Background: The outcomes of social inclusion and skills training programmes for pupils with special educational needs have been mixed. Programmes are generally based on research with mainstream samples yet it has been suggested that the social skills important for sociometric acceptance may be different for children who have special educational needs. Aims: The study aims to compare peer-assessed behavioural characteristics associated with sociometric status for included children who have moderate learning difficulties and their mainstream classmates. Sample: Mainstream classmates (N=867) of all the children ascertained as having Moderate Learning Difficulties (MLD) and included in mainstream middle schools (8?12 years) in one English county. Subgroups of rejected mainstream (n=38) and pupils with MLD (n=32) were further investigated. Method: Discriminant function analysis was carried out with peer assessment items as predictors of sociometric status group membership in the mainstream sample and contrasted with previously reported data from pupils with MLD in the same classes. Cluster analysis was used to identify behavioural subtypes within the rejected groups drawn from each sample. For pupils with MLD subtypes were validated using teacher assessments of social behaviour. Results: Systematic differences were identified across different analyses between the peer-assessed ehavioural characteristics associated with rejected sociometric status for pupils with MLD and for mainstream pupils. Conclusions: The appropriateness of generic social skills training programmes for promoting the social inclusion of pupils with MLD should be questioned and consideration given to rejected pupil subtypes.