The relationship between seed size and abundance within plant communities has received considerable research attention as ecologists have sought to understand the fundamental determinants of species abundance (Rabinowitz 1978, Mitchley and Grubb 1986, Rees 1995, Guo et al. 2000, Leishman and Murray 2001, Coomes et al. 2002, Levine and Rees 2002, Guo 2003, Murray and Leishman 2003). In addition to understanding patterns of plant abundance, finding predictable relationships between life-history traits such as seed size and plant rarity can yield valuable information for the conservation and management of species (Murray et al. 2002a). Seed size has been investigated in relation to plant rarity at several spatial scales, from within local communities (Rees 1995) to regional assemblages (Eriksson and Jakobsson 1998, Thompson et al. 1999, Bruun 2001) and across geographical ranges (Murray et al. 2002b). Qualitative assessment of the seed size-abundance relationship within communities has revealed that rare species are neither consistently small nor large seeded (Leishman and Murray 2001, Murray et al. 2002a, Guo 2003, Murray and Leishman 2003). However, previous studies have shown that the measure of plant abundance used is critical for determining the direction of seed sizeabundance relationships (Leishman and Murray 2001, Murray and Leishman 2003). This is consistent with recent work that has shown that plant size and longevity are important factors contributing to the seed size/ number trade-off (Moles et al. 2004). In communities where plant abundance has been measured as density, there is a tendency for negative relationships between seed size and abundance (Rees 1995, Guo et al. 2000, Coomes et al. 2002). Conversely, there is a tendency for positive relationships in communities where abundance has been measured as plant cover (Rabinowitz 1978, Leishman and Murray 2001). Cover is frequently used as a measure of plant abundance by ecologists because counting individuals for density estimates can be problematic. The sessile and clonal nature of plants means that determining whether a plant has arisen from a single seed or is simply a ramet is too complicated for most studies of whole plant communities. Plant cover (and biomass) can be negatively related to plant density (Guo 2003), so the dichotomy of seed size-abundance relationships might be explained in part by the different measures of abundance employed in the different studies. If the magnitude of the effect of seed size on abundance is low, and if sample sizes in communities are small, then qualitative syntheses may not detect the true effect of seed size on plant abundance. This may be important for an assessment of the seed size-abundance relationship within plant communities, because (i) relationships between life-history traits and plant rarity tend to be weak (Murray et al. 2002a), and (ii) communities in which the seed size-abundance relationship has been investigated vary widely in species richness from four to 157 species (Appendix 1). Indeed, there are less than 30 species in half of the communities studied. In contrast to qualitative syntheses, meta-analysis can provide a quantitative, statistically defensible summary of the relationship within communities (Hedges and Olkin 1985, Gurevitch et al. 2001). To achieve this, a test statistic of the relationship in each community (e.g. correlation coefficient, r) is standardised using a weighted average effect size (weighted by sample size). The effect sizes of the relationship in each community are combined to produce an aggregated effect size that can be tested for statistical significance using 95% confidence intervals (Arnqvist and Wooster 1995). If the intervals of the aggregated effect size do not overlap