Chapter XVIII 



CHOOSING PHYSIOLOGICAL VARIABLES 



§ 151. As the facts and relations have accumulated in this in- 

 vestigation, certain rules for their selection gradually appeared. I 

 was not consciously formulating such rules; rather the materials 

 themselves seemed to move into place, and I then began to wonder 

 by what tokens they gravitated. Given the objective of finding 

 regulatory processes of kinds that could prevail in all animals, the 

 steps by which that search was accomplished became partly auto- 

 matic. Now that the objective is achieved, I believe it is worth 

 while to turn aside long enough to examine the general import of 

 these rules of procedure. 



This is not a treatise on methods. Yet, every scientist has 

 methods. They are equally effective and decisive in the substance 

 of his results, whether or not he recognizes and classifies them. 

 Methods are usually gained by unconscious experience and by the 

 examples of others, almost never by precept. 



Scientists find strange those procedures they do not habitually 

 use. Sometimes suspicions are aroused by them. To understand 

 all investigations, it is necessary to admit that ways of doing things 

 other than those sanctioned by custom may be justified, even be 

 successful. For, to limit the solution of inquiries to the sort of 

 contribution to science for which tastes are already cultivated 

 would be to exclude all efforts to see beyond the limitations of 

 present outlook. 



<^ 152. Peoceduees 



How did this investigation actually proceed? Though most of 

 the steps taken were unforeseen, there was a pattern among them 

 (§104). What ones of the procedures might illustrate general 

 usages in scientific work? Would they be useful in planning 

 investigations, in so far as investigations are explicitly planned? 



A basis of this inquiry consisted in distinguishing among physi- 

 ological variables. Each variable was some quantity that could be 

 identified in a defined and reproducible manner. Since the identity 

 of the quantity depended upon the operations by which it was mea- 

 sured, each variable had dimensions that in part at once offered a 



422 



