206 DOGMATISM AND EVOLUTION 



tion of older and simpler agencies. Our thoughts direct our 

 conduct, and it is in this service that their meaning ultimately 

 consists; but every concept means both more and less than any 

 particular application of it contains. 



To this we have added that the reference of a concept to a 

 mode of conduct is never direct. The concept never directly 

 bridges the gap between stimulus and response. On the con 

 trary, thought is a long-circuiting of the connection, and its 

 whole character depends upon its indirectness, its involution, if 

 we may use the term. Though concepts, apart from the conduct 

 which they prompt, mean nothing, yet their meaning : s never 

 analyzable except into other concepts, indirect like the first in 

 their reference to conduct. 



But does not this really do away with the reference altogether? 

 It certainly would, if concepts were ever (in the rationalist s 

 sense) perfectly clear, if their implications ever became perfectly 

 explicit. But as thought always arises as a problem, so it always 

 remains more or less problematic, for that is what lack of clear 

 ness amounts to. Every concept involves an indefinite number 

 of problems; and these cannot be stated except in terms which 

 themselves in turn involve indefinite series of problems. No 

 where is there an absolute given, a self-sufficient first premise. 

 From this, as well as from the indirect and equivocal nature of 

 the reference of thought to conduct, it follows that the confirma 

 tion or invalidation of a concept by the result of the conduct 

 which it serves to guide can itself be no more than tentative. 

 But this does not mean that it is unreal or unessential to the 

 nature or development of thought. 



These considerations, however, have a decided bearing upon 

 the pragmatist contention, that apart from its reference to con 

 duct thought has no form. This is naturally understood to imply 

 that the nature of thought may be exhaustively described in the 

 statement of its relation to conduct. Now |t is very probable 

 that the statement of the relation between two terms may be 

 indefinitely developed, so as to include any assignable attribute 



