MICHIGAN ACADEMY OP SCIENCE. 85 



selection of ancestry practiced in the experiments the environmental 

 conditions were as uniform as it was possible to get them. For sexu- 

 ally reproducing forms Pearson' has shown that the limit of reduction 

 of variability in offspring when the parents are selected with exactly 

 similar characters is 89.44 per cent of the variability of the general 

 population. This limit is reached if the selection be carried on for 

 5 generations ; selection for any number of generations beyond that will 

 not further reduce the variability. Our selected Paramecium samples 

 apparently show somewhat less than this 11 per cent reduction of varia- 

 bility. We are not prepared as yet, however, to draw any general con- 

 clusion regarding the reduction of variability with selection of an- 

 cestry in this form. There are several factors which must be taken 

 into account, the data on which are now being reduced. For one thing 

 the question arises as to whether we have obtained a true value for 

 the ''general population" variability. A laboratory culture of Parame- 

 cium probably in ordinary- cases consists of at most a comparatively 

 few fraternities, and it is easily possible that by taking a sample from 

 one part of a culture we might get a series of individuals nearly all 

 descended from a single ancestor in that same culture. We should 

 expect then relatively low variability. The point which we wish to 

 emphasize here is that there is a considerable lowering of variability 

 manifested in our Paramecium series when we deal with pure strains 

 from a single ancestor. 



(b) The amount of variation is, as we should expect it to be, sen- 

 sibly the same in each of the control series. The extreme values for 

 the coefficients are 8.772 ±.189 and 8.292 ±.178. giving a difference of 

 .480 with a probable error of ± .260. The difference evidently cannot be 

 considered significant. In other words, though the individuals become 

 smaller the longer they stay in the culture vials, yet in proportion 

 to their size, the variability does not change. This is of course the 

 result we ought theoretically to get. The agreement between the results 

 actually obtained is good evidence both of the accuracy of the measure- 

 ments of these 1,.500 individuals, and of the uniformity of the experi- 

 mental conditions in the three series. * 



(c) In the case of the sugar series we have an altogether different 

 result. Here we find for the 100 hour group a coefficient of variation 

 of 7.85i9 ±.1G9, for the 200 hour group 9.080 ±.194, and for the 300 hour 

 group 9.345 ±.201. The longer the stay in the culture fluid to which 

 sugar has been added the greater becomes the variation until in the 

 300 hour series we have a group of individuals not sensibly less variable 

 than a random sample from a "wild" culture. Yet the individuals of 

 this 300 hour group are the direct descendants of a single ancestor, 

 and have all lived in the same environment. The difference between 

 the coefficient of variation for our "wild" culture sample (D sample) 

 and that for the 300 hour sugar series is ,018 ±.349, an obviously in- 

 significant one. 



This result appears to be of considerable interest as showing the 

 importance of environmental conditions in determining the amount of 

 variation in this form. It is evident that this increase in variation 

 in the sugar series cannot be attributed to less uniformity of environ- 

 ment, because there is no reason to suppose that after the sugar solu- 



'Proc. Roy. Soc. Vol. 62, p. 399. 



