Why Do Colours Look Grey When I Dim the Wall Light?

Colours look grey on a dim wall light because your eye has swapped receptors, not because the bulb is dying or the paint was mixed wrong. Cone vision, the kind that carries hue, needs a working level of light. The Illuminating Engineering Society defines photopic vision as vision mediated essentially or exclusively by the cones, generally associated with adaptation to a luminance of at least 5 cd/m². Turn the sconce right down and most of what you are looking at slides under that figure.

The direction of the damage is the tell. Red goes first and blue holds on, which is why the curtains can look brighter than they did at breakfast while the red spine on your book reads almost black. Houseland’s Globe Wall Sconce lists a G9 socket with a 7W LED maximum and a beaded pull chain for its switch, on or off, with no dimmable rating published, so the dimming is happening in your bulb and your circuit, not the fixture. Swapping the bulb changes none of it.

Why do the colours look grey when I turn the wall light right down?

Two receptor systems hand over between roughly 5 cd/m² and 0.005 cd/m², and the handover changes which wavelengths look bright. The CIE has a name and a formal definition for precisely what you are watching. Its entry for the Purkinje phenomenon reads: a reduction in the brightness of a predominantly long-wavelength colour stimulus relative to that of a predominantly short-wavelength colour stimulus, when the luminances are reduced in the same proportion from photopic to mesopic or scotopic levels. Note the same proportion: the light is cut evenly across the room, and the reds still lose.

The note under that definition gives the machinery: in passing from photopic vision to mesopic or scotopic vision the spectral luminous efficiencies change, and the wavelength of maximum efficiency is displaced towards the shorter wavelengths. Your peak sensitivity walks toward blue as the dial goes down. Nothing inside the lamp moved.

The band this happens in is not exotic. IES TM-12-12 and CIE 191-2010 formally define the mesopic range as 0.005 to 5.0 cd/m², and a 2018 review in Annual Review of Vision Science describes mesopic vision as running from moonlight to dusk and including most artificially lit nighttime environments, spanning 20 to 30% of the light levels over which vision operates. A quarter of your visual life sits in the register where hue is unreliable, and most of that quarter happens indoors after dark.

Why does the blue curtain look brighter while the red book goes black?

Rods are far more sensitive to blue than to red, so as they take over they hand blue the brightness they take from red. Anstis and MacLeod, in Journal of Vision in 2015 on why hearts flutter in dim light, built their experiment on that asymmetry: the effect needs a red that is lighter than the blue as seen by the cones, but darker than the blue as seen by the rods. That is your bedroom at eleven at night.

They also timed the two systems. Using red and blue spots they measured the rod response lagging the cone response by about 52 ms, which is why dim moving objects seem to wobble out of register. You will not notice 52 ms beside the bed. You will notice the yellow cushion going beige, because yellow is long-wavelength too and loses brightness for the same reason.

Which surfaces fall out of colour first, and at what light level?

A dark saturated surface drops below the 5 cd/m² line while the white page beside it is still comfortably in cone territory. The usual advice misses that, and arithmetic settles it. Cornell’s lighting notes give the standard relation for a matte surface: luminance = illuminance × reflectance / pi. Paint chips already print the reflectance as light reflectance value, 0 for black and 100 for white. Four surfaces, one bedside lamp, three levels, in cd/m²:

At 30 lux the page is photopic and the curtain and the spine are not, so half the room is being read by a system that reports no hue. At 100 lux the deep red is still under the line on its own. Only in the 300 lux column does every row clear 5 cd/m², and that number is worth holding onto, because separate research lands on it from another direction entirely.

A warmer bulb vs more light: which one brings the colour back?

More light wins by a distance, and the warmth of the bulb barely registers. The rule everyone repeats here, that low light has to be warm light, traces to a 1941 graph by Kruithof. Steve Fotios rebuilt it from the empirical record in LEUKOS in 2017 and it did not survive: of nine studies judged credible, five conclude that they do not support Kruithof, one gives partial support, the rest are unclear. None supports it.

What survives has no colour in it. Fotios concludes that variation in CCT has a negligible effect on ratings of brightness and pleasantness, that illuminances under approximately 300 lux may be perceived as unpleasant, and that 500 lux is sufficient. He quotes Boyce and Cuttle, 1990: once the subject is fully adapted to the conditions, the CCT of good colour rendering lamps in the range 2700K to 6300K has little effect on people’s impressions of the lighting of the room. Dim is the problem, not warm.

The third row is the one to act on. The CIE definition of colourfulness notes that for a stimulus of given chromaticity it typically increases as luminance rises, except at very high brightness; the IES calls this the Hunt effect, colourfulness of chromatic objects increasing with luminance even though chromaticity stays unchanged. Dimming buys atmosphere and pays for it in colour. That is a trade, not a fault.

Does a high CRI number protect the colour at low brightness?

No, because CRI describes the lamp’s spectrum and has no term at all for how far down you have turned it. The CIE defines Ra as the mean of the CIE 1974 special colour rendering indices for a specified set of eight test colour samples. Eight patches, averaged against a reference. Nothing in that calculation knows whether the surface in front of you lands at 80 cd/m² or 0.8. A CRI 95 lamp at 10% output greys out much like a CRI 80 one; the failure is behind your own cornea.

Compensating for the level rather than the spectrum is hard enough to be patentable. Kevin Houser at Penn State built a variable-spectrum technology on the observation that objects that appear vivid and colorful at high light levels become muted and shift toward gray at low light levels, retuning the LED mix as output falls. No bulb on a shop shelf does that.

Houseland publishes no CRI figure, no lumen output and no Kelvin rating for this sconce. The spec table says only that the bulb is not included and that a 3-colour-temperature G9 LED gives white, natural and warm modes as photographed. Those numbers belong to the bulb you buy, so read them off that box, and treat any spectral claim about a fixture sold without one as guesswork.

From the shop

Globe Wall Sconce with Switch, Pull Chain Glass Wall Lamp, 4 Shapes 4 Colors

The Globe Wall Sconce takes a G9 bulb, not included, with a 7W LED maximum, and its switch is a beaded pull chain with a small wood teardrop pull. Every design stands 5.9 in (15 cm) tall around a frosted opal glass globe of about 4.7 in (12 cm).

View the pull-chain globe sconce — 17 colourways & current price

How do I get the colour back beside the bed?

Put light on the thing you want to see rather than raising the whole room, and stop buying bulbs to solve it. Five steps, in the order that saves money:

1. Measure before you spend — hold a phone light meter flat on the open page at your reading level. Under 100 lux, the table above says the dark colours in that room are already out of cone range.
2. Aim, do not brighten — a lamp pointed at the page puts far more lux on the book per watt than one washing the wall. Our notes on height and angle for reading in bed cover where it has to sit to land on the page rather than your shoulder.
3. Split the job across two lights — one reaching 300 lux on the page, one staying under 5 lux for the last ten minutes before sleep. Making one lamp do both is what created the problem. The wall lamps collection is where the second one comes from, and a low second point of light is the job the bear sconce in a toddler bedroom already does.
4. Choose the shade colour for the level you will actually use — a saturated shade for a room you light properly, something above LRV 50 for a corner you keep dim, because a high-LRV surface stays photopic longer. The same logic runs through whether a rust brown disc reads orange or red.
5. Give your eyes a minute — Fairchild and Reniff, in JOSA A in 1995, found chromatic adaptation at constant luminance was 90% complete after approximately 60 s. Walk in, wait a minute, then look. Half the “this paint is wrong” verdicts are delivered inside ten seconds.

What I'd do

Leave the bulb alone and add a second light. The room you want at bedtime and the room you can judge a colour in are different rooms, and one dimmer cannot be both: below about 5 cd/m² the rods are answering, and they carry no hue.

My own bedside runs two levels: a pull-chain sconce by the pillow for the last stretch of the evening, and a second lamp aimed at the page that gets me past 300 lux on the paper. Houseland’s Globe Wall Sconce suits the first job, standing 5.9 in (15 cm) tall in every design with a frosted opal glass globe of about 4.7 in (12 cm), and its listed light coverage of 3 to 5 square meters, roughly 32 to 54 sq ft, is a bedside footprint rather than a room-lighting one.

If the pink still bothers you at low output, change the shade colour, not the bulb.