A 2-year study of optimal structures for multimedia instruction is being conducted to provide experimentally validated guidelines for the design of computer-based instruction generation systems and for human instruction in a multimedia setting. In order to obtain for analysis a significant range of the possible discourse structures that occur in instruction, the project's first phase elicited explanations of a demonstration device from experienced community college engineering instructors. The outcome of this phase was a set of variables and a set of hypotheses about relationships among variables that lead to effective instruction. The project's second phase will test these hypotheses on groups of students. Four major results were achieved in the first phase: (1) the development of a framework for discussing optimal discourse structures and/or visual presentations in multimedia instruction, based on the notion of a mapping between semiotic systems; (2) the discovery that the command and control speech act chain is used in "hands-on" instruction; (3) the development of a rich set of experimental hypotheses; and (4) a demonstration of the viability of a methodology combining linguistic analysis with experimental research. This report describes the first year's work, with sections on discourse analysis, semiotics, other analytic concepts, and project variables and hypotheses. Also presented are appendices on project methodology and the grammar of the command and control speech act chain, a 34-item bibliography, and a report distribution list. (ESR)