E. R. Saunders 367 



in shape that a lumpy seed can only be found now and again. In the 

 glabrous red, on the other hand, the pods are so short that though very 

 much fewer in number the ripe seeds are crowded together. Yet the 

 same proportion of doubles is obtained from each of the three strains. 



We may therefore conclude that no system of selection based on 

 the shape of the seed will enable us to obtain a proportion of doubles 

 which is constantly above the average. In the case of certain 

 sulphur-whites however it is quite possible by sorting the seeds 

 according to colour to separate almost completely those giving rise 

 to singles from those producing doubles. The present experiments 

 have shown that there are at least two types of sulphur-white on the 

 market, one in which the seeds are small, brown, often irregular in 

 shape, and indistinguishable in ajjpeai-ance from those of a true-breeding 

 white ; the other in which the seeds are very regular, larger, of a lighter 

 j'ellowish colour, and similar to those of the cream race Princess May. 

 These two types have no doubt a different origin, and are different in 

 constitution (see later, p. 370). In the case of the type with the yellow 

 seeds it was found possible in well ripened pods to sort the very yellow 

 seeds which give rise to the creams which are all double from the less 

 yellow seeds which give rise to whites of which nearly all are single. 

 The following result will show the degree of accuracy which can be 

 reached by this method. 



Of 81 seeds taken from one pod 



48 were expected to give rise to creams 33 to whites 



38 germinated 27 germinated 



34 flowered 26 flowered 



30 were cream doubles 24 were white and all single 



4 were white and all single 2 were cream doubles 



Of 72 seeds taken from another pod 



44 were expected to give rise to creams 28 to whites 



28 germinated 25 germinated 



28 flowered 16 flowered 



27 were cream doubles 16 were white (15 single, 1 double) 



1 was white and single were cream 



Thus of the 60 doubles which were obtained 57 were recognised by 

 the seed-colour ; of the whites 5 were wrongly classed as probable 

 creams, but the remaining 40 were correctly identified, and 39 proved 

 to be single. A very slight error must however always remain in 

 sorting the singles from doubles, since the rare double white is not 

 distinguishable in seed-colour from a single white. 



