It is often suggested that to make progress in consciousness science we need a theory of consciousness—one that tells us what consciousness is and what kinds of systems can have it. But this may be putting the cart before the horse. There are currently a wide range of very different theories all claiming to be theories of consciousness. How are we to decide between them if we do not already know which systems are conscious and what they are conscious of? (...) In this paper I aim to do two things. First, I explain why a theory‐driven approach to consciousness science, even one sensitive to Lakatosian norms and updating in a Bayesian manner, may lead to divergence rather than convergence. Second, I draw on the history of thermometry and natural kind reasoning to sketch a theory‐neutral route towards progress in consciousness science. (shrink)