A Rasch-scaled working memory task was used to estimate the working memory capacity of adolescents with moderate mental retardation. The working memory task was calibrated using a sample of 412 children from regular mainstream classes and special education classes in primary school. The main issue of this paper concerns the usefulness of the constructed scale for children with moderate mental retardation. The Item Response model used is the Verhelst and Glas generalized Rasch-model (see e.g., Verhelst & Glas, 1995), a strong measurement model with attractive properties, such as measurement on a true interval scale on which both items and persons can be represented and the estimation of personal capacities independent of the items actually used. Because the model can be tested and the response patterns of individual persons can be analyzed, inferences can be made about the usefulness of the model for persons with moderate mental retardation. When the estimated capacity of a person is not in line with the responses given to individual items, this implies that the measure is not reliable for this person. When such a situation arises for many persons with mental retardation, this indicates that the present model is not applicable to this population; however, when the capacity estimates of persons with moderate mental retardation are reliable, this implies that working memory capacity is measured on the same dimension. In this study, the performance of adolescents with mental retardation was generally comparable to those of six- to seven-year-old normally developing children, but never surpassed complexity level 4. Their response patterns also conformed to the model. We concluded that working memory was measured on qualitatively the same dimension in both groups.