In recent decades a number of authors have relied on the multiple realizability argument in order to reject methodological individualism. In this paper I argue that this strategy results in serious difficulties and makes it impossible to identify social entities and phenomena.