A growing body of research supports the suggestion that the relationships which children form with their teachers and classmates have an impact on learning (Roorda, Koomen, Spilt, & Oort, 2011). Largely built on studies with typically developing children, the current understanding of the relationship-learning association is that these relationships can impact upon learning either by directly improving the quality of pedagogy or through mediating factors such as increased pupil motivation (Martin & Dowson, 2009).The aim of this study was to expand the discussion and evidence base surrounding relationship-learning association to include pupils with Autism Spectrum Disorders (ASD) and Behavioural, Emotional and Social Difficulties (BESD). Previously these groups were not directly addressed by this literature, yet there is evidence from both government-compiled (Department for Education, 2011d) national statistics and independent research (e.g. Symes & Humphrey, 2010) that these groups are at risk of poor academic and social outcomes. While classroom relationships might be an ingredient of superior teaching of children with special educational needs (Dyson, Farrell, Polat, Hutcheson, & Gallanaugh, 2004), it was hypothesised that difficulties commonly associated with either or both of these groups, such as communication problems (Cashin, 2005; Lindsay, Dockrell, & Strand, 2007) may serve to influence the nature and salience of the relationship-learning association.The study used a mixed methods design, incorporating a multiple regression analysis to determine whether changes in teacher or peer relationship quality over 18 months predicted attainment relative to other plausible predictors at the end of that period (N= ASD:143 BESD: 648) and an embedded, multiple case study (Yin, 2011) analysis around two children from each group to determine how the facets of their individual educational needs and other contextual factors influenced the importance and nature of the relationship-learning association in their education. Multiple regression models indicated that relationship change was not a statistically significant predictor of attainment other than peer relationship change for pupils with BESD, where the effect size implies that some academic benefits may accrue at group level from successful relationship interventions. Case study analysis suggested that many of the factors reducing the likelihood of relationships directly improving attainment also make them more important to the effectiveness of teaching and the classroom functioning of pupils. Implications and directions for future research are also discussed.