This paper argues that studies of middle leaders have focused too much on their functions and characteristics, taking insufficient account of the influence of social and political contexts on leaders' choices of actions. Use of the analytical framework of Communities of Practice (Wenger, 1998) can address this oversight, but it pays insufficient attention to the dynamic processes by which people interact and make meaning of their interactions. The paper concludes that combining existing theories of power with the framework of Communities of Practice gives a more complete view of middle leaders at work than current approaches.