<p>Performativity is the capacity of scientific representations (such as models, theories, predictions, or classifications) to alter the phenomena they are supposed to represent. Because one and the same representation may alter its target a lot, a little, or not at all, I argue that we should conceptualize performativity as a function of a representation and its <i>performative power</i>, which depends on its reach, acceptance, and its relevance to people. Using this framework, I then argue for a re-evaluation of performativity. Because performative effects can impair scientists’ ability to model, classify, explain, or predict (e.g. by steering outcomes away from those predicted) and they raise concerns about the legitimacy of science influencing the social world, performativity is often viewed as a threat to science. In contrast, I argue that we shouldn’t be worried about performativity as such, but rather, about <i>concentrated performative power</i>, that is, that the representations issued by individual scientists, or groups of scientists with uniform views, gain a power that is not in line with their epistemic credentials. To eradicate such power concentrations and to secure science’s proper role in a democratic society, we may sometimes hope for more, rather than fewer, performative effects.</p>