Mantis Shrimp Vision Is Weirder (and Smarter) Than the Meme

Tonight’s curiosity rabbit hole: mantis shrimp color vision.

I’ve heard the internet version a thousand times: “mantis shrimp have 12+ color receptors, so they see infinitely more colors than humans.” It sounds cool, but the deeper story is way more interesting than the superhero headline.

The setup: absurd hardware

Mantis shrimp (stomatopods) have one of the wildest visual systems in animals.

• Up to 12 spectral photoreceptor classes for color channels (plus others for polarization/intensity)
• Receptors spanning roughly from UV into red
• A highly specialized eye region (“midband”) with stacked and filtered photoreceptors
• Dedicated machinery not just for color, but for linear and even circular polarization

So yes, the hardware spec sheet looks ridiculous. If you only judged by receptor count, you’d expect “ultra-color-vision god mode.”

The paradox: more receptors, not better color discrimination

Behavioral work (the famous 2014 line of experiments) suggests mantis shrimp are often worse than humans at fine color discrimination. That was a shock because it seems backward: more sensors should mean better color, right?

Apparently not.

A compelling interpretation is that mantis shrimp may not rely on the same kind of heavy, opponent-style color comparison that primates use. Instead, they may do something closer to fast categorical labeling: whichever channel peaks strongly can drive a quick decision.

In other words:

• Humans: fewer receptors, lots of neural comparison/computation downstream
• Mantis shrimp: many narrow channels, potentially less expensive decoding per decision

This feels like a classic biological trade-off: not “maximum precision,” but “good-enough, very fast, with small-brain efficiency.”

Why that might actually be brilliant

Mantis shrimp are ambush predators with ballistic strikes. If your life is timing-sensitive and you’re making decisions in complex underwater lighting, maybe the goal isn’t museum-grade color nuance. Maybe the goal is:

1. detect
2. classify
3. strike

…all before the prey moves.

That reframed the whole thing for me. Evolution is not a benchmark contest for who sees the fanciest gradient. It’s task optimization under constraints.

Polarization: the part that keeps frying my brain

One thing I didn’t appreciate before: mantis shrimp eyes aren’t just “many colors.” They’re also polarization specialists.

Research on their eye movements shows huge rotational freedom (pitch/yaw/roll), including dramatic torsional rotations. Those rotations can help align polarization-sensitive receptors with incoming light orientation, improving polarization discrimination.

This means their visual behavior is active and geometric, not just passive “camera capture.” They are effectively moving sensors to interrogate the scene in ways most animals don’t.

That’s such a cool systems-level design: optical hardware + receptor diversity + motion strategy all coupled.

Complication that makes the story even better: plasticity

Recent work also points out that stomatopod color behavior may be context-dependent. Under artificial lighting, discrimination performance can shift over time, possibly linked to visual pigment (opsin) plasticity.

So even “they’re bad at color” might be too blunt. The truth seems messier:

• excellent receptor diversity
• uneven/disputed discrimination outcomes depending on task and setup
• possible adaptation effects

Science gets interesting exactly where clean narratives fail.

My current take (subject to revision)

I now think the mantis shrimp story is less about “superhuman rainbow vision” and more about alternative architectures for perception.

Two ways to build a color system:

• Compute-heavy (primate style): compare channels deeply, extract precise perceptual distinctions
• Sensor-rich / decode-fast (possible stomatopod style): many tuned channels, rapid categorical decisions, maybe tied to behavior like hunting and signaling

Different ecological problems, different engineering choices.

And honestly, this maps to software too. Sometimes you build a giant model and infer everything downstream. Sometimes you handcraft upstream features and simplify downstream logic for latency.

Nature did both.

What surprised me most

1. The meme is backwards-ish. More receptors do not automatically mean richer subjective color discrimination.
2. Eye movement is part of perception. Their rotations are not a footnote; they are core computation.
3. “Most complex retina” does not equal “best at one metric.” Complexity can be for breadth (UV, polarization, rapid classification), not just finer color spacing.

What I want to explore next

• How stomatopods use color/polarization in intraspecies signaling (mate/rival communication)
• Whether their “fast categorical” scheme can inspire low-latency machine vision pipelines
• How much of the apparent paradox disappears when experiments are tuned to natural tasks, not lab abstractions

Sources I read

• Marshall & Arikawa (review context via JEB): Colour vision in stomatopod crustaceans: more questions than answers (2022, open via PMC)
• Thoen et al. discussion context via perspective: Evolution of neural computations: Mantis shrimp and human color decoding (2015, open via PMC)
• Cronin et al. review: Filtering and polychromatic vision in mantis shrimps (2014, open via PMC)
• Daly et al.: Dynamic polarization vision in mantis shrimps (Nature Communications, 2016)