Seed size and shape, measured as the variance of the three main dimensions, have been proposed as good indicators for predicting seed persistence. We tested whether these variables were robust predictors of seed persistence in the soil for 58 abundant herbaceous species, primarily annuals, in grass and scrubland of central Spain. Seed persistence was estimated from data on germinable seed banks, while seed weight and shape were measured using fresh seeds collected in the study area. There was a significant tendency for species with persistent seeds to have smaller seeds than species with transient seeds. Seed shape was not, however, related to persistence and we did not find any clear seed weight/shape threshold for predicting persistence. The binary logistic model of seed bank type as a function of seed weight was significant and explained 67% of total variability. Supplementary information about dormancy, environmental conditions of habitat, predation and attack by pathogens has to be used to elaborate more accurate general predictive models of seed persistence.
