An important issue in the usability of software is the quality of the manuals that are provided with the product. Some manuals describe the rules of a programming language to the user, and the syntax notation or metalanguage used to describe these rules can take various forms. In this study, we tested a notation which uses brackets and braces (designated "Signs") against a method which uses syntax diagrams (designated "Maps"). A two-day experimental design was used in which the subjects were trained and tested on one metalanguage the first day, and then trained and tested on the second metalanguage the next day in a Latin square design. Four phases of this experiment were conducted, with 12 subjects in each phase. The procedure remained the same in each phase, but the subject population (programmers or non-programmers) or training material (full manual or one-page instruction) was changed from one phase to the next. No performance differences were found between Maps and Signs when the subjects were given extensive training on the metalanguage. However, when training was restricted to one page of instruction, performance was consistently superior with Maps for both programmers and non-programmers.