Two methods are commonly used to elicit production data for prosody research. The first, in which participants read out a series of written sentences, gives good control over what data are elicited. The second, in which participants perform a task designed to elicit the speech of interest (e.g., a Referential Communication Task), is suitable for studying speech in context. However, certain research topics require the combination of these qualities. We developed an elicitation paradigm, Scripted Simulated Dialogue, that (a) gives precise control over the data that are elicited and (b) is suitable for studying speech in context. In addition, it allows the researcher to control or manipulate the preceding discourse, whereas a Referential Communication Task provides discourse that may be analysed afterwards. The paradigm simulates a series of short dialogues, in which the participant reads her text from a screen and the 'interlocutor' is a recorded voice. The participants are not made aware of which speech turn in the dialogue contains the target sentence. We illustrate how Scripted Simulated Dialogue may be used to manipulate the context and make the E-Prime script available to other researchers.