290 Mlt. G. UDNY YULE ON THE ASSOCIATION 



5 (owing presumably to the waning protection of the vaccination made in infancy) 

 iu the older age group. 



The constancy of the association in towns with widely different attack rates is 

 a point worthy of notice. Sheffield, Warrington, and Leicester exhibit practically 

 identical associations, although the attack rates vary from 7 '9 to 2 '5 and 67 '6 to 35 '3. 

 Not having the original figures for these cases I cannot state the probable errors. 



45. (2.) From Mr. GALTON'S " Natural Inheritance." 



Assortative Mating according to Temper. On p. 231 of "Natural Inheritance" 

 Mr. G ALTON gives the data, based on 1 1 1 marriages : 



Good-tempered husbands with bad-tempered wives . . 24 per cent. 



Bad-tempered good-tempered . . 31 

 Good-tempered ,, ,> >. 



Bad-tempered bad-tempered . . 23 



Here 



22 x 23 - 24 x 31 



22 x 23 + 24 x 31 



_ 



= 



for the association between temper in husband and wife, i.e., on the whole bad- 

 tempered husbands have good-tempered wives, and vice versd. But the probable 

 error of the association = 



^..c 

 6745 



so that only very slight stress can be laid on the sign of the association. 



The advantage of having the whole question of the association thus compressed 

 into one figure, with a definite probable error, is here very clearly marked. Com- 

 paring the actual figures with the distribution in the case of no association, 

 Mr. G ALTON concluded "that the figures taken from observation run as closely 

 with those derived through calculation as could be expected from the small number 

 of observations." * 



46. Association of Temper in Fraternities. On p. 235 of " Natural Inheritance " 

 Mr. GALTON gives, in the same investigation, data for the association of temper in a 

 fraternity (group of brothers and sisters). Thus in sixty-six fraternities of three 

 members there were eleven cases reported in which all were good-tempered ; fifteen 

 in which one was good and two bad ; twenty-one in which one was bad and two good ; 

 and eight in which all were bad. From data of this kind I formed all the possible 

 pairs (permutations). Thus in the above case, using G for good, y for bad, I find the 

 number of pairs as below : 



h This conclusion is in part affected by a slight error (owing to accumulation) in the " non-associated " 

 figures given. The (constant) difference between the observed and " non-associated " frequencies is 2 per 

 cent, to the nearest unit. 



