100 
A Study of Tri/2)anosome Strains 
opinion. But on pp. 92 — 3 percentage frequency curves are drawn for the three 
strains, and the following remark is made : 
" A survey of the curves obtained by plotting out in percentages the various 
lengths of trypauosoraes encountered in each of the three strains is of interest. 
It will be observed that in the case of rats the curves of each of the strains corre- 
spond fairly closely." 
Now what do the authors mean by " fairly closely " ? In their conclusions 
they identify B and G and diffei'entiate A. Unfortunately they have not given 
their actual frequencies, and I have had to endeavour to reconstruct them from 
the percentage curves. There results for the rat-data : 
Microns. 
16 
17 
IS 
19 
SO 
SI 
:2S 
S3 
S4 
S5 
26 
27 
28 
29 
30 
31 
32 
33 
34 
35 
36 
Totals 
Berlin Strain A 
1 
1 
10 
9 
12 
17 
17 
22 
28 
48 
47 
57 
55 
42 
39 
37 
28 
13 
8 
6 
3 
500 
Frankfurt Sti'ain B ... 
1 
3 
5 
1 
4 
10 
20 
22* 
18* 
25 
24 
35* 
23 
15 
18 
15 
8 
3 
250 
East Prussian Sti'ain C 
1 
4 
3 
6 
12 
15 
22 
24 
27 
28 
37 
31 
16 
10 
7 
5 
2 
250 
We obtain the following results : 
Strains ^ and i? : ;^= = :31-11, P = -0627, 
Strains A and C: x' = ^3-37, P = 0034, 
Strains B and G: x' = 72-72, P = < -000,001. 
Thus to judge from rats only, B and C are far more divergent from each 
other than either is from A ; in other words the strain A is intermediate between 
B and G and closer to B, from which it is not immensely divergent; two such 
samples as A and B might, as far as the length distributions go, be drawn from 
common material once in 16 trials. 
Now of course no one suggests that a conclusion drawn from this rat-material 
is to replace one drawn from guinea-pig material, but the statistician cannot agree 
that for rats "the strains correspond very closely " ; and he finds it illogical to place 
the evidence of the rat-data on one side and proceed to draw conclusions from the 
ocular inspection of the guinea-pig curves, without noticing that the conclusion is 
markedly opposed to the proper deduction from rat-data. Indeed while the guinea- 
pig-dataf give a relatively high degree of relationship between .Band C(P = '0157) 
it is not as high as the rats give between A and B (P = -0627) ; and while the 
* The values given by the percentage graphs in these cases are respectively 21, 17 and 34, and 
the total appears to be 247 and not 250 as stated. Either 247 were used or the graph is in error. 
The three individuals were introduced in a way calculated not to increase divergence. 
t The frequency distributions for the guinea-pigs have had to be reconstructed from the percentage 
curves, the necessary data not being published by the authors. 
