Causes and Consequences of Apparent Timescaling Across All Estimated Evolutionary Rates

Abstract

Evolutionary rates play a central role in connecting micro- and macroevolution. All evolutionary rate estimates, including rates of molecular evolution, trait evolution, and lineage diversification, share a similar scaling pattern with time: The highest rates are those measured over the shortest time interval. This creates a disconnect between micro- and macroevolution, although the pattern is the opposite of what some might expect: Patterns of change over short timescales predict that evolution has tremendous potential to create variation and that potential is barely tapped by macroevolution. In this review, we discuss this shared scaling pattern across evolutionary rates. We break down possible explanations for scaling into two categories, estimation error and model misspecification, and discuss how both apply to each type of rate. We also discuss the consequences of this ubiquitous pattern, which can lead to unexpected results when comparing rates over different timescales. Finally, after addressing purely statistical concerns, we explore a few possibilities for a shared unifying explanation across the three types of rates that results from a failure to fully understand and account for how biological processes scale over time. Expected final online publication date for the Annual Review of Ecology, Evolution, and Systematics, Volume 52 is November 2021. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.

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

“…A potential limitation of the DOT is that contrasts are computed using ancestral state estimates that are, themselves, inferred under a time-homogeneous model of evolution (Sallan and Friedman, 2012). Furthermore, the DOT explicitly assumes that adaptive radiation results from decelerating rates of phenotypic evolution, whereas similar patterns may result from myriad other process, including evolution on a rugged adaptive landscape (e.g., Revell et al, 2005; Harmon et al, 2021). Unlike the DOT, the DTT center of gravity does not explicitly assume an underlying evolutionary process but, instead, provides a means of describing and statistically evaluating whether the temporal structure of morphological variation within and among clades is consistent with the hypothesis of early variance partitioning among clades (Foote, 1996).…”

The scaling of diversification rates with age is likely explained by sampling bias

“…It is in principle possible that fundamental model misspecification-the inability of birth-death models to capture important aspects of the true cladogenic process or of our (imperfect) tree estimates-could contribute to the ARS. Harmon et al (2021) reviewed a number of ways by which biological processes could generate such a mismatch. For example, if there is substantial heterogeneity in diversification rates among clades (Mooers and Heard, 1997), or speciation events are clustered in time (Foote, 2005), or diversification is highly constrained by ecological factors (McPeek, 2008;Rabosky, 2009a), or the diversification process is hierarchical (Etienne and Rosindell, 2012;Rosenblum et al, 2012), estimating rates as the parameters of a birth-death process may be inappropriate in ways that lead to an ARS.…”

“…To date, the causes of this pervasive age-rate-scaling (ARS) and its implications for macroevolutionary analyses remain largely unresolved. A number of potential causes for this ARS have been suggested over the years, recently reviewed by Harmon et al (2021). These can be broadly grouped into three categories: (i) a true acceleration of diversification rates over geological time scales; (ii) model misspecification, which we define broadly as a serious mismatch between the assumptions involved in rate estimates and biological reality; and (iii) statistical artifacts that arise due to estimation error or sampling bias, even when the models used to describe the evolutionary process are appropriate.…”

“…These can be broadly grouped into three categories: (i) a true acceleration of diversification rates over geological time scales; (ii) model misspecification, which we define broadly as a serious mismatch between the assumptions involved in rate estimates and biological reality; and (iii) statistical artifacts that arise due to estimation error or sampling bias, even when the models used to describe the evolutionary process are appropriate. Among these, explanations that fall into the latter two categories are the most likely (Harmon et al, 2021), given the lack of a clear mechanism for a secular rate increase (accelerating especially over the last few million years) and the fact that analogous patterns are also commonly seen in phenotypic (Gingerich, 1983;Hendry and Kinnison, 1999;Uyeda et al, 2011) and molecular (Ho et al, 2005(Ho et al, , 2011 rate estimates; hence, in the following we focus on categories (ii) and (iii).…”

“…Note that r is approximately equal to r if λ varies only slowly over time, but will differ substantially from r if λ changes rapidly. The implications of the above identifiability issues for the ARS are currently unknown; this uncertainty was recently also pointed out by Harmon et al (2021). Since the majority of published phylogeny-based estimates of λ, μ, and r through time are unreliable, this raises the question as to whether an ARS would still be observed in identifiable variables such as the PDR, or whether the ARS is a mere artifact of the non-identifiability of birth-death phylogenetic models.…”