THE COLOR VISION OF BIRDS 23 



This suggests a change in the character of the light between 

 530 and 565 a 1 /*, corresponding roughly to the change from 

 yellow to green for man. 



After training with yellow and blue-green Chick D was given 

 the following tests. 



Wave-length, pi.fi. Chose 



(1/18 standard) 



There is a change between 535 and 565 w. This agrees with 

 the results obtained from A. The change in the energy of 

 the yellow was made as a slight control of thejprightness element. 



Chick A was offered green 530, violet 460 \ni. The use of the 

 yellow-green was due to the difficulty of obtaining lights nearer 

 together in the spectrum. In an earlier experiment after train- 

 ing with red and green Chick A had been offered a choice between 

 white and red, white and blue, and red and blue (impure color 

 niters) . He chose accurately red and blue in preference to white, 

 red in preference to blue and all in preference to green 500 fifi. 

 I expected him to choose the violet in preference to the green 

 and refused him food at the latter. Instead,' however, he chose 

 the green five times in succession, then, getting no food, de- 

 veloped a position habit. He was then fed at the green and 

 immediately chose it five times in quick succession. 



Chick D was offered a choice between 500 and 460 w. He 

 showed no choice and developed a position habit immediately. 



It is rather difficult to interpret the results of these two tests 

 with the assumption of identical color divisions for the chick 

 and man. In the beginning of training with yellow and blue 

 the shift from 520 to 500 w. seemed to confuse the chicks (page 

 19). If we suppose that there is a marked change in the re- 

 action value of the spectrum between these wave-lengths, there 

 seems to be a general agreement in the results; for Chick A 

 530 M/i and 580 /x/x. resembled each other more than they did 

 500 At/*, and 460 A 1 /* perhaps resembled 500 A 4 /*; for Chick D 

 500 w and 460 a 1 /* were more nearly alike than the original 

 580 /j>/j> and 500 pp. The accord of the animal's behavior with 

 this interpretation justifies the assumption, pending more accu- 



