This investigation is about the observation that certain relative clauses in German show peculiar appendices, involving an obligatory subject gap and lacking syntactic integration. The question raised here is whether these strings should be conceived of as an output of mental grammar or rather as systematic mistakes. This second notion would be plausible if a range of examples is poorly accepted in a controlled setting. Therefore, I conducted an experiment, more precisely: a pseudo response time experiment, based on real instances of the structure. In fact, the examples achieved quite high acceptance rates so that they cannot easily be dismissed as mere mistakes. They might, however, still represent grammatical illusions. It will be argued that they are not, leaving an inner-grammatical status of the structure as the only possibility.