92 Social Environment 



that weigh so heavily in individual cases. But 

 after all, the question is not one requiring a 

 precise answer. With proper safeguarding any 

 large list of prominent persons will do, even 

 though exceptions might be taken to a consid- 

 erable percentage of the names included. For 

 there must be borne in mind in this connection, 

 as well as in the discussion to follow, a 

 principle applicable to correlation work. The 

 principle is that the erroneous inclusion or ex- 

 clusion of material from the data, arising from 

 chance and not from any selective guidance, 

 must always operate in the long run to lower a 

 correlation.^ As an illustration of the prin- 

 ciple, suppose that one has picked out the large 

 and small grains from a given sample of wheat. 

 The large and small being kept separate, stand 

 out as two distinct types. Let a few handfuls 

 of wheat like the original sample be thrown 

 into the selected lots, and the contrast begins 

 to be blurred. Let a larger quantity be thrown 

 into each, and the contrast practically disap- 

 pears. Just so in the handling of data to dis- 

 cover correlations. Some miscellaneous errors 

 may be included in the data, but any correlation 



1 Whipple, G. M., Manual of Mental and Physical 

 Tests, pp. 41-42. Warwick & York, Baltimore, 1910. 



