This paper presents the findings of an experiment testing native speakers' intuition about the stress of disyllables. The experiment has involved reading tasks where nonce words were embedded. The 53 nonsense words display different phonological and morphological structures forced by the spelling and allow us to test the validity of hypotheses based on dictionary data. The results of the experiment prove rich enough to be comparable with dictionary and corpus studies. Speakers seem to have strong feelings about the prosody of our tokens since they agree on Ÿ of words. Quantity-sensitivity indeed plays a determining role for verbs and surprisingly enough, to some extent even for nouns. When quantity does not help, other factors come into play: the quality of the final consonant or the morphological make up.