Humans have three types of color sensors in their eyes. Some mantis shrimp have twelve types.

Many people thought these animals could see more colors than we can. In 2014, scientists tested this idea. They taught mantis shrimp to choose one color for food.

Then they showed them similar colors. The shrimp could not tell similar colors apart well. Humans are much better at this. Having more sensors does not always mean better color vision.

Humans see color with three types of cone cells, but some mantis shrimp have as many as twelve types of color receptors, covering light from ultraviolet to red. For years, many people believed that these animals must see a world of colors we cannot imagine.

In 2014, Hanne Thoen and her team at the University of Queensland tested this belief. They trained mantis shrimp to choose a certain color to get food, and then showed them colors that were slightly different. The shrimp could only tell colors apart when their wavelengths differed by about 12 to 25 nanometers, while humans can usually notice differences of 1 to 4 nanometers. The researchers suggest that mantis shrimp may recognize colors quickly by pattern instead of comparing them carefully.

The eyes of mantis shrimp are among the most complex in the animal kingdom. In 1989, Thomas Cronin and Justin Marshall reported at least ten spectral types of photoreceptors in their retina, and later work showed that some species possess twelve, spanning ultraviolet to red. In 2008, researchers also found that these animals can detect circularly polarized light, an ability that is extremely rare. It seemed natural to assume that such equipment would produce extraordinary color discrimination.

However, a 2014 study published in Science challenged this assumption. After training the animals to associate a color with food, Thoen and colleagues found that they could reliably distinguish only wavelengths roughly 12 to 25 nanometers apart, whereas humans typically detect differences of 1 to 4 nanometers. The team proposed that mantis shrimp scan scenes and identify colors from the pattern of activated receptors, a fast but coarse system rather than the comparison-based processing used by humans. This model remains a hypothesis, though: direct neural evidence is limited, only a few species and individuals were tested, and the design of the learning task, such as control of brightness cues, may have influenced the results.