382 Measurement of Internal Capacity of Skull 
statement must be made with regard to the size of the mastoid processes. To 
assert that the mean vahie of the difference between the intermastoid arc and the 
bi-auricular arc is 2'3 inches is in our opinion wild guesswork, and even if it were 
true it neglects entirely the immense variability of this difference. We cannot 
expect any formula to give good results under such treatment, but as Dr Beddoe 
has chosen such a test we consent to accept it, especially as it enables us to indicate 
to the craniological reader the sort of statistic©- metrical treatment which seems 
to have passed hitherto as science. In the following table we give first the 
supposed measured value of the capacity, i.e. Barnard Davis' values less 1/20 and 
reduced to cm.^; then we give the predicted values and differences of B^, P and L, 
and G. F., and finally the mean errors of the three formulae. 
TABLE VI. 
Comparison of Cranial Formulae on Barnard Davis Measurement 
of Mean Racial Capacity. 
Race 
Capacity 
A 
PandL 
A 
G. F. 
A 
Romans of Britain... 
1385 
1490 
+ 105 
1431 
+ 46 
1463 
+ 78 
Australians 
1277 
1347 
+ 70 
13.30 
+ 53 
1366 
+ 89 
New Hebridians ... 
1362 
1366 
+ 4 
1309 
-53 
1376 
+ 14 
Kanakas 
1459 
1544 
+ 85 
1462 
+ 3 
1499 
+ 40 
Marquesans 
1491 
1507 
+ 16 
1454 
-37 
1473 
-18 
Netherlanders 
1502 
1515 
+ 13 
1441 
-61 
1477 
-25 
Finns 
1497 
1559 
+ 62 
1454 
-43 
1499 
+ 2 
Russians 
1529 
1570 
+ 41 i 
1492 
-37 
1522 
- 7 
Thais 
1483 
1539 
+ 56 
1430 
-53 
1487 
+ 4 
Mean ... 
1443 
1493 
+ 50 
1416 
-27 
1462 
+ 19 
Mean Error ... 
50 
43 
31 
In deducing the results P and L we have, owing to the mixture of races 
involved, used Dr Lee's most appropriate formula (10) his, that in the footnote 
on p. 247 of her memoir*, based on the best results for the means of 10 races. 
It will be seen at once that P and L and 0. F. present much better results than 
^3 which is far worse than either in determining the mean value of the whole 
series, and is worse than our formula by at least 50 per cent, in the mean error. 
If Dr Beddoe considers that his formula has given all plus errors and therefore 
can be approximately bettered by reduction in his multiplier, we must call his 
* Phil. Trans. Vol. 196, A, p. 247. We have used this formula throughout for race means because 
it is based on racial means and not on individual sliulls. Experience has shown that in the long run it 
works better than (10). The latter is much better for this series, giving a mean value 1446, only three 
in excess, and a mean error of only 35 ! We have not, however, taken advantage of this superiority to 
use it. 
