THE COLOR VISION OF BIRDS 15 



advisable therefore to repeat the experiment using still more 

 thorough controls. 



After a long delay, while smoked wedges were substituted for 

 the Aubert diaphragms used in equating light energies, the work 

 was resumed. Three cocks, A (now three years old), and two 

 younger ones, D and E, were trained for discrimination with 

 red 650 w. and green 520 m*. uncalibrated, just as given by 

 the carbon arc spectrum. Chick D was trained to avoid the 

 red (middle stimulus plate) and choose the green; Chicks A 

 and E to choose the red. After about 400 trials the chicks 

 began to react perfectly. They were given an additional hun- 

 dred trials to make the reaction more nearly automatic — then, 

 as the lights could not be equated immediately, a series of 

 rough control tests for brightness and secondary criteria was 

 begun, partly to accustom the chicks to a changing problem 

 and partly as a test of their sensitivity to wave-length. Three 

 types of tests were employed to distinguish between brightness 

 and wave-length; 1, first one, then the other light was reduced 

 to threshold intensity while the other remained at the full en- 

 ergy of the spectrum; 2, a white light of constant intensity 

 was substituted for each of the colored lights in turn; 3, each 

 light was exposed alone with one passage completely dark. 

 Table 4 shows the records of the chicks in the first of these 

 tests. It will be noted in the results of this experiment that 

 reduction of the intensity of the negative color produced no 

 disturbance of the reaction while reaction of the positive color 

 led to many errors. Had the reaction been due to the relative 

 brightness of the stimulus patches Chick D should have been 

 disturbed when the red was reduced, as were A and E, since 

 this must then have represented the condition in which the 

 difference in brightness approached threshold. Instead of this, 

 however, Chick D reacted inaccurately only under conditions 

 where A and E remained undisturbed. This seems to limit the 

 effective stimulus either to the wave-length or to the absolute 

 intensity of the positive stimulus patch. 



