Our goal was to investigate whether Lombard Italian speakers reliably use prosodic features to distinguish canonical wh-questions (CQs) and non-canonical rhetorical wh-questions (RhQs). We hypothesized that such speakers could identify each question type by using prosody alone, and we designed both a perception and a production task to test this hypothesis. The former consisted of a forced-choice identification task with 32 target stimuli. To elicit the production data, participants performed a discourse completion task with the same number of stimuli as in the perception study. Utterances were intonationally labeled and acoustically analyzed for pitch change (prenuclear pitch accent and nuclear contour), initial and final pitch, and relative duration. Results revealed that participants were highly accurate at identifying RhQs but less so at identifying CQs. Results of successful productions showed that initial but also final cues were systematically used, with pitch level being lower in RhQs than in CQs, and final boundary tones reliably differentiating RhQs from CQs. Finally, RhQs were significantly longer than CQs. We conclude that speakers do rely on both intonation and duration cues to distinguish CQs from RhQs.