Results
By analyzing worldwide variation in several plant traits, we found strong and statistically significant phylogenetic signal of seed mass, plant height, genome size, leaf area, leaf N and growth form across the plant species (λ = 0.976, P < 0.001; λ = 0.964, P < 0.001; λ = 0.956, P < 0.001; λ = 0.883, P < 0.001; λ = 0.771, P < 0.001; D = -0.190, Prandom = 0, PBrownian = 0.982; Fig. 1 and 2), showing that plant traits covary in direct proportion to their shared evolutionary history.
General linear model showed that seed mass and plant height were higher in the woody plants than in the non-woody species (t = -6.676, P < 0.001; t = -38.42, P < 0.001; Fig. 3a, b). Although there was a trend for woody plants to have larger genome size and leaf area than non-woody species, this was not significant (t = 0.735, P = 0.462; t = 0.595, P = 0.552; Fig. 3c, d). However, leaf N was lower in the woody plants than in the non-woody species (t = 8.782, P < 0.001; Fig. 3e).
Without controlling phylogeny, we identified statistically significant relationships between seed mass and plant height across 1071 species based on multi-variable generalized linear model (t = 3.299, P = 0.001; Table 1). Although seed mass and leaf area were positively correlated (t = 6.961, P < 0.001; Table 1), we detected significant interactive effect of plant height and leaf area on variations in seed mass (t = 2.204, P = 0.028; Table 1). Significant relationship was found between seed mass and genome size (t = 5.034, P < 0.001; Table 1) but not between seed mass and leaf N in the absence of phylogeny (t = 1.646, P = 0.099; Table 1). Growth form, however, well predicted variations in seed mass across the plant species (t = 5.784, P < 0.001; Table 1).
In analysis that controlled for the phylogeny, plant height, leaf area, and genome size alone appeared to be a reliable predictor of variations in seed mass (z = 4.268, P < 0.001; z = 2.549, P = 0.011; z = 4.514, P < 0.001; Table 1). After correcting for phylogenetic effects, growth form and leaf N failed to predict variations in seed mass (z = 0.838, P = 0.402; z = 0.850, P = 0.395; Table 1). There was no significant interactive effect of plant height and leaf area on seed mass in the presence of phylogenetic considerations (z = 0.437, P = 0.662; Table 1).
Phylogeny explained the vast majority of seed mass variation across the plant species (partial R2lik = 50.89%, ΔlogLik = 380.8, P < 0.001), while plant height, leaf area, genome size explained a minority of variation (R2lik = 7.39%, ΔlogLik = 41.1, P < 0.001; R2lik = 0.58%, ΔlogLik = 3.1, P = 0.01; R2lik = 1.85%, ΔlogLik = 10.0, P < 0.001; Fig. 4). Leaf N and growth form failed to explain the variance in seed mass (R2lik = 0.06%, ΔlogLik = 0.3, P = 0.41; R2lik = 0.09%, ΔlogLik = 0.5, P = 0.33; Fig. 4).