tv'O rectangular 6 x 10 -inch observation windows the size of which 

 could be varied to a narrow horizontal slit by means of sliding doors. 

 Later, in August, the observation windows were equipped with one- 

 way glass o As nearly as could be determined, the fish were gener- 

 ally unaware of the presence of the observer as long as strict silence 

 was maintainedo Occasionally, however, before the glass was used 

 it seemed as if the fish were aware of the small part of the observer's 

 face which was visible through the small slit-like aperture. 



By means of a 3/8=inch centrifugal pump driven by a 

 1/30 ho p. near-silent electric motor, water was sucked from the 

 surface of the tank at one side near the center eind was injected into 

 the southern or "downstream" one-third in a continuous stream dur- 

 ing both control and test periods. When desired, the test material 

 was introduced into this stream from within the booth through a glass 

 T connection in the rubber tubingc 



The method of measuring the reaction was modified from: 

 that of van Weel (1952). Each test consisted usually of five succes- 

 sive 2-minute control periods, followed by at least five successive 

 2- minute test periodSo For each period, the time spent by the fish in 

 the downstream one-third of the tank, marked off with string, was 

 measured by an electric clock operated by a push switch, and the 

 number of loops (half»circles) nnade by each fish in this "attraction" 

 area was recorded on a hand counter. By means of fluorescein dye 

 it was found that a liquid quickly spread throughout the downstream 

 ore-third of the tank, remained there for about 10 minutes, and then 

 gradually dispersed up the tank along the far wallo If attracted by 

 the material, the fish would be expected to spend a longer period of 

 tiine in the attraction area as compared with control conditions. If 

 excited by the nnaterial the fish would be expected to swim more 

 rapidly and to circle more times in the attraction area than during 

 control conditions. If a fish entered, turned, and left the area, the 

 count would be one loopo If the fish entered, made a complete circle 

 within the area, and then left, the count would be three loops. If 

 two fish performed: in this way, either together or independently, the 

 counts would be doubled. Later the method of counting was modified 

 to conform with that in the pond (see below)o 



The tests in the tank were conducted for the most part on 

 one established tunny during the sunnmerj aund on two established 

 tunny thereaitero Unestablished fish did not react to the test sub- 

 stances auid were ignored in timing and countingo A typical series of 

 reaction graphs is shown in figure 3. 



