84 GEOLOGY OF THE HIGH PLATEAUS. 



nite number of tentative classifications, having less and less of the artificial 

 character, and approaching nearer and nearer to the natural. Each classi- 

 fication represents its author's coordinated knowledge of the category of 

 which he treats, and the classifications which are generally accepted at any 

 time represent the stage of knowledge and induction then prevailing. No 

 system is permanent and none ought to be permanent, but they ought rather 

 to change progressively as knowledge and induction progress. Least of all 

 ought any system to attempt to represent anything more than we actually 

 know. The best system at any time is that which represents most accu- 

 rately the state of knowledge and rational induction at that time. 



The progress of classification, then, is from the simple or artificial sys- 

 tems which take account of one set or scale of characters and relations, to the 

 natural systems which take into account the totality of characters and rela- 

 tions. Hence the classification is gradually growing more and more com- 

 plex and difficult. The present conditions of most systems of classifications, 

 viewed with reference to their respective stages of progress, eeern to be 

 much nearer the artificial than to the natural. Even in those categories of 

 natural objects which sometimes are claimed to be classified according to 

 natural systems, the progress from the purely artificial has often been small 

 and the approach to the natural very distant. Though recognizing that a 

 natural classification must embrace the totality of characters, naturalists 

 still employ and are compelled to employ in many cases only a single set 

 of characters for the grouping of a given category. On the other hand, we 

 are often able to recognize correlations between the various properties or 

 characters of a group of natural objects, such that, when we arrange them 

 according to one set of characters, we find that we have also arranged them 

 (in consequence of those correlations) in logical harmony with the others. 

 But this rarely happens except in verv small groups with a narrow range 

 of variation; our knowledge is rarely equal to a full and sufficient recog- 

 nition of such correlations in large groups. Most of the later classifications, 

 however, assume the existence of such correlations while using a single 

 chai-acter as a criterion. Although this course is far from being wholly 

 satisfactory, it appears to be the only practicable one. Sometimes this 

 assumption holds true to a remarkable extent ; much more frequently the 



