💻 computer science

What Color is the Sky (for a non-human) ?

This paper proposes a computational method to determine the monochromatic metamer (dominant wavelength) of skylight for various non-human visual systems, revealing that while metamers exist across diverse species, their perceived "sky color" often differs drastically from the human perception of blue.

Original authors: Yair Weiss, Ofer Springer

Published 2026-06-30

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Yair Weiss, Ofer Springer

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Sky's True Color: A Tale of Different Eyes

Imagine looking up at a clear day and seeing a vast, blue canopy. You might think, "The sky is blue." But this paper asks a fascinating question: Is the sky actually blue, or is it just blue to us?

The authors, Yair Weiss and Ofer Springer, argue that the sky isn't inherently blue. Instead, the sky is a chaotic soup of many different colors mixed together. It only looks blue because of the specific "cameras" inside our eyes.

Here is the simple breakdown of their research, using some everyday analogies.

1. The Sky is a Smoothie, Not a Single Flavor

Think of sunlight as a giant bucket of white paint. When it hits the atmosphere, the air molecules act like a sieve. They let the "heavy" colors (like red and orange) pass through easily, but they bounce the "light" colors (like blue and violet) all over the place.

This is why the sky looks blue. But here's the twist: The light reaching your eye isn't just pure blue paint. It's a smoothie made of violet, blue, green, and even a little bit of red, all blended together.

2. The Human "Camera" (The Cone Filter)

Our eyes have three types of light sensors (called cones) that act like colored filters.

The Analogy: Imagine you are looking at that sky-smoothie through a pair of sunglasses that only let blue light through. Even though the smoothie has red and green in it, your sunglasses filter them out, and you only see blue.
The Result: Because of how our eyes are built, that complex sky-smoothie triggers our brain exactly the same way as a single, pure beam of blue light (around 480 nanometers). In the world of color science, these two different lights are called metamers. They are physically different, but they look identical to us.

3. What About the Animals?

The paper asks: If a dog, a bee, or a butterfly looks at the same sky-smoothie, what do they see?

The authors built a computer program to simulate the eyes of different species. They treated the sky-smoothie as a fixed input and asked, "What single color of light would trigger this animal's eyes to feel the same as the sky?"

The results were surprising:

For Humans: The sky matches a pure blue light (approx. 479nm).
For a Honeybee: The sky matches a deep violet light (approx. 395nm).
For a Gecko: The sky matches a different shade of violet (approx. 409nm).
For a Dog: The sky matches a blue-green light (approx. 473nm).
For a Chicken: The sky matches a violet-blue light (approx. 431nm).

The Big Reveal: The "color" of the sky changes depending on who is looking at it. To a bee, the sky isn't blue; it's violet. To a chicken, it's a different shade of violet. There is no single "true" color of the sky; there is only the color the sky appears to be for a specific pair of eyes.

4. The Rainbow Test

The authors use a rainbow to explain this. In a rainbow, the "blue" part is actually a single, pure color (monochromatic). The sky, however, is a mix.

For us: The mixed sky light tricks our eyes into thinking it's the same as the pure blue part of the rainbow.
For others: That same mixed sky light might trick a bee's eyes into thinking it's the same as the violet part of the rainbow.

5. The "Knock-Out" Experiment

To prove that the eye's design is the deciding factor, the researchers created fake, imaginary eyes with random sensors.

They made eyes with 2 sensors, 3 sensors, and 4 sensors.
They randomly shifted the sensors to look at different parts of the light spectrum.
The Result: For some of these fake eyes, the sky looked like a deep violet. For others, it looked like a bright yellow or even green!

This proves that the physics of the sky (the light itself) doesn't decide the color. The hardware (the eye) decides the color.

The Bottom Line

The paper concludes that asking "What color is the sky?" is like asking "What is the volume of a song?" without specifying the speaker.

If you play the song on a tiny phone speaker, it sounds quiet.
If you play it on a massive concert system, it sounds loud.

The song (the sky's light) is the same, but the experience (the color) depends entirely on the speaker (the eye). So, for a non-human, the sky might be violet, green, or yellow. It is only blue because we are wearing human "sunglasses."

Technical Summary: What Color is the Sky? (for a non-human)

Problem Statement
The paper addresses the discrepancy between the physical composition of skylight and its perceived color. Daylight skylight is a mixture of many wavelengths, heavily weighted toward shorter wavelengths due to Rayleigh scattering ( $1/\lambda^4$ ), yet human observers perceive it as "blue." This perception arises because skylight and a specific monochromatic blue light (around 480 nm) are metamers for the human visual system; they produce identical excitations in human cone cells.

The central question posed is whether a "color" can be defined for non-human visual systems in the same way. Specifically, for arbitrary visual systems (bichromats, trichromats, and tetrachromats), is it possible to find a monochromatic light source that is metameric to skylight? If so, does the dominant wavelength of this metamer remain consistent across species, or does it vary significantly based on the species' spectral sensitivity functions?

Methodology
The authors propose a computational method to determine the dominant wavelength of a monochromatic metamer for any given visual system. The visual system is modeled as a set of $N$ linear sensors (cones) with spectral sensitivity functions $c_i(\lambda)$ .

Metamer Definition: Two light sources are metameric if they yield the same vector of cone responses. The goal is to find a mixture of white light ( $w$ ) and a monochromatic light source ( $p(\lambda_0)$ ) that matches the response of the input skylight ( $l$ ).
Linear Constraints: The problem is formulated as finding positive coefficients $\alpha$ and $\beta$ such that $C_l = \alpha C_w + \beta C_{p(\lambda_0)}$ , where $C$ represents the cone response mapping.
Algorithmic Approach:
- Trichromats ( $N=3$ ): The system of equations is overconstrained. The algorithm performs an exhaustive search over a range of wavelengths $\lambda_0$ to find the specific value where a solution with positive $\alpha$ and $\beta$ exists.
- Dichromats ( $N=2$ ): The system has two constraints and two unknowns, allowing for multiple solutions. To define a unique "dominant" wavelength, the authors constrain the solution to a pure monochromatic metamer ( $\alpha = 0$ ), seeking a $\lambda_0$ that satisfies the equations exactly.
- Tetrachromats ( $N=4$ ): An exact match with only white and monochromatic light is often impossible. The method minimizes a perturbation vector $n$ to find the "closest" metamer, solving for $\alpha$ , $\beta$ , and $\lambda_0$ that minimize the norm of the residual error.
Data Sources: The study utilizes published spectral sensitivity functions for various species (humans, butterflies, honeybees, geckos, dogs, gerbils, and chickens) and generates synthetic visual systems with randomly distributed Gaussian cone responses.

Key Results

Human Validation: The algorithm successfully replicates the known human result, identifying a monochromatic metamer for skylight at approximately 479–480 nm, confirming the method's validity.
Species Variation: When applied to other species, the dominant wavelength of the skylight metamer varies drastically:
- Butterfly (Lycaena Heterona): ~457 nm.
- Honeybee (Apis Mellifera): ~395 nm (violet range).
- Gecko (Gecko gecko): ~409 nm.
- Dog (Canis Familiaris): ~473 nm.
- Gerbil (Meriones unguiculatus): ~413 nm.
- Chicken (Gallus gallus domesticus): ~431 nm.
Synthetic Systems: Experiments with randomly generated trichromatic systems showed that the same physical skylight could be metameric to monochromatic lights ranging from 400 nm (violet) to 580 nm (yellow).
Evolutionary Model: A simulation of 26 synthetic visual systems with varying receptor combinations (UVS, SWS, MWS, LWS) demonstrated that the perceived "color" of the sky can range anywhere between violet and green depending on the specific receptor configuration.

Significance and Claims
The paper argues that the perceived color of the sky is not an intrinsic property of the physical light but is fundamentally determined by the response functions of the observer's photoreceptors. The authors claim to have demonstrated two main points:

It is possible to define a "color" for non-human observers without relying on linguistic color naming, by constructing monochromatic metamers.
The "color" of the sky is not universal; the dominant wavelength of the monochromatic metamer depends heavily on the form of the photoreceptors. Consequently, the same physical skylight can appear as violet, blue, green, or even yellow to different species.

The work challenges the assumption that physics alone dictates color perception, showing that the mapping from physical spectra to perceived color is highly variable across the animal kingdom.