Morning Light: The Lux Number Is Measured in the Wrong Place

It is 7:40 on a February morning and it is not properly light yet. The lamp you bought in November is on the desk, angled at your face, running its thirty minutes because the box said thirty minutes. You are sitting about sixty centimetres back, because that is where the keyboard is. The lamp’s rating was measured at twenty. Behind you, over your left shoulder, is a window you have your back to.

I went looking for where the thirty minutes and the ten thousand came from, expecting a judgment call. What I found was a compliance workaround from a depression clinic, a correction published in the same field twenty-six years ago, and a set of null results from 2018 that appears in no popular coverage I could locate. The one thing that did survive with a registered trial behind it is a desk lamp, and its effects are real and small.

One thing up front. Almost everything below has circadian biomarkers or self-reported sleepiness as its endpoint rather than measured cognition, and several studies that did measure cognition came back null. This is on a site about working with your head because morning light sets the timing of tonight’s sleep, which is upstream of everything — not because there is a trial showing it makes you better at your job.

A number with a birthday

The 10,000 in “10,000 lux” is not a dose. It is a compliance fix, and it is about thirty-six years old.

In the first NIMH seasonal-depression studies of the early 1980s, fixtures emitting 2,500 lux beat control fixtures emitting 300. That reaches me through Rosenthal’s own blog post rather than the primary papers — the weakest kind of source, and it points at the primary work rather than standing in for it. A pooled analysis of 332 patients across 14 centres, which I found inside a state legislative technology assessment, put the working protocol at 2,500 lux for at least two hours daily.

Two hours. Every day. In front of a box.

Compliance was the problem and brighter-for-shorter was the fix. Terman and Terman are credited with showing thirty minutes at 10,000 lux worked about as well; I reached that through the introduction of a 1998 side-effect paper citing it, not the 1990 study, so treat it as a citation rather than something I verified. The substitution even came hedged — one clinical summary says thirty minutes at 10,000 may be as good as two hours at 2,500, then adds in the same sentence that such a linear relationship does not necessarily apply. That hedge did not survive transmission.

Then the geometry, which almost nobody states. Research-grade figures carry a distance: the BROAD trial’s preprint specifies control lamps delivering 10,000 lux at 20 cm, with illuminance falling rapidly further back, and clinical guidance puts the head 30–46 cm from the box. Consumer light-therapy lamps, on the other hand, are commonly advertised at “up to 10,000 lux” for roughly $60 to $86, and across the listings I looked at on general retail sites not one stated a distance at all. That is a category observation from a sweep of listings rather than an audit, and it is the kind of thing you can check in a minute on any retailer you like: find the lux figure, then look for the distance it was measured at.

Do that arithmetic yourself, because it is the transferable part. Idealised as a point source, illuminance falls with the square of distance: three times further back is about nine times dimmer. A lamp rated at 20 cm, met by someone sitting at their keyboard, delivers something far closer to 1,000 lux than 10,000. A panel is not a point source and the real falloff is gentler, so treat the ninefold loss as an upper bound on the loss and the 1,000 as a floor, not a measurement — no manufacturer I found publishes a distance–illuminance curve. A number with no distance attached is not a specification.

The premise broke in the same field, twice

If 10,000 were a dose, brighter would buy more. It does not.

Zeitzer and colleagues gave 23 healthy volunteers a single 6.5-hour light pulse in the early biological night and fitted the dose–response curve. Half the maximum phase shift produced by roughly 9,000 lux was obtained at about 100 lux — ordinary room light. Chang’s group, re-describing that curve in 2012, put it as roughly linear between about 50 and 500 lux and asymptotic above about 550. I read Zeitzer at abstract level; the 550 figure is Chang’s description of the earlier work, not Zeitzer’s own number.

Both are nocturnal, phase-delaying exposures. Hold that.

In 2011 Dewan and colleagues ran the experiment that should have settled it: 56 healthy adults aged 20 to 40, nine conditions crossing duration of one, two or three hours with intensity of 2,000, 4,000 or 8,000 lux. Two-way ANOVA — duration significant at P = 0.01, intensity not significant, no interaction. Each participant’s own dim-light admission — awake, for the same hours — was subtracted from their light admission before analysis, a tighter control than most of this literature manages. A fourfold increase in brightness changed the size of the clock shift by nothing.

The study was NIH-funded and states explicitly that it was not industry-supported. Zee, the senior author, has consulted for Philips/Respironics, which manufactures light-therapy devices, and in a later consensus document reports research funding from Philips — a conflict pointing squarely against this paper’s own finding, which is a direction conflicts almost never point.

Two things fence that result in, and I want them stated before anyone repeats it. The exposure was centred three hours before each participant’s core body temperature minimum — in the paper’s own worked example, a midpoint around half past two in the morning, with four hours of scheduled sleep on either side. Nobody was having a morning. And every intensity Dewan used sat above roughly 1,000 lux, which Zeitzer’s curve says is already past saturation for phase resetting; a later paper citing Dewan says exactly that. So the study did not show that intensity does not matter. It showed that intensity does not matter above the point where it had already stopped mattering — which is what Zeitzer predicts, not an independent confirmation of it. The abstract’s own conclusion is scoped to phototherapy for circadian rhythm sleep disorders — not daytime light, and not morning.

Eastman wrote the commentary in the same issue of Sleep. Her opening verdict: “increase the duration of the light exposure, not the intensity.” She notes the investigators were surprised not to get the dose response they expected. That is her reading of the same night-time, above-saturation protocol, and it travels with the same fence. I retrieved roughly the first 350 words and nothing more, so I can report her opening position and not her overall one.

Fifteen years later the listings still say 10,000.

December 2018, when two labs checked and one was checking itself

The field named its own problem first. Souman and colleagues reviewed 68 publications on acute alerting effects of light: plenty reported people feeling more alert under bright white light, very few of the reaction-time studies found significant effects, and small samples throughout made firm conclusions difficult. What was needed, they concluded, was better-powered work capable of producing real dose–response curves. Read that author line: Souman was at Philips Lighting Research, most co-authors were at Philips Research, and the paper states the work was largely funded by Philips Lighting. A lighting-industry review concluding that the evidence for lighting’s alerting effect is inconsistent and that blue-enrichment shows no systematic advantage — the second conflict in this article running against its own funder’s interest, and the popular layer picked up neither.

Then the December 2018 issue of the Journal of Biological Rhythms ran three papers, opened by a review the two groups wrote together.

Lok’s group at Groningen took 50 participants through a 07:30–17:30 protocol, alternating 1.5 hours of dim light under 10 lux with an hour of experimental light between 24 and 2,000 lux, each participant fixed at one intensity — ten per level. Time of day was a significant predictor of alertness on every parameter. Light intensity produced dose–response relationships on a few parameters at some times of day, and none survived correction for multiple testing. Smolders’ group at Eindhoven, same issue: 60 participants, 20 to 2,000 lux at eye level, an hour in the morning and again in the afternoon on separate days during their normal routine — no clear dose-dependent relationship with alertness or executive control, only subtle decreases in sleepiness and subtle increases in vitality.

Ten people per intensity level, in Lok’s design, is thin for curve-fitting, and Smolders’ paper attributes the field’s inconsistency to low power outright — so the honest reading of both is “no dose–response detectable at this power,” not “no effect exists” — the difference between an underpowered study and a negative one, and it carries the rest of this article.

The lead author of that Eindhoven paper is also the lead author of the 2012 study finding 1,000 lux at eye level beat 200 lux on alertness, sustained attention and heart-rate variability in 32 people, performance effects strongest in the morning and at the end of the hour — the best morning-timed, eye-level, working-hours positive in everything I read. Six years later, with nearly double the sample and a hundredfold intensity range, the same lead author published the null and attributed the literature’s inconsistency partly to low sample sizes and low power. The two groups also co-wrote the review opening that issue, which concludes that the daytime literature on white-light intensity and alertness is inconclusive, particularly for objective measures.

That qualifier keeps recurring, and one study shows the whole shape of it on its own: a 57-person experiment crossing illuminance with colour temperature found reaction speed significantly better on two executive-control tasks at ten times the light at the eye, while the psychomotor vigilance task and a working-memory task came back null. Same people, same session, opposite answers depending on the measure. A 2022 systematic review found that among studies of higher-intensity morning bright light, 64.7% (11 of 17) reported benefit on subjective alertness against 31.3% (5 of 16) on objective alertness, and 33.3% (3 of 9) on reaction time — with effects depending on homeostatic sleep drive, time of day, and the properties of whichever control light each study used, which leaves the studies barely comparable with one another. People reliably report feeling more awake in bright light. Whether they perform better is a much thinner literature, and that gap between what a measure asks and what a claim implies is the same trap that ran through the blue-light glasses trials.

Eight years on, across seven retrieval passes, I found no popular coverage of any of it.

What a lux meter cannot see

Here is where it stops being a demolition.

In 2021 Grant and colleagues put 39 young adults — mean age 24.5, holding a seven-hour time in bed for the preceding week — through eight hours of daytime light in one of four white-light conditions. Every condition was about 50 lux at the eye in the vertical plane: identical photopic illuminance. What differed was melanopic content, 25 to 45 melanopic EDI — melanopic equivalent daylight illuminance, the light as melanopsin weights it rather than as the eye’s brightness sensors do — produced by varying colour temperature and spectrum. A lux meter could not have told those conditions apart, and they produced a 5% difference in accuracy on a two-minute addition task (p = 0.004) and faster motor sequence learning (p = 0.02) — at illuminances a tenth of a well-daylit room, far below the 250 melanopic EDI that current consensus recommends.

That is the cleanest demonstration in anything I read that lux is the wrong unit, and it is worth being exact about how much weight it carries, because the abstract reports three positive findings while the full text is null on six objective task measures and on all but one of the subjective scales. Reaction time on the psychomotor vigilance task: p = 0.78. Attentional failures: p = 0.42. Concentration on the d-2 task: p = 0.91. Declarative memory: p = 0.84. Both positives held only at the extreme contrast; neither intermediate condition separated from the control. The design is between-subjects with eight to ten people per light condition — the same thinness I used against Lok a moment ago.

The subjective sleepiness scale also ran backwards — the highest-melanopic condition of the four reported more sleepiness than the next one down (p = 0.02) — and I am going to resist doing anything with that. The authors call it unexpected and wonder whether spectral composition beyond melanopic content is involved, which is a hypothesis they are entitled to and not a finding. It comes out of the same eight-to-ten-person cells as the positives above, and a small sample does not become reliable when its result happens to be inconvenient for the claim under test. Their own summary is the fair one: the light did not improve every domain tested, and there was no evidence that lower-melanopic light was better.

The study was funded by an investigator-initiated grant from Seoul Semiconductor, which also supplied the luminaires. The paper states the funder had no role in design, analysis or writing, and the senior author separately discloses honoraria and travel funds from the same company.

That 250 figure comes from an 18-author consensus statement published in 2022, the modern replacement for “10,000 lux,” and its geometry is worth as much as its number: minimum 250 melanopic EDI, at the eye, in the vertical plane, at about 1.2 metres — the light arriving at your face while seated, not the light landing on your desk. It is a consensus reached by iterative voting among workshop participants rather than a GRADE-rated guideline, and its competing-interests statement runs to roughly 1,200 words: patents licensed to a light-therapy manufacturer, several authors funded by or consulting for lighting companies. None of that makes it wrong. All of it belongs in the same sentence. And the document itself reports more than tenfold variation between individuals in sensitivity to melatonin suppression, which means any single recommended number is an average response across a population whose members differ by an order of magnitude.

Grant’s group ran the applied follow-up in 2023, and it is the best practical study here. Sixteen adults, three days inpatient, two eight-hour simulated workdays, randomised crossover. One condition was fluorescent room light at about 30 melanopic EDI and 50 photopic lux, which is what a great many offices actually are. The other was the same room plus a single LED task lamp on the work surface, at about 250 melanopic EDI and 210 photopic lux. It is the only prospectively registered trial in the set (NCT04745312), analysed by linear mixed models with a false-discovery-rate correction — which is why the numbers below are q rather than p: the correction charges each result for the fact that the battery tested many outcomes at once, so a q that survives has already paid that toll. Improvement in addition-task accuracy from baseline: 3.15% ± 1.18% with the lamp against 0.93% ± 1.1% without, q = 0.005. Psychomotor vigilance reaction time and attentional failures improved, all q ≤ 0.030. Sleepiness, alertness, happiness, health, mood and motivation improved, all q ≤ 0.036. No difference in mood disturbance, affect, declarative memory or motor learning, all q ≥ 0.308, reported by the authors as prominently as the rest. The lamp made people sharper and happier. It did not make them learn better. The one thing I should add, because I read the full text and the abstract alone would not tell you: raw motor speed did improve, q = 0.045, at an effect size of 0.06 that the authors themselves call very small.

Now the sizes, because they are the difference between a finding and a promise. The addition-task gain was 2.2 percentage points, Cohen’s d = 0.48. Reaction time was 11 milliseconds faster, d = 0.84. Nine-tenths of an attentional lapse fewer, d = 0.44. Half a point on the sleepiness scale, d = 0.28. On the visual-analogue scales, alertness moved six millimetres out of a hundred, d = 0.10; happiness three millimetres, d = 0.12. These are real, they point consistently one way, and most of them are small. Eleven milliseconds is not a different afternoon.

This study was funded by the lamp’s manufacturer too — an investigator-initiated grant from Biological Innovation and Optimization Systems, which also supplied the task lamp, with the same statement that the funder had no role in the work. That makes three of the applied positives in this piece — the glazing study further down and both Grant trials — funded by the maker of the thing being tested. All three disclosed it. None of it is hidden, and all of it is worth knowing before you buy anything.

So: I wanted to end here — melanopic content is the variable, photopic lux is noise. Five things stop me, and I found the last of them late.

One. That 2023 trial changed both spectrum and illuminance — 50 to 210 photopic lux as well as 30 to 250 melanopic. The authors say so themselves: the relative contribution of the two requires further investigation. Strong evidence that the consensus recommendation works; weak evidence about which variable does the work.

Two. A 2025 crossover from Bergen ran the psychomotor vigilance task in 39 young adults across four morning light conditions on separate days: narrowband blue at high melanopic illuminance, narrowband red at low melanopic illuminance, bright white at roughly 1,000 melanopic lux and 8,000 K, and a dim control. Blue and red both cut lapses and shortened reaction time against dim. Red also improved the fastest responses. Blue was not better than red. And the bright white condition — the one that most nearly matches what the consensus recommends for daytime — did nothing significant on either measure, which the authors themselves call surprising given its melanopic content. Their own systematic review, they note, finds that most studies report no significant effect of bright or blue-enriched white light on this task. I have the abstract and two paragraphs of their discussion, no statistics and no effect sizes, so take those as directions rather than magnitudes. Within one within-subjects design, melanopic illuminance did not order the results. This study, incidentally, is funded by the Research Council of Norway with no competing interests declared — of the well-powered daytime studies here whose funding I could retrieve, the only one with no manufacturer attached.

Three. Souman’s review found no systematic pattern for blue against longer wavelengths, and noted studies reporting no detriment from filtering the short wavelengths out entirely. Two independent sources seven years apart, both finding that wavelength specificity does not hold cleanly — and wavelength specificity is the mechanistic basis of the whole blue-light-in-the-morning protocol.

Four. A collision the group cannot resolve either. The 2021 study found motor sequence learning 3.2× faster under high-melanopic light; the 2023 study, same group, largely overlapping authors, found it null — while reaction time went the other way, null in 2021 and significant in 2023. The 2023 discussion addresses this directly and lists the candidates: between-subjects versus within-subject design and the different prior light history that follows from it, the size of the lighting contrast, light geometry, the interval between learning and recall, and ordinary differences between laboratories. It settles none of them. A group publishing its own inconsistency and failing to explain it is the state of the evidence, not a footnote to it.

Five, found late. In 2016 Segal and colleagues — Lockley on that author line as well — restricted 60 young adults hard — five hours in bed the night before, three in the laboratory — and randomised them to three hours of narrowband blue light (23 people) or green light (25) at equal photon density, against a darkness control (12). Subjective sleepiness, vigilance, working memory, selective attention, polysomnographic and ocular sleepiness. No significant effect of wavelength on any of it, bar a single mood subscale. I have that at abstract level. It matters because it is larger than the 2021 study, it is in severely sleep-restricted people, and it is the cleanest wavelength comparison anywhere in this piece.

What survives is narrower than I wanted and still worth having: a lux meter cannot distinguish light conditions that produce measurably different cognitive outcomes, so a recommendation specified in lux is specified in the wrong unit — but melanopic EDI has not been shown to be the right one either.

Before you change anything

Bipolar disorder is a real contraindication for self-administered morning bright light. A 2018 review by Benedetti and colleagues, which reaches me only through the published correspondence replying to it and a repository abstract, pooled 41 studies and 799 patients treated with light therapy: 0.9% switched into mania and 1.4% into hypomania — but detection depended on whether anyone looked: 0% where no assessment method was reported, and only 12 of 43 studies used a mania scale at all (the replying authors’ own count is 43 where the review says 41), which biases the pooled rate downward. A meta-analysis of four RCTs and 190 participants found manic switch at 1.1% with light against 1.2% in controls, no excess; four trials and 190 people is a small base, and a wide confidence interval on remission makes that endpoint uninformative rather than negative. Against this, Sit and colleagues argue in published correspondence that the safety case is overstated, report from their own case series that women with bipolar illness are highly sensitive to morning bright light with mixed-state induction a substantial risk, and advise starting with fifteen minutes at midday — case series plus correspondence, the weakest design here carrying the most cautionary claim. If this is you, it is a conversation with your clinician, not a protocol you download.

The side-effect rate everyone quotes has no control group. A survey of 70 SAD patients on short-term 10,000-lux therapy found 45.7% reporting side effects, mostly headaches and eye complaints, almost all mild and transient. That design can show how often people on light therapy report symptoms; it cannot show whether the light caused them. The placebo-controlled version exists: healthy young adults, one 30-minute session at 10,000 lux against dim red light under 500 lux, no significant group-by-time interaction on any side effect. Eye strain and blurred vision rose significantly in both arms, placebo included. Both numbers are real and the difference between them is the control group. Neither tested months of daily use.

Morning specifically produces more activation. In an 83-patient pre-post study across 88 symptoms, being overactive, excited or elated emerged more under morning light and remitted more under evening light. Nausea emerged at 15.9%, jumpiness 8.8%, headache 8.4%, eye irritation 6.0% — uncontrolled, in SAD patients, and for most symptoms remission equalled or exceeded emergence. In a healthy person that is not a disorder; it is a plausible mechanism for feeling wired and unable to settle.

On eyes: the only long-duration follow-up examined 50 SAD patients before and after treatment, and 17 of them again after three to six years of seasonal use totalling 60 to 1,250 cumulative hours. No ocular abnormalities either time. The authors state that current knowledge is insufficient to specify definite ocular contraindications, and recommend periodic examination for anyone with pre-existing ocular abnormalities or on photosensitising drugs. Clinical guidance citing that study names lithium, phenothiazines, St John’s wort and melatonin among them — the last of which this audience takes more than most. Seventeen self-selected people with no control group is reassuring; it is not a safety clearance. The only modern comfort data sit in the two Grant trials: the brighter task-lamp condition was not associated with adverse visual experience or discomfort, and the 2021 study found no difference in headache or eye strain between conditions.

Finally, brighter is not monotonically better. The 2022 consensus itself reports a care-home study whose brightest intervention, around 900 melanopic EDI, produced reduced sleep efficiency and quality against a roughly 100 melanopic EDI comparison, and an office study where raising an already-bright environment further prevented the normal seasonal advance in sleep timing. A 57-person study crossing illuminance with colour temperature found that the higher colour temperature gave no significant benefit and increased negative affect. The consensus states that adverse effects of lighting greatly exceeding its recommendations warrant future research — a document arguing for brighter indoor light, saying that about itself.

One thing I am not answering here

Whether a window counts needs its own piece, so here is only what is cheap to state. No trial has ever compared daylight through glass against the same daylight unglazed, on any circadian or cognitive outcome — seven passes found no such trial. Nor could I find a primary source for the widely repeated figure putting a window at fifty times less effective. And the popular mechanism, that glass strips out the wavelengths that matter, runs against the transmission figures for ordinary clear glazing that appear in patent and fenestration literature, which cut off sharply below about 400 nm and above that passes visible light close to evenly; melanopsin peaks near 480 nm, on the transmitted side of that line. That holds for clear glazing; the coated and low-emissivity glass in most offices is a different product with a different curve, and nobody has published a melanopic transmittance figure for any window. What a window really costs you is quantity and geometry — a rectangle of sky instead of a hemisphere of it. That is a different argument with different advice attached, and it is coming.

What survives

1. Timing, and you can compute it on the back of an envelope. Lok’s null study found time of day significant on every alertness measure it took, while light intensity was significant on none that survived correction. The clock outperformed the lamp. For where to put the exposure, the most precise anchor I found comes from a 2001 study of 42 SAD patients that timed each person’s light against their own melatonin onset: the antidepressant effect was potentiated by early morning in circadian time — optimally about 8.5 hours after melatonin onset, or about 2.5 hours after the midpoint of sleep. You cannot measure your melatonin onset; you can compute your sleep midpoint. Sleep 00:30 to 07:30, midpoint 04:00, target around 06:30. Sleep 02:00 to 10:00, midpoint 06:00, target around 08:30 — after that person wakes, not the instant they wake. Two caveats travel with this permanently: the same paper’s headline null is that depression ratings were similar with light at either time of day, and it was validated against depression scores in patients, not against cognition in anyone. Its senior author also founded and directs a nonprofit that distributes light-therapy guidance, and the paper’s own funding line I could not retrieve.

2. Face the window instead of sitting near it. In a two-week crossover of 30 knowledge workers — mean age 34, full-time computer work, the best population match in this literature — horizontal illuminance at the desk was effectively identical between conditions, 234 against 223 lux, while vertical illuminance at eye level facing the window differed 5.5-fold, 51.4 against 286. Outcomes tracked the eye-level number: 36.96 more minutes of sleep per night (p < 0.001) and 42% higher scores on a decision-making simulation (p < 0.0001). The qualifications are large. It was funded by View Inc., the electrochromic-glazing manufacturer whose product was the intervention, with four View employees among the co-authors and data analysis among their declared roles; participants could not be blinded; the authors say part of the effect may be placebo and that they cannot separate daylight from view; 37.3% of the variance was unexplained. And in the stratum most like a well-rested reader — the good sleepers — the sleep effect was 18.08 minutes at p = 0.214, not significant. Turning your chair is still free.

3. Longer probably beats brighter — and I am demoting this one. It was going to rest on Dewan, and Dewan will not hold it: a night-time protocol centred three hours before the temperature minimum, across an intensity range already past saturation. Nothing I found varied duration in the morning, in the advancing direction, at intensities below saturation. So “twenty minutes of ordinary daylight beats five minutes of something very bright” is a guess with a plausible mechanism behind it, and you should treat it as one. And know what the ceiling on a single hour looks like across the whole curve. In the one phase-response curve built with a proper dim-light control, an hour of roughly 8,000 lux produced a fitted peak-to-trough amplitude of 2.2 hours, against more than 3 hours for a 6.7-hour exposure in a separate earlier study under similar conditions. One hour buys most of what nearly seven hours buys — but that is delays and advances together, and the morning share of it is not in anything I read. Two numbers in a figure legend suggest the dim control arm produced most of the morning advance by itself, and I am not going to use them: the paper’s own fitted analysis found no discernible curve in the dim arm at all, which makes individual points in it noise rather than evidence. That correction runs against my argument and it is still the correction.

4. If your office is genuinely dim, the desk lamp above is the best applied evidence I found — not a therapy box, not a renovation. Sixteen people, spectrum and illuminance both changed, memory and motor learning null, gains real but mostly small, and the trial paid for by the company that makes the lamp. A much smaller claim than the internet’s, and the only one here with a registered primary outcome behind it.

What none of this supports: that a 10,000-lux box at arm’s length delivers what its label implies, that brightness is the dial, or that any of this has been shown to make a well-rested person better at their job.

I had a tidier ending than that and gave it up. The tidy version was that the nulls come from well-rested people and the positives from the tired, which would tell you exactly who this is aimed at, and several authors do float sleep pressure as the moderator. But the 2016 study above is sixty severely sleep-deprived people and a flat null, and nobody has ever randomised well-rested and sleep-restricted participants to the same light protocol. There is no test of it — only a pattern that a large study in the right population fails to fit. Sleep pressure is a live hypothesis, not an answer.

Tomorrow, stand at the window facing out for as long as you were going to sit in front of the lamp. It costs nothing, and it is measured in the right plane.

About this article

Written by Drew Anton. Drew covers behaviour, sleep, and the compounds people take for both — doses, timing, and what the label does and does not tell you. Not a physician or research scientist — reads the primary literature closely and refuses to round up.

Medical review: None. NeuriFuel does not currently have a licensed clinician on the editorial team, and this article has not been medically reviewed. We state this rather than implying an authority we do not have. See our About page for our full methodology.

Sources: 25 numbered references — one expert consensus statement, three systematic or narrative reviews, one meta-analysis of RCTs, eighteen human experimental, trial or observational reports, one journal commentary and one published correspondence. Secondary and non-peer-reviewed material is listed after the references and labelled where it appears in the text. Four of the 25 were read at full text; the rest were read at abstract, index-record or secondary level, and wherever a reading level limits what may be claimed, the text says so in the sentence that makes the claim. Funding disclosures could not be retrieved for seventeen of the twenty-five studies cited here — silence in a reference line below means not retrieved, not clean. Of the eight we did retrieve, four changed what this article says: the two Grant trials, the glazing study, and the Philips-funded review that concluded against its own funder’s interest. Where disclosures were retrieved they appear inline, including for studies whose findings support our conclusion. Missing values are marked in text and conflicts are flagged rather than silently resolved.

Corrections: Found an error? Write to hello@neurifuel.com with a source and we will fix it and log the correction.

Last updated: 18 September 2026

References

Zeitzer JM, Dijk D-J, Kronauer RE, Brown EN, Czeisler CA. Sensitivity of the human circadian pacemaker to nocturnal light: melatonin phase resetting and suppression. J Physiol 2000;526(Pt 3):695–702. doi:10.1111/j.1469-7793.2000.00695.x · PMID 10922269
Chang A-M, Santhi N, St Hilaire M, et al. Human responses to bright light of different durations. J Physiol 2012;590(13):3103–3112. doi:10.1113/jphysiol.2011.226555 · PMID 22526883
Dewan K, Benloucif S, Reid K, Wolfe LF, Zee PC. Light-induced changes of the circadian clock of humans: increasing duration is more effective than increasing light intensity. Sleep 2011;34(5):593–599. doi:10.1093/sleep/34.5.593 · PMID 21532952 · PMC3079938 — NIH-funded; the paper states it was not industry-supported. Co-author Zee has received research support from and consulted for Philips/Respironics. A night-time phase-delay protocol centred 3 h before the core body temperature minimum, at intensities all above the phase-resetting saturation point — see the body for why that fences the finding.
Eastman CI. How to get a bigger dose of bright light. Sleep 2011;34(5):559–560. doi:10.1093/sleep/34.5.559 · PMID 21532946
St Hilaire MA, Gooley JJ, Khalsa SBS, Kronauer RE, Czeisler CA, Lockley SW. Human phase response curve to a 1 h pulse of bright white light. J Physiol 2012;590(13):3035–3045. doi:10.1113/jphysiol.2012.227892 · PMID 22547633 — read at full abstract including the fitted result, plus figure legends; results text not retrieved. Funding not retrieved.
Boubekri M, Lee J, MacNaughton P, Woo M, Schuyler L, Tinianov B, Satish U. The impact of optimized daylight and views on the sleep duration and cognitive performance of office workers. Int J Environ Res Public Health 2020;17(9):3219. doi:10.3390/ijerph17093219 · PMID 32384634 — Funded by View Inc.; four co-authors are View Inc. employees with declared roles including data analysis.
Brown TM, Brainard GC, Cajochen C, et al. Recommendations for daytime, evening, and nighttime indoor light exposure to best support physiology, sleep, and wakefulness in healthy adults. PLoS Biol 2022;20(3):e3001571. doi:10.1371/journal.pbio.3001571 — Extensive declared interests: patents licensed to a light-therapy manufacturer; multiple authors funded by or consulting for lighting companies.
Kogan AO, Guilford PM. Side effects of short-term 10,000-lux light therapy. Am J Psychiatry 1998;155(2):293–294. doi:10.1176/ajp.155.2.293
Terman M, Terman JS. Bright light therapy: side effects and benefits across the symptom spectrum. J Clin Psychiatry 1999;60(11):799–808. PMID 10584776
Botanov Y, Ilardi SS. The acute side effects of bright light therapy: a placebo-controlled investigation. PLoS ONE 2013;8(9):e75893. doi:10.1371/journal.pone.0075893
Siraji MA, Kalavally V, Schaefer A, Haque S. Effects of daytime electric light exposure on human alertness and higher cognitive functions: a systematic review. Front Psychol 2022;12:765750. doi:10.3389/fpsyg.2021.765750 · PMID 35069337
Terman JS, Terman M, Lo ES, Cooper TB. Circadian time of morning light administration and therapeutic response in winter depression. Arch Gen Psychiatry 2001;58(1):69–75. doi:10.1001/archpsyc.58.1.69 · PMID 11146760 — funding not retrieved; co-author M. Terman founded and directs the Center for Environmental Therapeutics, which distributes light-therapy guidance.
Gallin PF, Terman M, Remé CE, Rafferty B, Terman JS, Burde RM. Ophthalmologic examination of patients with seasonal affective disorder, before and after bright light therapy. Am J Ophthalmol 1995;119(2):202–210. doi:10.1016/s0002-9394(14)73874-7 · PMID 7832227 — funding not retrieved; co-author M. Terman founded and directs the Center for Environmental Therapeutics, which distributes light-therapy guidance.
Sit DK, Wisner KL, Hanusa BH, et al. Light therapy and risk of hypomania, mania, or mixed state emergence: response to Benedetti et al. Am J Psychiatry 2018. doi:10.1176/appi.ajp.2018.18020231r — the Benedetti et al. review it responds to was read only through this correspondence and a repository abstract; its primary citation was not retrieved.
Hirakawa H, Terao T, Muronaga M, Ishii N. Adjunctive bright light therapy for treating bipolar depression: a systematic review and meta-analysis of randomized controlled trials. Brain Behav 2020;10(12):e01876. doi:10.1002/brb3.1876 · PMC7749573 — funding and conflicts not retrieved.
Souman JL, Tinga AM, te Pas SF, van Ee R, Vlaskamp BNS. Acute alerting effects of light: a systematic literature review. Behav Brain Res 2018;337:228–239. doi:10.1016/j.bbr.2017.09.016 · PMID 28912014 — Largely funded by Philips Lighting; most authors were Philips employees.
Lok R, Woelders T, Gordijn MCM, Hut RA, Beersma DGM. White light during daytime does not improve alertness in well-rested individuals. J Biol Rhythms 2018;33(6):637–648. doi:10.1177/0748730418796036 · PMID 30191761
Smolders KCHJ, Peeters ST, Vogels IMLC, de Kort YAW. Investigation of dose-response relationships for effects of white light exposure on correlates of alertness and executive control during regular daytime working hours. J Biol Rhythms 2018;33(6):649–661. doi:10.1177/0748730418796438 · PMID 30198360 — funding not retrieved; the Eindhoven group has institutional lighting-industry ties.
Lok R, Smolders KCHJ, Beersma DGM, de Kort YAW. Light, alertness, and alerting effects of white light: a literature overview. J Biol Rhythms 2018;33(6):589–601. doi:10.1177/0748730418796443 · PMID 30191746
Smolders KCHJ, de Kort YAW, Cluitmans PJM. A higher illuminance induces alertness even during office hours: findings on subjective measures, task performance and heart rate measures. Physiol Behav 2012;107(1):7–16. — DOI/PMID not retrieved; participant age, occupation and funding not retrieved; the Eindhoven group has institutional lighting-industry ties, not verified for this paper.
Smolders KCHJ, de Kort YAW. NIF effects of illuminance and correlated color temperature of office light on alertness, mood, and performance across cognitive domains. Building and Environment 2019;149:253–263. — DOI not retrieved; the Eindhoven group has institutional lighting-industry ties, not verified for this paper.
Grant LK, Kent BA, Mayer MD, Stickgold R, Lockley SW, Rahman SA. Daytime exposure to short wavelength-enriched light improves cognitive performance in sleep-restricted college-aged adults. Front Neurol 2021;12:624217. doi:10.3389/fneur.2021.624217 · PMID 33692742 — Funded by an investigator-initiated grant from Seoul Semiconductor Co., Ltd., which also supplied the luminaires; the paper states the funder had no role in the work. The senior author discloses honoraria and travel funds from the same company; co-author Lockley declares extensive lighting-industry interests.
Grant LK, Crosthwaite PC, Mayer MD, et al. Supplementation of ambient lighting with a task lamp improves daytime alertness and cognitive performance in sleep-restricted individuals. Sleep 2023;46(8):zsad096. doi:10.1093/sleep/zsad096 · PMID 37026184 · ClinicalTrials.gov NCT04745312 — Funded by an investigator-initiated grant from Biological Innovation and Optimization Systems, LLC, which also supplied the task lamp; the paper states the funder had no role in the work. Co-author Lockley discloses consulting fees from View Inc. (see reference 6) among many others; the senior author discloses research support from Biological Innovation and Optimization Systems and other lighting companies.
Segal AY, Sletten TL, Flynn-Evans EE, Lockley SW, Rajaratnam SMW. Daytime exposure to short- and medium-wavelength light did not improve alertness and neurobehavioral performance. J Biol Rhythms 2016;31(5):470–482. doi:10.1177/0748730416659953 · PMID 27474192 — abstract level only; funding not retrieved.
Bjerrum LB, Visted E, Nordhus IH, et al. Light of both high and low melanopic illuminance improves alertness and attention during daytime. Sci Rep 2025;15(1):45238. doi:10.1038/s41598-025-29154-4 · PMID 41286402 · PMC12749685 — Funded by the Research Council of Norway (grant 275305); the authors declare no competing interests. Read at abstract and discussion level; the results section with statistics was not retrieved, and the text states directions only.

Secondary and non-peer-reviewed material, labelled where it appears above: Rosenthal NE, “On the frontiers of SAD: how much light is enough?” (personal blog, 2012) for the NIMH origin narrative; a state legislative technology assessment for the 332-patient pooled analysis; the introduction of reference 8 for the Terman 1990 attribution; clinical summaries for the reciprocity hedge and the 30–46 cm placement guidance; the BROAD trial preprint for the 20 cm specification; retail listings observed at Walmart, Newegg and eBay; patent-literature and fenestration sources for clear float glass transmission.