PREMISE OF THE STUDY: Numerous studies have treated the mass of a whole seed as an integrated unit, although the two components, seed kernel and seed coat, play different roles and suffer from different evolutionary selection pressures. In this study, we provided the first global-scale quantification of the relative biomass investments in seed coats and seed kernels. We tested the following hypotheses: there is a negative allometry between seed kernel mass and seed coat mass, and therefore seed coat ratio is negatively correlated with seed mass. METHODS: We compiled a global-scale dataset from the published literature, including 680 plant species from 420 genera and 108 families. The relationships between seed components were quantified using standardized major axes regression, ordinary least squares regression and phylogenetic independent analyses. KEY RESULTS: We found a weak, but significantly negative allometry between seed kernel mass and seed coat mass, which resulted in a negative relationship between seed mass and SCR. Similarly results were found after accounting for the phylogeny. CONCLUSIONS: The finding that smaller seeds invest more in protective tissues but less in stored reserves, may explain the general prediction that larger seeds suffer greater predation than smaller seeds. Furthermore, this weak allometry may also explain, at least in part, why so many studies failed to identify a clear pattern of the effect of seed mass on many ecological processes. Our study suggests that the allometry between the two seed components must be considered when evaluating the ecological significance and evolutionary history of seed mass. In compliance with data protection regulations, please contact the publication office if you would like to have your personal information removed from the database.