THE COLOR VISION OF BIRDS 19 



lights of equal intensity exposed under the same conditions 

 as the monochromatic. Chick A chose the stimulus patch to 

 which he usually reacted positively 7 times, the other 8 times. 

 Chick E chose the central patch 6 times, the other 4. Chick 

 D could not be tested. A was further offered a choice of two 

 white lights of relative intensities 1 to 90. He chose the brighter 

 7 times, the other 3, an accuracy less than that with any in- 

 tensity of monochromatic lights. These tests, with the other 

 controls mentioned in the discussion of technique limit the 

 efficient stimulus to the illumination of the stimulus plates. 

 Summarizing table 6 we find that with the red certainly brighter 

 than the green the chicks made 3 errors in 72 trials or 4.1 per 

 cent, error. In 325 trials with the green brighter or equal to 

 the red in brightness they made 23 errors or 7.7 per cent. 

 Further, in trials where the intensity of both lights was greatly 

 altered there was no reduction in the accuracy of choice, so 

 that the possibility of reaction to the absolute brightness of 

 either stimulus is eliminated. As between brightness and wave- 

 length, then, these experiments speak conclusively for the latter. 

 The poss bility of other characters of the stimulus lights them- 

 selves furnishing the basis of discrimination will be considered 

 after the work with yellow and blue light has been recorded 

 (page 20). 



The same three fowls were trained with yellow at 588 PP. 

 and blue-green at 500 pp. with energies equal. Chicks A and 

 E were trained to choose yellow, D to choose green in order 

 to take advantage of the previous training. Strangely enough 

 the chicks were confused by the green at 500 pp, not greatly 

 different from 520 pp. to my eyes, and failed to discriminate 

 in the first trials. They had previously reacted to 520 pp when 

 white light, darkness, or yellow were substituted for the red. 

 However, after 120 trials they gave less than ten per cent, error 

 and were considered ready for control tests. Tests for the 

 Purkinje effect had previously shown that the following bright- 

 ness values held for these wave-lengths for the moderately light 

 adapted eye of the fowl. 



Yellow Green 



4 1 yellow brighter. 



2 1 yellow perhaps brighter. 



1 1 neither brighter. 



1 45 green certainly brighter. 



