Researchers have persisted in framing leadership as the driver for change and performance improvement in schools despite convincing theoretical commentary that proposes leadership as a process of reciprocal interaction. Although conceptualizing leadership as a reciprocal process offers leverage for understanding leadership effects on learning, methodological constraints have limited empirical tests of this model. This report focuses on understanding the contribution that collaborative leadership and school capacity for improvement make to growth in student learning in elementary school mathematics. We compare unidirectional and reciprocal-effects models focusing on change in leadership and learning in 195 elementary schools over a 4-year period. The results support the efficacy of a reciprocal-effects model that conceptualizes leadership within a changing, mutually reinforcing system of organizational relationships. Copyright © 2010 The University of Chicago.