Abstract: The effectiveness of word stuffing for synchronization depends upon the ability to distinguish the stuff words from the data words at the destination and, thus, delete correctly the stuff words. If all input sequences are permitted, the data words must be encoded before stuffing occurs so that the stuff word can be distinct from the data words. Previously, a code was given in which a single transmission error could change a data word into the stuff word. In this paper, we give the code that maximizes the minimum distance between the stuff word and any data word. It follows that the probability of character synchronization loss because of confusion between data and stuff words due to transmission errors is considerably reduced. Some implementations are given which suggest that the increase in distance can be achieved at a small cost in hardware.