BIPARENTAL INHERITANCE IN PARAMECIUM 457 



into slow and rapid linos (see the preceding paper ('13), pages 351-354), 

 so that such assortative mating is impossi])le; (2) it is not consistent with 

 the result for the entire period in the present experiment, where the corre- 

 lation for the split pairs is practically 0. We are therefore driven, as the 

 only possible alternative, to the exi)lanation that this negative correla- 

 tion is merely a chance result, such as would occur now and then if the 

 experiment were repeated many times. Comparison of the coefficient 

 (— 0.3175) with the probable error (0.0554) given in the table would 

 seem to raise difficulties for this, but just what the probable error should 

 be is extremely doubtful. The probable errors given in the table are 

 based on the number of entries or cases compared, these being given in 

 the first and fourth columns of table 50, for the coefficient under discus- 

 sion this number is 120. But whether we should use in such cases the 

 number of pairs of entries (60 in place of 120), or the number of actual 

 lines compared in the experiment (44), or the number of actual pairs 

 (22), for computing the probable error, appears not to be established. 

 If the last named figure is the correct one, the probable error for the co- 

 efficient — 0.3175 would be =•= 0.1293, which, being more than one third 

 the coefficient, would readily reconcile the latter with the explanation 

 given. Such a change in the method of computing the probable error 

 would not, however, cast any doubt on the validity of the high correla- 

 tions found for the members of the pairs. For the pairs the probable 

 errors would have to be based on the number 16; this would give for 

 the entire period the coefficient 0.9238 ± 0.0247; for the first half; 

 0.6325 ±0.1012; for the second half, 0.9517 ±0.0161— so that the security 

 of the main results is not altered. 



Summary. The results of this experiment with a pure strain 

 therefore complete the demonstration, given in former experi- 

 ments, that conjugation results in bringing about a resemblance 

 between the progeny of the two members of a pair. If any doubt 

 was possible in the case of conjugants derived from wild cul- 

 tures, there remains none what ever with the results from a pure 

 strain. Here all the prospective conjugants are alike before con- 

 jugation; and there is no positive correlation whateA'er between 

 the progeny of the prospective members of pairs. But after con- 

 jugation such correlation comes into existence and rises to an 

 extraordinarily high figure, the progeny of the two members show- 

 ing a most surprising correspondence in rate of fission. 



Question may be raised as to the applicability of the term 

 'biparental inheritance' to this result. If employed, one of course 

 must not understand b}' it the production of a resemblance to 

 both parents, for in the present experiment all the parents were 



