M. A. Lewenz and K. Pearson 
383 
attention to the following table, which is a reconstruction* of his Table III for 
the eight skulls out of his ten, for which we could find the requisite data : 
TABLE VII. 
Eace 
Capacity 
A 
G. F. 
A 
Mountain Chinese . 
1473 
1423 
- 50 
1407 
- 66 
$ Cossack 
1530 
1406 
-124 
1371 
-159 
? Hindu 
1133 
1050 
- 83 
1115 
- 18 
(j Fatuhivan 
1720 
1700 
- 20 
1538 
-182 
$ Bodo 
1644 
1443 
-201 
1394 
-250 
(j Australian 
1313 
1243 
- 70 
1292 
- 21 
5 Papuan 
1313 
1228 
- 85 
1263 
- 50 
(j Dutchman 
1644 
1613 
- 31 
1543 
-101 
Mean Error 
- 83 
-106 
Now B3 is certainly better here than G. F. and the answer is perfectly simple 
if the reader will turn to the diagram on p. 375. For small skulls with more than 
the average capacity our result must be too small, but better than Dr Beddoe's ; 
for large skulls with moie than the average capacity Dr Beddoe's will give a 
better result than the line of mean values. In the above table Dr Beddoe has 
only taken three small skulls with more than the average capacity and we are 
naturally in all those cases better than he is. In the other cases of big skulls 
he is better than we are. and roughly better as the skulls are bigger. If he had 
only had a majority of small skulls v^^ith big capacities the tables would have 
been turned upon him. The fact is that nothing can be determined at all by 
Dr Beddoe's Tables II and III for ' thick ' and ' thin ' crania ; by judicious selection 
of small thin skulls and big thick skulls we can make give execrable results 
as compared to G.F. or P and L, the selection being made before the predictions 
have been calculated. It all depends upon looking at the diagram on p. 375 
and choosing skulls which fall on the other side of the Beddoe line to the 
regression line. What is quite clear from in Table VII is that Dr Beddoe 
cannot improve the results given in Table VI without making his position in 
Table VII much worse ; i.e. he must sacrifice his claim to be better than us 
on selected individual crania. For whereas B^ gives far too big values on 
Table VI, it gives far too small values on Table VII. To really test the corre- 
lation formulae against the guesswork formula of Dr Beddoe, the individual 
skulls must be sampled at random and not selected. To do this we started at 
p. 20 of Barnard Davis' Thesaurus and extracted the first male undeformed 
skull with the necessary data on this page, or if there happened to be none on 
this page, the nearest male skull on the preceding page or pages. We tlien passed 
* Dr Beddoe's Table exhibits a number of misprints or arithmetical slips, and is invalid as far as 
P and L is concerned, because he has used male formula for female crania, and an erroneous value for 
the auricular height. 
