the variation in response within and between daySo Both the 

 experimental plan and the observational data are included in 

 table 3o Two tunny were present in the pond throughout the 3-week 

 periods Various difficulties were experienced which prevented the 

 completion of all experiments in the series ei'her erratic behavior 

 of the fish or other difficulties as listed m the footnotes to table 3« 



The great variation in response is apparent from the data 

 of table 3„ For Tj^ and T2 it ranged from 1/2 to 4„ and for T3 from 

 1/2 to 2»l/2„ A superficial analysis of that portion of the data com- 

 prising complete blocks (ao nio onlyl showed that none of the main 

 effects (order of testings daySp and treatments) was significanto 

 Order of testing contributed the largest me^n squares days the next 

 largestj and treatment the smallest „ The last was less than the 

 residual mean squares In these and other series of experiments 

 often it has been noted that the first experiment of the day gives the 

 strongest response; the fish are suddenly aroused from a quiescent 

 condition and give a strong response relative to both the preceding 

 control periods and the succeeding test periodSo This contributes 

 to the variation between means for order of testing (respectively 

 2o4j, lo3, and 1<,3 for the observational data above)o Variation in 

 reactivity from day to day^ apart from erratic behavior, has been 

 noted repeatedly by both observers, and is believed real despite 

 the non~ significance of the data aboveo The reason is obscure--it 

 is unlikely that it is related to the state of hunger of the fish; it has 

 been suspected at times that it is related to the weather, tides k or 

 turbidity, but these factors cannot be adequately investigated with 

 the present datac 



There seenaed to be a decrease in the response and an 

 increase in its variability between the summer of 1951 and the 

 spring of 1953o Thus, in six pond experiments conducted with the 

 same material (each equivalent to about 25 go of tuna flesh) during 

 August, the mean difference between timing for test and control 

 periods was 32o6 seconds and the range was lie 2 to 52^2 secondSo 

 In the test series conducted in April, in which the material was 

 equivalent to about 350 g^ of tuna flesh (14 times as much as before) 

 treated in a different manner but one which did not significantly de- 

 crease its attractive properties, the mean difference between the 

 timing for test and control periods was 7o 1 seconds with a range of 

 - 13o0 to 6la8 seconds. The decreased response was reflected to 

 only a slight extent in the observational dat-i which dveraged lo9 for 

 August and lc7 for April; apparently there was a change in the ob- 

 server's judgement of the absolute strength of o reaction. The 

 reason for the decrease in response and the increase in its varia- 

 bility may be related to a decrease m the pond population from 1 3 to 



25 



