healthy adults. Stress-related symptom scores fell relative to placebo on the Self-rating Depression Scale, the State-Trait Anxiety Inventory and the Pittsburgh Sleep Quality Index, with sleep latency and daytime dysfunction subscales improving specifically. Two of seven authors are employed by the manufacturer of the branded compound. All outcomes are self-reported.

• Moulin and colleagues (2024): randomised, double-blind, placebo-controlled, 400 mg daily for 28 days in 30 healthy adults with Perceived Stress Scale scores of 14–26, registered on ClinicalTrials.gov and reported to CONSORT standards. The pre-specified primary outcome, salivary cortisol, was null. On the Depression, Anxiety and Stress Scale the direction reversed: placebo improved and theanine worsened slightly, significantly so at day 28 (−9.20 against +1.07, p = 0.017), with the stress subscale following (−4.67 against +0.53, p = 0.012). On the Profile of Mood States the theanine group changed on nothing while placebo improved significantly on total mood disturbance, fatigue, tension–anxiety and depression. The reported 12.92 and 17.98 per cent decreases in perceived stress are within-group changes that placebo matched.
• McAllister and colleagues (2024): 80 participants randomised to 200 mg L-theanine, 2,000 mg L-tyrosine or placebo 40 minutes before a mental stress challenge combining a Stroop task, mental arithmetic and a virtual-reality active shooter drill. No treatment-by-time interaction on any of the four stress measures. The authors state the conclusion without qualification. No commercial funding and no conflict of interest declared. This is the trial most closely resembling the situation the compound is marketed for, and it found nothing.
• Noah and colleagues (2022) reported a positive result on stress — but the intervention contained magnesium, vitamins B6, B9 and B12, rhodiola extract and green tea extract of which only 50 mg was L-theanine, below the level at which any isolated trial has found an effect. Magnesium and B6 hold authorised European health claims of their own. The trial was single-blinded because active and placebo tablets lacked structural similarity. MHES §17 excludes it from supporting this claim, and the authors themselves attribute the effect to the combination.
• Rizzo and colleagues (2022) tested L-theanine with vitamin B6 in 34 children with tic disorders, open-label, against psychoeducation rather than placebo. Tic severity improved; anxiety, the stated rationale for using theanine, did not differ between groups (30.9 against 31.5, p = 0.85). Excluded from this claim by population, design and MHES §17.
• One trial in this literature did move an objective marker, and it belongs on the record even though it cannot support this claim. White and colleagues (2016) ran a randomised, double-blind, placebo-controlled balanced crossover in 34 healthy adults aged 18–40 and reported the salivary cortisol response to a multitasking stressor reduced three hours post-dose on active treatment, with the primary subjective stress measure reduced at one hour. The intervention was NeuroBliss, a commercial nutrient drink containing 200 mg of L-theanine together with 25 mg of alpha-GPC, 1 mg of phosphatidylserine and 10 mg of micronized chamomile, funded by a grant from the manufacturer. MHES §17 excludes it from supporting a claim about L-theanine alone: the trial tested the product. It is recorded here rather than omitted because it is the one objective-marker result in the field that points the other way, and a reader is entitled to weigh it.

• No mechanistic finding in this evidence set explains an effect on physiological stress markers, and none is needed to interpret the results: the markers were measured directly and did not move. MHES §20 and §21 apply in reverse here — the absence of a mechanistic story is not the weakness, the null objective outcome is the finding.

Mixed, and the pattern is instructive. The two null trials are the better-controlled ones: Moulin is double-blind, CONSORT-reported and registered, with a pre-specified primary endpoint; McAllister has no commercial funding and no declared conflict. The positive trials are progressively less isolated and less blinded — Hidese is well-designed but entirely self-reported with the manufacturer among the authors, Noah is single-blinded with five active ingredients, Rizzo is open-label with no placebo. Formal RoB 2 domain assessment has not been performed by us and is outstanding.

Moulin's trial is a case study in how a null result travels as a positive one. The pre-specified primary endpoint did not move; both arms improved on the questionnaire that did; the second questionnaire favoured placebo significantly; and the reported percentage decreases are within-group changes that placebo matched, with the placebo result at day 28 carrying the stronger p-value. The conclusion that supplementation significantly decreased perceived stress describes a within-group change. MHES §15 applies directly: a failed primary endpoint with a positive secondary finding is not an unambiguously positive study.

400 mg for 28 days and 200 mg acutely both produced null results on objective markers. 200 mg for four weeks produced a positive result on self-report. No dose has been shown to move physiology.

Healthy adults scoring in the moderate range on validated stress scales, and healthy adults under acute experimental stress. Nothing here applies to clinical anxiety or to chronic stress as a medical presentation, which belong with a doctor rather than a supplement.

EU authorised health claim: None identified.

No authorised health claim exists for L-theanine and stress under Regulation (EC) No 1924/2006. We would regard such a claim as unsupported by the recent evidence regardless of what the regulations permitted. One disclosure is owed: the Noah formulation — magnesium, B vitamins, an adaptogen extract and L-theanine — resembles our Clarity closely enough that quoting its positive result would be commercially convenient for us, and we are not going to.

• A double-blind trial isolating L-theanine that moves a pre-specified objective physiological marker — cortisol, alpha-amylase, heart rate variability — against placebo.
• Independent replication of Hidese's four-week self-report result by a group with no manufacturer relationship.
• A trial adequately powered to detect the effect size implied by the frequently cited 200–400 mg range, rather than citing it as settled background.