RELAPSE AND RETROGRESSION. 85 



set up, producing the violet-blue or purple florets cf 

 the salsify {Tragopogon porrifolius), the deep blue 

 Sonchus alpimis, and the bright mauve succory, 

 CicJiorium intybus. As a whole, however, the Ligu- 

 lates are characterised by what seems a primitive 

 golden yellow, only occasionally rising to orange-red 

 or primrose in a few hawk weeds. 



That this hypothetical explanation may be the 

 true one seems more probable when we examine the 

 somewhat similar case of the StellatcB. Here it seems 

 pretty clear that mere dwarfing of the flowers, by 

 throwing them back upon earlier types of insect 

 fertilisation, has a tendency to produce retrogression 

 in colour. Even in the more closely allied DipsacecEy 

 ValeriaiiecBy and CanipanulacecBy we see a step taken 

 in the same direction, for while the large-flowered 

 Campanulas and Scabiosas are bright blue, the smaller 

 flowered teasel {Dipsaais silvestris) is pale lilac, the 

 Valerianas are almost white, and the Valcriandlas 

 are often all but colourless. In the StcllatcCy the 

 same tendency is carried even further. As a whole, 

 these small creeping weeds of the temperate regions 

 form a divergent group of the tropical Rubiacece (in- 

 cluding CiucJioniaccce)y from which they are clearly 

 derived as a degraded or dwarfed sub-order. Their 

 square stems, their leaf-like interpetiolar stipules, and 

 their usually lessened number of corolla-lobes, all 

 point them out as derivative forms, not as survivals 

 from an early ancestral type. Now, the tropical 

 Rubiacece have tubular blossoms with long throats, 

 and as x rule with five lobes to the corolla ; but many 

 of the Stellates have lost the tube and one corolla 

 lobe. Sherardia arvensis, which has departed least 



