J. Blakeman 
145 
TABLE XIX. 
Increase of Brain-weight in grs. and Diametral Product in cvi.^ above their 
■means for each unit increase iti the other observed Characters. 
Regression Coeflficient of Character with Probable Error. 
One mm. 
in L 
One mm. 
in B 
One mm. One mm. 
in if in U 
One Year 
in A 
One inch 
in S 
Mean of Character 
190-36 
149-34 
132-20 
555-79 
49-26 

67-16 
? 
183-20 
144-55 
129-21 
533-25 
45-90 
63-05 
Brain-weight 
Increase in -< 
grs. 
Mean 
1327-69 grs. 
11-0728 
± -92.54 
9-3936 
± 1 -0463 
12-5628 
± 1 -2074 
4-0895 
±-3204 
-2-1974 
± -5245 
9-8277 
±2-0338 
. ? 
1224-90 grs. 
9-7599 
± 1 -0446 
10-5593 
± -9607 
15-5037 
±1-1841 
3-7344 
±-3224 
-2-7245 
± -4721 
13-1934 
±2-3271 
Diametral 
Product 
Increase in 
cms. 
' 6 
3782-91 cm. 3 
15-1711 
± -6424 
-4-1097 
± 1 -4730 
29-6394 
±5-6361 
. ? 
3427-03 cm. 3 
13-5666 
± -6913 
-5-9478 
± 1 -3769 
40-5999 
±6-4775 
A man of his age would have lost 78-5 grs. of his original mean age braiu-weight, 
or have on this account only 1249 grs. instead of 1328 grs., the average at age 49'26. 
He has an excess of 1-64 mm. in head length, an excess of 3-66 mm. in head breadth, 
and a defect of 5-20 mm. in auricular height. Also an excess of 4'21 mm. in 
horizontal circumference. Thus on the first count his brain-weight would be 
18-15 grs. above the average, on the second 34-38 grs. above the average, on the 
third 65-33 below the average, and on the fourth 17 22 grs. above the average. 
But it would not be proper to club all these together and say that on account of 
his head measurements we should expect Bentham to be 4-42 grs. above the average 
when at the mean age. For all these characters are correlated, and we have seen 
that they, especially the auricular height, are influenced by age. Hence it is 
absolutely needful to use multiple regression formulae. These will be applied to 
Bentham's measurements later, but the above will suffice to show how a single 
character may have its influence appreciated. 
(b) We now pass to the multiple regression equations. These were found in 
the usual way, namely, by the calculation in this case of 25 constituent determinants 
and the minors of their first row of constituents. The work is straightforward, but 
laborious. For purposes of comparison the prediction formulae based on diametral 
Bjonietrika iv 19 
