How Mantis Shrimp See: The Strange Truth About 16 Color Receptors

The mantis shrimp has acquired a reputation in popular science writing as the animal with the best color vision on Earth. Trichromatic humans have three types of color receptor; the mantis shrimp has 12-16, depending on the species and how you count them, plus four channels for ultraviolet vision and another four for polarization. The intuitive conclusion — more channels means better discrimination — turned out to be wrong. The animal apparently does worse than humans at distinguishing similar colors. The story of how that paradox was discovered, and what it implies about how nervous systems can process color information, is more interesting than the popular version.

The eye anatomy

Mantis shrimp (Stomatopoda) are not actually shrimp; they're a separate order of marine crustaceans that diverged from the rest of Crustacea around 400 million years ago. The 450 or so species are predators, ambush hunters that strike with appendages capable of producing some of the fastest movements in the animal kingdom — up to 23 m/s in peacock mantis shrimp (Odontodactylus scyllarus) — and producing cavitation bubbles whose collapse adds a second strike of force. To find prey to strike, the animal needs accurate spatial vision in addition to color discrimination.

The eye itself is a compound eye divided into three regions: a dorsal hemisphere, a ventral hemisphere, and a midband of six rows of specialized facets between them. The dorsal and ventral regions handle spatial and movement vision. The midband handles color. The midband's facets contain photoreceptors with different opsins (the proteins that absorb light and trigger neural responses), each tuned to a different wavelength. In the most thoroughly studied species, peacock mantis shrimp, the midband contains receptors with peaks at roughly 320 nm (UV), 340 nm, 360 nm, 380 nm, 400 nm, 420 nm, 450 nm, 478 nm, 500 nm, 525 nm, 550 nm, and 580 nm — 12 distinct color receptor types. Other species have somewhat different distributions; the total number across the order is around 16.

For comparison, humans have three color receptor types (peaks around 420 nm, 530 nm, and 560 nm) plus rod cells for low-light vision. Most birds have four. Bees have three. Among reasonably well-studied animals, only the mantis shrimp comes close to a dozen color receptor types.

The expected behavior, and the surprise

The natural assumption, given the receptor count, is that the mantis shrimp should be able to discriminate fine color differences much better than humans. The discrimination ability of trichromatic humans — being able to distinguish two wavelengths differing by 1-2 nm in the optimal middle range — is the result of comparing the relative responses of the three receptor types. With twelve types, the relative comparison should be vastly more sensitive.

The 2014 paper by Thoen, How, Chiou, and Marshall in Science directly tested this. The researchers trained mantis shrimp to associate a specific color with a food reward, then presented the trained color alongside other colors at varying spectral distance and measured discrimination ability. The result was unambiguous: mantis shrimp can reliably discriminate colors that are about 25 nm apart, which is roughly an order of magnitude worse than human discrimination. They are not extraordinary color discriminators by the conventional measure.

This produced something of a crisis in the popular interpretation. If mantis shrimp have all these receptors and don't discriminate well, what are the receptors for?

The wavelength labeling hypothesis

The hypothesis that emerged from the Thoen et al. paper, and from subsequent work by Marshall's lab and others, is that mantis shrimp use a fundamentally different color coding strategy than vertebrates and most other animals. Vertebrates use opponent processing: the brain compares the relative responses of three receptor types to compute a continuous color signal. The brain doesn't directly know which wavelengths are present; it knows ratios, and from ratios it computes color.

The mantis shrimp's hypothesized strategy is wavelength labeling: each receptor type signals presence of light in its specific wavelength range, and the brain reads out which receptors are active to identify the wavelengths present. This strategy doesn't require comparing across receptors; it requires only knowing which receptors are firing. The cost is poor fine-discrimination — a 25 nm difference might fall entirely within one receptor's range — but the benefit is fast recognition of specific wavelengths without complex neural processing.

The fast-recognition framing fits the mantis shrimp's ecological niche. The animal is an ambush predator that strikes in milliseconds and lives in colorful reef environments where prey, mates, rivals, and threats may be characterized by specific spectral signatures. Knowing whether a particular wavelength is present is more useful than knowing the exact wavelength to within 1 nm. The animal isn't optimizing for color discrimination; it's optimizing for fast color identification.

The polarization and UV components

The midband also contains receptors for linear polarization (sensitive to the orientation of light polarization) and circular polarization (sensitive to the handedness of polarization). Circular polarization detection is essentially unique to mantis shrimp among animals; humans cannot perceive it at all without instruments. The function appears to be communication — some mantis shrimp species have body parts that reflect circularly polarized light in patterns invisible to predators but visible to conspecifics. Marshall's lab has documented this in several species, suggesting that the polarization channel is partly a private communication channel.

The UV component, with four distinct UV receptors, also seems to function partly for fluorescent body markings that are visible in UV but invisible in normal vision. Reef predators that don't have UV vision can't see the markings, but mantis shrimp can. The general theme — using channels that are private to the species — is a useful evolutionary strategy because the signaling can't be intercepted.

The scanning behavior

One of the curious features of mantis shrimp visual behavior is that they actively scan their eyes across the visual scene rather than acquiring the whole scene at once. The midband, which does the color processing, is a narrow strip; to assess color across a scene, the animal moves its eyes back and forth so that the midband sweeps across the whole field of view. Each eye can move independently — left and right, up and down, even rotating — which gives the visual system a complex temporal structure that has only recently begun to be characterized.

This scanning behavior may be related to the wavelength labeling hypothesis. If color identification depends on which receptors fire, the animal needs to bring the relevant region of the scene under the midband to identify its color. The eye movements are part of the visual algorithm, not a separate behavior.

The unresolved questions

Several questions about mantis shrimp vision remain open. The wavelength-labeling hypothesis is a good fit for the discrimination data and the ecological context, but direct neural evidence is limited. We don't yet have detailed recordings from the visual processing centers of the mantis shrimp brain that show how the receptor signals are integrated. The animal is small and difficult to do electrophysiology on, and the relevant neural circuits are still being mapped.

The 2014 study has been replicated and extended, and the basic finding — surprisingly poor color discrimination given the receptor count — is robust across species and experimental designs. But there's some suggestion from later work (How et al. 2014, Thoen et al. 2018) that performance varies with task structure: discrimination is poor in the abstract two-color discrimination task used in the Science paper but may be better in more naturalistic tasks involving recognition of specific patterns. The question of what color resolution the animal actually uses in its natural ecology isn't fully settled.

The other unresolved question is the evolutionary trajectory of the receptor system. Most animals with multi-receptor color vision have three or four receptors; the mantis shrimp's twelve is anomalous. The lineage leading to modern stomatopods has had hundreds of millions of years of independent evolution, but the specific selection pressures that produced the receptor diversity aren't well understood. The wavelength-labeling strategy may be a stable solution to a particular ecological problem, or it may be a transitional state on the way to something else.

The deeper observation

The mantis shrimp is a useful corrective to the assumption that more channels means better discrimination. Vertebrate color vision is built around the principle that comparing relative responses of a small number of receptors produces fine-grained discrimination through computation; mantis shrimp color vision is apparently built around the principle that having many receptors at known wavelengths produces fast wavelength identification through direct readout. Both strategies work — they just optimize for different things. The vertebrate approach prioritizes discrimination at the cost of computational machinery; the stomatopod approach prioritizes speed at the cost of discrimination. Neither is better in any absolute sense; they're both well-fit to their respective ecological niches.

The broader lesson is that biology does not converge on single solutions to perceptual problems. The same input — light at various wavelengths — can be handled by very different neural strategies, each with different trade-offs. The popular science framing of mantis shrimp as having superhuman color vision was wrong, but the corrective truth is more interesting: a major animal lineage has been doing color vision in a fundamentally different way for hundreds of millions of years, and we have only recently begun to understand it. The reef contains aliens, and we are the ones with the impoverished receptor count.