482 JENNINGS— HEREDITY IN PROTOZOA. [April 24, 



the same time and in the same proportion in all the individuals, then 

 at any given moment the coefficient of correlation will, of course, 

 not be altered by it. But if for any reason the changes occur more 

 quickly or strongly in certain individuals than in others (as is usually 

 the case), then, of course, the coefficient of correlation will be 

 decreased. Or, if we throw together individuals taken at different 

 stages of the process, the correlation becomes greatly decreased ; it 

 may even become negative. For examples, see rows 15 and 20, 

 Table XVIII. 



(c) Even if a given agent causes a change in the same direction 

 (e. g., an increase) in both length and breadth, the inclusion of 

 different stages in the process may reduce the correlation (if it is 

 already high). This will occur (i) if the two dimensions are not 

 changed proportionately to each other, and (2) if the change in a 

 given dimension varies at different stages of the process. Both these 

 conditions, as we have seen, are fulfilled in the changes in dimensions 

 induced by the environment. Under almost any environmental 

 change breadth is altered more than the length. Furthermore, when 

 nutrition is decreased, breadth decreases more rapidly at first than 

 later. The inclusion of different stages of the process in a collection 

 therefore results in the inclusion of various different proportions of 

 breadth to length — lowering the correlation. 



(d) If the correlation is already low, indicating the presence of 

 many different ratios of length to breadth, then varied changes in 

 these ratios may compensate some of the existing differences, causing 

 an increase in the correlation. Wliether this shall or shall not occur 

 depends upon the condition of aft'airs before the changes are made, 

 and on the nature of the changes themselves. A special case of this 

 comes up in the next. 



(e) When a culture containing thin, poorly fed individuals is 

 given added nutriment, the correlation between length and breadth 

 increases (compare, in Table XVIII., rows i and 2; 8 and 9; 10 and 

 ir ; 13 and 14; 17 and 18; 17 and 19, etc.). This is because, when 

 fresh nutriment is added, the thinnest, poorest-fed individuals nat- 

 urally take more food than do the individuals that are already plump 

 and well-fed ; they therefore increase most in breadth. As a result, 

 existing difi'erences in breadth are compensated ; all the animals take 



