The biproportion of S on margins of M is called the intern composition law, K: (S,M) -> X = K(S,M) / X = A S B. A and B are diagonal matrices, algorithmically computed, providing the respect of margins of M. Biproportion is an empirical concept. In this paper, the author shows that any algorithm used to compute a biproportion leads to the me result. Then the concept is unique and no longer empirical. Some special properties are also indicated.