452 
Association of Finger-Prints 
skeleton hands were available. The writers of the latter paper state that in con- 
sequence of the comparatively small number of bones measured they look upon 
that study " as one of suggestion rather than of definite statistical proof," and it 
is possible that with more adequate data their results might have been somewhat 
modified and exceptions less numerous. 
There appears to be no direct connection between finger bones and the patterns 
of finger-prints, but it is distinctly interesting to find that some of the most 
striking relations discovered amongst the former also exist in the latter. In 
particular, my conclusion (a) agrees with (vii), (c) with (iii 6) and with (viii), and 
(e) with (ii) but not with (v). 
11. Concluding Remarks. The most important conclusions reached in this 
paper have been summarized on p. 432, and in the Remarks on Tables 8 and 9, 
pp. 443 — 445, and on Tables 14 — 16, p. 449 ; it scarcely seems necessary to re- 
capitulate them, but a comparison will show an almost perfect agreement although 
the sets of results have been obtained by entirely different methods. 
The essential results of the present paper are that finger-prints are not scat- 
tered at random over the fingers; certain types are more or less peculiar to certain 
fingers, and further the appearance of one type is correlated with the appearance 
of a second. In this respect certain fingers are more closely related to each other 
than to any third finger, and the distribution of this relationship is in general 
similar to what is known of the like distribution of the correlations of the bones of 
the same fingers. 
It has been already stated that the material used is taken entirely from adult 
males of the lower type of the artisan and labouring classes; it would be of 
interest to compare the results obtained with those found from the finger-prints 
of females of the same grade of society, and also when the material is drawn 
from the professional classes. 
Tables I to XX, and XL VI to LV, are of a type which I have not previously 
met with ; novel methods have accordingly been employed in calculating coefficients 
of contingency and correlation ratios from those Tables. The general investigation 
of Tables of this type offers an interesting problem, demanding further study. 
I am deeply grateful to Professor Pearson for placing at my disposal the 
necessary material together with a number of books and memoirs bearing on 
the subject, and for much valuable assistance given during the course of the 
investigation. 
It can scarcely be expected, with such a mass of numerical calculation involved, 
that the work should be entirely free from inaccuracies, but I trust that no serious 
errors have escaped detection. The laborious arithmetic has been much lightened 
by the use of a calculator, for the loan of which my thanks are due to the Govern- 
ment Grant Committee of the Royal Society of London. 
The Tables on which the preceding calculations are based are given in the 
Appendix, pp. 453 — 478. 
