106 FREDERICK ORPEN BOWER 



is perhaps the most prevalent source of fallacious conclusion in 

 dealing with organic life. It is the rock upon which the Linnaean 

 System split. But how many are there today who, while they 

 would speak with scornful superiority of that system as rightly 

 displaced by the so-called Natural System, would straightway set 

 themselves to write down comparative arguments based solely 

 upon some minor detail of internal anatomy, exhibited in some 

 isolated and relatively primitive organism? None have erred in 

 this respect more frequently than the anatomists. The exact 

 position of protoxylem will often weigh more with a writer (who 

 could not tell you what circumstances determine it) than the 

 whole sum of characters of the external form of the organism. 

 The corrective to this perilous method is to be found in the trac- 

 ing of parallels in as many characters as possible — in fact the adop- 

 tion of exactly the same practice in the modern comparative study 

 as was introduced by the founders of the Natural System of classi- 

 fication of Flowering Plants. 



Having thus recognized some of the sources of weakness which 

 are apt to detract from the worth and permanence of phyletic 

 comparisons, we may in conclusion state positively what are those 

 factors which would conduce to correctness, and therefore per- 

 manence in the quest of lines of descent. In the first place we 

 must be certain that we are dealing with phyletic unities. Parallel 

 development is not known to have resulted so often from similar 

 adaptations in distinct lines of evolution that it is necessary to 

 be certain that the forms compared, as illustrating a supposed pro- 

 gression are really akin. The tests of this will consist in the com- 

 parison on the basis of as many characters as possible, and in the 

 recognition in as continuous sequences as possible of intermediate 

 steps between extreme forms. Towards the latter the fossil 

 record with its sequence of horizons may be expected to supply 

 the most direct aid, especially in the case of relatively primitive 

 types. In any case the position will be the more assured in pro- 

 portion to the number of related species. If these can, on ade- 

 quate investigation, be themselves ranged into short series, show- 

 ing by parallel progression of several distinct characters that they 

 constitute true sequences, it may then become possible to link 



