Interdisciplinary research is endorsed by every funder and disliked by a good many of the people who evaluate it, and the tension is not hypocrisy. The problems considered most important — climate, health inequality, the effects of new technologies — do not respect departmental boundaries, and work that crosses them is genuinely necessary. Assessment, however, is conducted by disciplinary specialists, each of whom sees a portion of a project and, quite reasonably, judges it by the standards of their own field, where it will usually appear thin. A proposal that satisfies two disciplines fully is rare; one that satisfies each of them partially is common and scores badly. The incentive structure therefore rewards the rhetoric of interdisciplinarity and penalises the practice, which is roughly what is observed. Nobody in the chain is behaving badly, which is what makes the pattern durable.