Narrativist philosophy of history popularized a constructivist view arguing that historical works are not simple depictions of the past but rather are complicated constructions. According to narrativists, historians engage in a creative activity of proposing points of view, interpretations, or theses on the past that do not straightforwardly reflect past events. Although this is a broad constructivist view behind the theorizing of a number of authors, it is possible to distinguish within this line of thinking at least two general proposals about how to understand historical works. The first, defended for instance by Frank Ankersmit, maintains that historical works are representations of the past. Nevertheless, these representations are not descriptions of past events—they represent in a special way that could be characterized via a certain complexity, indirectness, holism, and a retrospective approach. The second proposal, presented in the work of Paul Roth and Jouni-Matti Kuukkanen, discards the epistemic framework of representation and understands historical works as the outcome of specific practices. In this article, I focus on these two constructivist versions, which could be called representationalism and non-representationalism. I analyze their crucial features, discuss their differences, and dispute the accusation that the latter view formulates an extreme theory of history. I argue that non-representationalism does not erase the notion of the past from its account of history; it merely attributes to the past a function different from the one it has within the representationalist paradigm.