Acquisition of cognitive and socio-emotional skills in the classroom involves repeated interactions between everyone in the room. Combinations of personalities and abilities and behaviours continually develop and affect each other. We will characterize these social interactions by way of new and existing psychometric tests, questionnaires and administrative registers and relate these to outcomes by way of developing and extending theories of education production combined with causal reduced form econometric models. Our identification strategy builds on elements of randomness and selection in the allocation of pupils to schools and classes. Danish municipalities differ in availability of private school alternatives (selection) and public schools differ in special education policies for the needy. There has also been a rapid and differential expansion of special-needs education during the last decade across Denmark. Together, these developments will be used to provide quasi-experimental variation in classroom composition – where needy pupils are placed in classes and schools, and how exposed they are to special teaching – which will enable us to measure effects of contexts and interventions respectively. Results obtained from this project will inform policy makers by identifying combinations of behaviours and abilities which produce optimal classroom environments and learning outcomes.