ESTIMATION AND TESTING OF FINER GRADATIONS 113 



ranks high in the discrimination of line-lengths, in 

 memory for nonsense syllables, etc., while Pupil Z 

 ranks low in each of these functions. For, if the re- 

 sult were conditioned by specific abilities, a given 

 pupil would occupy very different ranks in the dif- 

 ferent tests. Spearman has, indeed, found ex- 

 traordinarily high correlations in some instances: 

 on this account he holds it to be demonstrated that 

 there really is such a thing as general intelligence, 

 and that its grade can be experimentally determined 

 by tests that correlate to a high degree with one an- 

 other. 



We can agree with the first of these conclusions. 

 We have already alluded frequently in what has 

 gone before to the 'general' and ' formal' character 

 of intelligence, whose influence is operative in activi- 

 ties of widely differing character. Of course, we 

 must admit that this influence of intelligence is 

 never more than approximately uniform, that within 

 the " general intelligence " of every person there 

 exist certain specially strong and certain specially 

 weak points, so that a truer picture of the total in- 

 telligence of the individual is given by the idea of a 

 mutual balancing or compensation of different capac- 

 ities than by the idea of their equality or corre- 

 spondence. 



But just here does the value of Spearman's 

 method for the testing of intelligence become dubi- 

 ous. If we select four or five tests that show very 

 high intercorrelations in order to use their totals as 

 a measure of intelligence, there exists the danger 

 that we may be testing by them only a very restricted 

 portion of the field of intelligence and leaving en- 



