450 
Association of Finger-Prints 
(c) It follows from (6) that if in any of these Tables we start from a first order 
association and pass in any direction through those of other orders we find a 
continuous and rapid fall ; that is, a finger is always more closely related to a 
consecutive finger than to one more remote (but see (a)); and the greater the 
difference in rank between two fingers, whether on the same or on different hands, 
the less close is the association between them. i 
yd) The association between any pair of fingers in one hand is, in general, 
closer than either of the corresponding associations between a finger of the right 
and one of the left hand. There is one exception to this rule in associations of the 
second order, one in the third and one in the fourth. 
(e) The associations of the left hand are in every case closer than the corre- 
sponding associations of the right. 
(/) The associations of either thumb with any finger all fall below those of 
the fourth order of (6), and the range of the sixteen coefficients is only from ■424 
to -508. As it is difficult to base any conclusions on these figures as to the 
relations between the thumb and the various fingers, I have carefully checked 
them by reworking the whole of the calculations involved, but have in every case 
arrived at the same result. I have also found the probable error* for the largest 
and for one of the smallest coefficients of the set. As the contingency coefficients 
are all of the same order of magnitude and the number of individuals the same in 
all cases, the probable errors of all will be of about the same magnitude and it is 
unnecessary to calculate more. The probable errors in the two cases being of the 
order -Oil the differences in the contingency coefficients may be regarded as 
insignificant. Although in three cases out of the four the contingencies of the 
thumb with the middle, ring and little finger respectively are in ascending order 
of magnitude, the differences are so small in comparison with the probable errors 
that no conclusion can be drawn as to the relations between the thumb and the 
various fingers. We may notice, however, that the rule {d) holds good for the 
thumbs with but two exceptions. 
The contingency coefficients given in Tables 14 6, 15 6, and 16 6, are all smaller 
than the corresponding results of the other series, but a careful study will show 
that the remarks {a) to {g) almost invariably apply to these Tables also. 
Note. In some preliminary work on this paper I classified the types as 
follows: — (1) Arches and loops with 1 — 3 ridges, (2) Loops with 4 — 10 ridges, 
(3) Loops with 11—14 ridges, (4) Loops with 15 or more ridges, (5) Whorls, 
(6) Composites. With this classification the following contingency coefficients 
were found for corresponding fingers of the two hands: — Thumb '686, Fore- 
finger -642, Middle finger -686, Ring finger -730, Little finger -738. These results, 
which were not corrected for grouping, are seen to agree very closely with those 
* The method employed is that given in Biometrika, Vol. v. Parts i. and ii., "On the Probable 
Error of Mean-Square Contingency," by John Blakeman and Karl Pearson. 
