208 SO. E. KOEHLERI. 



and the prickles on the midrib and petiole are few and 

 weak. The panicle has a remarkable appearance ; for its 

 branches (which are few) are erect, the uppermost alone 

 spreading so as to be erect-patent, and their lengths are such 

 as to place the flowers in an irregular convex corymb. On 

 my specimen there are only two branches which do not form 

 part of this corymb and which are not themselves corymbose ; 

 their lower half is long and naked and the upper forms a 

 raceme of flowers. It seems not impossible that this may be 

 a state of R. pallidus, but my materials are not sufficient 

 from which to form a satisfactory oj)inion. 



The R. glmululosus of Smith is very difierent from that 

 of BeHardi, and is unquestionably referable to R. Koehleri 

 a verus. It is the var. cuspidatus (an ill-chosen name) of my 

 Synojysis. Its leaflet is of an unequally rhomboidal form 

 (the lower half of the rhomb being longer than the upper), 

 with its upper part very regularly narrowed to the point, 

 but having its edge lobate-serrate j the lower part likewise 

 nan'ows gi-adually until close to the base, where it is rounded 

 and slightly notched. The upper part of the leaflet in tiTie 

 R. Koehleri is often very similar, but the tip projects slightly 

 more from the general outline ; also, the base is rather 

 broadly or truncately cordate. The panicle of Smith's plant 

 is almost exactly that of R. pallidus. This plant seems to 

 lie between those two marked forms of the species, and has 

 helped to convince me of their specific identity. It is cer- 

 tain that this is the R. glandidosus of Smith, for Mr D. 

 Turner (who originally sent it to Smith from !R,ydal in 

 Westmoreland) identified with his plant the specimens 

 gathered at the same place by Mr Borrer, who kindly pre- 

 sented some of them to me. His words were that "-S. 

 glandulosus (Sm.) is wholly this plant of this place." The 

 other plants which I placed under var. cuspidatus approach 

 XiXOVQ nearly to the true R. Koehleri, 



