New classes of codes with similar application to Gilbert's code [1] are defined. Weakening of his synchronizing condition gives higher efficiency in the new codes. One class (the F codes) is shown to be the most efficient possible among those codes which use Gilbert's method of achieving synchronization, i.e., the ``position fixing'' method. The other classes use a basically different method of achieving synchronization, and have even higher efficiencies than the F codes. The codes are compared mainly with regard to efficiency, but some idea of the hardware required is also given. The efficiency is compared to that of the bound on Golomb's comma-free codes [2], but evidence is supplied that codes constructed to carry information will never reach this limit.