XX . PROCEEDIJTGS OF THE 



character as well as more or less of fixity. We may suppose this to 

 be going on through millions of ages, innumerable branches, whether 

 near the centre or more or less distant from it, ceasing to grow or 

 to branch out, leaving gaps in the upper part of the tree, partially 

 filled up, perhaps, in a few instances by returning branches from the 

 circumferential ones, and all decaying at the base, leaving only their 

 upper extremities to continue the process in future ages. We should 

 then have the present races represented by the countless branchlets 

 forming the flat-topped summit of the Dicotyledonous tree — a hun- 

 dred to a hundred and fifty thousand perhaps if we take into ac- 

 count species only, ten times as many if we go into subspecies and 

 varieties; the branches which immediately bore these present 

 branchlets, as well as the lower more general ramifications, will 

 have wholly disappeared from our view, or left only here and there 

 the most fragmentary traces ; and the surviving branchlets them- 

 selves will be most irregularly placed. Here we should see thou- 

 sands crowded into compact patches definitely circumscribed at every 

 point (Compositse, Orchidese, Graminese, &c.) ; there we should meet 

 with enormous gaps, either quite unoccupied or a few solitary 

 branchlets or small clusters isolated in the middle {Moringa, Aristo- 

 loclda, Nepenthes, &c.). In other parts, again, irregular masses 

 may be more or less connected by loosely scattered branchlets or 

 clusters, obliterating all boundaries we might be disposed to assign 

 to them (many of the bicarpellary gamopetalous orders, the several 

 curvembryous orders, &ic.). In the imaginary construction of such 

 a tree, all we can do is to map out the summit as it were from a 

 bu'd's-eye view, and under each cluster, or cluster of clusters, to 

 place as the common trunk an imaginaiy type of a genus, order, or 

 class, according to the depth to which we would go. If we believe 

 that this type, or original trunk-branoh, is exactly represented by 

 (has descended unchanged to) one of the present branchlets, we 

 place it immediately under that branchlet, as having been directly 

 continuous with it, and regard the remainder of the cluster as the 

 persistent summits of lateral ofi'sets. If we consider that the direct 

 trunk-race of a cluster has become extinct in its precise form, and 

 has left descendants only from its branches, we place it under one of 

 the gaps in the cluster or under a vacancy outside the cluster, ac- 

 cording to the conjectures we may think the most plausible, as de- 

 rived from the relative structures, geographical relations, &c. of the 

 present branchlets or other evidences we can bring to bear upon the 

 question. Such circumstantial evidence will always be exceedingly 



