<p>Legitimate gatekeeping in science that is based on the standards constitutive of science presupposes a distinction between science and non-science and thus a solution to the demarcation problem. However, the latter has often been considered as unsolved, if not unsolvable. This paper aims to trace the consequences of discussions about the demarcation problem for gatekeeping. I argue that an approach that relativizes scientific status to time, research field, and level allows for a meaningful demarcation and thus gatekeeping. Particular emphasis is put on the last point, i.e. the level comprising the units to which we wish to attribute scientific status, e.g. ideas, experiments, or publications. I conduct an empirical analysis to identify the kinds of things that are called “scientific”, “unscientific”, or “pseudoscientific” in ordinary talk. To classify the results, I propose a systematic coding scheme based on the idea that science is a practice that leads from starting points to results. I consider a few units, such as ideas and experiments, and discuss the criteria to be applied for these units and the consequences for gatekeeping. Such a unit-sensitive account has the advantage of being more lenient concerning the starting points of research and more stringent regarding the results of scientific inquiry.</p>