I consider the ‘inferentialist’ thesis that whenever a mental state rationally justifies a belief it is in virtue of inferential relations holding between the contents of the two states. I suggest that no good argument has yet been given for the thesis. I focus in particular on Williamson (Knowledge and its limits, 2000) and Ginsborg (Reasons for belief, 2011) and show that neither provides us with a reason to deny the plausible idea that experience can provide non-inferential justification for belief. I finish by pointing out some theoretical costs and tensions associated with endorsing inferentialism.