ORIGIN AND DEVELOPMENT OF THE SECTIONS. 237 



found to be too variable to serve as taxonomic criteria. Other revisions, covering por- 

 tions of North America, will be found in the various floras. The most notable of the 

 recent contributions of this nature will be found in Jepson's Flora of California (pp. 433 

 to 442, 1914) and in Rydberg's Flora of the Rocky Mountains and Adjacent Plains 

 (pp. 245 to 250, 1917). 



The present authors have greatly benefited from the monographs and revisions just 

 enumerated. Especial attention, however, has been given to the assembling of pre- 

 viously unknown facts regarding the plants, more particularly through field studies 

 and cultural experiments. The result is that the forms can now be more accurately 

 described and their phylogeny worked out in a manner much more satisfactory than 

 was previously possible. 



ORIGIN AND DEVELOPMENT OF THE SECTIONS. 



The American species of Atriplex can not be assembled into natural groups and the 

 origin of these satisfactorily determined until those of other parts of the world have 

 also been studied in some detail. Therefore, aside from the two subgenera almost 

 universally accepted, no formal classification into sections will be here presented. The 

 group names proposed by Standley (N. Am. Fl. 21:33-72, 1916) will be sometimes used 

 for convenience, as will also other sectional and subgeneric names which have been 

 proposed by European workers without much regard to American forms. 



The primitive stock from which Atriplex has developed was almost certainly of a 

 type in which the radicle pointed downward in the embryo and in which the essential 

 organ of the pistillate flower was subtended by some sort of a perianth. The inferior 

 position of the radicle is so common in the Chenopodiales that any modification from this 

 indicates a divergent phylogenetic line. Similarly the absence of a perianth is looked 

 upon as a case of suppression and therefore a mark of advance. With these postulates 

 in mind it becomes impossible to select any one Atriplex as the common ancestor. The 

 one most nearly meeting the requirements is A. hortensis, the taxonomic type of the 

 genus. Therefore, this species is placed at the beginning of the subgenus Euatriplex, 

 i. e., the branch in which the radicle is always inferior and the perianth either present 

 or absent in the pistillate flowers. In the only other subgenus, namely Obione, the 

 embryo always is inverted, so that the radicle points upward (superior) and a much 

 reduced perianth is found only in a few primitive species. 



Of the American species of Euatriplex only hortensis, itself introduced, exhibits a 

 true perianth in the pistillate flowers. Even here a partial suppression has taken place, 

 since only a portion of the flowers possess this structure. This species, together with 

 a few minor derivatives, comprises the section Dichospermum Dumortier. Next to 

 hortensis, but probably not a derivative from it, comes the section Teutliopsis Dumortier, 

 which includes A. patula (section Hastatae Standley) with a long series of subspecies 

 and minor variations widely distributed in both hemispheres. It has a perianth only 

 in the staminate flowers and has developed features of the bracts which are quite dis- 

 tinctive. The absence of fusion between the bracts in some of the forms and the lack of 

 complete fusion in any of them may be taken as indicating a low stage of development, 

 while other considerations, especially the wide distribution and the abundance in a large 

 variety of ecologic habitats, indicate this as a very successful type. 



Atriplex calif ornica is placed near to patula in the phylogenetic chart, but probably 

 its connections are much more primitive. Its position is very doubtful as will be explained 

 in discussing its relationships (p. 257). Also, A. semibaccata and A. halimoides, two species 

 introduced from Australia and doubtless of a very different alliance, are given place in 

 the taxonomic sequence as a matter of convenience. In the former the radicle is lateral, 

 rather than inferior, suggesting a connection with the subgenus Obione. 



