Two types of modulations of the auditory N1 event-related brain potential component are often discussed as expression of auditory predictive processing. First, the sound-related N1 component is attenuated for sounds generated by the listener compared to the N1 elicited by externally produced sounds (N1 suppression). Second, when participants are asked to generate sounds by pressing a button, an omission-related component in the N1 time-range is elicited if the sound is occasionally omitted (omission-N1). The phenomena of N1 suppression and omission-N1 were explained by specific forward modeling that takes place when the sensory input is predictable: Prediction error is reduced if forward model and sensory input match (N1 suppression) and enhanced if they mismatch (omission-N1). This common theoretical account for both phenomena is appealing, but it has not yet been directly tested. We manipulated the predictability of a sound in a self-generation paradigm in which in two conditions either 20% or 50% of the button presses did not generate a sound, inducing a strong or a weak expectation for the occurrence of the sound. In agreement with the forward modelling account, an omission-N1 was observed in the 20% but not in the 50% condition. However, N1 suppression was highly similar in both conditions. Thus, our results demonstrate a clear effect of predictability for the omission-N1, but not for the N1 suppression. This implies that the N1 suppression and the N1 omission phenomena rely on (at least partly) different mechanisms and it puts prediction related accounts for the N1 suppression in question.