,go8.] JENNINGS— HEREDITY IN PROTOZOA. 455 



are lacking in these. There is no evidence of an unusually low 

 degree of congenital variation in the conjugants, for the non-conju- 

 gating specimens beginning fission show a still lower variability 

 (Table XVL). 



It appears highly probable that if we could examine a large 

 number of individuals, derived from the same parent, cultivated 

 under identically the same conditions, and all in precisely the same 

 stage of growth, we should find coefficients of variation considerably 

 smaller than the smallest we have found, which is 4.521 (row 8, 

 Table X.). Indeed, if we could further exclude all inaccuracies of 

 measurement, it is quite possible that the coefficient of variation 

 would approach closely to zero, if it did not reach it completely. 

 This would, of course, mean that the variations observed among the 

 progeny of a single individual are not congenital, but are all due to 

 growth and environmental action. Further evidence of this will 

 come out later in this paper. 



4. Effects of Growth on the Observed Correlation Between 

 Length and Breadth. 



As Diagram 5 shows, the curves of length and breadth diverge 

 at the beginning, then run for a considerable distance 'nearly parallel, 

 then finally approach each other. That is, at first the breadth decreases 

 while the length increases ; later they increase together ; and still 

 later the breadth increases while the length decreases. If a collec- 

 tion of specimens includes individuals in various different stages of 

 growth (as is usually the case), then these various relations of 

 breadth to length will deeply affect the amount of correlation observed 

 between the two dimensions. 



Thus, if we take a collection composed of various ages under one 

 hour, when the length is increasing while the breadth is decreasing, 

 then on the whole greater length will be associated with less breadth, 

 so that the correlation between them will tend to be negative. This 

 is the explanation of the negative correlation shown in Table X., 

 rows 2, 4, 5, 7, II, 13, 14. Next follows a period (from about the 

 end of the first hour to the fourth) in which the inclusion of indi- 

 viduals of different ages tends to cause a certain degree of positive 

 correlation, since the two dimensions are increasing together. Then 



