Two training procedures were constructed to teach children with learning deficiencies to solve simple addition and subtraction word problems. Each training procedure was evaluated in a separate multiple baseline design across subjects. The experiments were designed to prepare for the development of computer assisted training programs. Two female mildly retarded children (age 14) participated in the first experiment. They were taught to work systematically through the problem task and to pay attention to several important components of the problem text. The training procedures turned out to be effective. Research was continued with a computer assisted training experiment. The computerized training procedure is described in relation to the findings of the first experiment. The second experiment was run with three male disabled children (age 10-11). They were taught to draw a diagram as an external representation of the problem. The diagrams reflected the properties of "change,""combine" and "compare" problem types. This training method appeared to be effective as well. The differences in the nature of the efficiency of these two training methods was discussed. The results were analyzed by graphing correct problems, mean response time, number of errors, and mean percentage of correct steps over training sessions. (Author/JM)