INHERITANCE IN ASEXUAL REPRODUCTION 191 



has shown that in these pure lines there is likewise a correlation 

 between parent and progeny. His explanation of this seemingly 

 anomalous condition and the criticism which has been urged 

 against selection experiments in general where Vilmorin's method 

 is employed is based upon the small numbers selected. Pearson 

 ('10), discussing Hanel's results, says: 



In selecting a few isolated individuals in each generation, where 

 non-hereditary influences are so influential, we may break the influence 

 of heredity at each step, and since such influences are equally efi"ective 

 with heredity, the chances are that we will do so onoe in every hvo 

 selections. Only by taking large numbers of the high and large num- 

 bers of the low would it have been possible to average out the effects 

 of environmental changes. 



The only ways in which this criticism can be met are by very 

 extensive selections involving large numbers of individuals, or 

 by a study of variation and environment which shall make 

 possible a differentiation between heritable and non-heritable 

 variations, upon some other basis than that of ancestral correlation. 



Resemblance of close relatives within the clone 



Number of tentacles. In all breeding experiments the polyps 

 have been so recorded that the number of tentacles of the buds can 

 be compared with that borne by their parents, both at the time 

 when the buds were produced and w^hen the parents themselves 

 were buds. Comparisons of both numbers must be made in 

 order to answer two distinct questions: (1) do the offspring 

 resemble their parents when the latter were in the same stage of 

 development: (2) are the variations of the parents during their 

 development transmitted? 



To test the first question the correlation between the initial 

 number of tentacles of the parents and that of their buds was 

 computed. Since there is reason to believe that Clone A is 

 impure, it was divided into the two subordinate clones mentioned 

 earlier and the constants were computed for these. If there is 

 a cumulative inheritance of slight variations this division should 

 affect only the probable errors of the correlation constants. The 



