166 Bulletin American Museum of Natural History. [Vol. XXXIV, 
is known of the habits of the latter, but the chickarees are well known to be 
less arboreal than the larger and longer tailed squirrels. The ratio of tail 
length to the total length varies in different groups from about 40 to 52%, 
and is very constant in groups of squirrels that are nearly related. It is 
not, however, closely correlated with the size of the animal, although all 
large squirrels are long-tailed, but so also are the Guiana pygmy squir- 
rels and the guerlinguets of eastern Brazil and the Guianas. It therefore 
seems proper to give weight to the relative length of the tail as a taxonomic 
character. 
Value of averages and ratios in diagnosis. 
In view of what has just been said respecting the wide range of indi- 
vidual variation, it is evident that ratios between different parts, as the 
length of the tail to the total length, and of parts of the skull to each other 
or to the skull as a whole, are of slight value if based on single specimens, 
they being necessarily as variable as the measurements on which they are 
based. On the other hand, ratios based on averages may have a high value 
as a convenient formula in diagnosis, and as bases of comparison of allied 
forms. But the number of specimens on which such ratios are based should 
be large enough to approach closely the normal for the form in question — 
at least 10, and preferably more, selected to include only suitable specimens, 
a. €., excluding adolescent and senile individuals. In the case of the skull, 
ossification continues after the period of sexual maturity is reached, and in 
old age there is often an excessive deposit of bony matter at peripheral points 
much beyond normal adult conditions. It is not usual, however, to find at 
hand sufficient material of a given form to furnish a satisfactory basis for 
ratios, and one must make discretionary use of whatever may be at hand. 
Three or four specimens afford of course more trustworthy results than a 
single specimen, however normal it may seem to be. Ratios based on 
single specimens, or on specimens that have not reached middle-life, are not 
to be trusted, as they may be very misleading. Nor can trustworthy ratios 
be obtained from photographs, as ordinarily taken. In the case of squirrel 
skulls, for instance, the depressed parts, as the rostrum and occipital region, 
become so much foreshortened in the photograph that ratios based on the 
flat surface of the photograph will sometimes vary 10 to 20% from the same 
ratios based on the skull from which the photograph was taken! This was 
to me a surprising result, learned only after considerable work had been 
expended in computing ratios from the flat surface of photographs. 
Another point to be mentioned in connection with averages and ratios, 
brought out strongly in computing skull ratios, is the difficulty of taking 
correct measurements of convex surfaces, or where the boundaries of parts 
