A preview of this full-text is provided by Springer Nature.
Content available from Nature
This content is subject to copyright. Terms and conditions apply.
502 | Nature | Vol 580 | 23 April 2020
Article
Extant timetrees are consistent with a
myriad of diversification histories
Stilianos Louca1,2 ✉ & Matthew W. Pennell3,4 ✉
Time-calibrated phylogenies of extant species (referred to here as ‘extant timetrees’)
are widely used for estimating diversication dynamics1. However, there has been
considerable debate surrounding the reliability of these inferences2–5 and, to date, this
critical question remains unresolved. Here we clarify the precise information that can
be extracted from extant timetrees under the generalized birth–death model, which
underlies most existing methods of estimation. We prove that, for any diversication
scenario, there exists an innite number of alternative diversication scenarios that
are equally likely to have generated any given extant timetree. These ‘congruent’
scenarios cannot possibly be distinguished using extant timetrees alone, even in the
presence of innite data. Importantly, congruent diversication scenarios can exhibit
markedly dierent and yet similarly plausible dynamics, which suggests that many
previous studies may have over-interpreted phylogenetic evidence. We introduce
identiable and easily interpretable variables that contain all available information
about past diversication dynamics, and demonstrate that these can be estimated
from extant timetrees. We suggest that measuring and modelling these identiable
variables oers a more robust way to study historical diversication dynamics. Our
ndings also make it clear that palaeontological data will continue to be crucial for
answering some macroevolutionary questions.
A central challenge in evolutionary biology is to reconstruct rates of
speciation and extinction over time5. Unfortunately, the majority of
taxa that have ever lived have not left much trace in the fossil record,
and the primary source of information on their past diversification
dynamics therefore comes from extant timetrees. Many methods
have been developed for extracting this information; most methods
fit variants of a birth–death process
1,6
. Despite the popularity of these
methods, which collectively have been used in thousands of studies
7–9
,
their reliability has been called into question by comparisons with
fossil-based estimates1,3,5,6,10. The reasoning behind these critiques
is that there may be insufficient information in extant timetrees to
fully reconstruct historical diversification dynamics. However, this
critical issue has remained unresolved; it is unknown precisely what
information on speciation and extinction rates is contained in extant
timetrees.
Here we present a definite answer to this question for the general
stochastic birth–death process with homogeneous (that is, lineage-
independent) rates, in which speciation (‘birth’) rates (λ) and extinc-
tion (‘death’) rates (μ) can vary over time, that underlies the majority
of existing methods for reconstructing diversification dynamics from
phylogenies
1
. We mathematically show that, for any givencandidate
birth–death model, there exists an infinite number of alternative birth–
death models that can explain any extant timetree equally as well as
can the candidate model. These alternative models may appear to be
similarly plausible and yet exhibit markedly different features, such as
different trends through time in both λ and μ. This severe ambiguity
persists for arbitrarily large trees and cannot be resolved even with an
infinite amount of data; it is thus impossible to design asymptotically
consistent estimators for λ and μ. Using simulated and real timetrees
as examples, we demonstrate how failing to recognize this issue can
seriously mislead our inferences about past diversification dynamics.
We present appropriately modified variables that are asymptotically
identifiable and that contain all available information on historical
diversification dynamics.
Lineages through time
An important feature of extant timetrees is the lineages-through-time
curve (LTT), which counts the number of lineages at each time in the
past that are represented by at least one sampled extant descending
species in the tree. The likelihood of a tree under a given birth–death
model, the LLT of the tree and the LTT that would be expected under
the model are linked as follows. Any given combination of (potentially
time-dependent) speciation and extinction rates (λ and μ, respectively)
and the probability that an extant species will be included in the tree
(‘sampling fraction’) (ρ) can be used to define a deterministic diver-
sification process, in which the number of lineages through time no
longer varies stochastically but instead according to a set of differ-
ential equations
6,11
(Supplementary Information section S.1). Given
a number of extant sampled species (M
o
), the LTT predicted by these
https://doi.org/10.1038/s41586-020-2176-1
Received: 14 September 2019
Accepted: 10 March 2020
Published online: 15 April 2020
Check for updates
1Department of Biology, University of Oregon, Eugene, OR, USA. 2Institute of Ecology and Evolution, University of Oregon, Eugene, OR, USA. 3Biodiversity Research Centre, University of British
Columbia, Vancouver, British Columbia, Canada. 4Department of Zoology, University of British Columbia, Vancouver, British Columbia, Canada. ✉e-mail: louca.research@gmail.com; pennell@
zoology.ubc.ca
Content courtesy of Springer Nature, terms of use apply. Rights reserved