The invention relates to an error correction coding method, using at least two distinct sections of a predetermined elementary code, associating an arrival vector (s.sub.2, s.sub.3) with a starting state vector (s.sub.0, s.sub.1), according to a vector of branch labels (b.sub.0, b.sub.1, b.sub.2, b.sub.3), defining a code word, two sections of said elementary code being distinct when the order and/or the role of the elements of said branch label vector are changed.