M. Greenwood and J. D. C. White 
385 
TABLE 111 bis. 
Fr'equency Distributions. 
Bacilli per cell. 
Observer 
Key 
Number 
0 
1 
2 
3 
4 
5 
6 
7 
8 
9 

12 
13 
Ik 
i5 

16 
T 
i 
ziy 
zo / 

ziy 
izy 
iV 

ZD 

io 
o 
£* 
TT SJ p 
ii o. u. 
ZlV 
oio 
no 
on 
Q 
o 
7 
o 
z 
TT n ,?T C! 
iyo 
oUO 
zzu 
Ql 
ol 
ol 
1 7 
io 
4 
o 
1 
J 
1 
IV S. C. 
243 
290 
220 
134 
54 
26 
15 
9 
6 
3 
— 
Ill 
188 
261 
226 
153 
92 
44 
19 
10 
2 
3 
1 
1 
V 
192 
248 
202 
165 
95 
47 
21 
19 
4 
5 
1 
1 
VII 
240 
251 
212 
150 
70 
38 
23 
12 
2 
2 
VI 
207 
263 
223 
153 
87 
35 
21 
6 
3 
2 
IV 
495 
571 
445 
255 
124 
55 
27 
13 
9 
4 
2 
Fleming 
T. C. 
152 
239 
262 
159 
96 
54 
22 
11 
4 
1 
10 Norm. 
111 
218 
210 
189 
144 
73 
35 
11 
11 
4 
4 
N. S. A. 
41 
126 
154 
164 
121 
62 
36 
35 
5 
2 
3 
1 
No. 2 
99 
227 
208 
134 
78 
34 
9 
7 
3 
1 
T. A. 
58 
107 
203 
207 
174 
138 
90 
49 
31 
19 
10 
4 
4 
1 
2 
1 
2 
N. S. B. 
19 
59 
98 
88 
65 
37 
17 
8 
5 
2 
1 
1 
Dr Strange ways and his colleagues adopted this plan. As we desired to study the 
effect on the distributions of raising the mean value, Dr Fleming, at our request, 
worked with emulsions thicker than those generally employed, so that the effect of 
the change could, to some extent, be gauged. 
We have used the customary method of moments and fitted the curves 
indicated by the value of for each distribution. 
Since the actual distributions did not point to high contact at both ends of the 
range we have not felt justified in applying Sheppard's corrections, and have in all 
cases worked with raw moments. 
Table IV. includes the principal constants for each distribution, together with 
the values of P*. Table V. contains the equation of each curve, and Graphs 1 — 15 
give the fitted frequency-curves for each worker's material. 
We shall first consider Dr Strangeways' data. 
Before referring to details, however, we desire to point out the completeness 
with which our results justify the use of Pearson's type curves for biological 
material of this class. We are now speaking to the practical man with no special 
knowledge of statistical theory. For him, the real justification of a statistical 
method must be its capacity adequately to describe samples of material. 
This point may be emphasized by reference to a criticism of the methods 
employed by biometricians. 
* The non-statistical may regard P as a measure of agreement between theory and experiment, the 
value lying between 1, which marks absolute harmony, and 0. See Pearson, Philosophical Magazine, 1900, 
Vol. 58, p. 175. Elderton, Biometrika, 1902, Vol. i. p. 156. 
Biometrika vi 49 
