544 Mr. F. Galton. [May 28, 



The whorls include circles, ellipses, and spirals, both simple and 

 compound, whatever may be the direction or closeness of their twist. 

 These are so apt to be confounded together unless the impression is 

 from a rolled finger and is afterwards scrutinised and outlined (as 

 explained in my previous memoir) that it seems best for the present 

 purpose to group them all, with few exceptions, under the one 

 symbol h. The exceptions are these. When two streams of ridges 

 proceed from opposite sides of the finger and interlock, the symbol t 

 is used, regardless of all other details. Again, when the whorl is 

 crozier shaped, as in j and k, it is necessarily enclosed in a loop, but 

 the loop is here ignored. If the crozier approaches very nearly and 

 mistakably to either of the plain eyes t, , it is dotted for a warning, 

 as in I and m. 



The loops in their simplest and common forms are shown by n and 

 o. Frequently they have an internal offset which may be variously 

 feathered or bent, short of being a whorl ; all such cases are expressed 

 by p, q. They have sometimes a conspicuous eye due to an internal 

 curvature of the ridges upon themselves, or even to an eye in the 

 central ridge ; these are all expressed by t or it, in which the sur- 

 rounding loop is left out in order to avoid multiplicity of lines. When 

 the eye approaches nearly to a crozier as in Z, m, the dotted symbols 

 r, w are used. 



In making a large and complete index, the symbols would, of course, 

 be cast as movable types, and be printed with the letterpress. It 

 will be seen from fig. 2 that there is space for 20 entries in one 8vo 

 page. 



I do not expect from my own reiterated experiences that there 

 would be much trouble due to transitional cases, after a standard 

 collection of doubtful forms had been establised so as to ensure that 

 different persons should abide by a common rule. I find much 

 uniformity in my own judgment. 



I give an index of 100 cases ; they are the first that occurred in my 

 catalogue of impressions, which are pasted in two rows on each page, 

 and are consequently numbered 1, 1' ; 2, 2', in order; but there are a 

 few blanks, so the numbers in the index happen to run from 1 to 56', 

 with some omissions, and not from 1 to 50'. 



These cases afford data for roughly measuring the increase in 

 power of discrimination obtained by basing indexes on the patterns 

 of 1, 2, 3, 6, and 10 digits respectively. It appears from Table III 

 that when all 10 digits are used, the number of different patterns 

 observed in the 100 cases was 83 ; therefore the average number of 

 references required to pick out a single well-defined case from among 

 these 100 would be equal to 100 divided by 83, that is, to about 1^. 



It will also be seen from Table III that, owing to the large effect of 

 correlation, an index based on all the ten digits is not much superior 



