The paper is a discussion of research on direct instruction and its implications for effective instruction in verbal math problem solving for elementary learning disabled students. Characteristics and problems of conventional programs are reviewed, as are the major features and logic of direct instruction programing. Components of instructional design are noted, including specification of objectives, identification of necessary preskills, and maintenance of acquired skills. The relationship of the direct instruction model to other promising models is analyzed, the feasibility of direct instruction in special education is considered, and obstacles to its implementation are noted. (Author/CL)