PLAN OF THE INVESTIGATION. 27 



STATISTICAL EXPRESSION OF THE MEASUREMENTS. 



In our previous discussion of the general methodological considera- 

 tions we have pointed out certain limitations in the outside control of 

 our subjects. It is in several respects unfortunate that the 3 hours of 

 confinement in the laboratory determined the practical limits of strict 

 experimental procedure. To have predetermined for all our subjects 

 the antecedent ingestion of foods and fluids, antecedent voiding of 

 feces and urine, antecedent amounts and kinds of mental and physical 

 activity, and antecedent periods of rest, with fixed waking and sleeping 

 hours, would have been difficult, if not impracticable. But even these 

 precautions, valuable as they might be, must have failed to provide 

 strict similarity of conditions on successive days. Experimental inter- 

 ference with the spontaneous reactions of intelligent and busy men 

 in their routine demands for food and drink, work and rest, might 

 easily produce mental and even purely physical disturbances which it 

 would be difficult or impossible to measure. At best we could not 

 control the immediate and remote effects of " colds," intestinal disturb- 

 ances, and other slight infections of the mucous membrane. It would 

 be obviously impossible to take account of all these and numberless 

 similar variations, and at the same time provide for similar phases of 

 the weekly and yearly rhythms, possible climatic changes, etc. A pre- 

 liminary exploration of the possible disturbances to discover then* re- 

 spective significance for each of our subjects would have been entirely 

 impracticable. Admitting the desirability of enforcing stricter experi- 

 mental conditions outside of laboratory hours, it seems that the best 

 practical procedure in the use of subjects whose outside activities are 

 not strictly regulated is to treat all except obvious or gross disturb- 

 ances as chance variations, which can not obscure any really significant 

 tendency in the group, provided the measurements are numerous 

 enough and their statistical treatment is adequate. 



In thus emphasizing the group system of comparison we are merely 

 following established precedents in this laboratory since its beginning. 

 The resources and facilities of the laboratory are such as to make it 

 practically obligatory that conclusions should not be drawn from exper- 

 imental data until there are sufficiently large groups of individuals 

 on whom the special observations have been made, and a sufficiently 

 large number of normal experiments for adequate comparison. In the 

 previously published studies from this Laboratory on diabetes and 

 on the metabolism of athletes, vegetarians, and normal and atrophic 

 infants, the group system has been invariably applied. 



Our main statistical requirement, after provision is made for a suffi- 

 cient number of accurate measurements of significant and systematic- 

 ally related processes, is to provide for the best basis of comparison of 

 the normal and alcohol experiments. 



