Physicalism is usually understood as the claim that every empirical entity is or is determined by physical entities. The claim is however imprecise until it is clarified what are the physical entities in question. A sceptical argument in the form of a dilemma tries to show that this problem of formulation of physicalism cannot be adequately met. If we understand physical entities as the entities introduced by current physics, the resulting claim becomes most probably false. If we instead understand physical entities as those entities introduced by some future ideal physics, the claim then becomes indeterminate in content. Both horns seem equally bad. In the first part of the paper, I survey the strengths and weaknesses of different proposed solutions to this problem of formulation. In the second part, I lay out a new formulation of physicalism, partly based on a mereological principle, which overcomes the dilemma, and argue that it is a correct formulation of physicalism to the extent that it rules out clear antiphysicalist scenarios and is compatible with clear physicalist scenarios.