The effects of internal model variability on the simulation of Arctic sea-ice extent and volume have been examined with the aid of a seven-member ensemble with a coupled regional climate model for the period 1948–2008. Beyond general weaknesses related to insufficient representation of feedback processes, it is found that the model's ability to reproduce observed summer sea-ice retreat depends mainly on two factors: the correct simulation of the atmospheric circulation during the summer months and the sea-ice volume at the beginning of the melting period. Since internal model variability shows its maximum during the summer months, the ability to reproduce the observed atmospheric summer circulation is limited. In addition, the atmospheric circulation during summer also significantly affects the sea-ice volume over the years leading to a limited ability to start with reasonable sea-ice volume into the melting period. Furthermore, the sea-ice volume pathway shows notable decadal variability that varies in amplitude among the ensemble members. The scatter is particularly large in periods when the ice volume increases, indicating limited skill in reproducing high-ice years.