The peculiar biogeography of the so-called "Lost World", i.e. the summits of sandstone tableaux ( tepuis) in the Neotropical Guyana region of Venezuela, has generated a debate regarding the factors that are thought to account for modern vegetation patterns in the region. Some argue that plant communities on these high-elevation summits reflect a long history of evolution in isolation, while others surmise that there has been substantial biotic interchange with the surrounding lowlands during glacial times. Until now, these apparently competing hypotheses have not been tested using palaeoecological methods. I used pollen analysis of Quaternary sediments and documented past vertical migrations of vegetation in response to climate changes, which supports the second hypothesis. Physiographical analysis, however, shows that about half the tableaux summits are too high for their flora to have reached the lowlands during the last glaciation, suggesting that a portion of the tableaux vegetation has always experienced some degree of biotic isolation. Thus, a component of the summit vegetation evolved in isolation, whereas other taxa experienced interchange, as reflected in endemism patterns among the tableaux summits. Biogeographical patterns on the summits are the result of complex evolutionary processes. The two hypotheses invoked to explain the vegetation patterns are not mutually exclusive, but instead complement one another.