C. D. Fawcett 
423 
Now it will be noticed that the numbers for the later period are quite 
insignificant, so small indeed as to be largely affected by age at death of the 
individuals. We can hardly get a real random age distribution in 5 to 9 skulls ! 
Turning first to the length, the means and standard deviations of I, II and 
III were calculated and hence the probable error of the differences of the means 
was found. The results were as follows : 
J'Mi-Mni= -2413 ± -9882 
c/" Mn - Mui = - 1-1933 ± 2-0817 
J'Mj^-Mu = 1-4.346 + 2-2534 
Thus in no case was there a sensible difference in length, within the limits of 
random sampling, of the three groups. 
For the females 
? ill - if„i = -8943 ± -5570 
? if„ - ilfni = - 2-0840 ± 1 -2746 
$ ilf I - i/n = 2-9783 + 1-3589 
The differences here are all larger than their probable errors, but only in one 
case slightly more than twice the probable error. It is impossible again to assert 
that there is a real class difference. 
31152 + -7221 
3-0938 ± 1-3378 
6-2090 + 1-4703 
Now these differences appear significant ; in two cases they are four times and 
in the third case twice their probable errors. But on closer examination of the 
individual crania we doubted whether the apparent sensibility of the differences 
in the first and third lines is not largely due to the existence of two or three 
rather juvenile skulls in the first series. 
For the females 
? ifi - ifni = - 1-7674 + -6367 
$ lf„ - if III = -2278 ± -8208 
? if I -if„ - - 1-9952 + -9831 
The last two results are hardly to be classed as sensible differences, the first is 
possibly such. 
If the reader will now turn to the frequency diagrams for the breadth of skull 
in $ and cT (p. 445) he would expect to find in the former modes at about 129-5 
and 131-5 and in the latter modes at 132 and 138, if there be a real difference in 
the two series of early and late skulls. The conspicuous female maxima are at 
Turning now to the breadth we have 
