The Collaborative Forest Landscape Restoration Program (CFLRP) has created the conditions in which collaborative adaptive management (CAM) approaches in forest restoration can be tested.  In theory, CAM engages stakeholders to collectively: define forest landscape restoration goals, assumptions, uncertainties, and options; develop and implement a science-based monitoring strategy to assess restoration effects; deliberate the effects of actions on goals, assumptions, and uncertainties; and recommend changes in goals, assumptions, and actions based on the “best available science”. In reality, since people are involved, things get messy, fast. My presentation addresses two questions: 1) In what ways do multi-stakeholder group dynamics interact with federal agency institutional and organizational processes?; and 2) To what extent do these interactions facilitate or frustrate collaborative adaptive management on federal landscapes? Drawing on experiences and perspectives of colleagues involved with the Colorado Front Range CFLRP, the Four Forest Restoration Initiative, and the Uncompahgre Plateau CFLRP, I discuss social group dynamics and institutional/organizational factors affecting CAM across these cases. A key take-away is that in order for CAM to realize its potential for integrating science into forest restoration decision-making, CAM participants need to be intentional and reflexive about social group dynamics and institutional/organizational processes.