Karl Pearson 
129 
of even numbers. No curve whatever could fit the data satisfactorily under the 
circumstances ! Either they used a scale graduated to 2 microns only, and had a 
prejudice in favour of the scale markings, or else their even numbers were in some 
way more conspicuous than their odd. Whatever the source of this peculiarity 
may be, there can be no doubt of the bias*. 
The only way to obtain a reasonable measure of the goodness of fit in Stephens 
and Fantham's results for T. rhodesiense is to group from 10 to 12, 12 to 14 and so 
on in comparing the observed and calculated frequencies. If this be done we find 
X' = 5'03 for 13 groups and P = '957, a splendid fit. The frequencies are as 
follows : 
1 
/ 
10-u 
IJ1.-I6 
16-18 
18-20 
20-22 
22-2Jt 
2Jf-26 
26-28 
28-30 
30-32 
32-3Jf 
36-38 
Totals 
Observed 
Calculated 
i 
9 
7-17 
38-5 
34-67 
83-0 
92-99 
133-5 
1.32-91 
134-0 
124-79 
127-0 
124-28 
135-5 
146-35 
139-5 
145-56 
112-0 
106-55 
60-5 
56-22 
22-0 
21-36 
4- 0 
5- 84 
1-5 
1-17 
1000 
999-85 
(ii) T. brucei. The data for this trypanosome were taken from Sir David 
Bruce and colleagues' diagram-]-. I have not come across the original publication 
with the measurements involved in this diagram. Describing this species in 
July 1910|, the authors speak of its well-marked dimorphism. This is very 
obvious in the graphs for length given for the Uganda 1909 and Zululand 1894 
strains, but the numbers given are far too slender (160 and 200 respectively) to 
justify any attempt at analytical resolution. Graphically we may take it that 
roughly the following are the means of the components : 
T. minus. 
20 microns 
18 microns 
T. majus. 
28 microns 
29 mici-ons. 
Uganda 1909 
Zululand 1894 
These are not very widely divergent from the values 
19-8 microns 26*0 microns 
we have found from the seven resolutions. 
In May 1911 § the two curves for Uganda and Zululand appear to be added 
together to give a T. brucei curve of length distribution. This is again markedly 
bimodal with one component mean at 18-75 microns and the other at 27-5 microns, 
both approximative. Thus far T. brucei appears qidte well to fit in with our other 
material. But in September 1911 appears the diagram of T. brucei said to be 
* Bias of this or of a similar character is not uncommon — even in the pages of this Journal. 
I remember once pointing out to a Scotch anthropometer liis prejudice in favour of whole centi- 
metres. He loolsed at his results, recognised the bias, and then gravely told me that it was not 
due to any personal bias, but that the Creator must have designed Scotsmen on the metric scale ! 
t R. S. Proc. Vol. 84, B, p. 331. 
t R. S. Proc. Vol. 83, B, p. 2. 
§ R. S. Proc. Vol. 84, B, p. 186. 
Biometrika x 17 
