in the morning and three in the afternoono The sky was overcast, with 

 a few rain squalls in the nnorringo Twelve tunny and four yellowfin 

 were present, but only the former were timed and counted. The na- 

 ture of the test substances is not pertinent to this discussion. 



The basic data are shown graphically in figure 6, 



It will be noted immediately that there is a high correla- 

 tion between the times and number of passes for successive test 

 periods (r = 0, 874; P less than OoOr)^ Both variates reflect the re- 

 actions alnnost equally well m this particular series. It will also be 

 noted that in all experiments except No, 98j the times and counts dur» 

 ing the test periods are generally higher than during the preceding 

 control periodSo 



Analysis of variance was employed to determine if there 

 were significant differences between the means of control and test 

 periods (equivalent to a "t" test) for each separate experiment. The 

 results are shown in table 2o It may be concluded that there is a 

 statistically significant response to the test substances of experiments 

 Nos, 95, 97, 99, and 100 (P less than 0,05 or OoOUo 



In experiment Noo 98 the average time and count during 

 the test periods were both slightly less thdn during the preceding con- 

 trol periods. Obviously^ however, this was due to unusual activity 

 during the controls. To overcome this difficulty a more mfornnative 

 method of analyzing the data is to group all control periods for all 

 experiments as Tqp determine the mean square between means for 

 experiment Tq to Ty, determine the pooled nnean square within ex- 

 periments, and use this last to determine a least significant difference 

 (fiducial interval) of a treatment mean for comparing not only Tj with 

 Tqj T2 with Tq, etCj, but also for comparing T^ with T^t etc. The 

 analysis, for time data only^ is as follows" 



** P less than 0„0I 



19 



