Aim: Higher-elevation areas on islands and continental mountains tend to be separated by longer distances, predicting higher endemism at higher elevations; our study is the first to test the generality of the predicted pattern. We also compare it empirically with contrasting expectations from hypotheses invoking higher speciation with area, temperature and species richness. Location: 32 insular and 18 continental elevational gradients from around the world. Methods: We compiled entire floras with elevation-specific occurrence information, and calculated the proportion of native species that are endemic ('percent endemism') in 100 m bands, for each of the 50 elevational gradients. Using generalized linear models, we tested the relationships between percent endemism and elevation, isolation, temperature, area and species richness. Results: Percent endemism consistently increased monotonically with elevation, globally. This was independent of richness-elevation relationships, which had varying shapes but decreased with elevation at high elevations. The endemism-elevation relationships were consistent with isolationrelated predictions, but inconsistent with hypotheses related to area, richness and temperature. Main conclusions: Higher per-species speciation rates caused by increasing isolation with elevation are the most plausible and parsimonious explanation for the globally consistent pattern of higher endemism at higher elevations that we identify. We suggest that topography-driven isolation increases speciation rates in mountainous areas, across all elevations, and increasingly towards the equator. If so, it represents a mechanism that may contribute to generating latitudinal diversity gradients in a way that is consistent with both present-day and palaeontological evidence.

Summary1. Priority question exercises are becoming an increasingly common tool to frame future agendas in conservation and ecological science. They are an effective way to identify research foci that advance the field and that also have high policy and conservation relevance. 2. To date, there has been no coherent synthesis of key questions and priority research areas for palaeoecology, which combines biological, geochemical and molecular techniques in order to reconstruct past ecological and environmental systems on time-scales from decades to millions of years. 3. We adapted a well-established methodology to identify 50 priority research questions in palaeoecology. Using a set of criteria designed to identify realistic and achievable research goals, we selected questions from a pool submitted by the international palaeoecology research community and relevant policy practitioners. 4. The integration of online participation, both before and during the workshop, increased international engagement in question selection. 5. The questions selected are structured around six themes: human-environment interactions in the Anthropocene; biodiversity, conservation and novel ecosystems; biodiversity over long time-scales; ecosystem processes and biogeochemical cycling; comparing, combining and synthesizing information from multiple records; and new developments in palaeoecology. 6. Future opportunities in palaeoecology are related to improved incorporation of uncertainty into reconstructions, an enhanced understanding of ecological and evolutionary dynamics and processes and the continued application of long-term data for better-informed landscape management. 256-26750 priority research questions in palaeoecology 257 7. Synthesis. Palaeoecology is a vibrant and thriving discipline, and these 50 priority questions highlight its potential for addressing both pure (e.g. ecological and evolutionary, methodological) and applied (e.g. environmental and conservation) issues related to ecological science and global change.