This study investigated the longitudinal stability of sociometric classification in two groups of children aged 8 to 10 years over a 2-year period. Two simple, but sensitive and valid forced-choice sociometric measures were completed by 41 children with moderate learning difficulties (MLD) and 375 mainstream children. On both occasions and on both measures (one using a ‘‘work with’’ and one a ‘‘play with’’ sociometric criterion) MLD pupils were significantly less likely to be classified as popular and more likely to be classified as rejected. Over the 2-year period just over half of the children, whether MLD or mainstream, were classified in the same sociometric category. However, the stability coefficients for the rejected and popular sociometric status groups were found to be higher in the included sample than in the mainstream sample. The results of this study are compared with previous longitudinal studies in the area that have focused on mainstream pupils, and implications for the use of sociometric measures in evaluating the outcomes of inclusion are discussed.