ARTICLE
Light therapy: a protocol guide for circadian phase shifting
What intensity, what timing, what duration, and why most people using a light box are using it at the wrong hour.
Most people who own a light box use it for less than the effective dose, at the wrong time relative to their circadian phase, and stop after two weeks because nothing happened. The protocol works. The execution is what fails.
What follows is the operational version: the lux value that matters, the timing math relative to your biological phase rather than the clock on the wall, the duration that the response curve actually saturates at, and the failure modes that quietly waste months for people doing nearly the right thing.
What light therapy actually does
A specific subset of cells in the retina — melanopsin-containing retinal ganglion cells, distinct from the rods and cones that handle vision — signal directly to the suprachiasmatic nucleus, the master clock in the hypothalamus. Bright light early in your subjective biological morning advances the clock, making you sleepy earlier the next night. Bright light in your subjective biological evening delays the clock, pushing sleepiness later.
Indoor lighting doesn't work for this. Typical indoor brightness is 100-500 lux at the eye. Effective phase-shifting begins around 2,500 lux and is calibrated at 10,000 lux. Daylight outdoors is 10,000-100,000 lux. From the SCN's perspective, normal indoor lighting registers as continuous dim twilight — not enough signal to move anything.
Staring at a screen doesn't substitute either. Phones and laptops emit roughly 50-100 lux at the eye, even on the brightest blue-rich settings. A 6500K display looks white but can't deliver the photopic dose required to shift the SCN. Blue-light glasses or 'blue light' marketing on monitors is unrelated to therapeutic dose.
If you've ever felt instantly alert stepping out into a sunny morning after a bad night, that's the protocol working at full strength. We're trying to reproduce a slice of that, every day, on schedule.
If you don't know whether your pattern is DSPD or another phase issue, start with delayed sleep phase disorder first. The protocol below is correct for DSPD and the phase-advance side of shift work and eastward jet lag. It is the wrong protocol for advanced sleep phase, and the wrong protocol entirely for pressure-driven insomnia.
The protocol: morning light for phase advance
This is the most-used application. Used for DSPD, for some shift-work recovery, for eastward jet lag.
Equipment
A 10,000 lux light box, full-spectrum white, designed for SAD or circadian use. Reputable categories include light boxes from companies like Verilux, Carex, and Northern Light Technologies — we don't recommend a specific product. Distance from face: 30-60cm (1-2 feet). Eyes open but NOT looking directly at the bulb. The light enters via peripheral vision — you read, eat breakfast, or work at a desk facing the device. Looking directly at it is uncomfortable and provides no extra benefit.
Timing — the calculation that matters
Light therapy timing is relative to your current circadian phase, NOT to your clock time. The biological 'morning' for someone sleeping 3am to 11am is approximately 30 minutes AFTER their natural wake time. For them, 11:30am is biologically equivalent to 6:30am for someone sleeping 11pm to 7am. The protocol works on the body's clock, not on the wall's clock.
Worked example
If you naturally wake at 11am and want to phase-advance, your first light session is 11:00-11:30am, NOT 7am. Using a light box at 7am — when you're biologically still in subjective night — can actively DELAY your phase further. This is the single most common mistake in DIY light therapy: applying clock-wrong timing and concluding the protocol doesn't work.
Duration and dose
Thirty minutes at 10,000 lux is the standard effective dose. Shorter sessions (15-20 minutes) work proportionally less. Longer sessions (60-90 minutes) are not proportionally better — the response saturates. Eye strain and headaches scale linearly with duration, so doubling time roughly doubles side effects without doubling benefit. Stick with thirty minutes.
Titration
After 5-7 days of light at your natural wake time, advance the session by 15 minutes earlier per day, and advance your wake time by 15 minutes earlier per day. The two move together. Most people achieve 1-2 hours of phase advance over 3-4 weeks of consistent practice. The titration is slow because the SCN resists faster shifts and overshooting causes regression.
The protocol: evening light avoidance
The second half of phase advance, and the one most home users skip. Light advances OR delays the clock depending on when it hits the SCN. Light in your biological evening — roughly 3 to 5 hours before your natural sleep onset — actively pushes the phase later. Bright morning light without bright evening avoidance is half the protocol; doing both is what compounds the gains.
Practical implementation: from 3 hours before your target bedtime, the lighting at your eye should drop below 50 lux. That means dimmer switches, lamps instead of overhead lighting, 'warm' bulbs under 3000K, and either dimming screens aggressively or using blue-blocking glasses if you have to use them.
The light-blocking glasses look ridiculous and probably aren't necessary if you'd just turn off the overhead light at 9pm. The cheaper fix is the better fix. The intervention with the highest compliance is the one you'll actually do every night.
Morning protocol gives you maybe 60% of the available phase shift. Evening protocol gives you the other 40%. Skipping either one halves the result. The two have to be done together to compound.
Reverse protocol: phase delay for chronic early waking
Less common but real: some people wake at 4am chronically and can't get back to sleep, falling asleep at 8pm. The mirror image of DSPD — called Advanced Sleep Phase Disorder, more common with age — uses the same biology in reverse.
Inverted timing: bright light in your biological evening (not morning), avoid bright light in your biological morning. For someone waking at 4am: schedule a light session at 9-10pm, and avoid bright light from 5-9am. The mechanism is identical; the sign is flipped. Readers with ASPD can apply the entire morning-light logic above in reverse without rewriting it.
What kills the protocol
Four failure modes account for almost all DIY light therapy that doesn't work.
Inconsistent timing
Light at 11am one day, 9am the next, 1pm the next — the SCN can't lock onto a phase that's moving. The protocol requires the same time every day for at least two weeks before titration begins. Drift more than 30 minutes day-to-day and the entrainment fails.
Weekend off
Skipping the protocol on weekends is functionally equivalent to flying west and then east every Saturday-Sunday. The phase drifts back the same way an entrained schedule drifts during vacation. The fixed timing is seven days a week or it's nothing. There is no version of this protocol that succeeds with weekend cheating. None. The biology doesn't take weekends off.
Clock-time instead of biological-morning time
The mistake from the worked example, mentioned a second time because it's the largest source of failure. A DSPD patient at 7am clock-time is biologically at 3am. Light there delays the phase, not advances it. If you set the alarm for the clock-wrong hour and use the light box there, six weeks later you're worse and you blame the protocol.
Dose stacking
People assume more is better — 90 minutes at 10,000 lux, twice per day, or 'I'll really push this week.' It doesn't compound. The response curve saturates at 30 minutes. The extra time produces eye strain and dropout, not extra phase shift. Save the energy for weeks three through eight, when the protocol's working but feels boring.
When light therapy alone isn't enough
For most DSPD patients, the strongest evidence is for the combination of light therapy plus timed low-dose melatonin: 0.3-0.5mg of melatonin taken 5-7 hours before current natural sleep onset. The two interventions produce phase shifts of roughly 1.5 to 2 times what light alone produces in published DSPD work. Combined with strict behavioral discipline around the new wake and bed times, the package is the standard of care.
If you're six to eight weeks into consistent protocol — morning light timed to biological morning, evening light avoidance, consistent wake time, and timed melatonin — and there is no measurable phase shift, the pattern may not be light-responsive. Possibilities to investigate with a clinician: non-24-hour sleep-wake disorder, primary mood disorder driving timing changes, undiagnosed sleep apnea masquerading as a phase problem. See our medical disclaimer for when to escalate.
If you haven't already, the pattern article delayed sleep phase disorder explains who this protocol is for and who it isn't.
Our pillar guide on circadian misalignment covers DSPD alongside the other circadian patterns — advanced phase, shift work, jet lag — and where each fits.
The same protocol applied to jet lag works on a compressed timeline — days instead of weeks. Our jet lag protocol guide covers the day-by-day eastward and westward versions.
For rotating-shift workers, light timing is the central harm-reduction lever — shift work disorder covers the protocol around the rotation.
Melatonin is the chronobiotic complement to morning bright light — dose-and-timing in melatonin deep dive.
Morning bright light advances the cortisol curve; if your 3am wake pattern is partly phase-related, 3am cortisol awakening covers the mechanism.
Age-related phase advance (going to bed earlier than wanted) responds to evening bright light — see insomnia in older adults for the older-adult pattern context.
If anxiety is part of the picture, morning light still helps but the dose response is smaller — anxiety insomnia covers the modifications.
If you're tracking phase shifts via wearable, sleep tracker comparison covers which devices give usable phase data.
Anything that doesn't respond to six to eight weeks of careful protocol, or shows up with concerning symptoms, belongs with a clinician. See our medical disclaimer.
More circadian pieces at the circadian hub.
Frequently asked questions
Can sunlight replace a light box?
Yes, if you can get 30 minutes of direct outdoor exposure at the right biological time. Outdoor daylight at 10,000-100,000 lux is more than the box delivers. The catch is reliability: most people can't commit to outdoor light at a specific time every day, year-round, regardless of weather. If your geography and lifestyle allow it, sunlight is the better tool. For most people in most climates, a box is the reliable version of the same intervention.
What about light therapy glasses like Re-Timer or Luminette?
They work in principle — light delivered closer to the eye at lower nominal lux can still reach an effective dose. The evidence on dose adequacy is mixed and the cost is higher than a comparable box for a similar effect. The reasonable use case is portability — if you travel often or can't sit in front of a fixed box, glasses are a fair option. As a primary at-home tool, a box is simpler and cheaper.
Does the color of the light matter?
Full-spectrum white at 10,000 lux is the most-tested format. Blue-enriched white can produce equivalent phase shifts at lower nominal lux but tolerability varies (more headaches, more eye strain). Red and amber are not effective for therapeutic phase work and are appropriate for evening avoidance rather than morning therapy. If you have a choice, full-spectrum white is the conservative default.
Can I get too much light therapy?
Side effects are real but mostly mild: headaches, eye strain, occasional nausea, mainly in the first week. They typically subside as you adapt. Light therapy can trigger hypomanic episodes in people with bipolar disorder — if you have any history of mood instability, see our medical disclaimer and consult a clinician before starting. Otherwise the safety profile is favorable; dose-stacking produces diminishing returns rather than danger.
How fast will I notice changes?
Subjective alertness shift in 3-7 days. Measurable phase shift (earlier sleep onset, easier wake time) in 2-4 weeks. Stable new phase that holds without conscious effort in 6-8 weeks. The first week often feels like nothing is changing — the SCN resists. Most people who quit do so at days 5-10, exactly when the curve is about to start bending. Hold through that window.