218 Opsonic Index — "Mathematical Error and Functional Error" 
and White's record. How this result is reached I am unable to say. It cannot 
apparently be due to a more thorough mixture on the slide, because Strangeways' 
working with a Wright-Laboratory slide and using practically the same method 
of counting get variations close to Greenwood and White. It would appear to 
me that Fleming has, even when counting 1000 leucocytes, a method of more 
evenly distributing his large and small number of bacilli per leucocyte within 
his 50. How he manages this I am unable to say ; it might possibly be achieved 
by picking each 50 leucocytes at random from the slide, but in this case it is 
difficult to understand how, if he does not use a mechanical stage, recounting 
the same leucocyte is avoided. But, whatever be the nature of the recording, 
this end is reached, that the mean number of bacilli in 50 leucocytes starting 
anywhere in the record is less variable than Greenwood and White's result. An 
appreciation of the extent of this lessened variation may be reached by comparing 
the curve for opsonic indices deduced from 400 non-independent means from 
Fleming's counts on Slide T. A. I. with that found for 400 on Greenwood and 
White's Slide B : see Diagrams VIII and IX. 
Diagram VIII. Opsonic Indices from dependent means. (Fleming.) 
400 
Opsonic Indices. 
