94 Colours of Sweet Peas [CH. 



instead of a light purplish blue. Similarly reds having the 

 factor for light-wing are Painted Lady, if without it are of 

 the type called Miss Hunt, with wings of a full pink. 



Another pair of subordinate classes is due to the 

 absence of a factor for whole-colour. When this factor is 

 absent the purples become purple picotees, with an 

 edging of colour on a white ground, and the reds become 

 what we have called "tinged whites" ; a type with a little 

 pink in the standard only. The purple picotees, when 

 the bud is just opening, have a purple edge to the 

 standard and a blue edge to the wings, these edges being 

 all that remains of the colours proper to these parts 

 in whole-coloured purple. The " tinged white," which is 

 really a red picotee, has the red of the Painted Lady 

 standard reduced to a few pink lines in the centre of the 

 standard. Both these picotee forms, like the picotees of 

 tulips and most other flowers, acquire a more general flush 

 of colour as the flower ages. These various types are 

 illustrated in the coloured plate (Plate V), from which their 

 peculiarities will be easily appreciated. Among the picotees 

 of course there are two types of purples and two types of 

 reds, differing in the presence or absence of the factor for 

 light wing, but in forms so nearly white these differences 

 are evasive, and we have not found it possible to classify 

 the picotee class in respect of these distinctions. 



It is to be observed that the simultaneous appearance 

 of the subordinate types in each of the classes is a charac- 

 teristic and necessary feature of the normal Mendelian 

 processes. It would be impossible, for instance, to make a 

 plant which could throw purple picotees in /% without 

 red picotees, if the family contain both purples and reds. 

 Conversely, if the plants throw purples and reds, and also 

 whole-colours and picotees, then both kinds must be repre- 

 sented in both classes. In any examination of /% families, 

 so soon as one pair of large classes can be recognized and 

 one subordinate type perceived in one of them, search 

 should be made in the other class in order to find the 

 corresponding subordinate type in it also. If in one class 

 a type exists which is never to be found in the other class 

 though a long series has been examined, some complication 

 is to be suspected, but whether the phenomenon is due to 



