Topographically distinct adaptive landscapes for teeth, skeletons, and size explain the adaptive radiation of Carnivora (Mammalia)

Abstract

Models of adaptive radiation were originally developed to explain the early, rapid appearance of distinct modes of life within diversifying clades. Phylogenetic tests of this hypothesis have yielded limited support for temporally declining rates of phenotypic evolution across diverse clades, but the concept of an adaptive landscape that links form to fitness, while also crucial to these models, has received more limited attention. Using methods that assess the temporal accumulation of morphological variation and estimate the topography of the underlying adaptive landscape, I found evidence of an early partitioning of mandibulo-dental morphological variation in Carnivora (Mammalia) that occurs on an adaptive landscape with multiple peaks, consistent with classic ideas about adaptive radiation. Although strong support for this mode of adaptive radiation is present in traits related to diet, its signal is not present in body mass data or for traits related to locomotor behavior and substrate use. These findings suggest that adaptive radiations may occur along some axes of ecomorphological variation without leaving a signal in others and that their dynamics are more complex than simple univariate tests might suggest.

Figure 1

Some theories of adaptive radiation posit that a generalized ancestral form (a) should diversify simultaneously along distinct axes of resource use (here, diet and locomotor mode) (b), establishing complete functional diversity early in clade history with little subsequent change (c). Models of staged adaptive radiation differ from the simultaneous model in suggesting that the same generalized ancestor (d) should first diversify along one axis (frequently substrate‐use) (e) and subsequently along another (e.g. resource‐use) (f), yielding a sequential pattern of diversification in morphological and ecological data.

Figure 2

When the mode of phenotypic evolution is consistent with a random walk, phenotypic variation accumulates gradually over phylogeny (a), leading to a pattern in which the average relative subclade disparity through time (ARSD) declines steadily from 1 (i.e. all variance within a single clade) to zero (all clades are single tips) as time progresses from root to tip (d) and a center of gravity (red arrow) that falls at approximately the midpoint. Under an “early burst” scenario, trait variation does not accumulate steadily but, rather, seems to accumulate early in the history of the clade (b), yielding an ARSD curve that drops precipitously and then levels out (e), and a center of gravity that falls below the midpoint of clade history. Evolutionary modes such as “late bursts” or time constant rates in a bounded space (e.g., constraints) yield a pattern in which phenotypic variation accumulates toward the tips of the tree (c) and an ARSD curve that declines very slowly until late in clade history (f), with a very high center of gravity.

Figure 3

Principal components analysis of log shape variables from the mandibulodental log shape variables (a) shows a strong phylogenetic partitioning. The shape space defined by PCs 1 and 2 of log‐shape variables from the post‐cranial skeleton (b) shows slight separation of some clades and functional groups along PCs 1 and 2, but with much more overlap than is found for the mandibulodental data.

Figure 4

Disparity through time plots for carnivoran mandibulodental (a), post‐cranial (b) and body mass (c) data. The solid line is the average relative sub‐clade disparity for the data set, the dashed line is the median from 9999 datasets simulated under a constant rates process, and the shaded area corresponds to the 95% quantiles of the simulated data. Shown on the right (d‐f) are the corresponding centers of gravity for the simulated datasets (solid bars) and the center of gravity for the trait data (dashed vertical line).

Figure 5

Carnivoran phylogeny showing the 25 adaptive peak shifts for mandibulodental traits identified by the $\mathtt{phyloEM}$ approach. Black circles are placed at the midpoint of each branch along which a peak shift is inferred to have occurred, with branches inheriting and retaining that peak colored distinctly. Clade names are provided to orient the reader to shifts described in the results and are not intended to be exhaustive.