This study compared the effectiveness of computer-mediated instruction (CMI) and teacher-mediated instruction (TMI) on the word problem-solving performance of students struggling in mathematics. Both conditions integrated cognitive modeling that focused on the problem structure using visual representations with critical instructional elements specifically targeting the needs of at-risk students. Participants were 25 third-grade students with mathematics difficulties who were randomly assigned to either a TMI or a CMI condition. Results indicated no statistically significant between-condition differences at posttest and on a 4-week retention test of word problem-solving. Furthermore, there was no transfer of the word problem-solving skills to a school administered, standardized mathematics achievement test. Implications for educational practice are discussed. (Contains 2 figures and 3 tables.)