A. BAKCLAYANA. 319 



those common in the genus and less highly modified from the original leaf-like structure. 

 The thickening of the bracts in lypica has resulted in the almost complete loss of the thin, 

 herbaceous borders. From these considerations it becomes necessary to place the pal- 

 men group first in a phylogenetic arrangement, reserving the highly specialized taxonomic 

 type of the species, that is, subspecies lypica, for treatment at the end. 



The subspecies palmeri exhibits no features indicative of a special development. On 

 the other hand, the remarkable width of the compressed bracts, mentioned above as 

 one of the characters of the palmeri branch of the species finds its greatest development in 

 subspecies dilatata (especially the minor variation called A. rosei Standley). Although 

 the shape fluctuates to such a degree that future studies may indicate the impracticabil- 

 ity of the distinction, it seems helpful at present to retain a subspecies dilatata for this 

 form. Possibly other substantiating characters will be found after the forms have 

 received a more thorough field study. As far as now known, dilatata is confined to two 

 islands ofT the west coast of Lower California (where true palmeri also grows), but one 

 collection from Raza Island, in the Gulf of California, has bracts up to 5 mm. wide and is 

 therefore to be classified as very close if not the same as dilatata (Johnston 3232, SF, 

 UC), and other collections from Raza Island (Palmer 155 and 156), here cited as sub- 

 species palmeri, were identified by Rose (Contr. U. S. Nat. Herb. 1 :80, 1890) as dilatata. 



The subspecies magdalenae is here maintained only because no direct intergradation 

 with palmeri is known, although the two come from the same island. The herbaceous, 

 prostrate habit and the less strongly compressed nature of the bracts will furnish char- 

 acters substantiating those used in the key, if the former are found to be constant, which 

 seems unlikely. Since the bracts are longer than broad and only lightly compressed, this 

 subspecies should perhaps stand ahead of palmeri, connecting this with typical barclay- 

 ana. Its rare occurrence, however, and this only on an outlying island, is evidence that 

 it is a divergent type that did not give rise to the other forms. 



The most easily distinguished subspecies and the one least subject to intergradation 

 is lurida. This is the only one thus far found to be generally distributed on the mainland 

 of Lower California, exclusive of the Cape Region the flora of which is more like that of 

 the southern islands. It has not been collected on any of the islands. Two collections 

 of typica have been made on the west coast, opposite Cedros Island. Aside from this, 

 the other subspecies occur only in Sonora and Sinaloa and on the islands off both sides 

 of the peninsula. In some cases a single subspecies is known both from the gulf and from 

 islands off the west coast, yet none of them are known to mingle with lurida. The near- 

 est approach is at Agua Verde and Los Dolores, where lurida grows on the mainland and 

 other subspecies on the adjacent islands. Further field studies and also experiments 

 are needed to determine whether lurida is an ecologic response to the mainland environ- 

 ment or an hereditary type held true by isolation. 



The typica branch has produced but two subspecies, typica itself and sonorae. This 

 latter is characterized by having the margin of the bract more fully developed around the 

 summit, where it is deeply cut into a number of sharp teeth. This commonly leaves the 

 lower portion as a spindle-shaped structure, the base of which is pedicel-like. In the 

 original diagnosis, Standley differentiated this form also on the failure of the bracts to 

 unite above the middle. However, this lack of union applies in most cases only to the 

 marginal wings and therefore may be looked upon as another expression of a single char- 

 acter. A careful examination of the much more copious material now at hand indicates 

 that the extent of fusion varies too greatly in both subspecies to be of taxonomic value. 

 After all of the available material has been classified into typica and sonorae on the basis 

 of the bracts, it is found that, while they have the same general area of distribution, they 

 seldom occur at the same station. This is indicated by the above citation of collections. 

 The distribution indicates that they may be genetic types preserved in their charac- 



