E. Schuster 
477 
With regard to the assumption that families which do not contain a deaf 
member are the same in point of size as those that do, I endeavoured to obtain 
evidence from the data at my disposal. It seems quite clear that, at any rate, a 
considerable proportion of those persons who are said to have acc^uired deafness 
must in reality have no sort of inborn tendency in this direction ; the families 
which contain them can therefore be taken as more nearly approaching a sample 
of families whose only common character is that they contain at least one married 
member, than those which contain a congenital deaf-mute. If then it is found 
that the mean size and the variation in size in both of these classes is approximately 
the same, it could be argued that this is to a certain extent evidence that the 
assumption is a reasonable one. The families were accordingly divided into these 
two classes, and the first two columns in Table XII. show the distribution of size 
for each class. Table XVI. gives a comparison of the mean size of family, and of 
TABLE XVI. 
Showing difference in mean size of families containing at least one congenital 
deaf-mute and of tJiose containing at least one member wJio Jias acquired 
deafness ; also the difference in the standard deviations. 
Mean 
Standard Deviation 
Congenital 
6-0821 
2-9862 
+ -0583 
+ -0412 
Acquired 
5-7704 
2-9070 
+ -0386 
+ -0273 
Difference 
-3] 17 
-0792 
+ -0699 
+ -0495 
the standard deviation. It will be seen that the mean size of families containincf 
at least one congenital deaf-mute is larger than those containing one of the 
" acquired " variety by an amount which is rather more than four times as great 
as its probable error. There is thus a certain amount of reason for believing that 
families containing a deaf-mute member are in reality somewhat larger than those 
that do not. If this had been allowed for in making the correlation table the 
number of pairs of hearing and hearing brothers and sisters would not have been 
so great, and there would have been a consequent reduction in the value of r. 
The difference in the value of the standard deviation is not twice as great as its 
probable error, and is therefore probably not significant. 
The value of r for fraternal correlation was found from Table XV, to be 
approximately "74, which is a rather higher value than is normal, though not very 
much higher than that for the amount of red in coats of greyhounds of the same 
litter, which was found to be -700 {Biometrika, Vol. ii. p. 390). 
Table XVII. explains itself The mean number of deaf members per family in 
each case is not insisted upon, owing to the probable gross inaccuracies of the 
classification into cases of congenital and acquired deafness. 
