Physical defence investment in seeds varies greatly among plant species and is associated with many potential factors. Exploring the factors explaining the interspecific variation in physical defence has long attracted particular attention in both ecology and evolution studies. However, the relative importance of the factors has not yet been quantitatively evaluated, which may lead to the misunderstanding of the main driver generating such interspecific variation.

Here, by compiling a global database of the seed coat ratio (SCR), a proxy of seed physical defence, for 1,362 species, we provided the first quantification of the relative explanations of six factors that have been commonly considered to be associated with the interspecific variation in SCR: seed mass, seed desiccation response (desiccation-sensitive vs. desiccation-tolerant), seed dormancy (nondormant, physical dormant or other dormant types), growth form (herbaceous vs. woody), fruit type (dry vs. fleshy), and climate (19 bioclimatic variables representing temperature and precipitation). Using partial R² calculation that can estimate the reduction in explained variance of a given factor after removing it from a full model containing all the other factors, we disentangled the relative contribution of each factor to explain variation in the SCR.

Seed desiccation response was the most important factor associating with SCR variation, followed by seed dormancy, with a partial R²_lik value of 9.70% and 1.41%, respectively. Plant growth form, seed mass, climate and fruit type showed little power to explain SCR variation, although they are often cited in the literature as potential factors.

Our results highlight the important roles of seed desiccation response in shaping interspecific variation in the SCR. These findings provide a detailed interpretation of ecological patterns underlying the physical defence mechanisms and allocation strategies of plant seeds.

Data on the SCR and seed mass were collected from the literature published up until June 5, 2022. We used ((diaspore$ or “seed* coat$” or (seed$ and (kernel$ or reserve$)) or (seed$ and “embryo* endosperm$”)) and (ratio$ or proportion$ or fraction$ or tissue$ or percent* or weigh* or mass*)) or (seed$ and (defen* or protect*) and (physical* or mechanical*)) as search terms in the ISI Web of Science and further restricted our search to be consistent with the ‘Study Field Categories’ of Wu et al. (2019), with the additional category ‘anatomy morphology’. Google Scholar was also searched with the same keywords to expand our dataset. Finally, 9,322 journal papers written in English were search out. We first screened all the papers based on the titles and abstracts and excluded 6,857 studies that were not relevant to our focused question. For the remaining 2,465 papers, we screened the full texts and finally yielded 85 papers containing available SCR data collected from 203 sampling sites. The nomenclature of the species was checked against the World Flora Online database (WFO, v.2022.07; www.worldfloraonline.org) using the function ‘nameMatch’ in the R package ‘U.Taxonstand’ (Zhang & Qian, 2023), and 12 species were mismatched which were further checked against the Plants of the World Online (POWO, https://powo.science.kew.org/). In total, 1,758 records of the SCR were collected, comprising 1,362 species from 724 genera and 150 families.