Alternative stable-state theory (ASS) is widely accepted as explaining landscape-level vegetation dynamics, such as switches between forest and grassland. This theory argues that webs of feedbacks stabilise vegetation composition and structure, and that abrupt state shifts can occur if stabilising feedbacks are weakened. However, it is difficult to identify stabilising feedback loops and the disturbance thresholds beyond which state changes occur. Here, we argue that doing this requires a synthetic approach blending observation, experimentation, simulation, conceptual models, and narratives. Using forest boundaries and large mammal extinctions, we illustrate how a multifaceted research program can advance understanding of feedback-driven ecosystem change. Our integrative approach has applicability to other complex macroecological systems controlled by numerous feedbacks where controlled experimentation is impossible.