BIPARENTAL INHERITANCE IN PARAMECIUM 455 



former being 2.125 fissions, the latter 1.895. The difference is 

 without significance. 



To obtain a precise expression of the similarity between a and 

 b in each case, and to compare accurately the pairs and split 

 pairs in this respect, we must determine for each the coefficient 

 of correlation between the number of fissions in a and that in h. 



The data for this are given in table 49. To determine the corre- 

 lation between a and b, we must enter, in the correlation table 

 both representatives (x and y) of a mated with both representa- 

 tives (x and y) of b, giving us (in the complete case) four entries 

 for each pair (or split pair). In the incomplete cases there will 

 be either one, or two, such entries for each pair. That is, for p^ir 

 1 (table 49) we should enter in the correlation table the matings; 

 33 X 31, 33 X 23, 37 X 31, and 37 X 23. Proceeding in this 

 way, we obtain, for the entire period, 43 entries for determining 

 the correlation between a and b of the conjugants (pairs); 56 

 such entries for the correlations of a and 6 of the split pairs. 

 (In practice it is simpler to compute the correlation by the differ- 

 ence method, without constructing a correlation table, as set 

 forth on page 429, but the above gives the guiding principles, 

 whatever the method used). 



The correlation between a and b was also determined, in this 

 manner, separately for the fissions during the first half of the 

 experiment; during the second half of the experiment, and for 

 the entire period. This gives us three independent determina- 

 tions, for both pairs and, split pairs. The data for these correla- 

 tions in the partial periods are given in tables 34 and 35 of my pre- 

 ceding paper (^13), so that they need not be repeated here. 



The coefficients of correlation between a and b for the pairs, 

 as compared with the same for the split pairs, are given in table 50. 



Table 50 expresses in figures the surprising difference between 

 pairs and split pairs, that is evident on inspecting table 49. 

 Between the two members of the pairs we ha\'e for the entire 

 period the extraordinarily high correlation of 0.9238, while be- 

 tween the two members of split paii's there is no correlation what- 

 ever (the coefficient being as near to as could be expected). 



THE JOURNAL OF EXPERIMENTAL ZOOLOGY, VOL. 14, NO. 3 



