90 
A Study of Tr Dpanosome Strains 
does not of coui'se assert dogmatically that any two strains are not of identical 
race. What be does assert is that no argument for the sameness of the strains can 
be based on the statistics provided ; for these actually show wide divergence, and 
he asks if the strains are a priori assumed to be "same," for a full a posteriori 
examination of the sources of the divergence. 
The scope of the present paper is not the complete investigation of all the data 
of the Royal Society Commission, nor an endeavour to obtain from the published 
data the full conclusions which may be legitimately drawn from them. Its purpose 
is to illustrate the statistical methods which ought to be applied to such material 
and to indicate the essential necessity of control experiments on strains known to 
be the same or accepted as different. A point should be noted here, namely, that 
I have only found two cases where the strains on the basis of the statistical 
evidence are said to be different. The first is in the case of Trypanosoma evansi 
and Trypanosoma hrucei. Sir David Bruce* gives (1911) the frequency distri- 
bution of lengths of 820 individuals of T. evansi and compares it by means of a 
graph of percentages with T. hrucei. The percentages of the latter appear to be 
deduced from the lengths for two series of 160 trypanosomes and 200 trypanosomes 
cultivated in a variety of animals (Uganda, 1909, and Zululand, 1894) and pub- 
lished in the preceding yearf, but no reference is given in the paper to the 
original of the percentages in the graph, nor is any demonstration given in the 
paj^er of 1910 of the statistical sameness of the Uganda and Zululand strains — 
there is merely said to be " marked resemblance j" where the trained statistician 
finds marked divergence §. Stephens and Fantham|| use the curve of 1911 to 
assert that there is a "general resemblance between the curves representing 
the measurements of these trypanosomes {T. gamhiense, T. rJiodesiense, T. hrucei)" 
and consider that this " general resemblance " shows that " the method is a 
trustworthy one." It is not clear what " the method " referred to really sig- 
nifies. The statistical comparison of means and maximum and minimum 
lengths without statement of probable errors, and the mere graphical exami- 
nation of frequency curves are wholly inadequate to determine sameness or 
* li. S. Proc. Vol. 84, B, p. 18(5, 1911. 
t R. S. Proc. Vol. 83, B, pp. 5 and 11, 1910. 
X R. S. Proc. Vol. 83, B, p. 12. 
§ The two distributions are as follows : 
15 
17 
18 
19 
£0 
SI 
32 
23 
24 
25 
26 
27 
28 
20 
SO 
31 
32 
S3 
34 
35 
Totals 
Uganda, 1909 
1 
4 
6 
10 
26 
u 
14 
12 
9 
12 
6 
Q 
12 
10 
7 
1 
2 
3 
3 
160 
Zuhilaiid, 1894 
4 
3 
11 
11 
20 
.32 
17 
4 
4 
3 
5 
3 
7 
7 
10 
13 
13 
8 
10 
8 
3 
1 
3 
200 
These give x^ = 101'18, leading to P<-000,001, or not once in a million trials would two so divergent 
distributions be obtained by sampling the same population. 
II li. S. Proc. Vol. 85, B, p. 233, 1912. 
