28 VICTOR E. SHELFORD AND W. C. ALLEE 



tion can be of two kinds. These are indicated by the following 

 questions: After entering the modified water a number of times 

 and turning back, did the fishes often turn back before reaching 

 the gradient ? When transferred to the control after exposure 

 to the gradient did they turn back, or show a preference for either 

 end of the control ? 



The first type of behavior could be shown only where the grad- 

 ient was confined to the central third of the tanks. Since some 

 species turn in any part of the tanks it is necessary to select a 

 particular species which does not show this trait. Hybopsis 

 was our best example of this for normally they went back and 

 forth symmetrically in the controls. The graphs indicate that 

 in the boiled water experiments this species turned back often 

 before the gradient was reached, which raises the question as to 

 whether they associate the center drain or difference in lighting 

 with stimulating water ahead. 



Fishes are able to form associations (Mobius fide Holmes, '11 

 and many others). There appear to be two ways in which 

 associations formed in the experimental tanks could be carried 

 over to the control tanks. One is through the kinaesthetic 

 sense, the other through differences in illumination of the sides 

 of the body when approaching the drain from the different direc- 

 tions. The data which we have upon this question comes from 

 comparing the responses in the first and second halves of the 

 double experiments described on page 6. 



It is obvious that if the fishes were able to associate the dif- 

 ferences in illumination upon the sides of the body when ap- 

 proaching the drain, with increasing stimulation further on in 

 the gradient and were able to retain this after being dipped out 

 of the water with a net and placed for a short time in very 

 different surroundings, they should show an apparent preference 

 for the end of the control tank opposite that in which they 

 spent most time in the experiment, (cf. Fig. 1). An examina- 

 tion of the records of twenty-eight controls indicates that the 

 fishes usually showed some difference in their relations to the 

 two ends of the control tanks, but in these exchanged controls 

 less than half of the fishes show an apparent preference for the 

 end that should have been favored if the light was depended 

 upon. This is as it might be if the kinaesthetic sense were 

 depended upon for the reaction, but the number of experiments 



