We discuss the question of the normalization of the correlation function an
d its consistency with the often used form C(p(1), p(2)) = 1+\f(p(1), p(2))
\(2) We suggest an event mixing method which allows one to obtain absolutel
y normalized correlation functions from experimental data.