This paper is about the idea of a moral distinction between doing harm and allowing harm to occurr. It discusses, and developes a general argument against, the version of the distinction often described as counterfactual, which I characterize in terms of making a moral difference between different ways of causing harm (in contrast to, e.g., the version famously discussed by Jonathan Bennett). The gist of the argument is that all variants of this version of the doing-allowing idea would have to (...) make factors out of the agent's control into wrong-making characteristics and that, therefore, this version violates a plausible extension of the principle that 'ought' implies 'can'. (shrink)