Sources of Error in Serological Work. 75 



experimental method, with the discovery of certain sources of 

 error not usually taken into account in serological work. 



The first source of error is the possibility of there being marked 

 changes in the chemical composition of serological substances as a 

 result of changes in concentration. The bactericidal substance 

 with which I was working is fairly stable. It can be heated to 

 6o° C. for an hour without loss of bactericidal power and can be 

 stored in the ice chest for weeks with but slight deterioration. The 

 substance can be passed through a series of chemical manipulations, 

 involving such processes as salting-out, dialyzing, evaporating to 

 dryness, and redissolving, and can be recovered quantitatively 

 from the final product of such manipulations, provided the volume 

 of fluid in which it is dissolved is at no time allowed to increase 

 much above the original volume from which the substance was 

 obtained. If the volume is allowed at any stage to materially 

 increase, there is brought about a rapid deterioration of the 

 bactericidal substance at that stage, giving a final product without 

 bactericidal action. 



The second source of error is the possibility of there being 

 marked changes in the specific properties of serological substances 

 as a result of variations in the amount of sodium chloride with 

 which they are mixed. The purified bactericidal substance from 

 horse leucocytes, dissolved in distilled water, has about half the 

 bactericidal power of the initial crude product. If dissolved in 

 physiological saline solution, instead of in distilled water, it is 

 without bactericidal power. 



55 (664) 



The relation of the virulence of the tubercle bacillus to its 

 persistence in the circulation. 



By ALFRED F. HESS. 



[From the Research Laboratory, Department of Health, New York 



City.} 



It seems as if the tubercle bacillus offered an exceptional oppor- 

 tunity to study the question presented in the title of this study. 

 As is well known, one type of the bacillus, namely, the human 

 type, is non virulent for the rabbit, whereas the bovine type causes 



