506 On the Distribution of the Means of Samples 
In undertaking this large count, those methods were adopted which are usually- 
advised in the preparation of films. The mixture, to be subsequently placed on 
the slides, was made up of three equal parts, (1) serum, (2) corpuscles, (3) bacterial 
emulsion. The first two were derived from the operator and, in the case of the 
corpuscles, care was taken that they should be equally distributed, the upper 
portion of the centrifugalised blood being taken off and thoroughly mixed in a 
small tube. The bacterial emulsion was prepared as usual and carefully mixed so 
as to secure the greatest possible homogeneity. Since we desired a large count of 
a homogeneous character, not a comparison of different mixtures, it was decided 
to use one pipette only, but of larger calibre than usual, in order to secure the 
greater amount of mixture requisite for the preparation of a large number of 
slides. This pipette was throttled for ease of manipulation. The proportions of 
the three materials were carefully mixed and the pipette placed in an incubator 
for fifteen minutes, the emulsion having been designed to give an average of three 
to four bacilli per cell. The pipette was slightly rotated from time to time in 
order to keep the corpuscles from settling. Sufficient slides being in readiness, 
the pipette was withdrawn from the incubator, the contents again thoroughly 
mixed, and films were prepared as quickly as possible. These were then stained 
with Aniline Fuchsin and Methylene Blue ; thereafter cover-slips were applied 
with Canada Balsam to preserve the slides during the process of counting. The 
general quality both of films and staining was exceedingly good, very few slides 
being of an inferior character. 
The counting of so large a number of cells as 20,000 was, needless to say, 
sufficiently laborious, especially in regard to the leucocytes containing more than 
10 bacilli apiece. Reasonable accuracy was found possible up to 15 and the few 
cells containing more than this were marked 16. A mechanical stage was used — 
the only method, in the operator's opinion, which can secure that the same cell 
shall not be counted twice over. The only cells omitted were those the outline of 
which was indistinct, or where for any reason, e.g. the clumping of cells or bacilli, 
it could not be determined how many bacilli lay in an individual cell ; the possible 
error introduced by this latter criterion will be discussed below. 
Definite fragments of bacteria were counted as bacteria ; so also were bacilli in 
definite contact with the periphery of cells. In order to avoid mental prejudice, 
addition of the rows of figures was postponed generally for days and always at 
least until the end of the day's work. In this way 20,000 cells were counted in 
the course of about a month, with occasional intervals. The sheets of figures were 
then taken and the totals of successive twenty-fives, fifties and hundreds were 
determined ; then the frequency of each number per cell on each sheet of 500 cells 
was tabulated, this latter process incidentally eliminating any arithmetical mistakes. 
The frequencies on each of the forty sheets were then added together and the total 
frequency obtained. 
The actual distribution of the twenty thousand cells is set forth in Table I. 
Proceeding to fit a curve, without using Sheppard's corrections, the constants of 
