I. Leitcii 187 



In measuring the beans both length and breadth were recorded. But, 

 as far as the problem in hand is concerned, the length is of uujst im- 

 portance. The range of difference being greater, it is obviously easier to 

 distinguish differences in length than in breadth. Therefore, in the data 

 accompanying this paper, only curves for length are included. Further, 

 since after 1914 it was impossible to continue measuring every bean, a 

 random sample of 25 beans was measured from each plant. 



In the analysis of the material two different methods of representation 

 were employed. The first was the plotting of the distribution curves for 

 the lengths and breadths of the beans of each row of plants. Except in 

 1914 each row contains plants from the seed of one parent plant only. 

 The curve of a row in 1915 will therefore be an expression of the geno- 

 typical nature of the plant from which its seed was selected. And so for 

 each year. The curves for the six rows in 1914 are not so distinguished, 

 and from them only general deductions can be drawn as to the general 

 nature of the 60 seeds sown. These curves of distribution were plotted 

 directly and not calculated to constant area ; for, plotted directly, charac- 

 teristic differences are more striking, such as the flatter form of typical 

 M curves and the typically lower fertility of the 31 type. 



This method affords little help in the selection of plants for further 

 breeding. P'or that purpose some method that would give a compre- 

 hensive graphic survey of the material is desirable, and is found in the 

 plotting of what I call charts. On these charts length and index ( lOOL/B) 

 were plotted, the x axis giving length and the y index, and each plant 

 being represented by a point whose co-ordinates were respectively the 

 average length and index of its beans. Such a chart was made each year 

 for the control plants of E and M on ordinary mm. paper. Over it the 

 limits of their distributions were sketched on transparent mm. paper 

 and the hybrid forms plotted there. This gives an immediate picture of 

 the position of every single hybrid plant in relation to the two parent 

 lines and the forms can be classified on inspection. The method is of 

 course cumbrous and ill-suited for reproduction, but it proved of greater 

 practical value than any other. 



As to the 1914 harvest then, the 60 hybrid plants, bearing seeds 

 containing Fo embryos in F^ seed-coats, show in their distribution curves 

 a range of variation from below the lowest limits of the line E to the 

 upper limit of 31. And this fact at once suggests a complication of the 

 problem. A distribution varying from the lowest limit of E to the upper 

 limit of 31 might have been expected in accordance with what Professor 

 Johannsen found in the hybridisation experiments already referred to. 



