If the world had only one colour, you would still see shape, depth, motion and most dangers, because the brain builds those mainly from differences in brightness, not hue. What you would lose is precise: about 2.28 million distinguishable colours would shrink to roughly 200 to 400 shades of one colour, so ripe and unripe fruit, healthy and pale skin, and red and green signals of the same brightness would look identical. People with complete achromatopsia, about 1 in 30,000 worldwide and about 5% on the Pacific atoll of Pingelap, already live in a world close to this.
- Under one wavelength, every surface gives the same cone ratios (the principle of univariance), so hue vanishes for every eye.
- About 2.28 million colours shrink to roughly 200 to 400 shades of one colour.
- That is a 6,000-fold cut in categories but only about a 60% cut in bits per point.
Calculated in this post from the 2.28 million colours counted by Pointer and Attridge (1998), as cited by Lee et al. (2024) and about 383 greys, computed below with these example inputs.
- Shape, stereo depth and motion survive, because they run on brightness.
- You lose ripeness, pallor, warning colours and any colour code not backed by position or pattern.
The question needs a precise meaning to be answerable. "Only one colour" could mean either of two things:
- every surface is painted the same hue in different lightnesses; or
- the world is lit by a single wavelength.
Physically, those two produce the same thing at the eye, so I'll use the second, because it can be built today: a street lit only by low-pressure sodium lamps, which emit almost all their light near 589 nm, is a one-colour world. Then the question becomes: what does vision do with one channel instead of three?
Why does one wavelength erase all colour, even for normal eyes?
Because each of the eye's three cone types reports only how many photons it caught, not which wavelengths they were. Under light of one wavelength, every surface changes the three cone signals in exactly the same proportion, so the ratios that the brain reads as hue never change.
This is the principle of univariance, set out by W. A. H. Rushton in 1972: a receptor's output depends on its quantum catch, but not on what quanta are caught. Here is what it means for colour.
- Cone signals. The response of cone type \( i \) (L, M or S) to a surface with reflectance \( \rho(\lambda) \) under illumination \( E(\lambda) \) is
\[ R_i = \int E(\lambda)\,\rho(\lambda)\,S_i(\lambda)\,d\lambda \]where \( S_i(\lambda) \) is the cone's spectral sensitivity.
- Make the light one wavelength. Put all illumination at \( \lambda_0 = 589 \) nm: \( E(\lambda) = E_0\,\delta(\lambda - \lambda_0) \). The integral collapses:
\[ R_i = E_0\,\rho(\lambda_0)\,S_i(\lambda_0) \]
- Look at the ratios. For any two cones,
\[ \frac{R_L}{R_M} = \frac{S_L(589)}{S_M(589)} \]which doesn't depend on the surface at all. Every object, whatever its pigment, produces the same cone ratios. Only the overall level, set by \( \rho(589) \), varies.
- Result. Chromaticity is constant everywhere; only lightness varies. A red sari, a green leaf and grey concrete differ only in how much 589 nm light they reflect.
That's why sodium-lit streets look eerie: every car, sari and signboard turns the same orange-grey. The same maths explains why white LED and white sunlight, which spread energy across the spectrum, give colour back. The design side of that, how colour is defined and measured, is in OKLCH and perceptual colour and white balance and colour temperature.
Knut Nordby, born in Oslo in 1942, was a complete achromat, as were his sister and brother, and he became a vision scientist himself. In a 1990 personal account he describes the morning he learned what univariance means. In his first school year, in Gothenburg, the teacher's alphabet cards had red vowels and black consonants. One late-autumn morning the room lights came on, and the vowels turned dark grey while the consonants stayed solid black. The same "colour" could match different greys under different light. He learned the world by rule instead: very dark glass was usually cobalt blue, lighter glass bottle green. In full sun he often could not tell which traffic light was lit, and once followed other pedestrians across a road against a red light.
Knut Nordby: Vision in a complete achromat, a personal account (in Night Vision, Cambridge University Press, 1990)
How much information would you lose?
You would lose a factor of several thousand in the number of distinguishable surface appearances, but only about 60% of the information in bits, because brightness alone still carries a lot. That gap between "colours lost" and "information lost" is why a one-colour world is still very usable.
Calculated in this post from the 2.28 million colours counted by Pointer and Attridge (1998), as cited by Lee et al. (2024) and the grey count worked out below, with these example inputs.
- Colours with normal vision. Pointer and Attridge (1998) counted colours within the limits of possible surface colours in a perceptually uniform colour space and got about 2.28 million; later estimates run from at least 1.7 million to about 2.8 million. In bits: \( \log_2(2.28\times10^6) \approx 21.1 \) bits per point.
- Greys in a one-colour world. Assume (my estimate) surfaces reflect between 2% (black cloth) and 90% (fresh white paint), a 45:1 range, and that you can tell apart brightnesses differing by a Weber fraction \( k \). Successive just-noticeable steps multiply by \( 1 + k \), so the number of levels is
\[ n = \frac{\ln(L_{\max}/L_{\min})}{\ln(1+k)} = \frac{\ln 45}{\ln 1.01} = \frac{3.81}{0.00995} \approx 383 \]With \( k = 2\% \), \( n \approx 192 \).
Calculated in this post from a 45:1 reflectance range and a 1% brightness step, both example inputs (my assumptions), compared with the 2.28 million colours counted by Pointer and Attridge (1998), as cited by Lee et al. (2024).
- Compare. \( 2.28\times10^6 / 383 \approx 6{,}000 \) times fewer distinguishable appearances. But \( \log_2 383 \approx 8.6 \) bits, so you keep about \( 8.6/21.1 \approx 41\% \) of the bits per point.
Sensitivity check: in a real scene with shadows, the luminance range can be 1,000:1 or more, which raises \( n \) to about 700 at \( k = 1\% \); colour counts also depend on the light source. The conclusion barely moves.
You lose most of the categories, not most of the information.
See the data as a table
| Bits per point ( bits) | |
|---|---|
| Full colour (2.28 million) | 21.1 |
| One colour, 1% threshold (383 greys) | 8.6 |
| One colour, 2% threshold (192 greys) | 7.6 |
Source: Colour count from Pointer and Attridge (1998), as cited by Lee et al. (2024); grey counts computed for a 45:1 reflectance range · Bits = log2(number of distinguishable values). The grey counts are my estimate; see the working.
How many shades would your one-colour world have?
Would you still see shape and depth?
Yes. Edges, shading, perspective, occlusion, motion and stereo depth are all carried mostly by brightness differences, and experiments that remove brightness differences while keeping colour show the reverse: depth and motion weaken badly when only colour is left.
Livingstone recorded evoked brain potentials in normal human subjects viewing two-colour random-dot stereograms that reversed in depth. When the two colours were set to equal luminance (equiluminance), the responses to depth reversals disappeared or were greatly reduced, while responses to other changes between the two eyes' images stayed, and for most subjects even grew. It extends her 1987 work with David Hubel, which proposed separate brain channels for form, colour, motion and depth. The implication for a one-colour world: stereo depth runs on luminance, which you keep. The limit: few subjects, specialised stimuli, and brain responses rather than everyday tasks.
Read the paper (free PDF)Photographers already rely on this. A black-and-white photograph keeps faces, buildings and landscapes perfectly readable, which is why it has never gone out of use. The test of good design is the same: a layout or logo that only works in colour usually has weak contrast, the point behind colour contrast, WCAG and APCA, which measures contrast in lightness, not hue.

What would you actually lose?
You would lose every judgement that depends on hue at constant brightness: ripeness, freshness, illness, warning colours, and colour codes. Some of these have real stakes.
| Task | Depends on | In a one-colour world |
|---|---|---|
| Reading shape, edges, faces | Luminance contrast | Kept |
| Stereo depth and motion | Luminance (equiluminant colour weakens both) | Kept |
| Spotting ripe mango or red chilli among leaves | Red–green contrast at similar lightness | Mostly lost |
| Noticing pallor, jaundice or a rash | Small hue shifts in skin | Mostly lost |
| Traffic signals | Hue, plus the position of the lit lamp | Kept by position; lost for lone flashing lights |
| Warning colours on insects, snakes, labels | Hue and pattern | Pattern kept, colour code lost |
| Seeing through colour camouflage | Texture, confused by colour | Sometimes easier |
| Wiring, charts, maps, metro lines | Colour coding | Lost unless also coded by pattern or label |
The last row is a design lesson that applies now, because many people already live partway there. In India, a survey of 13,492 schoolchildren in Uttarakhand found 3.0% of boys and 1.4% of girls affected.
India's transport ministry amended the driving-licence medical forms in June 2020 to allow people with mild to medium colour blindness to drive, with severe cases still restricted, which works because signals are also coded by position.
The team screened 13,492 students aged 6 to 16 (6,750 boys and 6,742 girls) in government and private schools, testing colour vision with Ishihara plates. Colour vision deficiency was found in 2.2% overall: 3.0% of boys and 1.4% of girls. The limit: one town, and a screening test (Ishihara plates) rather than a full diagnosis of type and severity.
Read the paper (free)Would anything get better?
A few things could: breaking colour camouflage, and seeing in low or harsh light if you depended on rods. The evidence on camouflage is mixed and depends on the scene.
In 1992, Morgan, Adams and Mollon showed in the lab that people with red–green colour blindness could find a texture-defined target hidden among randomly red and green elements better than people with normal colour vision, because the irrelevant colours did not distract them. Real scenes are more complicated, as a large online experiment found.
In two citizen-science games, players searched photos of camouflaged nightjars (39 images, 135,968 captures from 9,926 plays) and ground-nesting birds' eggs (88 images, 22,810 captures from 1,531 plays), seen either in normal colour or in a simulated two-colour (dichromatic) version. Normal colour vision found targets faster overall. But simulated dichromats learned to find egg clutches faster and were less thrown by background brightness, while missing eggs whose brightness was close to the background. The limit: simulated, not real, dichromats, on uncalibrated screens.
Read the paper (free)
What is life like for people who see no colour?
People with complete achromatopsia see in greyscale, but their bigger problems are not the missing colour. Their cones don't respond in daylight, so they rely on rods: visual acuity is typically 20/200 or worse, bright light is dazzling, and the eyes often oscillate (nystagmus). A one-colour world with normal cones would be far easier than that.
The best-known community is on Pingelap atoll in Micronesia, where about 5% of roughly 3,000 people are affected, traced to a population bottleneck after a typhoon and famine in the 1770s. Genes such as CNGB3 and CNGA3 are the usual causes. It is a reminder to separate two things the what-if mixes together: the eye's sensors and the world's colours. Remove colour from the world and your sharp cone vision stays. Remove the cones and you lose sharpness too.
A one-colour world is physically easy (light everything at 589 nm), and by the principle of univariance it erases hue for every eye: cone ratios become \( S_L(589):S_M(589):S_S(589) \) for every surface. You keep shape, stereo depth and motion, which run on luminance; you drop from about 2.28 million colours to roughly 200 to 400 greys, a 6,000-fold cut in categories but only about a 60% cut in bits per point. What you lose is specific: ripeness, pallor, warning colours and any colour code not backed by position or pattern.
Calculated in this post from the 2.28 million colours counted by Pointer and Attridge (1998), as cited by Lee et al. (2024) and about 383 greys at a 1% step, with these example inputs.
What's new
- Berkeley scientists stimulate only M cones with lasers and report a colour outside normal vision, "olo" UC Berkeley News, Apr 2025
- Leeds study maps how many discernible colours sit under each colour name Journal of the International Colour Association, Feb 2024
- Mild to medium colour blind citizens can now get a driving licence in India Deccan Herald, Jun 2020 (still the current rule)
What if next?
- What if you painted a rangoli a kilometre long on an Indian highway? The opposite world: as many colours as possible, seen from far away.
- What if you took a selfie from a billion kilometres away? Light so scarce that colour is the first thing to go.
- What if you could only shoot one focal length for a year? Another limit that sounds crippling and turns out to sharpen how you see.
The everyday version of this question, designing for people who see fewer colours, comes up in India's top brand colours. More from the studio at Kala.
Questions people ask
How do colour-blind people see the world?
Most colour-blind people are not colourless: the common red–green types are dichromats who confuse reds, greens and browns of similar lightness but see blues and yellows. Complete colour blindness (achromatopsia), seeing only greys, affects about 1 in 30,000 people and comes with poor acuity and light sensitivity.
Why does sodium street light make everything look orange-grey?
Low-pressure sodium lamps emit almost all their light near 589 nm. Under one wavelength, every surface stimulates the three cone types in the same ratio (the principle of univariance), so objects differ only in lightness, not hue.
Can you see depth without colour?
Yes. Stereo depth, shading, perspective and motion are carried mostly by brightness differences. Experiments show depth from two-colour stereograms fading when the colours are set to equal brightness, the opposite case.
How many colours can the human eye see?
The usual estimate is about 2.28 million distinguishable surface colours (Pointer and Attridge, 1998), with later estimates from at least 1.7 million to about 2.8 million depending on method and light source.
Sources
- Wikipedia: Principle of univariance (source for Rushton 1972 and the definition)
- Lee, Vazirian, Pan, Lai and Westland: A psychophysical analysis of the discernible palette for colour names, JAIC (2024) (source for 2.28 million colours, citing Pointer and Attridge 1998, and the 1.7 million lower estimate)
- Livingstone: Differences between Stereopsis, Interocular Correlation and Binocularity, Vision Research (1996) (source for depth responses vanishing at equiluminance)
- Mittal et al.: Ocular morbidities and color blindness among school children in Uttarakhand, Indian Journal of Ophthalmology (2022) (source for 2.2%, 3.0% and 1.4%)
- Troscianko et al.: Relative advantages of dichromatic and trichromatic color vision in camouflage breaking, Behavioral Ecology (2017) (source for the camouflage games)
- Morgan, Adams and Mollon: Dichromats detect colour-camouflaged objects that are not detected by trichromats, Proc. R. Soc. B (1992) (source for the camouflage advantage)
- Wikipedia: Achromatopsia (source for 1 in 30,000, Pingelap's 5%, acuity of 20/200 and the genes)
- Deccan Herald: Mild-medium colour blind citizens can obtain driving licence (2020) (source for the MoRTH rule change)
- Wikipedia: Sodium-vapor lamp (source for low-pressure sodium's near-monochromatic 589 nm light)
- UC Berkeley News: Scientists trick the eye into seeing a new color, olo (2025)
- Nordby, K.: Vision in a complete achromat: a personal account, in Hess, Sharpe and Nordby (eds), Night Vision, Cambridge University Press (1990) (source for the alphabet-card and traffic-light episodes)

