We investigate how users gradually adapt to a faulty system and how the system error rate influences this adaptation process. We present results of a study that verifies that a user's learning rate to compensate for system errors depends on how erroneous that system is – they learn to avoid erroneous actions faster if errors occur more frequently.