Initial recognition of an antigen as an object the organism must defen
d itself against is performed on the basis of an information not imply
ing the specific recognition of the antigen but the simple recognition
that it does not belong to the self. The application of statistical c
oncepts shows that this recognition must be performed on the basis of
analytical operations consisting of sectioning the field of antigen va
riability between the histocompatibility complexes and subcomplexes an
d placing the antigen into one of these sectors. It also implies stati
stical inference from the overall data provided by sensor elements con
sisting of lymphocyte clones belonging to the histocompatibility sub-c
omplex in which the antigen is placed. According to the laws of mathem
atical statistics the inference is affected by two types of errors whi
ch are dialectically interrelated: a decrease in one type leads to an
increase in the other and it is thus impossible to reduce both types o
f errors simultaneously. At least as far as endogenous antigens are co
ncerned, the organism's response implies a change in the ratio of the
two types of errors. In order to nullify the effects of this change th
e not self recognition requires a complex mechanism implying the prese
nce of opposed regulation factors.