The Trouble with Topology: Phylogenies without Fossils Provide a Revisionist Perspective of Evolutionary History in Topological Analyses of Diversity

Abstract

The decidedly asymmetrical architecture of the Tree of Life betrays the fact that some evolutionary lineages have diversified to a greater or lesser extent than have others. It has been a goal of evolutionary ecology to identify shifts in diversification rate (speciation rate minus extinction rate) and their causal bases. Historically, the question of relative diversity was addressed in terms of phylogenies of fossils (Simpson 1944) with focus shifting to taxonomies in the 1970s (Raup et al. 1973; Gould et al. 1977). The dramatic rise in the availability of molecular sequence data in the 1990s led to renewed interest in phylogenies and diversity from evolutionary biologists and attempts to explain patterns of diversity returned to phylogenies. For the first time, however, historical events inferred to have shaped current diversity were identified without recourse to palaeontological data (Harvey et al. 1994). Two principal approaches have been developed– “distance” and “topological” methods. Distance (aka “temporal”) methods represent a natural development of taxonomic methods that were initially aphylogenetic. They exploit branch length data or the temporal spacing of branching events within phylogenetic trees and while distance methods remain popular for analyzing rates of evolution, they have lost out in favor of topological methods for the analysis of the history of clade diversity. Topological (aka “tree shape”) methods were developed in parallel with distance methods but are distinguished in that they exploit only tree balance as a record of evolutionary history (Slowinski and Guyer 1989). They have been adopted widely because they eschew temporal and distance data and so they can be applied readily to phylogenetic trees, such as supertrees, where distance data are lacking (Moore et al. 2004). The purpose of these methods is to determine 1) if lineages within a given phylogenetic tree diversified under different rates and 2) which particular lineages within a given phylogenetic tree are more diverse than would be expected under a given model, such as a Yule model (Yule 1924). These methods were not, however, conceived to identify the causal bases of tree imbalance but merely to identify whether these phenomena might exist in the first place (Raup 1985; Slowinski and Guyer 1989). In this endeavour, topological methods perform their task admirably. However, in practice, the overwhelming majority of studies (Table 1) that have employed topology-based methods have sought to identify causality “to know if shifts in diversification rate are correlated with changes in some other variable (e.g., the origin of morphological or behavioral novelties, ecological associations, biogeographic events)” (Moore et al. 2004; p. 524). Intrinsic causal factors are identified through coincidence with diversification rate shifts in tree topology and extrinsic causes are identified through their temporal correspondence to the age of the node on which a diversification rate shift is identified. Increasingly sophisticated methods are being developed to test the coincidence between the putative causal factors and the diversification rate shifts to which they are attributed (Moore and Donoghue 2007, 2009). However, the modus operandum is invariably inductive—first identifying diversification rate shifts and then seeking their causal bases rather than deductive—testing hypotheses of causal association between innovations and their impact upon diversity (but see Moore and Donoghue 2009). Ultimately, hypotheses of causality rest on the intuitively reasonable assumption that the position of diversification rate shifts within the topology of a phylogenetic tree reflects their relative timing. Thus, diversification rate shifts identified near the root or the tips of a tree are considered to have occurred early or late, respectively, within the evolutionary history of the clade. In this contribution, we demonstrate that this assumption is not a natural expectation of phylogenetic trees of standing diversity. The topologies of census trees are in constant flux as taxa are added by speciation and pruned by extinction. Nonrandom imbalances in the diversity of sister clades need not be achieved by singular events or episodes of diversification, as topological methods presuppose. Nonrandom imbalances can also be achieved through temporally and causally unrelated episodes of random diversification (speciation and/or extinction), leading to spurious hypotheses of diversification rate shift. The contribution of extinction to standing diversity is especially critical, telescoping past flux in speciation and extinction into a single internal branch of a tree—an effect exploited in lineage through time analyses (Harvey et al. 1994). In this way, the summed