Sociology of the professions has faced difficulties in delineating and defining the core of its object of study, the concept and phenomenon of profession. Setting out from the so-called French epistemological tradition, this article tries out a new possibility. It is argued that if there are trans-historical elements in professions, they are found at the cognitive side, as a certain reasonably invariant cognitive structure. The structure implies that a specific type of (scientific) knowledge can be coupled with professional practice, that is, know-why is linked to knowhow. This is what the notion of profession primarily should be based on. Science, profession, and "object" together constitute a "truth-regime" in Foucault's sense. On the other hand, the social side of professions is historically variable. Indeed, the social attributes of a profession changes with social transformations and external interests, with its position in Bourdeauian professional fields, and with its relations in "the professional complex." The article concludes with a suggestion for a new definition of professions.