TIMOTHY H. GOLDSMITH 781 



("bee-white") = 15% (360 m^) + 85% ("bee-white" — UV) (1) 



That is, nionochionuitic ^500 ni/i, can substitute for the band of w^ave- 

 lengths between 300 and -iOO ni/x; a mixture composed of 15 per cent 

 360 n\fjL and 85 per cent white (xenon) light from which the ultra- 

 violet has been fdtered is matched for bees by the full xenon 

 spectrum. 



Similarly, the region 400 to 650 niju. can be replaced by monochro- 

 matic -190 lUjj,. In an experiment in which the training light was "bee- 

 white" and the test light a mixture of monochromatic ultraviolet and 

 monochromatic blue-green, Daumer observed that 



("bee-white") = 15% (360 m^) + 85% (490 m^) (2) 



Or, in other words, 360 m/x and 490 ni/x are complementary colors 

 for honeybees. 



In another experiment three ^vavelengths were employed in the test 

 light, and it Avas found that 



("bee-white") ^ 55% (588 m^u) + 30% (440 m/x) 



4- 15%(360m;x) (3) 



It follows from Equations (7) through (5) that 



65%(588m/z) + 35%(440m/x) = (490 m/.) 



= ("bee-white" — UV) (•/) 



and these relations, too, are confirmed by experiment. 



Mixtures of yellow and near ultraviolet lights produce new colors 

 which honeybees do not confuse with intermediate wavelengths. The 

 fact that these "bee-purples" exist indicates that the color vision of 

 the bee is based on more than two types of receptor. The available 

 behavioral data can be interpreted as indicating the presence of 

 three classes of receptor maximally sensitive in the near idtraviolet, 

 the blue-violet, and the green-yellow-red. The results of Daumer's 

 experiments can be summarized in a preliminary chromaticity dia- 

 gram (Fig. 3) . 



As the foregoing suggests, the narrow band of wavelengths between 

 480 and 500 ni/x, which is seen by bees as a distinct color, is analogous 

 in several respects to yellow in human color vision. For bees, blue- 

 green is the region of the spectrum in which colors are least saturated 

 (9) , and 480 to 500 nifi. is probably where wavelength discrimination 

 is sharpest — where hue changes most rapidly as a function of wave- 

 length. The color "matches" expressed in Eq. (4) suggest that the 



