Why Red Light Does Not Ruin Your Night Vision

This page contains affiliate links. If you buy through one we may earn a commission, at no cost to you.

Because you have two separate visual systems and the red light is only talking to one of them. Your cones, which run in daylight and carry colour, are still 6 percent sensitive at 660 nanometres. Your rods, which carry everything you can see in the dark, are 0.031 percent sensitive there, which is 3,200 times down from their own peak and is for practical purposes blind. So a deep red beam hands your cones enough to read a map by and hands your rods almost nothing, which leaves their dark adaptation intact.

Then the part that follows from it and is more useful than the answer itself. One lumen of deep red light costs your dark adaptation about 0.013 of what one lumen of green costs it. That is a factor of 183 between two beams that a listing describes only by colour name, and the lumen figure on the box cannot show it to you, because the lumen is defined by the cone curve and the rods are the thing at risk. We carry one light whose colour choice is its headline feature, and on this arithmetic it offers the three worst colours and not the one that works.

Two systems, not one dimmer

The retina runs two photoreceptor populations with different jobs and different spectral tuning.

Cones are the daylight system. There are around six million, concentrated in the central fovea, they give you colour and fine detail, and the combined sensitivity of the three types peaks at 555 nanometres, in the yellow green. They need a lot of light and they respond fast.

Rods are the night system. There are around 120 million, spread across the periphery and absent from the very centre of your gaze, they carry no colour information at all, and their sensitivity peaks at 507 nanometres, shifted toward blue green. They are enormously more sensitive than cones, to the point where a fully dark adapted rod system can respond to a handful of photons.

Two consequences fall straight out of that and both are worth knowing. The first is why you cannot see a faint star by looking straight at it and can see it by looking slightly to one side: the centre of your gaze has no rods. The second is the colour shift, which is the reason this page exists. Because the two curves peak at different wavelengths, the relative brightness of two colours changes as the light level falls. Reds darken dramatically and blue greens hold up. A red flower and a green leaf that look equally bright at noon do not look equally bright at dusk, and the leaf wins.

The number that makes red work

Put the two curves side by side at the wavelength of a deep red LED, around 660 nanometres.

The cone curve, which is the standard photopic luminous efficiency function, is at 0.061 there, so 6.1 percent of its peak value. The rod curve, the scotopic function, is at 0.000313, so 0.031 percent of its peak. Divide one by the other and the rods are getting 0.0051 of what the cones are getting, relative to their own maxima.

Now convert that into the thing you actually want to know, which is the rod cost of a published lumen. A lumen is 683 times the photopic function times the radiant watts. Rod drive is conventionally 1700 times the scotopic function times the same watts. So rod drive per lumen is 1700 over 683, which is 2.489, multiplied by the ratio of the two functions. Scale the answer so a 555 nanometre source comes out at exactly 1, which it does almost perfectly, since 2.489 times 0.402 is 1.001.

Run four beams through it:

Deep red at 660 nanometres: 2.489 times 0.000313 divided by 0.061, which is 0.0128.

Green at 530 nanometres, a typical green LED: 2.489 times 0.811 divided by 0.862, which is 2.34.

A cool white LED around 6000 kelvin: about 2.0, because a cool white is built from a blue pump and therefore carries a lot of short wavelength energy that rods see well.

A warm white around 3000 kelvin: about 1.3.

So deep red is 78 times gentler than a 555 nanometre source and 183 times gentler than a green beam of the same published lumens. That is the answer to the question and it is a very large number, which is why the technique survives in astronomy, aviation, submarines and darkrooms rather than just in folklore.

The correction: green is not a night colour

There is a claim that usually travels alongside this one, and I had it in my own notes for this page before working it through: that the same spectral shift explains why a green beam looks brighter than a white one of the same measured output at night, so green is the night-friendly alternative to white. The first half is technically true and far smaller than it sounds. The second half is backwards.

Look at the table. Green at 2.34 against cool white at 2.0 is a ratio of 1.2. That is a twenty percent difference, which is nothing you would notice and nothing worth designing a product around. The reason is that a cool white LED is not a neutral comparison: it is made by pumping a phosphor with a blue die, so it is already rich in exactly the short wavelengths the rods are most sensitive to. Its scotopic content is high. Against a warm white at 1.3 the green does win more clearly, by a factor of 1.8, and against an old incandescent source even more. But against the cool white that modern lights actually emit, the green advantage substantially disappears.

And in the direction that matters, green is worse than both whites. A green beam costs your dark adaptation 2.34 per lumen against warm white's 1.3. If your purpose is to keep your rods adapted, green is the wrong choice, and it is wrong by nearly a factor of two against the gentlest white you can buy.

So why do lights have green beams? Because green is not on the list for night vision reasons. It sits close to the cone sensitivity peak, which means it produces more lumens per watt than any other colour, and it penetrates haze and foliage better than white while reading clearly against vegetation. Those are efficiency and contrast arguments, and they are good arguments. They are not dark adaptation arguments, and no listing we carry claims they are. The confusion is the reader's to avoid.

The light whose colours are the feature, and what it does not have

An ERITOM rechargeable EDC flashlight with white, UV and green beam is the one product in our catalogue where the choice of emitter colour is the headline, so it is the one to test this against. Its record publishes a 1000 lumen white, a colour temperature of 5700 to 6200 kelvin, separate UV and green beams, seven modes, 36 selectable brightness levels across high, medium, low and ultra-low, a stated 22 hour ultra-low runtime, a hidden waterproof USB-C port and a magnetic base.

Read that against the table and three things follow. The 5700 to 6200 kelvin white is squarely the cool white case, so it is the costliest of the whites for dark adaptation at about 2.0 per lumen. The green does not improve on it meaningfully, coming in at 2.34, which is marginally worse. And the UV is not a night vision option at all in either direction, because the eye's lens absorbs most of it before it reaches the retina, so what you see from a UV source is fluorescence from whatever you point it at rather than the beam.

What the light does have, and what genuinely matters here, is the 36 level ultra-low mode. Output is the dominant variable and colour is the refinement, so a light that can go very dim is doing more for your night vision than a light with an extra emitter. Its record is candid that "lumen and runtime figures are maker-stated" and that "compact turbo output usually steps down as it warms", both true. The gap, for this subject, is that the listing publishes a maximum of 1000 lumens and no minimum, so the one specification that would tell you whether ultra-low is genuinely low is missing. Details in our ERITOM EDC flashlight review.

The light that does have the right colour is the smaller one. FOSFIS 500 lumen rechargeable EDC clip-on flashlight publishes a dual-bead LED with seven modes "including red and red-blue flashing", 500 lumens maximum, 120 metres of beam distance, IP68, about 53 grams and a stated 22 hours on low. The red mode is almost certainly there as a signalling and marker feature rather than as a dark adaptation feature, and it does not matter why it is there. On the arithmetic above, that red mode is the single most night-friendly setting across every light we carry, by roughly two orders of magnitude per lumen. Our FOSFIS EDC flashlight review has the rest. The honest caveat is that no listing publishes the wavelength of a red mode, and a red-orange emitter at 620 nanometres is roughly twenty times worse than a deep red one at 660, so "red" on a box could mean either.

The only listing on the site that publishes a wavelength

Which brings us to the exception, and it is worth dwelling on because it shows what the category could look like.

A 365nm rechargeable UV black light flashlight names a number. Its specification reads "UV wavelength: 365nm", its title carries it, and its summary explicitly contrasts that with "a typical 395nm black light". Across 1,412 published products this is the only one that tells you where on the spectrum its light sits, and the distinction it draws is a real one rather than marketing.

Here is why 365 beats 395 for the job, which the listing asserts without explaining. At 395 nanometres a meaningful fraction of the output is inside the visible range, so you see a violet glow from the beam itself, and that glow lands on the same surface you are trying to inspect and washes out the faint fluorescence you are looking for. At 365 nanometres the emission is almost entirely outside what the eye responds to, so you see the fluorescence against a dark background rather than against a violet one. Signal against background, not output. The listing's claim that 365 is "purer" is correct and for a better reason than purity.

There is also a point here that sharpens the whole page. Short wavelength sources are the case where the lumen figure is most misleading about rod effects, because the cone curve collapses faster than the rod curve does as you move toward violet. A violet source therefore produces very few lumens for a great deal of rod stimulation, so a light that reads as nearly zero lumens can still be doing real damage to your adaptation. That is the structural flaw in using lumens at night, in its most extreme form. Its record is properly firm that the light "must never be aimed at human or pet eyes", which is the thing to act on. Our 365nm UV black light review covers it.

Why the output number matters more than the colour, with the arithmetic

Colour is the refinement. Output is the whole game, and our two lanterns show how badly the published figures miss it.

Start from what you are aiming at. Full moonlight gives roughly 0.25 lux, and that is ample to move around by once you are adapted. A lantern radiates in roughly every direction, call it 12.6 steradians. To deliver 0.25 lux at a metre and a half you need 0.25 times 1.5 squared, which is 0.5625 candela, and 0.5625 times 12.6 is about 7.1 lumens. Seven lumens. That is the correct size of a light that will not undo your adaptation.

Now a rechargeable dimmable LED camping lantern 2 pack, which publishes 300 lumens, stepless dimming, three colour temperatures, a 4000mAh cell and up to seven hours on its brightest setting. At full output that is 300 over 12.6, which is 23.8 candela, giving 10.6 lux at a metre and a half, which is 42 times full moonlight. To reach the 7 lumen figure you need to dim it to 2.4 percent of maximum. So the stepless dimming is the genuinely important specification on this product, and whether it is useful depends entirely on a number the listing does not publish: the minimum output. Stepless down to 20 percent is a different product from stepless down to 2 percent, and both are described the same way. Our dimmable camping lantern review notes that its runtime figures are maker stated, which they are.

And the case that shows the scale of the mismatch. A RadiantFlux rechargeable solar camping lantern publishes 4000 lumens, a 6600mAh cell, three light modes, IP65, solar and Type-C charging and a USB output for charging a phone. Four thousand lumens into 12.6 steradians is 317 candela, which is 141 lux at a metre and a half, or 564 times full moonlight. To get to the 7 lumen target you would need 0.18 percent of its maximum, and it publishes three modes rather than a dimmer, so its lowest setting is almost certainly far above that. Its record is straightforward that "per-mode runtimes are not specified", which is the related gap. None of this makes it a bad lantern. It makes it a work light and a power bank, which is what its listing is really describing, and 4000 lumens at a campsite table after dark is a specification pulling against the purpose. See our RadiantFlux solar camping lantern review.

The clock you are protecting

One more set of figures, because they explain why this is worth any care at all.

Dark adaptation is not a dimmer sliding open. It is a chemical process: rhodopsin, the pigment in your rods, is bleached by light and has to be regenerated, and the regeneration has a half time of roughly five minutes. Four or five half lives gets you most of the way, which is why full dark adaptation takes 20 to 30 minutes. The cone branch of the curve finishes in about five to seven minutes and then the rod branch takes over, which is the kink people notice as their vision suddenly improving again after they thought it had settled.

What you gain across that half hour is about a millionfold increase in sensitivity, six orders of magnitude. What it costs to discard is one brief look at a white beam. Half a second of bright white light and you are back near the start of the curve, so you spend half a second and repay 1,800 seconds. A ratio of 3,600 to one is why the habit of handling a light carefully at night is not fussiness, and why the cost is asymmetric in a way almost nothing else is: being careful costs you nothing, and one lapse costs you the whole evening's adaptation.

The honest limit

Not one flashlight or lantern listing we carry publishes a wavelength for a coloured mode, only colour names, and the single exception above is a UV inspection light rather than a general purpose one. Nobody publishes a scotopic figure or an S over P ratio, which is the specification that would answer this question directly and which exists and is used in street lighting work. Nobody publishes a minimum output, which on the arithmetic above matters more than the maximum everybody leads with. So the numbers on this page are computed from the standard sensitivity functions and from published colour temperatures and lumen figures, not measured on any of these lights.

Our own post on whether more lumens mean a headlamp shines further had part of this right already and deserves the credit. Writing about a lantern's modest output, it says the job "is not to illuminate anything. It is to be a visible warm object on a table while leaving people's night vision intact, and 1000 lumens pointed at a dinner would be a fault rather than a feature". Both halves are correct, and the arithmetic above supplies the figure behind the instinct: the target is around seven lumens, and both of our lanterns publish maxima more than forty times that.

Which leaves the free techniques, and they are worth more than any purchase here. Turn the light down before you turn it on, not after, because the damage is done in the first instant. Point it at the ground or at what you are doing rather than anywhere near another person's face, since a beam that never enters an eye costs that eye nothing. Close or cover one eye when you have to use a bright white light, which preserves adaptation in the other and is the oldest trick in this subject. And if you have no red light at all, a layer or two of red cellophane or a red balloon stretched over the lens does the same job as a red emitter, because what you need is not a red source but the absence of everything below 600 nanometres, and a filter achieves that by subtraction for the cost of nothing.