334 WHITMAN CROSS 
well to state the principles which have guided him in his 
analysis of systematic propositions : 
1. A systematic classification of rocks has for its object the 
arrangement of all rocks in groups according to a method, pro- 
ducing a regular subordination of parts. While in all classifica- 
tions the groups must possess certain chosen characters in 
common, the system is natural or artificial according as the 
groups express community of fundamental, important character- 
istics, or of comparatively unimportant properties, chosen for 
convenience only. 
It is commonly admitted that rocks are incapable of truly 
natural classification, but that the nearest possible approach to a 
natural system is to be desired. It is, therefore, important to 
scrutinize systematic propositions to see if their factors of classi- 
fication have a fundamental connection, making the sequence 
of divisions natural in desirable degree. Each factor applied 
should have important natural relations to those preceding and 
following it in construction of the system. The writer thus 
believes that the classification of systematic petrography should 
be a hierarchical classification, in distinction to the cross-classification 
necessary in petrology. 
2. A systematic classification must be logical in construction. 
Whatever principles or criteria the architect of a system may 
adopt, he must be logically consequent and consistent in their 
application, else his structure is weak, and must fall. 
3. It is unscientific to use an assumption which is known to 
be untrue as a basis of classification. 
4. An adequate system of petrography must be stable and 
comprehensive. In order to fulfil these requirements it must 
be based upon knowledge, not upon theory or hypothesis ; it 
must be created for a// rocks, known or unknown, of the charac- 
ters now understood; its framework must be capabie of change 
in detail without injury to the structure as a whole. 
5. The use of ignorance concerning the constitution of some 
rocks as a factor for their classification is no longer permissible. 
If, in the comments to be made upon various systems, it 
