LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 45 



two sides may have aided the fish. Since the slit was used 

 chiefly on the red side, on which the fish were never fed, 

 they may have learned by seeing the reflection of the slit 

 on the lens to go to the opposite side for food. Check 

 tests were, therefore, made in order to learn whether any- 

 thing else than the stimulus patch served to guide the fish. 



1. Check Tests by Changes in Manipulation 



a. Shifts of slit and filters as a clue. — The fish were in- 

 variably shut into the retention compartment before the 

 apparatus was adjusted for the day's trials. Since they 

 were then unable to see the stimulus plates or other parts 

 of the apparatus, these could afford them no clue in making 

 their first choice for the day. To see whether their choice 

 in subsequent trials was possibly influenced by the shift- 

 ing of filter and slits from^ side to side, I often introduced 



make believe " shifts. These were interpolated in the 

 course of the regular series of the day at times when the 

 fish were discriminating accurately, and consisted in going 

 through the manipulation of shifting slits or filters but 

 without actually making the change. The responses of 

 the fish were in no way affected. 



b. Position of the experimenter as a clue.— To learn 

 whether a possible clue was afforded by the movement 

 of the string or curtain or of my eyes, when I was in the 

 usual position at the peek-hole outside the curtain, I fre- 

 quently made tests while inside the curtain or in some 

 other unusual position. None of these altered the responses 

 of the fish. 



2. Check Tests by Change in the Apparatus 



a. Food-bar. — Although, as already stated (p. 11), both 

 ends of the food-bar were kept as nearly alike in appear- 

 ance and position as possible, nevertheless, new food-bars 

 were introduced from time to time, or the bar was occa- 

 sionally removed and the fish were fed by hand if they 

 remained before the blue plate. These changes made no 

 difference in the percentage of correct responses of the 

 fish. 



