THE GENUS CROCUS. 



the least importance for specific diagnosis; nearly all the Cyanic species vary in 

 colour to white, and the Xanthic species occasionally vary to white, and even to 

 blue; but this is very exceptional. 



The Xanthic species are more constant in their colouring than the Cyanic; and 

 I know of no instance in which a blue species varies with orange flowers ; though 

 Herbert placed the orange C. susianus as a variety of the lilac C. reticulatus; and 

 C. chrysanthus and C. biflorus as varieties of one species; a view which cannot now 



be accepted. 



There are several special features in colour- variation which must be noticed. 

 Three or four species, as C. asturicus, C. versicolor, C. vernus, and C. aerius, are 

 essentially various in their colouring; and in these it is difficult to find two flowers 

 precisely alike in their colour and markings even in the same habitat: other species 

 are perfectly constant ; and again there are those which are uniform in their colouring; 

 in the same habitat which vary geographically, e.g. C. cancellous (Plate XXXI, fig. 

 i,) at its western limit, in the Ionian Islands, has white flowers, eastward, in Greece, 

 they are lilac, and still farther to the east, in Asia Minor and Syria, ^ the colour 

 deepens, and the flowers are invariably of a rich purple (Plate XXXP, fig. 2). 

 This tendency to change eastward from white to blue does not stand alone; and is 

 also noticeable in C. biflorus (Plate LIX). The Italian form is generally white, 

 varying occasionally to lilac; but in Georgia, in the variety of the species known 

 as C. Adami (Plate LIX", fig. 2), the flower is almost invariably lilac or purple. 

 There are also many instances of mimetic variation: two distinct species assuming 

 the same form of special marking or colouring, when associated in the same habitat. 

 I shall further refer to this subject in dealing with the geographical distribution of 

 the genus, and the special characters of species in relation to geographical association. 

 Filament. The Filament (PI. B) is generally white or yellow, and often partakes of 

 the colour of the throat ; in several deep purple species it is purple or lilac. In colour 

 it is not related to the colour of the anther, and in the five or six white-anthered 

 species it is distinctly orange. In C. cy P rius (Plate LVII, fig. 2) it is bright scarlet. 



The filament is generally of about half the length of the anther, and their 

 relative length is constant within each species. In C. Boissieri (Plate XX, fig. 4), 

 from Cilicia, the filament is notably longer than the anther, and there are two or 

 three species in which it is scarcely one fourth the length of the anther. 



In most species the filament is slightly papillose; and in C. pulchcllus and C. 



Tournefortii, it is densely covered with hairs; but the hairy appendages of the filament 



have no relation to the presence of the beard in the throat. 



Anthers (Plate B, fig. 1). There are few specific characters in the anthers that 



need be referred to, except that in seven or eight species the anthers are white, 



and in nearly all the remaining species orange. The colour of the anthers, unlike 



