In many languages, rhetorical questions (RQs) are produced with different prosodic realizations than string-identical information-seeking questions (ISQs). RQs typically have longer constituent durations and breathier voice quality than ISQs and differ in nuclear accent type. This paper reports on an identification experiment (Experiment 1) and an EEG experiment (Experiment 2) on German wh-questions. In the identification experiment, we manipulated nuclear pitch accent type, voice quality and constituent duration and participants indicated whether they judged the realization as ISQ or RQ. The results showed additive effects of the three factors, with pitch accent as strongest predictor. In the EEG experiment, participants heard the stimuli in two contexts, triggering an ISQ or RQ (blocked). We manipulated pitch accent type and voice quality, resulting in RQ-coherent and ISQ-coherent stimuli, based on the outcome of Experiment 1. Results showed a prosodic expectancy positivity (PEP) for prosodic realizations that were incoherent with ISQ-contexts with an onset of ∼120ms after the onset of the word with nuclear accent. This effect might reflect the emotional prosodic aspect of RQs. Taken together, participants use prosody to resolve the ambiguity and event-related potentials (ERPs) react to prosodic realizations that do not match contextually triggered expectations.