Phylogenetically conservative trait correlation: Quantification and interpretation

Abstract

Abstract Correlation across species between two quantitative traits, or between a trait and a habitat property, can suggest that a trait value is effective in sustaining populations in some contexts but not others. It is widely held that such correlations should be controlled for phylogeny, via phylogenetically independent contrasts (PICs) or phylogenetic generalized least squares (PGLS). A weakness of this idea is that a clade's traits tend to confer success in particular habitats or ways of life, and those niches in turn tend to select for the same traits to continue in the clade. This feedback mechanism can bind phylogeny and niche together as a unified cause for present‐day trait configurations. Accordingly, the phylogenetically conservative share of a trait correlation ought not to be excluded from consideration as potentially ecologically functional. Another weakness is that PGLS does not yield a complete or accurate breakdown of covariation between traits A and B because it corresponds to a generating model where B predicts variation in A but not the reverse, and phylogenetic signal in B is not modelled. Multi‐response mixed models using phylogenetic covariance matrices can quantify conservative trait correlation (CTC), a share of A‐B covariation that is phylogenetically conservative. Because the evidence is from correlative data, it is not possible to split CTC into causation by phylogenetic history versus causation by continuing reciprocal selection between A and B. Moreover, it is quite likely biologically that the two influences have acted in concert, through phylogenetic niche conservatism. Synthesis: The CTC concept treats phylogenetic conservatism as a conjoint interpretation alongside ongoing influence of other traits. CTC can be quantified via multi‐response phylogenetic mixed models.