CORRELATIONS IN NOTHOSCORDUM AND ALLIUM. 



109 



But before any great stress is laid upon these results some 

 real measure of the relative variability of the environmental 

 conditions in different habitats should be secured. Until 

 better data can be obtained these differences m the constants 

 for two collections of material in habitats a stone's throw 

 from each other may serve as a further warning agamst the 

 collection of biometric material without due regard to envir- 

 onmental conditions * . 



The coefficients of correlation are the constants which give 

 the direct answer to the chief question in hand. They are : — 



Plants from southwest slope, r = .549i.025 

 Plants from southeast slope, r = . 49 ±.0 25 

 Difference 059±.035 



These constants show at once that there is a very sub- 

 stantial interdependence between the length of the flowering- 

 stalk and the number of flowers per inflorescence. The two 

 constants appear somewhat unlike to the eye, but when the 

 probable error of their difference is taken into account it is 

 clear that their divergence is not significant. 



To render the relationship between the length of the stem 

 and the number of flowers per inflorescence more clear to 

 those unacquainted with biometric terminology we may 

 express the relationship in terms of regression instead of cor- 

 relation. The equations to the regression straight lines are:— 



For the series from the southwest slope, 



y = 2.501 + .19Ix. 

 For the series from the southeast slope, 



y = 2.375 + .170x. 



Here x denotes length of stalk and y number of flowers. 

 From these regression coefficients we see that an increase of 

 about two-tenths of a flower per inflorescence is on an aver- 

 age associated with an increase of a centimeter in the length 



of the stalk. 



The relationship between the length of the stalk and the 

 number of flowers per inflorescence may be represented 

 graphically by the slope of the regression line, as in Diagram I. 



* For one of the first discussions of this subject see the papers on the 

 eources of apparent polymorphism in plants in Biometrika. 1 : 304-319. 

 1902. 



