A curious problem arises in connexion with the notion of the participant observer, a problem partly ethical and partly methodological. It seems not to have been clearly seen and stated, although solutions to it exist — in practice, as it were. The problem arises like this. Standard accounts of the method of participant observation require, I would argue, an anthropological observer to be both a stranger and a friend among the people he is studying. Yet one person cannot be a stranger and a friend at the same time: the roles are mutually exclusive. This being so, it is a fortiori impossible to play either role in integrity while trying to combine them, with the result that an uneasy compromise is liable to be forged.