The theory underlying a translator program, which is needed on a computer whenever programs written in the algoritumic language ALGOL are to be used on that machine, is described. The approach followed is related to the recursive sequential methods. These methods have been simplified by the explicit use of recursive subroutines based on syntactic skeletons defining the ALGOL language. Specifications for a translator based on these principles are given in machine-independent form. They include substantial advances over those used in the design of the first ALGOL translator for the ORACLE. (auth)