<p>There is often an expectation that the public should exhibit strong, if not unconditional, deference to science, as exemplified by campaigns urging people to “follow the science”. This paper argues for a more nuanced stance toward such deference. It contends that non-scientists may, at times, have good reasons to withhold or qualify the extent of their deference to science, particularly when there is evidence of inadequate gatekeeping within scientific communities. Gatekeeping can fail in different ways: it can be excessively restrictive, excluding theories, methods, or other elements that deserve consideration; or unduly permissive, allowing elements into scientific discourse that do not merit serious attention. In both cases, these failures undermine the pluralistic exchange of ideas that is essential to scientific inquiry and progress, and can give rise to four degenerative patterns of consensus formation or revision: premature consensus, belated consensus, premature consensus revision, and belated consensus revision. When non-scientists have sufficient reason to believe that such dysfunctions are present, it may be rational for them to adjust their level of deference accordingly.</p>