138 PSYCHOLOGICAL METHODS OF TESTING INTELLIGENCE 



high correlations that Method A or Method B taken 

 alone are adequate for testing and ranking intelli- 

 gence. For, in the first place, Bies' results do not 



TABLE XIX 



Ries' experiments on 24 boys (Mittelschule, 2d class, ages 12-14). 



Test A : Method of word-pairs and right associates. 



Test B : Association of effect to a given cause. 



Correlation of Test A with Test B 0.61 



Correlation of Test A with Estimated Intelligence 0.85 



Correlation of Test B with Estimated Intelligence 0.94 



Correlation of Tests A and B (combined) with Est. Int 0.98 



present the requisite uniformity (in one class, 

 Method A correlated with estimated intelligence by 

 only 0.59), and it is very questionable whether repe- 

 tition of the tests in other places would furnish the 

 same high correlations. Again, each of his methods 

 tests only one phase of intelligence, and a comparison 

 of the two methods with one another shows how little 

 right we have to infer one phase from the other. 

 Thus, Eies gives for one class a table of the original 

 data from which I have been able to calculate some 

 results not mentioned by him (Table XIX). The re- 

 sult is that the two methods do not correlate at all 

 highly with one another, only 0.61; in other words, 

 the ranking of intelligence by Method A furnishes a 

 distribution of stations that is in some parts 

 quite different from the distribution furnished by 

 Method B. 



The example is, however, excellently adapted to 

 point out the way toward the method that is to be ap- 

 plied. 



What does it mean that both tests correlate so high 

 with estimated intelligence, but so low with one an- 



