122 
Miscellanea 
in the notation used by Prof. Pearson above, and from this formula o- was found to be "1459, 
where the term depending on the means contributes practically nothing. Actually the standard 
deviation calculated from the grouped sex-ratios was found to be •1546 ±'0085, so that again the 
difference is within the limits of the probable error. 
Turning now to the question of correlation, in the case of completed families of eight and 
upwards r was found to be "0101 + •0213, while in the case of the first four members only, r 
was found to be — •0064 + •0213. These coefficients are both zero within the limits of probable 
error so that no significant increase in correlation is obtained by raising the number of individuals 
on which the sex-ratio is based from 4 to an average of 11 •4. 
It should also be noted that Mr Cobb's corrective factor is positive and if the true correlation 
were at all comparable with its value found for other physical characters, which is of course 
always positive, the result of our calculation must give a positive correlation. The fact that it 
gives a small negative correlation is compatible with its being really zero but not with its 
being the reduced value of a true positive correlation found by using a positive reducing factor. 
