Etiikl M. Elderton 
63 
Dr Waite correlated* the number of" each type of finger-print on the whole 
hand with every other type and from his diagram it is clear that arches are 
positively correlated with small loops and negatively with every other type so 
that we must place arch next to loop. Large loops arc positively associated with 
whorls and composites and negatively with small loops and arches. Composites 
and whorls have a positive association. Composites seem to be more nearly 
associated with middle sized loops than with large ones. In fact what seems to 
interfere with obtaining a true natural order is the arbitrary divisiijn at 12 ridges 
of large and small loops; thi; sign in some of these tables is difficult t(j ascertain, 
a difficulty that never occurs if we deal with loops with no sub-division into small 
and large, or with arches or whorls. As I said before, arch, loop, "not plain" loop, 
composite, whorl is the order I favour, but we should want larger numbers of 
finger-prints than are available now and even then " not plain " loops would be too 
small a group comparatively for satisfixctory statistical treatment. Possibly in the 
end we shall have to combine " not plain" loops and composites, they certainly come 
very close together in any natural order. In the meantime there seem to be two 
possible practical orders (1) Arch, Small Loop, Large Loop, Composite, Whorl, 
and (2) Arch, Small Loop, Composite, Large Loop, Whorl. I worked out the 
corrective factors (class index correlations) for both orders and was pleased to 
discover that they were almost identical, and that the value of the factor in this 
data certainly depended on the number in each group rather than on the actual 
order. In six cases which I tried I obtained the following results : 
Arch, S. L., L. L., Arch, S. L., C. , 
C, W. L. L., W. 
Son's Right Foi'efinger '904 ■9.34' 
Father's Left Forefinger -936 -936 
Mother's Left Forefinger -941 -941 
Daughter's Left Forefinger -934 -934 
Daughter's Right Forefinger "944 '942 
Mother's Right Forefinger "943 -943 
Thus in only one case is there any difference in the third decimal place so that 
if we are prepared to accept type of finger-print as a continuous character we 
can apply the corrective factor for grouping even though we have not established 
an entirely satisfactory natural order. I have chosen the first of the two orders 
to use in every case, arch, small loop, large loop, composite, whorl. In the 
original tables I kept central pocket loops in a class by themselves, and thi'oughout 
I worked all contingency in two ways, (1) grouping C. P. L. with whorls, and 
(2) grouping C. P. L with composites, but the results are not significantly different 
* Dr Waite in his paper ou the association of finger-prints in tlie same individual has endeavoured 
to form a scale of relationship on the basis of association as measured by contingency in the prints of 
the several fingers. I think he is on the right path to deduce such a scale, but I believe that it cannot be 
done by merely comparing the numerical magnitude of the associations. It is true that contingency has 
no sign, but if we are to form a scale of next relationship, we must consider sign, a positive and 
negative association of the same magnitude being not of the same significance in a scalar relationship. 
I have therefore felt compelled to give signs to Dr Waite's contingency coefficients based on other con- 
siderations than those of contingency, namely by actual examination of the contingency tables them- 
selves. This has compelled me to modify bis scale of relationship. 
