David Bloor has claimed that Wittgenstein is best read as offering the beginnings of a sociological theory of knowledge, despite Wittgenstein’s reluctance to view his work this way. This leads him to dismiss Wittgenstein’s many self-characterizations as mere ‘prejudice’. In doing so, however, Bloor misses the import of Wittgenstein’s work as a ‘grammatical investigation’. The problems inherent in Bloor’s interpretative approach can be discerned in his attitude toward Wittgenstein’s use of imaginary scenarios: he demands that they be replaced by real natural history and real ethnography. This demand is misplaced. The very self-characterizations Bloor dismisses show how imaginary scenarios have a place in his philosophical project simply by being imagined. Three examples are examined and presented in such a way as to make Bloor’s demand for replacement increasingly more difficult to comprehend: while in the first case, the demand seems simply beside the point, in the second and third cases, it becomes difficult to say just what would count as replacements. Wittgenstein’s imaginary scenarios are thus best read not as suggestions for further empirical research, but as devices to aid in recovering the naturalness and familiarity of our concepts, which is precisely what one would expect from them as part of a grammatical investigation. © 1994 Taylor & Francis Group, LLC.