Seed mass is correlated with a number of other plant traits, including dispersal mode, growth form and specific leaf area. Specific leaf area is the main determinant of potential relative growth rate and an indicator of the site quality to which a species is adapted. The relationships with dispersal mode and growth form have consistent form in five datasets from three continents, and each account for about 20-30% of variation in log seed mass. Thus, there is also very substantial variation within growth form and dispersal categories. Much, but not all, of the 20-30% is associated with shifted family composition between growth forms or dispersal modes. Experiments have shown that seedlings of larger-seeded species are better able to survive hazards including deep shade, drought, physical damage and the presence of competing vegetation. If there is a common mechanism under these different hazards, it seemingly must be a `reserve effect', whereby during deployment and early growth larger-seeded species hold a bigger percentage of seed reserves uncommitted to seedling structure and available to support respiration or repair damage. A reserve effect has not yet been demonstrated directly. It remains possible that different mechanisms operate under different hazards. Under a reserve effect, advantages of larger seed size should be temporary, and temporary advantage has indeed been observed with regard to seedling survival under dense shade. Although larger seed mass confers benefits on seedlings, larger seeds must necessarily be produced in smaller numbers per unit of resource allocated. Seed mass is presumed to have evolved as a compromise between these counterposed pressures. Yet there has proved to be surprisingly little difference in average seed mass between very different vegetation regions, at least in temperate climates. Rather, there is startlingly wide variation in seed mass among species growing interspersed with each other. Recent applications of game theory may be capable of accounting for this wide variation between coexisting species, but at present these models are driven by competition among seedling species (as opposed to between seedlings and adults). It remains unclear whether competition among seedlings is a decisive influence on species composition in most of the world's vegetation types.