We report on the results of an experiment designed to test the phonological properties of declarative questions in American English. Previous work is not conclusive in whether a typical melodic contour exists for declarative questions and whether it is crucial for their recognition in spontaneous dialogues. We conclude that speakers are equally good at recognizing declarative questions with and without having access to prosodic information. However, certain contours are taken to be more likely to signal questions, especially the rising contour described in Gunlogson (2001).