This article examines some issues facing the fieldworker attempting to observe and record “natural” conversations, and it reconsiders the long-held sociolinguistic notion of the observer's paradox by recasting it within Allan Bell's framework of audience design theory. Style shifting in observed and recorded speech events is seen to be influenced by speakers' perception of the fieldworker's social role, and by the fieldworker's participant role in the speech event.An earlier and much briefer version of this article was presented at the 2002 Berkeley Linguistics Society conference. I am most grateful to Barbara Johnstone, Jane Hill, and two anonymous reviewers for their insightful and helpful comments. Funding for this fieldwork was provided by grants from the International Research and Exchange Board, the Academy for Educational Development, and the Berkeley Program in Post-Soviet Studies.