"Curricula-in-action" generally differ from "official" curricula. That is particularly true of clerkship curricula because the practising doctors who supervise medical students' clinical activities are only secondarily educators. Clerkship education is evaluated, however, according to benchmarks set by official curricula. As a result, clerkship evaluations are important points of contact between clinical teachers and medical schools. We reasoned that an evaluation instrument is part of a medical school's official curriculum discourse and clinical teachers' reactions to it are a discourse of curriculum-in-action. We set out to answer the questions: What are clinical teachers' discourses of curriculum-in-action and how do they relate to an official curriculum discourse? Nineteen clerkship placement leads from two hospitals contributing to a single undergraduate medical programme participated. The evaluation instrument was the Manchester Clinical Placement Index, for which validity evidence has been published. Respondents were asked to say how they would react to junior students giving their placements low or high scores for each of 12 items from the Index. After transcription, we conducted a critical discourse analysis (CDA) of their audio-recorded answers. We purposefully selected the six items that elicited the widest spectrum of responses for analysis because quantity of material can compromise the quality of CDA. A dominant discourse of curriculum-in-action defined how teachers should "really" teach and junior students should learn. It deconstructed the need for teachers to be present when students performed clinical tasks because teachers' role was to give critical feedback on case presentations that were coincidental to clinical care. It positioned students at the bottom of a power hierarchy so they had to "struggle" to be taught. It placed respondents in a powerful position relative to "the hospital" and "the university", though there were tensions between respondents, patients, and nurses. Respondents dismissed criticism that was invalid according to their curriculum-in-action, which included most items in an evaluation instrument. There was a contrasting, non-dominant discourse of responding reflectively to feedback, which generated realistic ways of improving students' learning. The strength of respondents' emotions shows just how committed doctors are to students' learning. The strength of their expressions of power, however, explains why many of them teach in their own way rather than according to official curricula. Changes to clinical curricula, our findings suggest, will not be successful unless they are carefully negotiated with practising doctors.