DEPARTMENT OF GENETICS. 



137 



material, which included only 404 children, was grouped into four classes of 

 equal grade intervals. These constituted the 'full table.' For the sake of 

 comparison, r was computed for both the Oxford and Harvard tables on the 

 assumption that the classes in each were of equal value. As a further test of 

 the results, a correlation table of the Harvard data, called the 'expanded' 

 table, was constructed on the grades divided to tenths. In all three cases r 

 was computed by Pearson's 'product-moment' method. 



Table 8. — Coefficient of contingency and correlation in ability between fathers and children. 



Ci = mean square contingency coefficient. 



C2 = mean contingency coefficient. 



r = coefficient of correlation or association. 



'Pearson's "Theory of contingency." Drapers' Co. Research Mem.; Biom. Series I., Math. 

 Contrib. to the Theory of Evolution XIII, 1904, as cited by Schuster and Elderton, op. cit., 7. 



^Pearson's "Product-moment" method, after West: Introduction to Math. Statistics, 83-84, 



'Pearson's "Correlation of characters not quantitatively measurable,'' Phil. Trans. Royal Soc. 

 of London, Series A. 195: 1-47, 1900, as cited by Schuster and Elderton, op. cit., 7. 



Whipple, Manual of Mental and Physical Tests, 1 :50,. 



*By Yule's formula; Q (or r) =cos 



'By "Method of unlike signs." 

 3d ed., 1914. 



"A further comparison of the percentage of children in different grades 

 whose fathers were of first or second grade gave a difference in the parallelism 

 of the fitted straight lines of the Harvard and Oxford data of only 4' of. arc, 

 i. e., a divergence of about 1 mm. in 1.75 meters. Various other statistical 

 values were computed which need not be discussed here. 



"The conclusion drawn was that the data, on the whole, were sufficiently 

 reliable to give significant indications. Although scholarship grades may be 

 considered a crude measure of intellectual ability and far from ideal for 

 scientific purposes, they are at present the best that we have covering a period 

 of more than one generation. In view of the results obtained and the lack of 

 any more reliable data, it seems very desirable to utilize such material in 

 further investigations. 



"One serious fault in both the Harvard and Oxford material as a basis for 

 studies in hereditj^ is the utter lack of data concerning the females. This is 

 especially unfortunate as pertaining to consorts, since it deprives us of data as 

 to the type of matings. Naturally, the suggestion was to secure the records of 

 coeducational schools, but the fact that coeducation in this countiy is of very 

 recent growth, as well as other considerations, discouraged the attempt to 

 obtain these records. The result of the Harvard studies, however, stimu- 

 lated the desire to see what might be available. 



"Inquiries were addressed to some 280 coeducational institutions all over 

 the United States, including about 200 colleges, as to the extent and character 



