The sensitivity of pollen as an indicator of elevation in neotropical lowland and Andean forests was measured using modern pollen samples collected from moss-polsters along a transect between 340 m and 3530 m elevation and from surface sediments in lowland swamps (240 m) of Madre de Dios, Peru. A blind study, using samples collected from the same transect in the following year, provided a test of reproducibility. The results show (1) clear elevational distribution patterns and (2) the ability of calibration data to predict the altitude of the blind samples. Characteristic associations of pollen taxa are found under differing hydrologies and elevations. The floodplain pollen assemblages are characterized by abundant Mauritia, Sloanea, Ficus, Iriartea and Arecaceae pollen types. At higher elevations, these lowland types decrease or are absent. Alchornea, Urticaceae/Moraceae, Bignoniaceae and Cecropia are dominant components of the pollen rain of the low-elevation zone (<1000 m). Acalypha, Alchornea, Cecropia, Rubiaceae and Urticaceae/Moraceae are important between 1000 m and 1600 m elevation. Pollen of Hedyosmum, Alnus, Poaceae and Combretaceae/Melastomataceae are abundant between 1600 m and 2000 m. Cecropia pollen dominates samples from low- to mid-elevation disturbed forests. Alnus pollen is most abundant, and Poaceae becomes rare, between 2000 and 2700 m. At high elevations above 2700 m, Asteraceae, Poaceae, Polylepis, Muehlenbeckia-type and Myrsine pollen are dominant. Statistical analysis of the data set using Detrended Correspondence Analysis (DCA) shows a precise correlation between community composition and elevation. The DCA axis 1 values are strongly correlated with sample elevation, exhibiting a linear relationship (r2=0.904). The results provide an estimate of the sensitivity of pollen analysis in the Neotropics as a proxy for measuring elevation and, by inference, temperature.