Donaldson, Groiuth of Central Nervous System. 351 



When the foregoing formula is apphed, the theoretical curve 

 gives a very good graduation of the brain and cord weights for the 

 larger values of x, but fails to adequately represent them for the 

 smaller values of ;c. 



The values obtained by the formula are too high for the brain 

 weight, and too low for the spinal cord weight. In order to meet 

 this difficulty, the constant /? empirically determined, has been 

 introduced, and the resulting formula becomes 



y = C\og{x + 13) + J 



in which /? is the new constant. 



This is the general formula which we have employed for the 

 present work, and it has been found very satisfactory, as will be 

 seen from the tables and charts. 



Arranging the rats examined in groups differing by ten grams 

 in body weight, and calculating the mean values of the observed 

 weights of the brain for the mid value of each of these groups, we 

 obtain the curve which is given in chart 3. The mean values 

 (M = the broken hne) obtained by so treating the observations, 

 are given in column C of table i. The table and chart show that 

 the curve based on the means, fits closely with the theoretical 

 logarithmic curve {C = continuous line). 



The coefficient of correlation between brain weight and body 

 weight in the case of the 680 records, was determined accordingly 

 to the formula [2] 



r = I ^ - v' v" 



(Davenport '04) and is high, being .7639 ±.0108. 



For comparison with this result, it may be noted that Pearl 

 ('05) in the case of the total series of Bavarian brains, weighed 

 by BiscHOFF ('80), found the coefficient of correlation between 

 brain weight and body weight to be as follows: 



Male 0.1671 ±0.0343 



Female 0.2260 ±0.0412 



In the case of Worcester school children 6 to 17 years of age, 

 in which the measurements are more accurate than they could 

 possibly be in the case of Bischoff's series. Boas ('05) found for 



