Districts play a key and relatively unexamined role in distributed leadership research. This article explores how leadership was distributed through a district structure designed to improve the quality of teaching by improving the quality of teacher evaluation. It examines peer assistance and review, a policy designed to address the key problems of traditional teacher evaluation by allowing administrators and teacher leaders to share accountability for evaluation processes and decisions. The article presents data from a peer assistance and review program in one urban district, detailing how the program distributed accountability for teacher quality across the district organization. The article extends previous work on distributed leadership by showing how the design of shared tasks can effectively distribute accountability. The article also extends previous work on distributed leadership by elucidating the democratic effects of that distribution. As such, the article addresses questions of instrumentality (i.e., how can district leaders design and implement a better teacher evaluation system?) and agency (i.e., what are the political implications of distributing formal authority for teacher evaluation out of the hands of administrators and into the hands of teachers?). (Contains 2 figures and 15 footnotes.)