1908.] 



JENNINGS— HEREDITY IN PROTOZOA. 421 



3. The mean index, or ratio of breadth to length, is considerably- 

 more than twice as great in the young as in the adults ; in the former 

 it is 63.136 per cent. ; in the latter 25.114 per cent. 



4. The variability in length is less in the earliest stages of the 

 young than in the individuals that are not dividing. In the former 

 the coefficient that measures the variability is but 5.368, while in the 

 latter it is 7.765. 



5. The variability in breadth is likewise much less in the youngest 

 stages— the coefficient being 9.082 in place of 12.877. 



6. The correlation between length and breadth is nearly the same 

 in the youngest stage as in the random sample, being .6546 in the 

 former, .6064 in the latter. 



From the other lots smaller numbers were examined. These 

 gave on the whole similar results, though with certain significant 

 differences. The facts are as follows : 



From lot 2 (descendants of D), 124 halves were obtained. On 

 account of the small number, I threw together all in which the depth 

 of the constriction was less than one fourth the breadth, and consid- 

 ered these the earliest stage (the depth of constriction and length are 

 given for the entire dividing specimens in Table XLIL, appendix). 

 There were thus obtained eighty young individuals (dimensions 

 for the entire dividing specimens in Table XLIII., appendix). 

 It is evident that this lot includes individuals varying more in 

 age and growth than in lot i, since in lot 2 we have included those 

 having a much greater depth of constriction. The results are shown, 

 in comparison with a random sample of the same lot, in rows 4 and 

 6 of Table VIII. The facts are in the main parallel with those for 

 lot I. As compared with the random sample, the mean length of 

 the young is less than one half, the mean breadth a little greater, the 

 mean ratio of breadth to length more than double, the coefficients 

 of variation for length and breadth much less. A striking dififer- 

 ence between this set and the young of lot i is that in the present 

 case the correlation between length and breadth has decreased to 

 such an extent that the coefficient computed (.1048) is without sig- 

 nificance, being less than its probable error (.1055). This is due, 

 as we shall clearlv see later, to the fact that we have included in the 



