Abstract Background Patients with depression demonstrate cognitive impairment on a wide range of cognitive tasks, particularly putative tasks of frontal lobe function. Recent models of frontal lobe function have argued that the frontal pole region is involved in cognitive branching, a process requiring holding in mind one goal while performing sub-goal processes. Evidence for this model comes from functional neuroimaging and frontal-pole lesion patients. We have utilised these new concepts to investigate the possibility that patients with depression are impaired at cognitive 'branching'. Methods 11 non-medicated patients with major depression were compared to 11 matched controls in a behavioural study on a task of cognitive 'branching'. In the version employed here, we recorded participant's performance as they learnt to perform the task. This involved participants completing a control condition, followed by a working memory condition, a dual-task condition and finally the branching condition, which integrates processes in the working memory and dual-task conditions. We also measured participants on a number of other cognitive tasks as well as mood-state before and after the branching experiment. Results Patients took longer to learn the first condition, but performed comparably to controls after six runs of the task. Overall, reaction times decreased with repeated exposure on the task conditions in controls, with this effect attenuated in patients. Importantly, no differences were found between patients and controls on the branching condition. There was, however, a significant change in mood-state with patients increasing in positive affect and decreasing in negative affect after the experiment. Conclusion We found no clear evidence of a fundamental impairment in anterior prefrontal 'branching processes' in patients with depression. Rather our data argue for a contextual learning impairment underlying cognitive dysfunction in this disorder. Our data suggest that MDD patients are able to perform high-level cognitive control tasks comparably to controls provided they are well trained. Future work should replicate these preliminary findings in a larger sample of MDD patients.