H. W. Acton and W. F. Harvey 
285 
The mean is now increased to 5,620,000 erythrocytes per c.mm. and the increase 
is chiefly due to the removal of a rather large number of anaemic persons contained 
in our original total. The probable error is now only + 43,000 and the standard 
deviation only 556,000, that is to say the exclusion of the above individuals causes 
a considerable decrease in the range and variability of our distribution. The above 
figures so contracted probably represent fairly accurately the mean and variability 
for an apparently healthy adult Indian male population living in the plains of 
India. The effect of altitude on the number of erythrocytes is well shown in 
TABLE III. 
Showing the Frequency Distribution of the Total First Counts 
with, the omission of Persons suffering from Anaemia, or 
Enlarged Spleens, and of Persons from High Altitudes. 
Number of 
Erythrocytes 
in 100, 000' s 
Frequency 
Number of 
Erythrocytes 
in 100,000's 
Frequency 
I 
61 
4 
48 
0 
62 
1 
44 
0 
68 
3 
¥ 
0 
64 
1 
. 1 
65 
0 
47 
0 
06 
0 
48 
1 
67 
2 
49 
2 
68 
0 
50 
5 
69 
2 
51 
6 
70 
0 
52 
7 
71 
0 
53 
4 
72 
0 
54 
11 
73 
1 
55 
3 
50 
2 
Total 
75 
57 
6 
58 
59 
00 
5 
2 
5 
Mean 
56 ± -43 
S. D. 
5-56 
Table IV and Fig. 5 (a) and (b). These exhibit the frequency distribution of 100 
first counts and the corresponding second counts. It was not necessary in this 
investigation to consider the exclusion of abnormal individuals (anaemics, etc). 
The second count was made at an interval of 18 days after the first, that is to say 
after an 18 days' residence at Kasauli. The counts were limited to cases free from 
any development of malarial fever during their course of anti-rabic treatment. 
The means of the first and second counts are respectively 5,328,000 + 69,600 and 
6,492,000 ± 66,500. The difference between the means is 1,164,000, and the 
probable error of the difference = ± 65,000, from which we may conclude that 
these two means are significantly different from one another, and that the second 
count represents an increase on the first. We might have applied our tests to the 
distributions themselves instead of to their means. 
