44 CORA D. REEVES 



of the three sunfish, are compared, it appears that with 

 all three fish discrimination was easier and more exact 

 when great brightness differences were present; but there 

 was no intensity of red light tested where some fish did 

 not show ten consecutive tests, with 90 to 100 per cent of 

 correct choices. The average of all the sunfish records 

 plotted, as in figure 16, fails to give a learning curve for 

 two reasons: (1) there is at first an instinctive avoidance 

 of the bright red plate; this gives a high initial percentage 

 of correct or blue choices; (2) although the curve of 552 is 

 similar in shape to that of the other fish, its successive high 

 and low points occur at different values of red light intensity. 

 The composite curve shown in figure 16 is, therefore, not 

 a curve of learning; it shows merely that under the condi- 

 tions of the experiment there is, at each intensity of red, 

 some individual in the group that discriminates, so that 

 the composite curve never drops to the 70 per cent level. 

 When the composite sunfish graph is compared with 

 that of the dace it is seen (1) that the percentage of correct 

 choices in the region of matched brightness for the fish is, 

 on the average, higher for sunfish than for dace (compare 

 figs. 14 and 16); (2) that the range of slit- widths where dis- 

 crimination is less accurate is shorter for the sunfish than 

 for the dace. Discrimination fails for sunfish at slit-widths 

 2-0.9 mm. and for dace at widths 5-0.9 mm. This is in 

 accord with the greater capability of sunfish to discriminate 

 white light intensities. The more exact aim of the sunfish 

 in capturing their food is perhaps evidence of their greater 

 visual acuity. 



D. Tests of the Apparatus and its Manipulation. 



In the preceding account of this experimental work it 

 is assumed that the fish are guided in their choice of stimu- 

 lus patches solely by the patches themselves. But the 

 operator must frequently shift filters and slit from side to 

 side and must at each test pull the door-string and operate 

 the electric switch while looking through the peep-hole. 

 It is, therefore, possible that the fish were guided to a 

 correct choice by some one of these manipulations. It is 

 possible that some difference in food-bars or filters of the 



