In the constructivist literature on student knowing in science and mathematics, what students say and do in classrooms and interviews is said to be the result of their conceptions. These conceptions are thought to be "mental models' and "internal representations' that are the outcome of constructive work. But does this have to be the case? Do students have to have conceptions when they talk about certain topics? Moreover, if we were to accept that there are mental models driving what someone says at some point, how would talk look like when a learner is between two mental models? In this chapter, I draw on interview data to show that people do actively participate in conversations without ever having thought or talked about a topic before, which means, that they could not have constructed a mental model. Rather, by participating in conversations, children learn to speak—without having to learn grammar, without having to learn conceptions—and from the way language is used, answers to questions can be derived even without prior knowledge-building processes to have occurred. I develop a position grounded in Bakhtin (Vološinov), Heidegger, and Derrida.