54 
Inheritance of the Duration of Life 
in which one variable was the niiinber of years' interval between the births of the 
elder and younger, and the second variable the difference in duration of life of 
elder and younger. Some remarks must be made on our method of procedure 
here. The greatest number of years' interval between births was found to be 29. 
The greatest difference of duration of life 98 years, and this might be either plus 
or minus ; thus if the year was used as unit the correlation table would contain 
5684 compartments, and would become very unmanageable. When the table was 
split into three — adults and adults, minors and minors, adults and minors, the last 
was still very unwieldy. It was thus necessary to use as unit a five year differ- 
ence in the duration of life. Using one year difference as unit in the birth 
interval we found for adults and adults : 
Brothers 
Sisters 
Mean excess in life of elder ... 
Standard deviation of excess... 
Mean interval between births 
Standard deviation of interval 
Correlation 
Number of pairs 
4-289 yrs. 
22-0053 yrs. 
6-462 yrs. 
4-3530 yrs. 
•1062 + -0206 
1051 
4-542 yrs. 
22-1325 yrs. 
6-7503 yrs. 
4-6856 yrs. 
•1201 + -0246 
733 
The data are not quite the same as for our pairs of adult brothers and sisters 
given above, but they show much the same advantage, i.e. four years to the elder. 
They further demonstrate that longevity is correlated with position in the family. 
This fact is suggestive for the source of other variations in the characters of an 
array of brethren. It may be that variability within the aiTay is not purely 
random but correlated, like variability in longevity, with the birth order. Our 
numbers show that on the whole the earlier born members of a family are the 
stronger, or at any rate fitted to survive the longer. 
The tables of adults and minors proving very unwieldy we determined to draw 
up one table only for all brothers and all sisters whatever their age at death, and 
for this purpose Miss Beeton returned to the Friends' material and extracted the 
dates of birth and death of 1606 pairs of sisters and 224>7 pairs of brothers. These 
were first classified according to 1, 2, 3 ... years' interval between their births, and 
then each of these classes subdivided into gnjups in which the character, life of 
elder less the life of younger, was nearest to 1, 2, 3 ... 90 or — 1, — 2, — 3 ... —90 
years. Cases in which the difference in duration of life was sensibly zero — very few 
numerically — were equally divided between the + 1 and — 1 groups. Finally each 
batch of five years was grouped under its mid-year, and thus Tables I. and II. 
were obtained. In the column A of these tables is given the number of cases in 
which for each year of interval between births the younger lives longer; in column 
B the number of cases in which the elder lives longer. It is only in the case of an 
interval of one or two years between births that A shows a majority over B. An 
examination of these colurnns demonstrates at once the great advantage of priority 
