880 On the Frequencij Distrihutions of Phagocytic Counts 
number of cells than a beginner would require. If a per.son counts simply mechanically every 
leucocyte he meets and fails to bring any intelligence to bear on the subject in hand, however 
much care he may have taken to avoid error, if his emulsion is not perfectly even (which is 
uncommon), he will have to count a very much lai'ger number of cells before he has a trustworthy 
count of the slide*." 
Now the statement that, given a " perfect " emulsion, it would be easy to say 
that a certain number of cells would furnish a result within ten per cent, of the 
true reading, can only be valid if the nature of phagocytic frequency distributions 
were exactly known, i.e. whether such distributions are "normal" or "skew," and in 
the latter event, of what type and possessing what degree of skewness. 
To the best of our knowledge, the present memoir contains the only attempt to 
answer these preliminary questions which has been made. 
Again, why "intelligent" counting of a few cells should give better results 
than " mechanical " counting of many, is not obvious unless we are furnished with 
some stringent definition of the word " intelligent." 
It is difficult for us to form a mental picture of a worker who, while failing " to 
bring any intelligence to bear on the subject in hand," at the same time, takes the 
greatest care to avoid error. We imagine that an intelligent worker, using the 
word intelligent in an ordinary way, will only reject cells so badly stained, broken 
up, massed together, or containing such clumps of bacteria that either the outline 
of the cell or the number of bacteria it contains cannot be accurately gauged. 
Of papers expressing an unfavourable opinion of the opsonic method, the most 
extensive is that by Drs Strangeways and Whiteman and Miss Fitzgerald +. 
These observers found very great differences between the means of phagocytic 
counts from batches of 50 or 100 cells taken from the same slide, i.e. mixed with 
the same serum. They also found that two capsules of the same blood, drawn at 
the same time and treated in the same manner, gave widely divergent readings. 
Similarly, opsonic indices of the same blood, estimated by different persons in terms 
of the same control, varied considerably. 
The authors of the paper just mentioned did not avail themselves of modern 
statistical methods of analysis. Their conclusion that opsonic determinations are 
subject to an error of 100 per cent, does not, using the word error in any exact 
sense, follow from their work. There is, for instance, an evident confusion between 
the extreme variations in the range of values exhibited by the separate counts 
and the mean variation. 
Further, a comparison of groups each of 100 cells from the same slide is not 
necessarily the same thing as comparing counts of 100 each from a separate slide. 
There is possibly here a fallacy analogous to that involved in the assertion that 
* Fleming, op. cit. p. 627. 
t "An Inquiry into the Value of the Opsonic Index" by M. P. Fitzgerald, R. I. Whiteman and 
T. S. P. Strangeways. (Bulletin of the Committee for the Study of Special Diseases, Vol. i. No. 8, 
Aug. 1907, Cambridge University Press.) 
