The need for better treatments for children and adolescents with mental health problems is universally acknowledged. While evidence-based practice provides a framework for identifying and disseminating effective treatments, therapeutic innovation is more likely to come from translational research in which basic science on the social and psychopathological bases of disorders. This will help to identify new treatment targets and promote the tailoring of new therapeutic agents and activities. For such an approach to be effective we first need to be able to isolate the underlying determinants of disorder. To do this we build causal models, set up hypotheses on the basis of these models and test predictions. Models are valuable as a basis or translational science if they (i) make distinctive and risky predictions and (ii) posit cause. Unfortunately simply observing the way a patient’s behavior (or brain) changes as environments change in naturalistic settings is of no value in testing causal models. First, even valuable models often make the same predictions as each other in naturalistic settings. This is because changes in a multitude of potentially significant factors co-vary under these conditions. It is therefore difficult to identify which is the actual determining factor. Complex multivariate approaches can only partially resolve this problem. Second, one cannot infer cause from a correlational analysis. Experimental analysis seeks to isolate the operating factor from the mix of possible covarying factors that alter performance/behaviour/brain and at the same time establish direction of causality. It therefore is a potentially effective way to pinpoint the source of disorder and to identify new treatment targets. In this talk I illustrate the value of the experimental approach for models of child and adolescent psychopathology by highlighting the way in it has been used to differentiate models of delay of gratification in ADHD. Further I will show how this can then provide a platform for diagnostic and therapeutic innovation.