<p>The referential status of theoretical terms has been at the center of the tension between scientific realism and theory change. On the one hand, descriptivist theories cannot account for referential success across theory changes; on the other hand, the causal theory of reference is accused of making reference too easy and fails to account for referential failure. Building on the work of Psillos (Perception, Realism, and the Problem of Reference, 2012), I develop a refined version of the hybrid causal descriptivist theory in this paper. I first show how an exactly true causal description that survives through theory change is possible if we include the relevant experimental conditions and contexts. I then argue that the meaning of a scientific term fixed through a causal description consists of the referent only, which allows us to account for the complex linguistic behavior in the scientific community. At last, I draw from Donnellan’s (The Philosophical Review 75(3):281–304, 1966) distinction between attributive use and referential use of definite description to account for scientific terms that are introduced without being associated with any phenomenon.</p>