Information for the National Tune Index data bank was gathered from a bibliographically complex body of material, including both printed and manuscript sources and representing ever3, genre of popular secular music in eighteenth-century England and America, as well as every level of humanistic expression, from stylized theatrical statements to simple entries in colonial commonplace books. The disparate categories of information to be gathered and manipulated included names of people, names of songs, verse incipits, tune incipits and other musical information, song titles and refrains, and bibliographic information (see fig. 1).The design for the input code and related programs to demonstrate the unity in all this diversity, was based on three guiding principles:(1) simplicity, to foster maximum understanding and participation in technical decisions by non-technical project staff;(2) flexibility, to permit continuing modifications and improvements as the project progressed; (3) efficiency, to stretch available funds. Since nearly every category of information to be gathered yielded variable-length data, and since not all categories were expected to be present in every source, a free-format code was clearly indicated as the most appropriate to represent such data, and the most efficient for processing its various components into a series of cross-referenced indexes. SNOBOL was selected as the most appropriate language for the free data format and for the types of data manipulations anticipated. ~ SNOBOL's one disadvantage, its slow operation, was overcome by the use of the SPITBOL compiler, which was approximately ten times faster.