RESULTS
The phylogenetic signals in seed mass and leaf size were moderate and statistically significant across palms, other monocots, dicots, and gymnosperms (Table 1; Fig 2), indicating that seed mass and leaf size have a common evolutionary history with species. We detected a strong phylogenetic signal in plant height of other monocots, dicots and gymnosperms but not palms (Table 1; Fig 2).
When controlling for phylogeny, seed mass showed significantly positive correlation with leaf size across plant species (Table 2; Fig 3). However, plant height was positively correlated with seed mass in other monocots and dicots rather than palms and gymnosperms (Table 2, Fig 3).
The first two axes of the Principal Component Analysis (PCA) together accounted for 87.3% of variability in the functional traits of plant species (P < 0.001, R2 = 0.154, permutations = 999). The principal axis (PC1) was determined positively by seed mass and plant height. The second axis was significantly and positively correlated with leaf area (Fig 4). Thus, dicots were ordinated in a triangle of multivariate space, while palms, gymnosperms and other monocots were ordinated in three separated spaces, with large-seeded palms at the positive extreme of PC1 and small-leaved gymnosperms at the negative extreme of PC2 (Fig 4).
The partial R2 for the logistic regression model showed that leaf size and phylogeny explained the vast majority of variation in seed mass across palm species (partial R2lik = 15.79%, ΔlogLik = 13.7, P < 0.001; R2lik = 16.92%, ΔlogLik = 14.8, P< 0.001; Fig 5), while phylogeny and plant height explained variation in seed mass of species of other monocots and dicots (R2lik = 42.91%, ΔlogLik = 78.2, P < 0.001; R2lik = 6.36%, ΔlogLik = 9.2, P< 0.001; R2lik = 33.81%, ΔlogLik = 330.5,P < 0.001; R2lik = 7.32%, ΔlogLik = 60.9, P < 0.001; Fig 5). Phylogeny rather than leaf size explained a majority of variation in seed mass in gymnosperms (R2lik = 16.35%, ΔlogLik = 3.6, P = 0.008; R2lik = 8.51%, ΔlogLik = 1.8, P = 0.059; Fig 5).