One battle line in the debate between constructivism and realism has been drawn between realistically inclined philosophers of science, on the one hand, and constructivist sociologists of science, on the other. Both camps have seemingly agreed that realism and constructivism are incompatible with one another. This paper investigates the supposed incompatibility of constructivism and realism from the perspective of scientific representation. The argument shows how constructivists, in their critique of scientific representation, stumbled over the constructed nature of their own accounts of science. This problem of reflexivity made the constructivists realize that any extreme constructivist claims tend to be self-defeating. That the problem of reflexivity was felt to be so acute by many constructivists in the field of science and technology studies shows, I argue, that in their wholesale attack on representation these "reflexivists' (and some representation-hostile neo-pragmatists alike) were themselves relying on a rather stringent notion of representation. Yet there are many ways to approach scientific representation, as well as many forms of constructivism, and thus constructivism and realism need not be irreconcilable. The pragmatic accounts of scientific representation, in particular, can fruitfully be combined with a more modest constructivist programme.