For recalcitrant seeds, mortality curves of germination versus water content typically imply a wide range of desiccation sensitivities within a seed population. However, seed to seed differences in water content, during desiccation, may confound our interpretation of these mortality plots. Here, we illustrate this problem for two batches of Vitellaria paradoxa (Sapotaceae) seeds collected in 1996 and 2002. Whole seeds were desiccated to various target water contents (TWCs) using silica gel. During desiccation, smaller seeds in the population dried most rapidly. Consequently, there was a significant linear relationship between whole-seed water content and seed mass during the drying process. In addition, following desiccation to low TWCs, only the largest seeds in the population retained viability. Taken together, this suggests that the larger seeds survived, not as a consequence of great relative desiccation tolerance, but as a result of taking longer to desiccate. Subsequently, the critical water content (CWC) for viability loss was calculated, based on the assumptions that in the seed population whole-seed water content during desiccation was normally distributed and the smallest, and hence driest, seeds were killed first. Using this approach, the driest seeds in the population that were killed, at each TWC, were always below a single CWC (c. 20% and 26% in 1996 and 2002, respectively). In subsequent experiments the effect of seed size variation on the response to desiccation was confirmed by conducting desiccation screens on seeds sorted into two discrete size classes, i.e. the seed-lot heterogeneity in mass was reduced. Using this approach, the mortality curves had a steeper slope. Furthermore, data for 24 tropical tree species from the Database of Tropical Tree Seed Research (DABATTS) revealed that seed lots with less variability in mass had steeper mortality curves. Thus, taken together, the data suggest that, at least for whole seeds, the wide range of desiccation sensitivities typically inferred is an artefact of seed to seed variation in mass, and hence water contents, during drying.