Abstract: A sequence of k-tuples of information elements is converted by means of an error-correcting convolutional code into a sequence of n-tuples (n greater than k) of code elements. In the processing of the code elements use is made of first of all a syndrome former to form a sequence of (n-k)-tuples of syndrome elements from the n-received sequence of n-tuples of code elements. From a segment of a number of successive (n-k)-tuples of syndrome elements one or more correction bits are formed in each case so as to be able to correct the received code bits and also, if a correction is applied, to be able to update a number of further syndrome bits from that same segment. Furthermore an (n-k)-tuple of secondary syndrome bits or residue bits is formed in each case from this updated segment. A first value of such an (n-k)-tuple indicates the state in which the decoder must be provisionally considered as having corrected all errors in a predetermined sequence of n-tuples of code elements.