Simple addition problems were presented using a true/false reaction time (RT) verification paradigm to 77 academically normal and 46 learning disabled (LD) subjects in the second, fourth, or sixth grade. The experiment was designed to determine the potential process deficits associated with a learning disability in mathematics achievement. Structural models representing alternative process strategies were fit to RT data. Across grade level and academic status, RT was best fitted by structural variables representing either an implicit counting strategy or a memory retrieval process. The majority of normal and LD second-grade subjects used the implicit counting strategy for problem solution; however, LD subjects required a greater amount of time to execute this process and appeared to be deficient in the ability to self-monitor the problem-solving process. A clear shift from reliance on the implicit counting strategy to the memory-retrieval process was evident from the second to sixth grade for normal subjects. No such shift was evident for LD subjects, as the majority of these subjects relied on the counting strategy in the second, fourth, and sixth grade. Subjects having a specific learning disability in mathematics achievement appear to differ from academically normal subjects in the developmental maturity of the component process used for problem solution, the temporal duration required to execute this strategy, and the ability to self-monitor the problem-solving process. Implications for remediation are discussed.