M. Greenwood and J. D. C. White 
507 
Table II were obtained and Graph 1 shows the theoretical curve and observations. 
It is evident that, although the curve represents the general trend of the observa- 
tions with fair accuracy, the goodness of fit as evaluated in the ordinary way is 
poor. This matter appears to require comment and explanation. 
TABLE I. 
Actual Frequencies of 20,000 Cells, together with Corrected 
Values determined on the Basis of a Random Count of 
2,000 Cells from the same Material. (See p. 509 infra.) 
Number of 
Bacilli per 
Cell 
Original 
Value 
Corrected 
Value* 
0 
1428 
1402-75 
1 
2633 
2636-57 
3351 
3355-55 
3 
3556 
3560-83 
2807 
2810-81 
5 
2038 
2040-77 
6 
14.30 
1431-94 
1031 
1032-40 
8 
658 
658-89 
9 
407 
407-55 
10 
271 
271-37 
11 
163 
163-22 
12 
103 
103-14 
13 
61 
61-08 
n 
37 
.37-05 
15 
17 
17-02 
16 
9 
9-01 
Totals 
20000 
19999-95 
It will have been noticed that for the purposes of this count we did not enumerate cells 
which contained clumps of bacteria. Our reason for this omission was that, to the best of our 
knowledge, workers ordinarily disregard cells the contents of which cannot be resolved into 
definite bacilli or fragments of bacilli, and we were most anxious to use data so far as possible 
identical with those employed in actual practice. It was, however, subsequently pointed out to 
us, that this process was open to serious objection for the following reasons. In the first place, 
some workers of great reputation never employ for diagnostic purposes mixtures containing an 
appreciable number of clumps, say more than two or three per cent. In the second place, the 
exclusion of clumps would tend arbitrarily to heighten the proportion of cells containing 0 bacilli, 
since cells containing clumps must be regarded as possessing a phagocytic power at least equal 
to those which contain one bacillus apiece; indeed, Harvey and McKendrick have adduced 
evidence t which supports a belief that it is rather the number of bacteria than the number of 
acts of ingestion which should be taken as a measure of phagocytic power. 
* 1428 cells with 0 bacilli per cell will occur not in 20,000, but in 20,360; hence in 20,000 we shall 
have not 1428 but 1402-75, the excess is distributed proportionally among the other groups : see p. 509. 
t Biometrika, Vol. vii. p. 64. 
