Abstract The largely unsolved problems in the theoretical analysis of differentiation and ageing involve a substantial component of linguistic (semantic) difficulties. Some of these are simple traps of ambiguity, resulting from metaphorical or analogical employment of established terms—for example, “terminal differentiation” (loss of division potential in vitro) as a borrowing from “differentiation” as used by developmental biologists, or “commitment” by analogy with “determination”. Some difficulties represent a failure to adopt (at least provisionally) an operational (empirical) view—for example, failure to ask what is the nature of the evidence for the view that a fertilized ovum is totipotential, or to scrutinize the evidence for the view that cells “terminally differentiated” in vitro in a conventional medium are in fact moribund under all conditions, or to examine more closely the view that the differentiated state and the cycling state are mutually exclusive. With respect to the problem of ageing, we review some of the critical experiments on ”terminal differentiation” or “clonal senescence”. We then proceed to consider some of the models that have been proposed, including a molecular model proposed by the author which appears to overcome some of the objections to other models. Some of the models exemplify the results of what are ultimately semantic vices. The problems with which these remarks began should indeed yield to the immense and novel resources of molecular biology. But the development of complete analyses demands not only good luck and delicate technique but also critical semantic clarity and severity. Given the best tools, we shall solve major theoretical problems only if we understand quite fully what problem it is that we are trying to solve—and the history of science illustrates that this is not as elementary a matter as it sounds. For the working scientist semantics (and indeed all philosophy of science) is not an indulgence or a frill, but a basic technical resource.