MEASUREMENT OF VARIABLE PROPERTIES 177 



are in danger of being deceived. In certain cases, however, 

 a comparison between two or several mean values oUaincd 

 tinder definite conditions may render good service. 



Example : Wheat is cultivated on two fields A and B, all the 

 conditions of existence being as exactly as possible the same for 

 both 1 — in such a way that the variation curve of any given 

 property of the wheat would be the same on A and B. A 

 certain quantity of potash manure is added to B. When the 

 wheat is ripe, one of its properties — for instance, the length of 

 the spike — is measured. If any difference exists between the 

 mean values of A and B it may be considered as an indirect 

 measure of the influence of the added potash. 



In general, it may be said that a comparison between two or 

 several mean values gives useful information as often as the 

 series of measurements have been collected under conditions 

 which were different only with regard to one factor. 



According to this principle, I use the mean values for the in- 

 vestigation of the gradation curves, the varying factor being 

 here a social cause. (See Part VIII.) In a similar way I have 

 used mean values for the investigation of the growth of certain 

 properties of a mixture (§ 114, pp. 155 and 157), the varying 

 factor being here the age of the measured specimens. 



' In other words, the same series of combinations of factors with the same 

 frequencies prevailing on both fields. 



