480 Assortatire Motinq in Man 



tested by examination of the Oxford records. For every married pair the foUowing 

 data were calciilated : (1) the excess of husband's agc over wifc'.s age at marriage, 

 "the husband's seuiority " ; this quaiitity, calloil A, was tabulated in one year 

 Units ; it is called positive whou the husband is older than the wife, negative when 

 the wife is the older; (2) the number of years during which the husband survives 

 the wife; this quantity, called A', is cle;irly positive if the wife dies tii-sl, negative 

 if the husband dies tirst. Caliing // the length of husband's life, W that of 

 the wife's life, we have as necessary relation betweon these four tiuantities 



//= A+ ir + A'. 



The niean value of A' was found to be — r986 years, showing that the wife in 

 the Oxford district dies on an average l'9.Sß years after the death of her husband*. 

 The environmental effect to be uieasurcd niust, if it exists, be measured by its 

 effect upon the coefficient of correlation bctween A' and the length of the wife's 

 (or husband's) life. Taking the correlation between the age of the wife at death 

 and the magnitude of A', the value found was 



,-^,„.= _OTr215 ±00139; 



the negative sign indicating that A' increases negatively with increa.sed age of 

 wife, or the older wives survive their husbands for a loiiger period than the 

 younger, as it is, of course, obviously likely that they should do. The chance 

 that a wife will survive her husband dopends upon mauy things too coinplex for 

 analysis ; but one factor, which the data euables us to measure, is the diÖerence 

 between her age and that of her husband at marriage, or the value of A ; if we 

 consider two women of the same age, one marrying a man a year older than herseif, 

 whilc the other marries a man twenty years her senior, clearly one will be much 

 more likely to survive her husband than the other. If, however, we consider the 

 array of husbands and wives, including all those pairs for which A has a constant 

 value, one factor, i.e. the expectation of tlie wife's snrvival, will be constant; 

 and if within each such arra}' the shock or tlie change in environment due to the 

 death of husband or wife produce any sensible efifect on the duration of life of the 

 survivor, this is something peculiar to each pair within the aiTay ; accordingl)', 

 if we takc all the husbands and wives in each such array, and form couples of mcu 

 and women by random selection from among them, the relation between the death 

 of one member of such a couple, and that of the other, will be independent of any 

 environmental influeuces resulting from actua! married life. If the Oxford records 

 be treated in this way, and the value of A' computed for every random couple, the 

 correlation between values of A' aud the length of life of the woraan placed in the 

 couple niay be determined ; the diffcrence between the value of r so obtaincd, and 

 that obtaincd from a study of actual husbands and wives, will be a measure of the 

 environmental effects referred to. The result of such a determination shows, 



* Tbe Standard dcviation of A' is lG-550 years. For Wenslcydale it is 17a8.3, or thcre is grcater 

 variability in tlie survival pcriud. 



