<p>According to the causal theory of reference, the references of (some) natural kind terms are fixed in baptisms. To wit, a so-baptized kind term refers to those things that share a certain inner constitution with the sample used in the baptism. I argue that this is incompatible with the claim that natural kind terms are <i>open textured</i>, i.e. that semantics can underdetermine reference. The two views, fixed reference and open textured reference, entail competing claims about the course of science. By examining an episode from the history of science, the discovery of isotopes, I conclude in favor of open texture for natural kind terms.</p>