M. Greenwood and J. D. C. White 
529 
control injections of Tuberculin, so that no two determinations are made under 
the same conditions, we do not think the method has more than a very restricted 
field of usefulness. This leads us on to refer to a cognate matter. 
Glynn and Cox use language* which seems to imply that we have attacked the 
opsonic index in the sense of denying that differences in phagocytic power as 
between normal and diseased persons occur. We have not asserted this nor does 
any statement of ours bear such an interpretation. We have merely pointed 
out, a contention not seriously challenged, that the claims advanced by certain 
enthusiasts to the effect that all variations of 10 from the normal are probably 
or necessarily significant, must be entirely unjustifiable. 
At the risk of being tedious, we desire to emphasise once more the distinction 
between vaccine therapy on the one hand and opsonic estimations on the other. 
If our work tends, as it certainly may tend, to unsettle the opinion of some readers 
as to the relative accuracy of the tuberculo-opsonic index, we wish in no way to be 
ranked among detractors of the claim of vaccine therapy to be regarded as a new 
and powerful weapon in the struggle against disease. On the contrary we are 
convinced of its present importance and look to the future for its still greater 
development. 
In conclusion we would again remark that the results here published are in 
essential agreement with those communicated by Harvey and McKendrick and 
ourselves in recent numbers of Biometrika. 
Conclusions. 
1. Large phagocytic counts extending to 15,000 or 20,000 cells are markedly 
skew and exhibit signs of heterogeneity. 
2. The exclusion of clumped bacilli has not been the main cause of the poor 
fit obtained. 
3. The heterogeneity may be due to difficulties in counting. 
4. The distributions of the means of samples of 25, 50 and 100 cells drawn 
from these populations are also markedly skew and excellently represented by 
Pearson's type curves. 
5. This skewness cannot be due entirely to homotyposis. 
6. The odds against a second sample of the same material giving an opsonic 
index, in terms of the true mean, outside of the limits 1'2 to -8 are, in the case of 
samples of 25, about 3 to 1, in the case of 50's about 6 to 1 and for lOO's about 
15 to 1. 
7. When the index is estimated on the basis of two small samples and without 
determining the mean of the normal series from a long count, the odds are reduced 
to I'o to 1, 2 to 1 and 4'6 to 1 respectively, for samples of 25, 50 or 100. 
* Glyun and Cox: BiochemicalJournal, Vol. iv. p. 300. 
