20 THE AMERICAN NATURALIST [Vol. XLIV 



modes on 4 and 6, and for the number of flowers, pro- 

 nounced modes on 1 and .*>. 



Bimodal and multimodal polygons have received so 

 much attention in the literature of variation that the in- 

 terest of a more detailed investigation was at once appar- 

 ent. Before considering further our data on this species 

 we may note the more important discussion of these 

 anomalous frequency distributions. 



Bateson 4 gives a bimodal variation polygon for the 

 length of horns in the beetle Xiflotrupcs gideoii and the 

 length of the forceps in the earwig, Forficula auricularia, 

 but the factors underlying the phenomena are not demon- 

 strated. Naturally the first suggestion concerning a 

 bimodal or polymodal polygon is that there is taxonomic 

 heterogeneity in the material examined. Davenport and 

 Blankensliip" have even suggested criteria for determin- 

 ing whether the component elements of a two-humped 

 curve are to be designated as "varieties" or "species." 

 So far as I am aware the first attempt to analyze a bimodal 

 polygon into its component elements is that of De Vries 6 



pp. 685-694, 1898. 



•De Vries, H., Archiv ftir EntwiclelungsemechanU; Vol. II, 1895; also 

 Ber. Deutsch. Bot. Ges., Vol. XVII, pp. 84-98, 1899; also "Die Mutation* 

 theorie," Vol. I, pp. 526-529; Vol. II, p. 349. 



