414 HALBERT L. DUNN 
values. 5) Determinations of the relative weights of the various 
parts of the central nervous system as compared to that of the 
encephalon. These several methods of analysis will now be con- 
sidered in more detail. 
1. Construction of field graphs and establishment of preluminary 
curves by inspection. The data secured for each measurement of 
the central nervous system were first plotted on a field graph in 
which the length of the body (C H) was used for the abscissa and 
the measurement in question for the ordinate. The material was 
then divided into classes on the basis of 5-em. intervals of crown- 
heel length and the arithmetic mean or average and the median 
were determined for each of these classes. ‘These means were then 
indicated in the graphs by special symbols, their exact position 
being determined by weighting for the distribution of the cases 
according to crown-heel length in the 5-cm. intervals. A pre- 
liminary curve was then drawn by inspection on the basis of the 
combined evidence furnished by the position of the arithmetic 
mean, the median, and the general distribution of the cases. This 
method has distinct advantages over the practice of establishing 
curves of inspection on the basis of the arithmetic mean alone, 
particularly in those instances where a curve is rising rapidly in 
a single interval or where the cases are not regularly distributed. 
The use of both the arithmetic mean and the median enables 
one to correct for chance deviations in the mean alone with con- 
siderable confidence, particularly when the cases are all spread 
before one on the field graph. 
2. Reduction of preliminary curves of inspection to numerical 
expression by means of empirical formulae. The curve of each 
value as determined by inspection was reduced to numerical ex- 
pression in the form of an empirical formula on the crown-heel 
or total length of the body expressed in em. This procedure in 
no way increases the reliability of the curve as determined by 
inspection nor does it increase the accuracy of the points upon 
which it is based, but it has several important advantages, for it 
permits an accurate and abbreviated expression of the curve, 
facilitates exact interpolation and the conversion of values into 
different scales and forms of expression, and, in some cases, aids 
in classifying the curves as to form. 
