ESTIMATION AND TESTING OF FINER GRADATIONS 115 



whether we can infer from the reactive intelligence, 

 however many tests of it we make, to this spontane- 

 ous intelligence. It is possible that certain tests or 

 certain combinations of tests have a fairly high cor- 

 relation with the spontaneous intelligence, but that 

 is something that can not be determined from the ex- 

 periments. 



Hence the mere comparison of tests with one an- 

 other affords us neither a clear insight into the neces- 

 sary compensations, nor a decision as to the sympto- 

 matic value of the testing; rather must we seek the 

 means of guaging the tests in some criterion that lies 

 outside of experiment. Such a criterion is supplied 

 by the estimation of the pupils made by the teacher. 



It follows that this estimation of intelligence by 

 the teacher thus comes to possess a methodological 

 significance of its own, for it can be set up as a stand- 

 ard for the comparison of other rank-orders, i. e., 

 those obtained experimentally, only when we have 

 first made sure of its own nature and its reliability. 

 There are two ways of going about this; the one is 

 by analysis of the procedure of the teacher when he 

 does the estimating of his pupils' intelligence, the 

 other is by determining objectively to what extent 

 his estimation is dependent upon what he knows 

 about the pedagogical rank-order of the children, 

 their places in the class or their examination marks, 

 etc. 



So, only when we have first discovered whether 

 the estimated intelligence is a useful means of con- 

 trol or under what precautions it is useful, can the 

 real experimental problem be attacked: the prob- 

 lem of finding out those combinations of tests that 



