While psychoanalysts claim that psychoanalysis is a science, many external critics denounce it as an unscientific system of untested (and in a Popperian view: untestable) beliefs. In part, this may be the effect of what Freud called the “resistance against psychoanalysis” (which cannot be cured simply with arguments). Beside that, however, the discussion of the epistemological status of psychoanalysis is strained by misunderstandings and misconceptions on both sides. The author hopes to improve this situation by placing psychoanalysis among those theories which deal with objects that are heterogeneous, i.e., always different, changing and developing. The shape of psychoanalytic theories is a result mainly of the functions they have to fulfil, and to do so, they must produce specific capacities which also involve certain risks. For instance: they easily divide into a plurality of models, they operate with open concepts and they tend to retain the form of “work in progress “.