vii NATURAL CROSSING 119 



to which it protrudes, varies in different strains and 

 species. Most American Uplands have a much shorter style 

 than Egyptians, and plants were noticed in F. 2 of Upland 

 Egyptian crosses, whose styles did not protrude beyond 

 the column. The inheritance was therefore investigated 

 statistically, and from a cross of 29 mm. style with 20 mm. 

 column, upon 24mm. style with 13mm. column; we bred 

 several strains with such flowers as the one shown in the 

 frontispiece. The plant from which this flower was photo- 

 graphed was an F 2 , and its descendants were grown up to 

 F 5 , still breeding true to this flower form. In F 4 they had 

 a mean style length of 18 J mm. with a column of 

 14J mm. : the anther filaments being 4j mm. long made 

 up the length of the brush to 19mm., which sufficed to 

 cover the style completely, so that it was visible only in 

 end view, whereas the style of the Egyptian parent had 

 projected for more than half a centimetre. 



The short-style flower was thus at our disposal, but 

 without avail. Four such plants from an F 3 , similar to 

 this respect, but otherwise differing, which had been 

 growing close together in 1908, gave the following 

 offspring in 1909 : A. 26 plants ; 7 rogues. B. 19 plants; 

 5 rogues. C. 47 plants; 16 rogues. D. 36 plants; 

 12 rogues. 



The percentage of rogues due to undoubted natural 

 crossing in the four families was thus respectively 

 26, 26, 34, and 33. The control families grown from 

 long-style brothers of these, gave the same range of 

 variation, from 25 to 35 per cent. These figures have 

 been extended since, but without improvement, and we 

 are driven to the conclusion that the accessibility of the 

 style is a minor factor in natural crossing, under the 

 conditions of our breeding plot. 



The figures for the breeding plot had always been so 

 much higher than the 5 to 10 per cent, on which we 

 decided for field conditions, that this last reverse led the 



