Although accurate reconstructions of Amazonian palaeoclimates are central to understanding the distribution and history of Neotropical biodiversity, current reconstructions based on proxy data are discordant and subject to intense debate. We review some current thinking in Amazonian climatology and incorporate some new ideas in an attempt to explain the apparently contradictory records of palaeoprecipitation from across the Amazon Basin. We suggest that palaeoecologists need to recognise that organised convective radiation, the process that induces most wet season rainfall in Amazonia, should be treated as a phenomenon related to, but separate from, the passage of the inter-tropical convergence zone (ITCZ). Decoupling ITCZ migration from larger convective activity provides a mechanism to account for observed palaeoclimatic heterogeneity. Patterns of observed precipitation change are consistent with the long-term persistence of closed forest across much of Amazonia, indicating that the greatest changes in precipitation during the last glacial maximum came during the wet season, which would have little negative impact on forest extent. Neotropical cooling at the last glacial maximum (LGM) is widely accepted, although the estimates of that cooling range between 1°C and > 5°C. In answering the basic question ‘What is meant by cooling?’ we observe that interhemispheric ice mass asymmetry may have caused cooling to be manifested differently according to location. A terrestrial cooling of ca. 5°C, as well as radiative cooling and event-based cooling combined to induce vegetation change. Probably, both absolute temperature and mean monthly minima were reduced by polar air incursions in the northern Neotropics. However, in the southern Neotropics, mean monthly minima were reduced by more frequent incursions of Patagonian air masses, but absolute minima may have been largely unchanged. Copyright © 2004 John Wiley & Sons, Ltd.