ESTIMATION AND TESTING OF FINER GRADATIONS 133 



trained psychologically and as a member of my semi- 

 nary understood perfectly the things to be kept in 

 mind in estimating intelligence. As he had already 

 taught these pupils the year before and had given 

 them during the current year ten hours of instruc- 

 tion a week (Latin and French), it may be taken for 

 granted that he possessed a really exact acquaint- 

 ance with the material before him. It can therefore 

 be said that his estimation of their intelligence was 

 made under specially favorable conditions. More- 

 over, he had two other teachers estimate the same 

 pupils. Of these, Teacher B, it is true, instructed the 

 class only two hours a week in history and Teacher 

 C four hours a week in religion and German. The 

 instructions were to judge the children not on the 

 basis of the particular ability that they might have 

 displayed in the subjects the teachers taught, but on 

 the basis of the impression of their general intelli- 

 gence. Beside these three estimated orders there 

 was also available the series of ' class-places ' of the 

 pupils. In making my calculations I excluded eight 

 pupils who were too old. There remained 23 

 enough to permit the reckoning of valid correlations. 

 Now a first glance at the table shows that the correla- 

 tion between intelligence and class-place is much 

 lower than those obtained in most of the elementary 

 school classes. The specially reliable estimation of 

 Teacher A gives a correlation of 0.43. Of those for 

 the two other teachers, the one is somewhat higher, 

 the other somewhat lower. When we combine the 

 estimates of all three teachers into an amalgamated 

 estimation-series, we obtain again a correlation with 

 the class-place of 0.45 a value that coincides almost 



