The mantis shrimp has sixteen types of photoreceptors. You have three.
The natural inference: mantis shrimp see more color. More channels, more resolution, richer experience of the spectrum. This is the inference nearly everyone draws, including many scientists before the behavioral studies were done. It seemed obvious. More detectors means more information. What else could it mean?
In 2014, Thoen et al. trained mantis shrimp to discriminate between wavelengths in a behavioral task — press the button on the side with the right color — and measured how close two wavelengths could get before the animal stopped distinguishing them. Humans, with our three cones, can tell apart wavelengths separated by about 1 to 2 nanometers. The mantis shrimp, with sixteen receptor types, started failing at around 25 nanometers. They weren't just slightly worse. They were an order of magnitude worse.
More hardware. Less resolution.
The way human color vision works is through comparison. You have three cone types — S, M, and L — with overlapping sensitivities peaking at roughly 420, 530, and 560 nanometers. You don't read color off any single cone. You compute it from the ratios between them. A spectrally pure yellow at 580nm and a mixture of red and green light that together produce the same cone activation ratios look identical to you, even though they're physically different stimuli. Color is not in the photons. It emerges from the opponent computation — L minus M, producing a red-green axis; S minus (L+M), producing a blue-yellow axis. Fine discrimination comes from the precision of those ratio comparisons near the neutral point, where small spectral shifts produce the largest changes in the ratio.
Three channels plus comparison logic equals 1-2nm resolution.
The mantis shrimp's sixteen channels are anatomically arranged in four specialized rows of the midband region of the eye. Each row covers a narrow slice of the spectrum. They span roughly 300 to 700 nanometers — well into ultraviolet, far beyond what you can see. Twelve of the channels handle visible and UV color; four more deal with polarization. Each channel has a narrow tuning curve. They don't overlap much. The natural use of a system like that is not comparison across channels — there's not enough overlap to make useful ratio comparisons — but identification by which channel fires.
The proposed mechanism: as the shrimp scans its eyes across an object (they move their eyes in coordinated scanning motions), each spectral row sweeps across the surface in sequence. The result is a temporal pattern of receptor activation, a kind of chromatographic readout. Which channels light up, in what order, at what intensity — this is the "color" of the object, experienced as a pattern rather than a computed value. A barcode, not a calculation.
This is a fundamentally different architecture for the same apparent problem.
The human system asks: what ratio of cone activations is this? It excels at distinguishing very similar stimuli — detecting the subtle spectral difference between a ripe and slightly overripe piece of fruit, or the color of a face in shadow versus in sunlight. It achieves color constancy, the ability to judge that a surface is the same color under radically different illuminants, because the opponent computation cancels out many illuminant changes. This is a measurement system optimized for fine discrimination and illuminant-invariant judgment.
The mantis shrimp system asks: which pattern is this? It can't distinguish wavelengths 10nm apart. But it can, in principle, rapidly identify any stimulus whose pattern of sixteen-channel activation is distinct — and it can do so faster, without needing to compare two stimuli side-by-side. It recognizes rather than measures. A scanner, not a ruler.
The ecological use case matters here. Mantis shrimp live in complex reef environments where color vision serves primarily intraspecific communication. The meral spot — a brightly colored patch on the appendages — is displayed during threat assessments before a fight. Its UV reflectance signals aggression level; a high-UV display means the opponent is escalating. The animal reading the display needs to know what pattern this is, not how different it is from a nearby wavelength. Recognition, not discrimination. And notably, mantis shrimp have five dedicated UV photoreceptors — an investment far beyond what any color-discrimination task would require, but directly suited to reading a UV signal with fine categorical specificity.
The sixteen channels aren't measuring the spectrum. They're reading a code that evolution has been writing in their signals for a long time.
A 2022 review ends on a note of deliberate uncertainty: the experiments "have not clarified this debate." More recent behavioral work from 2024–2025 found evidence of spectral opponent processing in mantis shrimp after all — they can distinguish color from gray under colored illumination, which you can't do with a pure identification system. The current best hypothesis is a hybrid: opponent processing and pattern recognition running simultaneously, each contributing to different behavioral contexts. Which adds another layer — not just two different architectures for color vision, but an animal whose visual system apparently runs both.
There's also plasticity. Animals kept under artificial lighting for ten or more weeks show degraded color discrimination. Their opsin expression adjusts to the ambient spectrum. The sixteen channels are not fixed — the system recalibrates based on the light environment, which means previous studies of captive animals may have been measuring miscalibrated visual systems, not the real thing.
What I keep coming back to is how the first framing — sixteen channels, you have three, therefore mantis shrimp see more — makes color vision sound like a single scale with a clear direction of improvement. The behavioral results forced a different question: more what? More discrimination? No. More recognition capacity for biologically relevant signals? Probably. More UV sensitivity? Substantially.
Asking "which sees color better?" is like asking whether a dictionary or a thesaurus is more useful. The answer is useless without knowing what you're trying to do with words. The mantis shrimp's visual system was not designed to outcompete human color vision at human visual tasks. It was designed to read the specific signals that mantis shrimp produce for each other in reef environments, quickly, reliably, and without requiring a comparison object to be nearby.
Three channels and a calculator. Sixteen channels and a barcode reader. Both are color vision. Neither is more.