224 MORRIS M. WELLS. 



constant by using an aspirator bottle as in Fig. I (A B), 

 This bottle was filled with treated water, corked at the top and 

 placed in the jar (J). The water was siphoned from J and the 

 pressure was constant as the solution escaped from A B, when 

 the level began to fall in /. The strength of the solution in the 

 aspirator bottle was always five times that desired in the treated 

 water end of the tank. 



The following variations of the simple graphing method of 

 recording were used by Shelford and Allee and have been intro- 

 duced here, (i) In experiments where the fish was decidedly 

 inactive and remained in one end of the tank, it was driven into 

 the opposite end with a rubber-tipped glass rod. The driving 

 was done at regular intervals and was repeated at similar in- 

 tervals in the controls. Dotted lines in the graphs indicate that 

 the fish was driven. Experiments of this sort were few in num- 

 ber and have for the most part been thrown out. In some ex- 

 periments, however, the fishes were active and yet remained 

 constantly in one end of the tank. Driving was again resorted 

 to, in some cases, to make sure that the selection of the given 

 end was a reaction to the gradient. A return to the original 

 end would indicate this to be the case. (2) A number of ex- 

 periments was performed with 410 small fishes in the tank at 

 the same time. These experiments were recorded by readings 

 taken 30 seconds apart. The readings indicate per cent, of 

 fishes in each third of the tank at the time of reading. (3) 

 Usually the fishes were not placed in the tank until the flow at 

 the ends had been going for some time. Thus a gradient was 

 established before the fishes were introduced. In some cases, 

 however, the fishes were either left in the tank when the ends 

 were reversed, or introduced before a gradient had formed. The 

 results of these experiments do not differ from the others ex- 

 cept in per cent, of time spent in the thirds of the tank. 



The controls were blank experiments, run with untreated 

 water flowing in at both ends, or with no flow at either end. 

 Experiments with the treated water first at one end and then 

 at the other, also served as controls. 



The gradient was determined by simple titration with stan- 

 dard acid or alkali, using phenolphthalein or methyl orange 



