Psychological interventions are first-line treatments of depression. Despite a rich theoretical background, the mediators of treatment effects remain only partially understood: it has been difficult to precisely delineate the targets psychological interventions engage, and even more difficult to differentiate amongst the targets engaged by different psychological interventions. Here, we outline these issues and discuss a surprisingly understudied approach, namely the study of cognitive and computational tasks to measure psychological treatment targets. Such tasks benefit from substantial advances in cognitive neuroscience over the past two decades, and have excellent face validity. We discuss two candidate tasks for backtranslation and conclude with a critical evaluation of potential problems associated with this neuro-cognitive approach. Highlights-Cognitive Behavioural Therapy (CBT) for depression has a rich theoretical background in cognitive and learning theories, but we still know little about the specific mechanisms engaged by different CBT modules.-Neuro-computational probes derived from modern computational cognitive neuroscience are specifically designed to characterise behaviours and cognitions that CBT aims to change-Cognitive computational tasks might be useful tools for research aimed at defining mechanisms and mediators brought about by different CBT modules.-We discuss metacognition as candidate mediators of Cognitive Restructuring, and Reward and Effort-based decision-making as candidate mediators for Behavioural Activation-The psychometric properties of tasks pose obstacles for their clinical use.