Results
Of the 9,370 species (combined dataset of the three floras), 2,635 species (28.1%) produced fleshy fruits. Temperate flora contained a lower proportion of fleshy-fruited species (22.5%) than both tropical flora (38.7%) and subtropical floras (32.6%; Fig. 2a, Table S1 in Supporting Information). Woody species contained a higher proportion of fleshy-fruited species (55.6%) than herbaceous species (9.6%) (Fig. 2b, Table S1).
Phylogenetic conservatism explained the vast majority of fruit type variation among floras (partial R2lik = 79.5%), while climate region and growth form explained little variation, with partial R2lik values of 0.3% and 1.7%, respectively (Table 2, Fig. 2c, d). To investigate the depth at which this conservatism occurred within the phylogeny, we repeated the logistic regression using a phylogeny in which nodes above the 0.67 threshold from the base of the phylogeny were collapsed (Fig. 1b). This threshold corresponds roughly to the taxonomic level of families, so this phylogeny removed phylogenetic information within families. The phylogenetic conservatism occurred both towards the tips and towards the base of the phylogenetic tree: when using the 0.67-threshold phylogeny, the partial R2lik for phylogeny was 41.6%, implying that the phylogenetic information towards the base of the phylogeny could explain roughly half as much variation as the full phylogeny. Furthermore, the partial R2lik values for plant growth form and climate region were 25.1% and 3.9% (Table 2), implying that plant growth explains more of the distribution of fruit type among species when the explanatory information provided towards the tips of the phylogeny is not available.
The low partial R2lik for growth form can be better understood by considering the regression of growth form on fruit type without the phylogeny or climate regions. In the model with growth form alone (full model GF, Table 2), the R2lik is 22.8% implying that growth form explains a moderate amount of variation in fruit type. This is expected, because 55.6% of the wood species had fleshy fruit, compared to only 9.6% of the herbaceous species (Fig. 2b, Table S1). However, in the model with growth form and phylogeny (full model GF+PHY, Table 2), the partial R2lik for growth form is 1.5%. Therefore, when taking account of phylogeny, growth form gives little additional information about fruit type. This suggests that the apparent association between fruit type and growth form (Spearman rank correlation = 0.503, Table S1, R2lik = 22.8%, Table 2) observed in this dataset cannot be distinguished from a spurious effect of both fruit type and growth form being strongly conserved phylogenetically. In contrast to growth form, the R2lik for climate region remains low even in the absence of phylogeny (R2lik = 4.2%, full model CR, Table 2). This implies that phylogeny does not play a role in the poor explanatory power of climate regions on fruit type.
To investigate lack of explanatory power of growth form and climate region on fruit type in more detail, we investigated the phylogenetic conservatism for each trait separately. Specifically, we performed analyses with each trait as the dependent variable and no independent variables. Both fruit type and plant growth form had very strong phylogenetic signal (low values of α of the Ornstein-Uhlenbeck model of evolution employed by phyloglm, Tung Ho and Ané 2014) and R2lik values of 86.3% and 86.8%, respectively, when the full phylogeny used (Table 3). When the 0.67-threshold phylogeny was used for the analyses, the R2lik values decreased to 46.8% and 53.2%, respectively. This implies that phylogenetic conservatism occurred both towards the base of the phylogeny (because the traits R2lik values were high) and also towards the tip (because the traits R2lik values were not as high as for the full phylogeny). The results for fruit type and plant growth form contrast those for climate regions. For tropical, subtropical, and temperate climate regions, the R2likvalues were 20.0%, 0%, and 12.2%, respectively. Furthermore, the R2lik values of the three climate regions computed for the 0.67-threshold phylogeny were similar to those for the full phylogeny, implying that most of the information available from the phylogeny occurred towards the base of the tree.
To obtain a more-complete picture of the pattern of phylogenetic conservatism of the three traits, for each trait we computed the R2lik across a range of threshold phylogenies, from 1 (full phylogeny) down to 0.2 (only 20% phylogenetic information near the basal was maintained) (Fig. 3). Both fruit type and growth form show very strong phylogenetic conservatism, and as the threshold above which nodes are collapsed increases, the values of R2likincrease roughly linearly. This implies that the phylogenetic information underlying the values of R2lik are even distributed across the phylogeny. In contrast, the values of R2lik for tropical and temperate climate regions stay roughly the same until the threshold drops below 0.8, implying that the phylogenetic relationships above this threshold provide little information about whether species occur in the two climate regions. The values of R2likonly drop substantially when they are below the 0.8 threshold, implying that phylogenetic conservatism occurs towards the base of the phylogeny. For the subtropical climate region, R2lik is always close to zero, implying that whether or not a species occurs in the subtropical climate region is not explained by phylogeny.