ESTIMATION AND TESTING OF FINER GRADATIONS 125 



group. The series thus secured is then noted down 

 with the proper number for each individual. 



It is an excellent plan to do the work all over again 

 after an interval of perhaps three or four weeks, and 

 without referring to the series first obtained. The 

 degree of correspondence between the two estima- 

 tions is then to be determined by the correlation 

 method. Only if the coefficient of reliability is high, 

 i. e., if the two series are very similar to one another, 

 should they be made the basis of further investiga- 

 tions. If they are then to be used, it is best to con- 

 struct an amalgamated series out of the two estima- 

 tions by taking for each pupil the mean of the two 

 rank-numbers that he has obtained and bringing to- 

 gether these means for the new rank-order. 



Distinct differences in method will appear when 

 the estimation of intelligence is made in elementary 

 schools (Volksschule) than when in higher schools. 



The elementary school teacher has the particular 

 advantage that he is usually the only teacher of the 

 class, and thus can observe the behavior of the chil- 

 dren in the most varied activities, in technical and 

 theoretical subjects, at play and at work. But just 

 this very breadth of information also renders him in 

 a certain sense less independent in his estimation. 

 Because his knowledge of his pupils extends in so 

 comprehensive a manner over their school perform- 

 ances; all their grades and the determination of 

 their class-places are the product of a single teacher, 

 so that it is psychologically easily intelligible that he 

 can not so easily free himself from this judgment 

 that he has himself worked out, even when he under- 



