"positive" end zone (first result 

 minus second result) plotted against 

 the difference in water temperature 

 (first measurement minus second 

 measurement) between the two times 

 a combination was tested on once- 

 shocked fish in the north block. 



Assuming that the electrical 

 field in the water (fig. 6) was not 

 changed by reversing the direction 

 of the electrical fields, the differ- 

 ence between north and south blocks 

 of tests probably represented a pre- 

 ference by the fish for the north 

 end (fig. 7). 



Effect of potential . — There was 

 a highly significant difference be- 

 tween the different levels of poten- 

 tial as shown by the three analyses 

 of variance (tables 4, 5, and 6). 



For the unshocked fish, the 

 overall potential effect was obscured 



rOO 



tu 



Z 90 

 O 



^ 60 

 §70 



D. 



Veo 



S 30 

 5 20 



C I - Nofth Block 



■i : South Block 



IklMn 



Onl 



D 



o -n o o 



Q o f 9 



ELECTRICAL COMBINATION (Frequency, durotion ond polenliol - respectively ) 



Figure 7. — Comparison between blocks of 

 the mean percentage of squawfish that 

 entered the ''positive" end zone. 



Table 6.— Analysis of variance of tests of unshocked and once- 

 shocked squairfish coaibi.ied. 



Pooled 

 Error *■ 



r'l .< 



31,623.71 

 < .05 



182 



173.76 



P < .01 



Pooled error is found by combining all suns of squares belcnr the 



double line since none of these are significant even at the 1C0^ level. 



The F values listed in the table are fo^jnd by comparing each mean 

 square (H.S.) with the pooled error mean square. 



by the significant interaction of potential 

 with pulse frequency (table 4). In particu- 

 lar, the rate of decrease in effect, with 

 increasing potential, is much less at 2 

 pulses per second than at the higher fre- 

 quencies of 5 and 8 pulses per second. 



For the once-shocked fish, a rapid rate 

 of decrease in effect with increasing poten- 

 tial occurred only at 8 pulses per second 

 (table 5). 



A more detailed examination of the 

 potential effect may be made using Tukey's . 

 Least Significant Difference (L.S.D. ) test.— 

 The L.S.D. is found by multiplying the 

 standard error by the Q value (tabulated in 

 table 10.6.1, p. 252, of Snedecor) and 

 dividing the result by the square root of 

 the number of observations in the means to 

 be compared. The standard error is 



V 



/ lO, 299.166 + 11,144.405 ^ j^^ qq. 

 108 ■ ' 



this 



is determined by pooling the error sum of 



_5/ Snedecor, Gteorge Waddel. 1956. Sta- 

 tistical methods applied to experiments 

 in agriculture and biology. 5th edi- 

 tion. Down State College Press, Ames, 

 Iowa. 



13 



