We investigated the potential of soil pollen analysis to provide information about recent land use and vegetation history in the lowland humid tropics. Our initial work at the La Selva Biological Station in Costa Rica revealed that pollen was surprisingly well preserved in mineral soil, with pollen concentrations in the upper 10 cm of our soil cores as high as values for some lake sediment samples. Pollen percentages in our initial core samples showed down-core variations that appeared to relate to what we knew about past episodes of forest clearance, pasture establishment, and secondary succession at our sampling sites. We carried out further tests of soil pollen analysis by collecting and analyzing additional soil cores from the La Selva Biological Station. Based on these new results, we conclude that soil pollen analysis in the lowland humid tropics does not hold the promise suggested by our initial work. Most pollen in our soil cores may be only a few decades old, and rapid downwash and/or bioturbation may prevent the development of a stratigraphy at some sites. Poor reproducibility at a scale of meters suggests that where a distinct pollen stratigraphy exists, it may be highly localized, as soil pollen profiles collected under tropical forest canopies may sample very small areas. Alternatively, small (1-cc) samples may be unrepresentative of soil horizons. A large number of samples will be needed in either case to characterize the land use and vegetation history of even a fairly small area. However, extreme locale-specificity could be an advantage for detection of evidence of the past occurrence of a particular plant taxon at a given location.