POSITION REGRESSION- 



fusion Series V and VI have not 

 been included, but examination of 

 the values in table 34 is sufficient 

 to show that they are not essen- "^ 

 tially different from the others, i^ 

 In the diagrams the ordinates give 

 the means as recorded in table 34 

 and the abscissas the position of "' 

 the whorls. c 



These diagrams at once make 

 clear the following points: 



(a) The regression is evidently '^ 

 not linear. Starting with a low ^ 

 value for the first whorls, the 

 curves all show a sharp rise, 

 amounting to almost exactly one - 

 leaf, to the second whorl. From - 

 the second to the third whorl there ^ _ 

 is approximately half as great a | '^ 

 rise as from the first to the second, i + 

 From the third whorl on, while °r: 

 the general trend of the lines is j _ 

 upward, their slope becomes more - "^ 

 and more gradual. The maximum ^ 

 towards which the lines tend is - 

 clearly 10-leaved whorls, though 

 on account of the small number 

 of entries in the outlying parts of s 

 the correlation tables at the upper ts> 

 end, the regression lines become 

 very irregular on these high "^ 

 values. 5 



(6) While the form of the re- n 

 gression line is clearly the same in 

 Series IV as it is in Series 1, II, '^ 

 and III, it differs in being practic- S 

 ally uniformly higher. It starts « 

 nearly a half leaf higher on the 

 first whorl, and m.aintains this °' 

 difference on the whole very even- s 

 ly out to the 15th whorl. From ^ 

 that point it becomes more irregu- 

 lar on account of the paucity of 

 observations. 



■PRIMARY BRANCHES. 



Mean leaf number 



63 



Fig. 11.— Regression line, stiowing ctiangeol mean 

 leaf-number with position. Primary ^branches. 



Series I, • ; Series II, o — ; 



Series III,^.. ; Series IV, ©-..— - — —©. 



