The adequate description of word stress is still a matter of discussion in phonological research. There are two types of approaches to explain German word stress: quantity-sensitive approaches (e.g., Giegerich 1985), on the one hand, claim that stress depends on syllable weight (the inherent structure of a syllable), quantity-insensitive approaches (e.g., Wiese 2000), on the other hand, claim that German word stress falls on a specific position in a word. There are some studies on the assignment of word stress by (language impaired) native speakers of German. Janßen (2003) found proof for the quantity-sensitive approach to German when the participants were urged to read out pseudowords. The present experiment is on perception: we presented spoken three syllable pseudowords to healthy participants and instructed them to: (a) remember as many items as they could (memory task), and (b) repeat the words (repetition task). Since regular word stress is assumed to make use of fewer cognitive resources than irregular word stress we expected participants to prefer one specific type of word stress in the memory task as well as in the repetition task. We found a preference for pseudowords stressed on the antepenultima (and penultima), supporting neither quantity-sensitive nor quantity-insensitive approaches, but an alternative approach connecting both approaches.