93 



In the elaboration of the results of the examinations of 3 

 and 4. IX. 18 at that time, no difference was made between 

 typical and atypical Pfeiffer's bacilli. The reason only 132 of 

 the 142 men examined are accounted for, is that information 

 about influenza was only available later, and in 10 cases it 

 could not be obtained. 



These figures show that Pfeiffer's bacillus is encountered to 

 a considerable extent also among persons who deny having 

 had influenza. Even if this information about previous in- 

 fluenza cannot be reckoned very reliable, it is incredible that all 

 the carriers of Pfeiffer's bacillus who said they had not suffered 

 from influenza, should actually have had this disease. It is 

 therefore justifiable to conclude from the figures that people 

 often become carriers of this microbe without having had an 

 attack of influenza. It is particularly interesting that in direct 

 response to the influenza epidemic in the beginning of 1920, 

 there has proved to be such a marked dissemination of Pfeif- 

 fer's bacilli among healthy persons the majority of whom stated 

 they had not been attacked during the epidemic . 



In order to investigate how far it would be possible from 

 the figures stated in above table to draw any conclusion at all 

 as to a possible connection between previous cases of in- 

 fluenza and the incidence of Pfeiffer's bacillus, the first three 

 groups (the first three lines in the table) have been treated 

 statistically. 



In the left half on the table below, the frequency of occur- 

 rence of Pfeiffer's bacillus expressed as a percentage, is re- 

 corded for soldiers who had had influenza and for those who had 

 not, separately. Then the difference between the two per- 

 centages is given. It will be seen that Pfeiffer's bacillus all 

 the way through occurred more widely among past influenza 

 patients than among others, whether we take all Pfeiffer's ba- 

 cilli into account or only the typical ones. 



In order to determine whether these divergencies are large 

 enough to be accounted as more than fortuitous", the mean 

 error of each difference is calculated. If p t and p, are the 

 two percentages, and the corresponding number of soldiers 

 examined, n x and n 2 , the mean error of p t — p 2 is given by 

 the equation u= |/P'P«>-Pi> + ftP°g=Pi> I fi is thus also 



