In this paper I question the view that realism must delineate the basic ontological furniture of the world rather than giving arguments in semantic or epistemic terms for the existence of a mind-independent world. I call this view of stating and defending realism the Ontological Defence of Realism (or ODR) and take Devitt's account of realism as a paradigmatic case of ODR. I argue that ODR cannot block ' verificationist antirealism' because the specific (physical) nature of what exists is not (...) enough to secure the mindindependence of what exists and, additionally, every element purported to achieve this, it compromises seriously the idea that realism is primarily an ontological issue. I also stress that ODR is in tension with a plausible realist insight namely the priority of the world over our theorizing. Because of this tension ODR weakens realism in several domains with no good reason. Specifically, I argue that in these domains ODR faces a dilemma: either to reject realism or to take realism to be dependent on a reductive account violating the realist insight. My point is that we should keep realism and ontology distinct and that compliance with the realist insight initiates a better strategy for the defence of realism. I address three possible objections thereby further clarifying my point. I conclude by presenting my view about the relation between ontology and realism. (shrink)