Systematic Biology Advance Access published January 30, 2012

Point of View

c The Author(s) 2012. Published by Oxford University Press, on behalf of the Society of Systematic Biologists. All rights reserved.
For Permissions, please email: [email protected]
DOI:10.1093/sysbio/syr124

Effects of Phylogenetic Signal on Ancestral State Reconstruction

G LENN L ITSIOS 1,2 AND N ICOLAS S ALAMIN1,2,∗

Downloaded from http://sysbio.oxfordjournals.org/ at Universidade Estadual Paulista Júlio de Mesquita Filho on March 6, 2012
1 Department of Ecology and Evolution, Biophore, University of Lausanne, 1015 Lausanne, Switzerland; and 2 Swiss Institute of Bioinformatics,
Quartier Sorge, 1015 Lausanne, Switzerland;
∗ Correspondence to be sent to: Department of Ecology and Evolution, Biophore, University of Lausanne, 1015 Lausanne, Switzerland;
E-mail: [email protected].
Received 28 July 2011; reviews returned 15 September 2011; accepted 11 November 2011
Associate Editor: Robb Brumfield

One of the standard tools used to understand the time (e.g., millions of years). In practice, time-calibrated
processes shaping trait evolution along the branches trees (or chronograms) are nearly always chosen over
of a phylogenetic tree is the reconstruction of ancestral trees depicting molecular changes (or phylograms; e.g.,
states (Pagel 1999). The purpose is to estimate the val- Crespi and Teo 2002; Jones et al. 2009a; Friedman et al.
ues of the trait of interest for every internal node of a 2009; Skinner and Lee 2010). The implicit argument for
phylogenetic tree based on the trait values of the extant this practical choice is that branch lengths estimated
species, a topology, and, depending on the method used, using the DNA markers employed for phylogenetic in-
branch lengths and a model of trait evolution (Ronquist ference should not be assumed to be correlated with the
2004). This approach has been used in a variety of con- rate of phenotypic evolution (Bromham et al. 2002). Evi-
texts such as biogeography (e.g., Nepokroeff et al. 2003; dence of substitution rate heterogeneity has been found
Blackburn 2008), ecological niche evolution (e.g., Evans in different taxonomic units (e.g., rodents: Spradling
et al. 2009; Smith and Beaulieu 2009), and metabolic et al. 2001; angiosperms: Smith and Donoghue 2008),
pathway evolution (e.g., Gabaldón and Huynen 2003; and changes in phenotypic evolutionary rate are not
Christin et al. 2008). uncommon between lineages (O’Meara et al. 2006).
Investigations of the factors affecting the accuracy However, it has recently been shown that molecular
with which ancestral character states can be recon- and phenotypic rates of change can be correlated, such
structed have focused in particular on the choice of as in flowering plants, where part of the variation in
statistical framework (Ekman et al. 2008) and the selec- phenotypic rates could be explained by the DNA substi-
tion of the best model of evolution (Cunningham et al. tution rate (Davies and Savolainen 2006). Furthermore,
1998; Mooers et al. 1999). However, other potential bi- herbaceous plant species exhibit faster substitution rates
ases affecting these methods, such as the effect of tree than tree and shrub species (Smith and Donoghue 2008),
shape (Mooers 2004), taxon sampling (Salisbury and which suggests an important effect of life-history char-
Kim 2001), as well as reconstructing traits involved in acteristics, and generation times in particular, on rates
species diversification (Goldberg and Igić 2008), have of molecular evolution. Similar patterns have also been
also received specific attention. Most of these studies found in mammals (Nikolaev et al. 2007) and inver-
conclude that ancestral character state reconstruction tebrates (Thomas et al. 2010). Therefore, if the rate of
is still not perfect, and that further developments are molecular evolution explains part of the phenotypic
necessary to improve its accuracy (e.g., Christin et al. variation between species, we argue that inferring an-
2010). Here, we examine how different estimations of cestral character states on time-calibrated trees could
branch lengths affect the accuracy of ancestral character have deceiving results as it may not appropriately
state reconstruction. In particular, we tested the effect of represent the evolution of species traits. However, no
using time-calibrated versus molecular branch lengths simple guideline to choose between chronograms and
and provide guidelines to select the most appropri- phylograms has been proposed so far.
ate branch lengths to reconstruct the ancestral state of
a trait.
Current maximum likelihood (ML) and Bayesian I NFERRING A NCESTRAL C HARACTER S TATES ON A
methods used to reconstruct ancestral character states S IMULATED D ATA S ET
incorporate information about branch lengths during In order to investigate the effects of molecular and
the inference. Branch lengths can be specified either in phenotypic rate heterogeneity on the accuracy of ances-
units of divergence (i.e., expected number of substitu- tral state reconstructions, we performed computer sim-
tions per site per unit of time) or in units of absolute ulations in R (R Development Core Team 2011) using the

1
2 SYSTEMATIC BIOLOGY VOL. 61

pling every 1000. The first 2000 trees were removed as

burn-in, and we constructed a maximum clade credibil-
ity tree with TreeAnnotator (Drummond and Rambaut
2007). The burn-in level and parameters of the analy-
sis were set based on a preliminary analysis to assure
an optimal convergence of the MCMC chains. Because
of the intensive computational effort required by the
Bayesian analyses, only 300 of the 5000 trees simulated
FIGURE 1. Schematic view of the tree used during the simulations were analyzed with BEAST.
with Branch B shown in dotted line subtending Clade C shown in gray. We generated continuous characters using a Brow-
a) The original tree is simulated using a birth–death process resulting

Downloaded from http://sysbio.oxfordjournals.org/ at Universidade Estadual Paulista Júlio de Mesquita Filho on March 6, 2012
in an ultrametric tree, which is scaled by setting the root node to 1. b) nian motion model of character evolution with a vari-
Mutation rates are then applied to each branch following an autocor- ance parameter of 1 for every node of the phylograms
relation process. c) A change in generation time is applied to Clade C (Fig. 1c) and chronograms (reconstructed with penal-
after the appearance of a key innovation along Branch B indicated by ized likelihood and BEAST). This created a data set of
a dotted line (see text for more details).
characters that evolved following either the molecu-
lar or ultrametric branch lengths. We also simulated
ape (Paradis et al. 2004), geiger (Harmon et al. 2008), and characters that evolved in a manner uncorrelated to
picante (Kembel et al. 2010) packages. We first simulated either the molecular or ultrametric branches to accom-
under a uniform birth–death process 5000 trees with 64 modate situations where the evolution of a character
terminal taxa whose branch lengths corresponded to rel- is linked to neither the substitution rates nor the time.
ative time after rescaling the root depth to 1 (Fig. 1a). We This was done by taking the original tree (Fig. 1a) and
then multiplied each branch length by a value drawn reassigning for each branch length a value drawn from
from a normal distribution with mean equal to the par- a normal distribution centered on the original branch
ent value and standard deviation equal to 0.2 (Fig. 1b), length. This was done 4 times for each simulated tree
the root value being set to one. By introducing substi- with 4 different standard deviations (0.05, 0.1, 0.3, 0.5) to
tution rate heterogeneity, we transformed the initial ul- create 4 trees having increasingly uncorrelated branch
trametric trees into phylograms where rate variation is lengths. We then reconstructed the characters on the
phylogenetically conserved among lineages (Sanderson phylograms and chronograms using a widely used ML
1997). algorithm (Pagel 1994) implemented in geiger (Harmon
To mimic a change in substitution rates, for exam- et al. 2008). For each simulated data set, we summed
ple, due to a change in life history, a Branch B (dotted over the differences between the inferred and the true
line in Fig. 1) was randomly selected and the substitu- ancestral states at each node to obtain one value per tree.
tion rates of every branch descending from Branch B This gave a relative measure of the error in the character
were modified as follows. We assumed that the change state reconstructed, with higher values meaning larger
in substitution rate was not instantaneous but moved differences between the inferred and the true values.
toward an optimum value (defined as 10 in all simula- We also measured the tree imbalance using the Colless
tions) following an Ornstein–Uhlenbeck process (Butler index (Colless 1995), the size of the clade originating on
and King 2004) with a selective force of 3 and a vari- Branch B, and the phylogenetic signal using both the K
ance of 0.2. In practice, this meant that the mean of the (Blomberg et al. 2003) and λ (Pagel 1999) indices.
normal distribution used to draw the substitution rate In total, we investigated 30 different scenarios. First,
assigned to each branch was not equal to the ancestor data were simulated on 6 different phylogenetic trees
rate but moved toward the optimum stochastically. This (1 chronogram, 1 phylogram, and 4 trees with uncor-
whole process converted the original tree (Fig. 1a) into related branch lengths; see earlier description). Second,
a phylogram with the Clade C originating from Branch the simulated data sets were analyzed on 5 different
B, showing an increase in substitution rate (Fig. 1c). phylogenetic trees (4 different chronograms and 1 phy-
Penalized likelihood (Sanderson 2002) was then used logram; see earlier description). Each scenario contained
to transform the phylogram into an ultrametric tree 5000 simulated data sets (i.e., one per tree created; 300
using different penalties corresponding to (i) complete selected at random for the BEAST reconstructions). We
parametric estimation (penalty = 0), (ii) some likelihood summarized the general trends in our simulations by
penalty on rate change (penalty = 10), and (iii) a global calculating the mean error overall trees for each sce-
molecular clock (penalty = 10,000). To account more nario. Our results show that the error was significantly
directly for molecular rate heterogeneity, we also used lower when characters were reconstructed on the same
BEAST (Drummond and Rambaut 2007) to infer chrono- tree as they were simulated rather than on another set of
grams. As BEAST, in contrast to penalized likelihood, branch lengths (t-test, t = −4.2, df = 8.3, P = 0.003). This
requires DNA data to perform the dating analysis, we is perfectly logical and demonstrates the potential dis-
simulated sequences of 1000 nucleotides using the phy- crepancy that can arise if a character is reconstructed
logram (Fig. 1c) of each simulation replicate and the F81 on a tree with branch lengths that do not represent
model (κ = 2.0). We then ran BEAST on this DNA ma- its tempo of evolution. For characters that evolved on
trix with the HKY model of nucleotide evolution, a root a tree with uncorrelated branch lengths, there was no
age of 1, and a fixed topology for 106 generations sam- difference in accuracy between reconstructing ancestral
2012 POINT OF VIEW 3

Downloaded from http://sysbio.oxfordjournals.org/ at Universidade Estadual Paulista Júlio de Mesquita Filho on March 6, 2012
FIGURE 3. Relationship between mean differences in ancestral
state reconstruction and phylogenetic signal in the primate data set.
Gray dots show for each pair of tree the difference in λ between the
phylogram and chronogram, whereas black dots are for K. The black
FIGURE 2. Relationship between mean error in ancestral charac- and gray lines show slopes of the relationship.
ter state reconstruction and phylogenetic signal. Closed circles show
the reconstructions done on phylograms; triangles, the reconstructions
done on chronograms (3 penalized likelihood estimations); and dia- analysis and we standardized the value of the charac-
monds, the reconstructions done on BEAST chronograms. Shades of ters. We then summed the absolute difference between
gray represent the mode of character evolution (black: simulated on the simulated and reconstructed character states for
the phylogram; dark gray: simulated on the chronogram; light gray: each node inferred on the two sets of branch lengths.
simulated on uncorrelated branch lengths).
This difference was divided by the total number of
nodes in the tree to obtain the mean difference per node.
character states on phylograms or chronograms (t-test, As we do not know the correct assignment of ancestral
t = −0.18, df = 4.6, P = 0.87). Furthermore, neither character states, we used this mean difference between
the size of Clade C nor the tree imbalance explained chronograms and phylograms instead of the true error
the reconstruction accuracy (slope =−1.15, adjusted as in the simulations.
R2 = 0.002, for clade size; slope =−0.0007, adjusted If, as shown in the simulations, phylogenetic signal is
R2 = 0.0003, for tree imbalance). linked to ancestral state reconstruction accuracy, recon-
Regardless of the tree, the phylogenetic signal, mea- structions done on pairs of trees having similar phylo-
sured by K or λ, always strongly explained the error genetic signals should have a smaller mean difference
in ancestral character state reconstruction (Fig. 2; K: per node than those done on a pair of trees showing
slope =−2.379, adjusted R2 = 0.53; λ: slope =−20.56, very different phylogenetic signals. Results showed
that both K and λ explain the mean difference in an-
adjusted R2 = 0.91). This result is highly intuitive as it
cestral character state reconstruction, with λ showing a
states that, by using branch lengths having a better fit to
stronger relationship (Fig. 3; K: slope = 0.014, adjusted
the character evolution (stronger phylogenetic signal),
the ancestral state reconstruction will be more accurate. R2 = 0.002; λ: slope = 0.21, adjusted R2 = 0.19). An-
other important factor was taxon sampling. The mean
difference between reconstructions was smaller when
I NFERRING A NCESTRAL S TATES ON A L ARGE more taxa were present in the analysis (slope =−7.3 ×
E MPIRICAL D ATA S ET 10−5 , adjusted R2 = 0.02). It has been shown that for
To test the effect of phylogenetic signal on ances- maximum parsimony methods, larger taxon sampling
tral character state reconstruction, we used a published does not necessarily lead to a more accurate reconstruc-
phylogeny containing 230 primate species (Arnold et al. tion (Li et al. 2008). On the contrary, our results confirm
2010) and reconstructed the ancestral states of 25 contin- previous findings that taxon sampling matters for ML
uous characters (Jones et al. 2009b) using an ML method and Bayesian methods for accurate ancestral character
on both published chronogram and phylogram. The state reconstruction (Heath et al. 2008).
phylogeny was originally built in a Bayesian frame-
work, allowing us to replicate the analysis by using 100
pairs of phylogram/chronogram drawn from the pos- C ONCLUSION
terior distribution of trees. For each character, species The results of the simulations and empirical analyses
with missing data were pruned from the tree before the show that using a phylogram or a chronogram can lead
4 SYSTEMATIC BIOLOGY VOL. 61

to a very different conclusion about the ancestral states Recently Skinner (2010) showed that taking pheno-
of a character. This was true regardless of the charac- typic rate heterogeneity into account improved recon-
ter evolutionary history. Usually, genes used to build structions of ancestral limb morphology in the scincid
phylogenetic trees are selected to reflect mostly neutral lizard clade Lerista. But as the model Skinner proposed
evolution, which is supposed to have no effect on the uses branch length to infer the probability of changes,
phenotype. For comparative analysis such as ancestral choosing the right sets of branches is still of primary
character state reconstructions, this would mean that us- importance. Nowadays, it is becoming more common
ing a phylogram or a chronogram should not matter. But to build large phylogenies to infer patterns of evo-
this reasoning misses one very important point. Sub- lution (Smith and Donoghue 2008). These data sets
stitution rate is a function of many variables, such as will, without doubt, encompass species with significant

Downloaded from http://sysbio.oxfordjournals.org/ at Universidade Estadual Paulista Júlio de Mesquita Filho on March 6, 2012
population size, generation time, DNA repair efficiency, genotypic and phenotypic rate heterogeneity, compli-
metabolism, and mode of reproduction (Ballard 2000; cating the choice between chronogram and phylogram.
Foltz et al. 2004; Fontanillas et al. 2007; Thomas et al. In this case, we recommend that inference of ances-
2010). With many of these variables under selection, it is tral characters should be performed on the tree that
expected that variation will appear between organisms. shows the strongest phylogenetic signal as it is expected
As mutations are needed for any novelty to evolve, a to be more informative and thus gives more accurate
higher substitution rate will increase the chance of ap- reconstructions.
pearance and subsequent evolution of a trait. Under this
paradigm, it is thus most probable that changes in many
phenotypic traits followed closely changes in the rate F UNDING
of DNA substitution and not a clock-like evolutionary This work was supported by the Swiss National Sci-
rate (Davies and Savolainen 2006; Smith and Beaulieu ence Foundation (grant numbers 3100AO-116442 and
2009; Seligmann 2010). In such case, ancestral character CRS 113-125240).
state reconstruction will without doubt be less accurate
if done on chronograms than on phylograms.
The phylogenetic signal of a character measures the A CKNOWLEDGMENTS
level of dependency in the trait value of species that is
due to their phylogenetic relatedness (Revell et al. 2008). This work received support from the Vital-IT facil-
For both indices used in this study, a value of 1 is ex- ities from the Swiss Institute of Bioinformatics. We
pected for a character evolving following a Brownian thank Pascal-Antoine Christin, Socrates Litsios, and
motion. The Brownian motion was introduced in phylo- two anonymous reviewers for valuable comments on a
genetics to calculate independent contrasts (Felsenstein previous version of the manuscript.
1985). It is now used extensively in comparative meth-
ods, including estimation of ancestral character states, R EFERENCES
due to its simplicity and its connection with the genetic
Arnold C., Matthews L.J., Nunn C.L. 2010. The 10kTrees website:
drift theory in population genetics (Martins 1994). Be- a new online resource for primate phylogeny. Evol. Anthropol.
cause of its inherent neutrality, it is increasingly seen as 19:114–118.
the standard null model for continuous character evo- Ballard J.W.O. 2000. Comparative genomics of mitochondrial DNA
lution (Butler and King 2004, Salamin et al. 2010). By in members of the Drosophila melanogaster subgroup. J. Mol. Evol.
51:48–63.
choosing to reconstruct ancestral states with the tree that Blackburn D.C. 2008. Biogeography and evolution of body size and
has the strongest phylogenetic signal one is simply fit- life history of African frogs: phylogeny of squeakers (Arthroleptis)
ting more accurately to the assumptions of the ancestral and long-fingered frogs (Cardioglossa) estimated from mitochon-
state reconstruction method. We showed that this was drial data. Mol. Phylogenet. Evol. 49:806–826.
also true for characters not evolving under Brownian Blomberg S.P., Garland T., Ives A.R. 2003. Testing for phylogenetic sig-
nal in comparative data: behavioral traits are more labile. Evolu-
motion (i.e., the characters simulated on uncorrelated tion. 57:717–745.
branch lengths). Although our simulations were done Bromham L., Woolfit M., Lee M.S.Y., Rambaut A. 2002. Testing the re-
solely using continuous characters, a direct link between lationship between morphological and molecular rates of change
discrete and continuous characters exists in comparative along phylogenies. Evolution. 56:1921–1930.
Butler M.A., King A.A. 2004. Phylogenetic comparative analysis: a
methods (Felsenstein 2005). We thus expect the trends modeling approach for adaptive evolution. Am. Nat. 164:683–695.
obtained in this study to hold for discrete characters as Christin P.A., Besnard G., Samaritani E., Duvall M.R., Hodkinson T.R.,
well. Savolainen V., Salamin N. 2008. Oligocene CO2 decline promoted
Two common indices to measure phylogenetic signal C4 photosynthesis in grasses. Curr. Biol. 18:37–43.
were used in this study. In both simulations and em- Christin P.A., Freckleton R.P., Osborne C.P. 2010. Can phylogenetics
identify C4 origins and reversals? Trends Ecol. Evol. 25:403–409.
pirical studies, λ had a stronger relation with the infer- Colless D.H. 1995. Relative symmetry of cladograms and phenograms:
ence accuracy. We believe that, as λ is optimized solely an experimental study. Syst. Biol. 44:102–108.
on the internal branches, it may relate more to ancestral Crespi B.J., Teo R. 2002. Comparative phylogenetic analysis of the evo-
state inference in contrast to K, which is measured on lution of semelparity and life history in salmonid fishes. Evolution.
56:1008–1020.
the whole tree. This suggests that using λ would be more Cunningham C., Omland K.E., Oakley T.H. 1998. Reconstructing an-
appropriate when choosing between a chronogram and cestral character states: a critical reappraisal. Trends Ecol. Evol.
a phylogram to infer ancestral states. 13:361–366.
2012 POINT OF VIEW 5

Davies J.T., Savolainen V. 2006. Neutral theory, phylogenies, and the Nepokroeff M., Sytsma K.J., Wagner W.L., Zimmer E.A. 2003. Recon-
relationship between phenotypic change and evolutionary rates. structing ancestral patterns of colonization and dispersal in the
Evolution. 60:476–483. Hawaiian understory tree genus Psychotria (Rubiaceae): a com-
Drummond A.J., Rambaut A. 2007. BEAST: Bayesian evolutionary parison of parsimony and likelihood approaches. Syst. Biol. 52:
analysis by sampling trees. BMC Evol. Biol. 7:214. 820–838.
Ekman S., Andersen H.L., Wedin M. 2008. The limitations of an- Nikolaev S.I., Montoya-Burgos J.I., Popadin K., Parand L.,
cestral state reconstruction and the evolution of the ascus in the Margulies E.H., National Institutes of Health Intramural Sequenc-
Lecanorales (lichenized Ascomycota). Syst. Biol. 57:141–156. ing Center Comparative Sequencing Program, Antonarakis S.E.
Evans M.E.K., Smith S.A., Flynn R.S., Donoghue M.J. 2009. Climate, 2007. Life-history traits drive the evolutionary rates of mammalian
niche evolution, and diversification of the ”bird-cage” evening coding and noncoding genomic elements. Proc. Natl. Acad. Sci.
primroses (Oenothera, sections Anogra and Kleinia). Am. Nat. USA 104:20443–20448.
173:225–240. O’Meara B.C., Ané C., Sanderson M.J., Wainwright P.C. 2006. Testing

Downloaded from http://sysbio.oxfordjournals.org/ at Universidade Estadual Paulista Júlio de Mesquita Filho on March 6, 2012
Felsenstein J. 1985. Phylogenies and the comparative method. Am. for different rates of continuous trait evolution using likelihood.
Nat. 125:1–15. Evolution. 60:922–933.
Felsenstein J. 2005. Using the quantitative genetic threshold model Pagel M. 1994. Detecting correlated evolution on phylogenies: a gen-
for inferences between and within species. Philos. Trans. R. Soc. B eral method for the comparative analysis of discrete characters.
360:1427–1434. Proc. R. Soc. Lond. B. 255:37–45.
Foltz D.W., Hrincevich A.W., Rocha-Olivares A. 2004. Apparent selec- Pagel M. 1999. Inferring the historical patterns of biological evolution.
tion intensity for the cytochrome oxidase subunit I gene varies with Nature. 401:877–884.
mode of reproduction in echinoderms. Genetica. 122:115–125. Paradis E., Claude J., Strimmer K. 2004. APE: analyses of phylogenet-
Fontanillas E., Welch J.J., Thomas J.A., Bromham L. 2007. The influ- ics and evolution in R language. Bioinformatics. 20:289–290.
ence of body size and net diversification rate on molecular evolu- R Development Core Team 2011. R: a language and environment for
tion during the radiation of animal phyla. BMC Evol. Biol. 7:95. statistical computing. Vienna (Austria): R Foundation for Statistical
Friedman N.R., Hofmann C.M., Kondo B., Omland K.E. 2009. Corre- Computing.
lated evolution of migration and sexual dichromatism in the New Revell L.J., Harmon L.J., Collar D.C. 2008. Phylogenetic signal, evolu-
World orioles (Icterus). Evolution. 63:3269–3274. tionary process, and rate. Syst. Biol. 57:591–601.
Gabaldón T., and Huynen 2003. Reconstruction of the proto- Ronquist F. 2004. Bayesian inference of character evolution. Trends
mitochondrial metabolism. Science. 301:609. Ecol. Evol. 19:475–481.
Goldberg E.E., Igić B. 2008. On phylogenetic tests of irreversible evo- Salamin N., Wüest R.O., Lavergne S., Thuiller W., Pearman P.B. 2010.
lution. Evolution. 62:2727–2741. Assessing rapid evolution in a changing environment. Trends Ecol.
Harmon, L.J., Weir J.T., Brock C.D., Glor R.E., Challenger W. 2008. Evol. 25:692–698.
GEIGER: investigating evolutionary radiations. Bioinformatics. Salisbury B.A., Kim J. 2001. Ancestral state estimation and taxon sam-
24:129–131. pling density. Syst. Biol. 50:557–564.
Heath T.A., Hedtke S.M., Hillis D.M. 2008. Taxon sampling and the Sanderson M.J. 1997 A nonparametric approach to estimating di-
accuracy of phylogenetic analyses. J. Syst. Evol. 46:239–257. vergence times in the absence of rate constancy. Mol. Biol. Evol.
Jones C.S., Bakker F.T., Schlichting C.D., Nicotra A.B. 2009a. Leaf shape 14:1218–1231.
evolution in the South African genus Pelargonium L’Her. (Gerani- Sanderson M.J. 2002. Estimating absolute rates of molecular evolution
aceae). Evolution. 63:479–497. and divergence times: a penalized likelihood approach. Mol. Biol.
Jones K.E., Bielby J., Cardillo M., Fritz S.A., O’Dell J., Orme C.D.L., Safi Evol. 19:101–109.
K., Sechrest W., Boakes E.H., Carbone C., Connolly C., Cutts M.J., Seligmann H. 2010. Positive correlations between molecular and mor-
Foster J.K., Grenyer R., Habib M., Plaster C.A., Price S.A., Rigby phological rates of evolution. J. Theor. Biol. 264:799–807.
E.A., Rist J., Teacher A., Bininda-Emonds O.R.P., Gittleman J.L., Skinner A. 2010. Rate heterogeneity, ancestral character state recon-
Mace G.M., Purvis A. 2009b. PanTHERIA: a species-level database struction, and the evolution of limb morphology in Lerista (Scinci-
of life history, ecology, and geography of extant and recently extinct dae, Squamata). Syst. Biol. 59:723–740.
mammals. Ecology. 90:2648. Skinner A., Lee M.S.Y. 2010. Plausibility of inferred ancestral pheno-
Kembel S.W., Cowan P.D., Helmus M.R., Cornwell W.K., Morlon H., types and the evaluation of alternative models of limb evolution in
Ackerly D.D., Blomberg S.P., Webb C.O. 2010. Picante: R tools for scincid lizards. Biol. Lett. 6:354–358.
integrating phylogenies and ecology. Bioinformatics 26: 1463–1464. Smith S.A., Beaulieu J.M. 2009. Life history influences rates of climatic
Li G., Steel M., Zhang L. 2008. More taxa are not necessarily better for niche evolution in flowering plants. Proc. R. Soc. Lond. B. 276:4345–
the reconstruction of ancestral character states. Syst. Biol. 57:647- 4352.
653. Smith S.A., Donoghue M.J. 2008. Rates of molecular evolution are
Martins E.P. 1994. Estimating the rate of phenotypic evolution from linked to life history in flowering plants. Science. 322:86–89.
comparative data. Am. Nat. 144:193–209. Spradling T.A., Hafner M.S., Demastes J.W. 2001. Differences in rate
Mooers A.O. 2004. Effects of tree shape on the accuracy of maximum of cytochrome-b evolution among species of rodents. J. Mammal.
likelihood-based ancestor reconstructions. Syst. Biol. 53: 809–814. 82:65–80.
Mooers A.O., Vamosi S.M., Schluter D. 1999. Using phylogenies to Thomas J.A., Welch J.J., Lanfear R., Bromham L. 2010. A generation
test macroevolutionary hypotheses of trait evolution in cranes time effect on the rate of molecular evolution in invertebrates. Mol.
(Gruinae). Am. Nat. 154:249–259. Biol. Evol. 27:1173–1180.