ESTIMATION AND TESTING OF FINER GRADATIONS 135 



despite the fact that the several teachers derived 

 their judgments from observations in quite different 

 school subjects. This decided correspondence of the 

 judgments of the teachers on their pupils' intelli- 

 gence, taken in conjunction with the fair degree of in- 

 dependence of their judgment from the class-place, 

 seems to me to be a forcible argument for the scien- 

 tific usefulness of the method of intelligence estima- 

 tion. But the result also teaches us, when we com- 

 pare with it the experience gained in the elementary 

 school, that only such estimations of intelligence are 

 useful as have been carried out by an exact method 

 and with special psychological knowledge. 



4. Rank-orders of Intelligence obtained by Tests 



We can now return once more to the starting point 

 of this whole section, the experimental testing of in- 

 telligence. For we may safely regard the estimation 

 of intelligence by the teachers, when undertaken with 

 the necessary precautions, as a suitable control-de- 

 vice by which we can measure the reliability of 

 experimental testing. 



The material just now available on the correlation 

 between rank-orders obtained by tests and those ob- 

 tained by estimations is, to be sure, very scanty, yet it 

 is already enough to indicate the direction in which 

 greater results are to be looked for. Here, too, do 

 we come upon that principle that we found univer- 

 sally applicable in intelligence testing : no single test 

 of whatever kind, but only a skillfully combined sys- 

 tem of tests yields a reliable gradation of intelligence. 



Burt in England and Eies in Germany have car- 

 ried on with normal children investigations pertain- 



