The avoidance of errors during learning, so-called errorless learning, results in increased memory performance. In the present study subjects had to learn items in an errorful and an errorless manner. After each learning session learned items were presented again, but now intermixed with items not learned before. Response-locked event-related potentials were used to investigate the neural underpinnings of cognitive control mechanisms during recognition of items learned under errorless and errorful conditions. Irrespective of the response’s correctness a typical error-related negativity (ERN) was observed for items classified as learned before. In contrast to the apparent difference in memory performance between learning modes, ERN amplitudes to hits and false alarms were not different. The present pattern of results can be explained neither in terms of error detection nor theconflict monitoring account. Instead we are trying to argue that the findings add support to a theoretical proposal which posits that variations of the ERN amplitude can be best explained by the subjects’ perceived likelihood of making an error.