Short answer
Colour is a response of a visual system, and other animals' visual systems differ from ours in the wavelengths they detect, the number of receptor types they compare and how they process the result. Most birds see ultraviolet and compare four cone types, so plumage that looks identical in both sexes to a person is frequently different to the birds themselves. No photograph or screen, built around three human cone responses, can show that difference directly.
A human eye samples the spectrum with three cone types and sees roughly 400 to 700 nm. Most birds have four cone types used for colour, one of them sensitive in the ultraviolet or violet, so their range extends below 400 nm and their colour space has an extra dimension: there are mixtures of ultraviolet and longer wavelengths that a bird can distinguish and that have no counterpart in human experience. Many insects, fish and reptiles also see ultraviolet. Most mammals have only two cone types. A given feather or petal therefore produces a different pattern of receptor signals in each viewer. The surface's reflectance spectrum is a physical fact; its colour is not one thing.
Ornithologists long classified species as sexually dichromatic, where males and females differ in colour, or monochromatic, where they look the same. In 2005 plumage reflectance from 139 passerine species regarded as monochromatic was run through a model of avian colour discrimination. More than 90% of the species were predicted to be distinguishable by sex to a bird. The differences were not confined to the ultraviolet; many lay within the human-visible range but below the threshold a human eye can resolve. The result is a model prediction, not a behavioural demonstration that the birds use the difference, but it removed the assumption that human categories describe what birds see.
It is tempting to rank animals by receptor count. Mantis shrimp are the corrective. Some species have 12 photoreceptor types sampling from about 300 to 720 nm, and were widely assumed to have extraordinary colour discrimination. Behavioural tests found the opposite: they discriminate between wavelengths poorly, far worse than humans do with three cone types. The explanation proposed is that they do not compare receptor outputs against one another in the usual opponent fashion, and instead recognise colours from the pattern of activation directly — fast, and coarse. Receptor count describes the sampling. What an animal can tell apart depends on what its nervous system does with the samples, and that has to be measured behaviourally.
The standard tool is the receptor-noise-limited model, which predicts whether two spectra can be told apart from the viewer's receptor sensitivities and an estimate of noise in each receptor channel. It reproduces measured thresholds in several di-, tri- and tetrachromatic species and underlies most statements of the form 'conspicuous to a bird'. Plotting plumage spectra in avian colour space this way shows that feathers occupy only a minority of the colours a bird could in principle distinguish, with structural colours reaching regions pigments cannot. The model needs a light source, a background and receptor data for the species, often borrowed from a relative. It predicts discriminability under bright light. It does not say what the animal attends to, prefers or remembers.
Receiver psychology, including how readily a viewer learns and remembers a signal, has been argued to shape how signals evolve. A model of the retina says nothing about learning or memory.
A display has three primaries chosen so that mixtures of them stimulate human cones in the ratios natural colours would. It emits no ultraviolet, and a bird's fourth cone would have nothing to respond to; even within 400–700 nm, a mixture that matches a feather for a human does not generally match it for a bird, because the match depends on the observer's receptors. 'Bird-vision' or 'bee-vision' images are false-colour maps: an ultraviolet channel is assigned to a visible primary so that a human can see where a pattern exists, not what it looks like. Angle-dependent and polarised signals are lost as well. What can be shown faithfully is a reflectance spectrum, and what can be stated is which viewers could tell two spectra apart.
Why: Human vision and consumer cameras are both blind to ultraviolet and tuned to three human cone responses.
Fix: Look for reflectance spectra measured from about 300 nm, analysed with a model of the relevant viewer's receptors.
Why: The image maps an ultraviolet channel onto a visible colour so humans can see it.
Fix: Read it as a map of where ultraviolet patterns are, and nothing about how they appear to the insect.
Why: Receptor count is equated with discrimination ability.
Fix: Check for behavioural discrimination data. Mantis shrimp, with 12 receptor types, discriminate wavelengths worse than humans.
Why: A camera match is a metameric match for the camera's three sensors only.
Fix: Compare the reflectance spectra themselves across the viewer's whole visible range.
Each statement is labelled by kind — established fact, a standard’s requirement, observed market data, a convention, or Colourwise’s own interpretation or analysis — with the strength of the evidence behind it.
FactStrong evidence
When plumage reflectance from 139 passerine species regarded as sexually monochromatic was analysed with a model of avian colour discrimination, more than 90% were predicted to be sexually dichromatic from an avian visual perspective.
Measured: 139 passerine species.
Caveat: A visual-model prediction from museum-skin reflectance, not a behavioural test of what birds discriminate or use.
FactStrong evidence
Most birds can see ultraviolet light, which humans with normal vision cannot, and ultraviolet-reflecting plumage is widespread among birds.
Source: Ultraviolet vision in birds: what is its function?; Human vision fails to distinguish widespread sexual dichromatism among sexually “monochromatic” birds
FactStrong evidence
Mantis shrimp with 12 photoreceptor types sampling from about 300 to 720 nm performed surprisingly poorly in behavioural wavelength-discrimination tests, ruling out a conventional colour-opponent system.
FactStrong evidence
Colour discrimination thresholds in di-, tri- and tetrachromatic animals can be modelled by assuming that they are set by noise in the photoreceptors, which allows discriminability to be predicted for species where receptor data exist but perceptual thresholds have not been measured.
Caveat: Developed for bright-light conditions and for discrimination of similar colours; it does not predict preference or attention.
Source: Receptor noise as a determinant of colour thresholds; Animal colour vision – behavioural tests and physiological concepts
Colourwise interpretationStrong evidence
A three-primary display cannot reproduce how a surface appears to a tetrachromatic or ultraviolet-sensitive animal; images labelled as showing another species' vision are false-colour visualisations for human viewers.
Based on: Follows from the principle that a colour match is defined relative to an observer's receptor sensitivities, and from the absence of ultraviolet emission in display primaries defined for human vision.
Source: Animal colour vision – behavioural tests and physiological concepts; CSS Color Module Level 4
Reviewed 6 October 2026. Colourwise summarises its sources in its own words and does not reproduce standards text or proprietary colour data. Spotted an error? Tell us.