How bees see color

Peitsch and colleagues (1992) measured single photoreceptor cells in 43 species of bees, wasps and their relatives. Peak sensitivities clustered at 340 nm (ultraviolet), 430 nm (blue) and 535 nm (green), and a color-vision model applied to these measurements predicts that most of the species have trichromatic (three-color) vision. Four of the 43 species, among them one solitary bee and one sawfly, also had a receptor peaking near 600 nm. We read only the summary (abstract) of the study.

Red flowers are often said to be invisible to bees. Chittka and Waser (1997) gathered evidence from physiology, behavior and ecology against that idea and concluded that red flowers are visible to bees. On the chart, the bee's green receptor still absorbs some light in the red part of the spectrum, though little at deep red.

Seeing red is not the same as telling colors apart in that range. The authors argue that bees will be unable to discriminate single-wavelength lights above about 550 nm, but should distinguish green-, yellow-, orange- and red-reflecting objects, «even if these do not reflect in shorter wavelengths», and that bumblebees can indeed perform this task. We read only the summary (abstract) of the paper.

A camera and a screen store three numbers per pixel, made for human eyes. Ultraviolet is not among them, so it is simply not in an ordinary photo. Pictures that do show ultraviolet in «bee vision» come from special ultraviolet cameras, and even those have to show ultraviolet as some color we can see: a false color, chosen by whoever made the picture.

For each pixel we guess a reflectance spectrum from its color with spectral.js, the same library our paint mixer uses, light it with standard daylight (CIE D65) and work out how much of that light the bee's blue (430 nm) and green (535 nm) receptors would catch, scaled so that white gives the same signal in both. The colors on the screen are our convention: the green receptor is shown as red, the blue receptor as green, and the blue channel, where ultraviolet would go, stays empty. That is why white looks yellow.

Read it as a map, not as a view. Two things are missing: ultraviolet patterns on petals, and differences between materials that look the same color to us but reflect light differently. A pixel's color does not tell which material it is, so the spectrum is a guess.

Scientists who measure what animals see from photographs do it differently: Troscianko and Stevens (2015) describe calibrated RAW photos, a grey standard, the camera's measured sensitivities and, for ultraviolet, extra ultraviolet photographs. An ordinary phone photo, they point out, is made for human eyes and is not reliable for such measurements.

A bee's eye is built from many small units, each looking in its own direction. Optical measurements cited by Rigosi and colleagues (2017) put neighboring units in the sharpest, front part of a honeybee's eye 2.1° apart horizontally and 0.9° vertically. Another study they cite argues from optical data that the spacing averages about 1.7° in both directions, and Rigosi and colleagues note that this agrees with behavioral tests of how fine a striped pattern bees resolve. An eye can resolve at best one dark and one light stripe per two units, so 1.7° means about 0.29 cycles per degree by our arithmetic. People in a two-choice stripe test resolved 32.1 to 44.2 (Lind and colleagues, 2017), more than a hundred times finer.

How blurry each point is comes from a different measurement. Rigosi and colleagues recorded from single light-sensing cells of honeybees adapted to bright light. In the front of the eye, each cell takes in a patch about 1.9° wide at half its peak sensitivity (their average; single cells went down to 1.6°). The simulator blurs the picture by that 1.9°. Stripes finer than 0.29 cycles per degree are then faded but not wiped out: the bee still could not resolve them, because its units are spaced too far apart to sample them.

Other measurements give somewhat sharper values. Older optical data listed by Land (1997) put the units of worker honeybees 1.0 to 1.3° apart, which would allow 0.38 to 0.5 cycles per degree, and bumblebees in a maze test resolved 0.35 to 0.36 (Macuda and colleagues, 2001).

Noticing a small thing is not the same as resolving a pattern: single light-sensing cells in the front of the bee eye responded to dark objects as small as 0.6° (Rigosi and colleagues, 2017). The blur shows how fine a pattern a bee can resolve, not the smallest spot it can notice.

How the blur is made: the simulator follows AcuityView, the method of Caves and Johnsen (2018), a bell-shaped blur; we set its width at half height to the 1.9° above. Such a blur leaves about a third of the contrast of stripes at 0.29 cycles per degree (our arithmetic), and coarse shapes and colors stay. A picture does not say how wide the scene was, so the simulator assumes that the long side of the picture covers about 70°, the view of a typical phone's main camera (the iPhone 16 main camera is 26 mm in 35 mm terms, about 69° across). If you zoomed in or cropped, the scene was narrower and the bee would see it blurrier than shown. The blur stands for the sharpest part of the eye only.