260 XATL'RAL HISTORY OF PLANTS. 



one mi<j^lit fancy them the pores of a Conifer. The woody fibres of 

 the roots present the same appearance. These roots are tuberculate 

 cyliiulers, Hke the subterranean swellings of a DnhHa. Their thick- 

 ness results from the great development of the cortical parenchyma. 

 The cells are all similar, and gorged with starch-granules, which we 

 also find in the pith and the numerous medullary rays that connect it 

 witli the cells of the herbaceous layer. 



The chief characters of this order once known, we can inquire into 

 its affinities, which are numerous. In the first place, it is more or 

 less closel}'- allied to the whole of Endlicher's class Pohjcarpica, 

 especially to MagnoViacea and Menispermacece ; and generally to those 

 orders which possess ternary flowers. As we have seen, the sole diffe- 

 rence between the true Magnoliacece and Anonacea is in the seed pro- 

 vided with an arillary thickening, generalized in the former, localized 

 or absent in the latter ;' while the albumen is not truly ruminate 

 in MngnoJiacea^ as it is in Anonacece. This character is no longer 

 sufficient to completely differentiate these from Meimpermacece, for 

 the albumen is deeply partitioned in plants of the latter order.^ But 

 the habit, the size of the flower, the inflorescence, the structure of 

 the stamens and fruit supply, as Bentham & Hooker^ have shown, 

 sufficient means to separate the two groups in practice. The Lardi- 

 zabalacecB, now placed near Berberidacece, have by this very fact 

 closer relations with Papaveracece than with Anonacece. Dilleniacere 

 have not the trimerous or dimerous flowers of the Anonacea. The 

 Nutmeg order has always been considered nearly allied to Anonacea 

 on account of the aril and ruminated albumen. These resemblances 

 must now-a-days be considered as only very specious. The ajDetalous 

 flowers, the mude of diclinism, the mouadelphous stamens, are the 

 chief reasons for removing the much-reduced type of the Myris- 

 ti caeca from Ano7iacecs. Together with MagnoliacecB the order most 



' The nril disappears in ScMzandrem, which '' Especially in Bttrasaia, whose albumen is 



liavc also been compiircd with Anonaeerp. on ac- deeply ruminated, and which we were the first 



count of the type of the flower, and the habit of to refer to 3Ienispermacece (Adansonia, ii. 316). 



Sageraa, Slelechocarptis, &c. The known Schi- * " Bene limitantur habitu, injlorescentia, 



zatid)-€<T have all unisexual flowers. Jforibus parv?s, sfamuiibns, ef prcBsertim semine 



• Spach {ShU. a Bnffon, vii. 493) does not {etiam in ill/x quilns albumen rectum et runii- 



admit this differential character in all its rigour, natiim) circa endocarpium intrumm peltato- 



for, says lie, the perispcrm "is anfractuose or ciirvatov. sulcata, et embryone elongato." (Gen., 



riiu'we ill several Mfifnolias." 30.) 



