Why extinction estimates from extant phylogenies are so often zero

Abstract

Time-calibrated phylogenies comprising only extant lineages are widely used to estimate historical speciation and extinction rates. Such extinction rate estimates have long been controversial as many phylogenetic studies report zero extinction in many taxa, a finding in conflict with the fossil record. To date, the causes of this widely observed discrepancy remain unresolved. Here we provide a novel and simple explanation for these "zero-inflated" extinction rate estimates, based on the recent discovery that there exist many alternative "congruent" diversification scenarios that cannot possibly be distinguished on the sole basis of extant timetrees. Consequently, estimation methods tend to converge to some scenario congruent to (i.e., statistically indistinguishable from) the true diversification scenario, but not necessarily to the true diversification scenario itself. This congruent scenario may in principle exhibit negative extinction rates, a biologically meaningless but mathematically feasible situation, in which case estimators will tend to hit and stick to the boundary estimate of zero extinction. To test this explanation, we estimated extinction rates using maximum likelihood for a set of simulated trees and for 121 empirical trees, while either allowing or preventing negative extinction rates. We find that the existence of congruence classes and imposed bounds on extinction rates can explain the zero-inflation of previous extinction rate estimates, even for large trees (1000 tips) and in the absence of any detectable model violations. Not only do our results likely resolve a long-standing mystery in phylogenetics, they demonstrate that model congruencies can have severe consequences in practice.