Humans are increasingly altering natural systems and at the same time desire to foresee the consequences and estimate future change. Many natural systems do not respond linearly to change of a forcing variable, but show delayed, threshold or non-reversible responses, which are difficult to predict. The terrestrial vegetation represents an important feedback mechanism for regional and global climate change and itself comprises systems with nonlinear responses. System change often occurs on long time scales, which makes it difficult to study, particularly where long-lived plants are involved. Here palaeoecological techniques and existing datasets provide a way to gain understanding of the resilience and potential tipping points of the terrestrial vegetation on time scales of decades to millennia. In my research supported by the Heisenberg program of the of the German Research Foundation (Deutsche Forschungsgemeinschaft) I analyse information on late Quaternary vegetation change as documented in pollen data to explore state changes in terrestrial vegetation, ecosystem resilience and threshold responses and uncover tipping points.