A. BARCLAYANA. 317. 



white-scaly, but the bract characters assign the plants cited to this subspecies. Among the collections listed 

 above as lypica and sonorae are some with leaves that are narrow and elliptic, but still white or gray with a 

 dense scurf. 



2. Atru'lkx BARCLAYANA TYPicA, but with dense subcylindric and pyramidal panicles 20 to 30 cm. long by 

 8 to 10 cm. broad, heavy with fruiting bracts. This form is from a protected draw on Patos Island, Johnston 

 3243. It is not even approached in density and size of the inflorescence by any of the other numerous col- 

 lections. 



3. A. iNsuLARis Rose, Contr. U. S. Nat. Herb. 1: 80, 1890. — ^The form or state of A. barclayana palmeri in 

 which the bracts are appendaged by two conspicuous toothed. crests. The type is from Raza Island, in the Gulf 

 of California. Since this character is unreliable in other species, it is not surprising to find that in nearly all of 

 the collections in which mo.st of the bracts are conspicuou.sly crested there are some (on the same plant) with 

 smooth sides. A number of such are cited under subspecies palmeri. An attempt was made in the North 

 American Flora to use also the more fruticose branches and the more nearly sessile leaves as criteria for the 

 separation of iiisularis. But the type specimens indicate that palmeri is perhaps as woody as insularix and the 

 original descriptions ("shrubby at base" for the former, "woody below" for the latter) furnish no support for 

 the distinction. Mr. Johnston, who collected abundantly on Raza Island, considered the plants as all clo.sely 

 similar in this respect, that is, all had a woody base and more or less herbaceous, ascending or decumbent 

 branches. His plants are all of the iiisularis form as to most of the bracts, except No. 3211, which is a staminate 

 plant with well-developed fruiting bracts in the leaf a.xils. These bracts are mostly smooth as in typical palmeri, 

 yet this was one of the most woody plants found. Other plants referred to insularis on the characters of the 

 bracts, such as Palmer's 862, are less than 3 dm. high and herbaceous nearly to the ba,se. The length of petiole 

 is equally unsatisfactory. On the type sheet of insularis .some petioles are one-third the length of the blade, 

 exactly the proportion stated for palmeri, and the large scries of specimens now at hand shows conclusively that 

 this feature is extremely variable, and that the variations do not parallel those of other characters. Therefore, 

 if insularis is to be retained in any rank, it must be on the single character of the appendages, and even these are 

 absent from some of the bracts. 



4. A. ROSEi Standley, N. Am. Fl. 21:60, 1916. — The extreme form of A. barclayana dilaUUa in which the 

 bracts are 6 to 8 mm. wide and with conspicuous crests on the face, the seeds correspondingly large. The typ)e 

 collection from Gaudalupe Island includes a splendid series of bracts, some of which are nearly twice as wde as 

 long. Others of the same collection are much narrower, and some, perhaps immature, are scarcely wider than 

 long. For the most part, they are much thicker than in other collections of subspecies diUUala and palmeri, 

 some apparently not at all compressed, but this is due to the exceptional development of the tubercles with 

 thickened bases, as indicated by other bracts, also of the type specimen, which are strongly compressed and 

 with only a few minute appendages on the face. Some of the leaves are slightly sinuate or subdentate. 



RELATIONSHIPS. 



This species stands near the beginning of that great phylogenetic branch of Atriplexes 

 characterized by a shrubby habit and dioecism. The evidence for this is found in the 

 only half-shrubby nature of many of the plants and in the almost constant occurrence of 

 at least a few pistillate flowers on the staminate plants. These female flowers are usu- 

 ally well formed and are protected by normal bracts which grow in clusters from the upper 

 leaf-axils, that is, just below the long and dense staminate inflorescence. In a few cases 

 no trace of fruiting bracts can be found, indicating that a complete separation has been 

 effected in these individuals. 



In assembling the numerous variations of A. barclayana, it is found that these fall into 

 two groups, which may be conveniently referred to as the palmeri and the typica groups. 

 These differ only in their fruiting bracts. In the former, the bracts are distinctly com- 

 pressed, so that the body is more or less strongly flattened, that is, they are as in most 

 Atriplexes only slight modifications from the reduced fohage leaf. These compressed 

 bracts are sometimes thickened by the enlarged bases of the appendages, or' tubercles, 

 but this is not to be confu.sed with the swelling or thickening of the body proper. The 

 bracts of the typica group are distinctly thickened and spongious, so that the body is 

 strongly convex on each face, that is, approaching globose. The bracts of the palmeri 

 group are usually wider in proportion to their length than are those of the typica group. 

 In each the length and width are sometimes equal, and the subspecies magdalenae is an 

 exception, but in the main the tendencies are as indicated. It seems that the loss of 



