120 
A Studij of Trifpanosoine Strains 
accoidiugly be prepared for some such change as this in the shifting of the mean 
when the host is varied. 
We have next to inquire what type of curve accui'ately describes the strains 
which we are fairly certain are homogeneous. 
If the reader will turn back to p. 110 he will note at once a marked difference 
between the distributions for T. caprae, T. pecorimi and T. simiae when compared 
with those entitled Mzimba strain, human strain, wild-game strain, T. brucei, 
T. rhodesiense, T. gambiense and the wild G. morsitans strain. The coefficients of 
variation of the former group are all under 9"5 (mean = 9"0()), the coefficients of 
variation of the latter group are all over 13"5 (mean = 17"29). We recognise 
therefore a totally different order of variability. Even in absolute variation as 
measured by the standard deviations we find the first group with its mean 
S. D. = 1'68 and the second with its mean 3"96. An examination of the graphs 
scattered through the trypanosorae papers to which we have referred will, we think, 
convince the statistician that we have to deal with heterogeneous and not skew 
homogeneous material*. It becomes of course important to ascertain whether in 
the pure strains a Gaussian curve will suffice to describe the frequency closely 
enough for statistical purposes, for, if it does, the analysis into at any rate two 
Gaussian components of the heterogeneous strains becomes relatively direct, if 
laborious. I will consider the T. pecorum, T. simiae, and T. caprae strains 
from this standpoint. 
(a) T. pecorum (see p. 110). 
Mean = 13"992 microns. S.D. = 12816 microns. 
Microns 
Observed 
Calculated 
Values 
Values 
,9 aad under 
2 
0-46 
10 
6 
5-98 
11 
42 
45-41 
12 
193 
192-52 
IS 
452 
456-70 
n 
618 
607-12 
15 
453 
452-49 
16 
178 
188-98 
17 
51 
44-16 
18 and over 
5 
G-20 
x2 = 7-630 
Hence in 57 out of 100 trials from material following the Gaussian distribution 
a more divergent sample than that observed would actually be obtained. We can 
therefore conclude that a simple Gaussian frequency adequately describes the 
distribution in size of T. pecorum. This is illustrated in Diagram II. 
* Note especially the bimodal graphs in li. S. Froc. Vol. 83, B, pp. 5 and 11, for both the Uganda 
and Zululand strains of T. hrucei, in YoY. 86, B, pp. 291 — '293, for human strains, in Vol. 86, B, 
pp. 395, 397 for wild-game strains and pp. 409, 411, 417 and 419 for G. morsitans strains. 
