476 



EEPOKT —1 889. 



both to weight and to statare, the simplest plan is to make a separate dia- 

 gram for each inch or second inch of stature, which is quite near enough. 

 I have, however, contrived to make a single page serve for the whole 

 process by using a sliding strip of paper. I submit it for inspection, but 

 do not care to describe it. 



A strong reason for giving prominence to relative rank is that it 

 affords the only feasible measure for many qualities, so that differences 

 in absolute performance have to be inferi-ed from it, according to a prin- 

 ciple now familiar to most anthropologists, by using the well-known table 

 of the Probability Integral. A small table based on the latter, but of a 

 totally different form, that I have lately more than once published (Op. cit., 

 p. 205), is very convenient for this sort of work. The following is a 

 brief extract from it : — 



Grades of Rank from 0° to 100°, together with the Deviations'^ from the Mean Values 



at those Grades. 



' The unit by whioli the deviations are measured is half the difference between the performances of 

 the persons who respectively occupy the grades 25° and 75°. 



Some of the consequences of marking separately the relative rank 

 and the absolute performance are seen by the table below. Here the 

 relative rank is in each case supposed to count between the grades of 50° 

 and 100°. Then, if it alone is considered; a man who stands at the grade 

 of 99° in a class that ranges within the limits of 0° and 100°, will be seen 

 to get very nearly the full amount of 10 marks, whereas if absolute per- 

 formance is alone considered, he would get no more than 7 marks, the full 

 number of 10 being never actually reached, but only closely approached 

 at some such high grade as 99-99 . . .° The figures in the table would 

 have run very differently if the marks for relative rank had begun after 

 90° and not after 50°. Still more so, if the lower limit had been 99°, and 

 more still if it had been 99'9°. It seems to me most reasonable, on the 

 whole, that they should usually begin after 50°, as in the following 

 table : — 



The general conclusion to which these remarks lead is, that before 

 arranging scales of marks, the first step is to measure a lai'ge number of 

 persons who are of the same class as the expected candidates ; this has 



