520 JENNINGS— HEREDITY IN PROTOZOA. [April .4, 



ing the length. Inclusion of different stages of this process in a 

 sample reduces the correlation ; it may make it zero or negative. 



2. Most environmental agents change the breadth more than the 

 length, even when both are changed in the same direction The inclu- 

 sion of different stages then reduces correlation. 



3. Samples in which some of the specimens are well-fed and 

 plump, others ill-fed and thin, of course, show low correlation, since 

 the ratio of breadth to length is not uniform. This is usually the 

 case in cultures where food is scarce. 



4. Addition of abundant nutriment causes the thin specimens to 

 increase in breadth, by taking food, while the plump ones change 

 little. As a result the proportion of breadth to length becomes nearly 

 uniform throughout the lot; the correlation is therefore increased. 

 As a rule, any agent which increases the mean breadth likewise ( for 

 the reason just set forth) increases the correlation between breadth 

 and length. 



Decrease of nutriment, for the converse reason, decreases the 

 correlation. 



5. Any agent that causes rapid multiplication decreases the cor- 

 relation between length and breadth for the period of multiplication. 

 This is owing to the inclusion in the collection of many stages of 

 growth, showing different proportions of length to breadth. 



6. Slight differences in one dimension may be produced without 

 corresponding differences in the other, so that in a collection varying 

 little in length the correlation may be low. But considerable changes 

 in one dimension are usually accompanied by corresponding changes 

 in the other. Hence, when two groups of differing lengths are 

 thrown together, the correlation may become higher than in either 

 one taken separately (for example, see page 437). 



In any ordinary sample of Paramecium all these varied factors 

 are at work in determining the observed correlation. It is clear that 

 no particular coefficient of correlation can be considered character- 

 istic for Paramecium or for any particular race of Paramecium, for 

 by various combinations of these factors we may get any coefficient 

 of correlation ranging from a pronounced negative value upward 

 through zero to a high positive value. In Tables X. and XVIII. we 



