The author investigated the effects of the concrete-representational-abstract (CRA) instructional sequence on the computation performance of students with specific learning disabilities and students identified as at risk for failure in mathematics. Researchers have showed the CRA sequence to be effective for teaching basic mathematics facts, fractions, algebra, and place value. However, they have not studied its effects when it is used to teach subtraction with regrouping. Therefore, the author examined the effects of CRA instruction on elementary school students' fluency in computing subtraction problems with regrouping and maintaining these skills. The author used a multiple-probe-across-groups design and demonstrated a functional relation between CRA instruction and subtraction with regrouping across all students. The author discusses the results and implications. (Contains 1 table and 3 figures.)