How Do Characters Evolve? Number of Species Are Allowed to Vary Inde- Arising From: R

brief communications arising

Evolution variance, one must analyse data, whether observed or simulated,in which time and the How do characters evolve? number of species are allowed to vary inde- Arising from: R. E. Ricklefs Nature 430, 338–341 (2004) pendently. Purvis’s simulations lack this independence: the first does not vary clade icklefs1 claims to show that morpho- both models. Further, both models predict age; the second holds the species number logical evolution in birds is associated that clade age too will correlate with variance, constant. The results are not surprising, and Rwith speciation events — that is, it is but Ricklefs found no significant association. they do not address the validity of my analy- punctuational — by inference from data on Why not? Ricklefs proposes a model in which sis. Although average divergence time and only species number, clade age and character speciation promotes gradual divergent number of species are correlated among variance from a range of passerine clades. He change. It is unclear what predictions his clades, multiple regression allows one to suggests that variance increases in propor- model makes about his data. Alternatively, identify unique contributions of time and tion with clade age under gradual change, the negative result may simply be a type II species number to trait variance among but in proportion to the logarithm of species error.Clade age was only weakly predictive of clades of different ages. number if change is punctuational. Here I variance in my simulations, and the phylo- Purvis’s comment about a type II error show that both clade age and the logarithm geny from which the clade ages are taken4 con- (failing to detect a true relationship) is of of species number independently predict flicts markedly with more recent evidence5. more concern, particularly because relative variance under both gradual and punctua- The mode of character evolution can age, based on DNA-hybridization phylo- tional change, rendering Ricklefs’ results sometimes be inferred using detailed geny6, is estimated less well than the number uninformative about his central hypothesis. phylogenetic information3,6,7.Clade ages and of species. In a test of the monophyly of 40 First,I simulated 100 clades that survived species numbers alone are not enough. out of 106 (from ref. 6) tribe-to-family-level to the same age (60 units) under a constant- Andy Purvis clades using a maximum-likelihood analysis rates birth–death process (speciation rate, Department of Biological Sciences, of over 4 kilobases of the RAG-1 and RAG-2 0.2; extinction rate, 0.16). Trees contained Imperial College London, Silwood Park Campus, genes7, only three were found that were from one species (excluded from analyses) to Ascot, Berks SL5 7PY, UK significantly paraphyletic compared with 27 238 species (mean,49.3).I evolved two char- e-mail: [email protected] strongly supported and 10 ambiguous clades8. acters on each tree under brownian motion doi:10.1038/nature03092 This does not “conflict markedly” with the — one gradual, the other changing only at 1. Ricklefs, R. E. Nature 430, 338–341 (2004). monophyly of clades used in my analysis. It speciation. Log(species number) is a highly 2. Felsenstein, J. Am. Nat. 125, 1–15 (1985). is of more importance, however, that the significant predictor of variance in the 3. Bokma, F. J. Evol. Biol. 15, 1048–1056 (2002). relative ages of clades in the sequence-based ǃ 4. Sibley, C. G. & Ahlquist, J. E. Phylogeny and Classification of evolved trait under both gradual (t96 5.42, Birds: A Case Study of Molecular Evolution (Yale Univ. Press, and DNA-hybridization phylogenies were 2 P<0.0001, r ǃ0.23) and punctuational New Haven, Connecticut, 1990). not compared. ǃ < 2ǃ 5. Barker, F. K., Cibois, A., Schikler, P., Feinstein, J. & Cracraft, J. change (t96 7.57,P 0.0001,r 0.37),even If the phylogeny in ref. 6 provides a among clades of the same age. Proc. Natl Acad. Sci. USA 101, 11040–11045 (2004). reasonably accurate view of clade age, then 6. Mooers, A. Ø., Vamosi, S. M. & Schluter, D. Am. Nat. 154, Second, I evolved 100 surviving clades 249–259 (1999). the absence of a significant time effect until they reached a fixed size (50 extant 7. Pagel, M. Nature 401, 877–884 (1999). on variance among different-aged clades species) by using the same process and would be sufficient to reject a model of evolved characters as before. Clade age gradual evolution that is independent of ranged from 23.7 to 176.2 time units and Ricklefs replies — Purvis1 states that, under species number. Gradual evolutionary significantly (though weakly) predicted random gradual change, clades accumulate divergence, whether fast or slow, driven by variance in the evolved trait under both variance in proportion to their total branch interactions among species in a clade (as ǃ ǃ 2ǃ gradual (t98 2.03, P 0.046, r 0.03) and length. Accordingly, at a given age, clades opposed to punctuated evolution associated ǃ ǃ 2ǃ 9 punctuational(t98 2.32,P 0.02,r 0.04) with more species should exhibit greater with speciation ) is also species-dependent, models,even among clades of the same size. variance. This is incorrect, as shown both rather than time-dependent, inasmuch as These results are not surprising. Under analytically2,3 and by simulation4,5, and the the pressure to diversify is in some way gradual change, variance accumulates along error underlines a misunderstanding of my proportional to species number. Increasing phylogenetic branches2. Larger clades have analysis. knowledge of phylogenetic relationships more total branch length within them, even Variance depends strictly on the average makes this an opportune time to examine in same-aged clades (across my trees, Spear- time of divergence between species within a more closely the generation of trait variance man’s rǃ0.98),and so have more variance in clade and not on the number of species or in diversifying clades. gradually evolved traits.With punctuational total branch length. The positive relation- Robert E. Ricklefs change,variance accumulates only at specia- ship between species number and variance in Department of Biology, University of Missouri tions, but older clades have experienced Purvis’s simulations with fixed time reflects at St Louis, St Louis, Missouri 63121-4499, USA more speciation (and extinction) events — the occurrence of earlier branch points in e-mail: [email protected] even in same-sized clades — to an extent what turn out to be larger clades. Had Purvis doi:10.1038/nature03093 3 that depends on the extinction rate .Clade simulated gradual evolution in an uncon- 1. Purvis, A. Nature doi:10.1038/nature03092 (2004). age is therefore a good predictor of number strained speciation–extinction process with 2. Slatkin, M. Paleobiology 7, 421–425 (1981). 3. Gavrilets, S. Proc. R. Soc. Lond. B 266, 817–824 (1999). of speciations across the same-sized trees varied clade ages, he would have obtained a 4. Raup, D. M. & Gould, S. J. Syst. Zool. 23, 305–322 (1974). (Spearman’s rǃ0.81). significant partial correlation between vari- 5. Valentine, J. W., Collins, A. G. & Meyer, C. P. Paleobiology 20, Ricklefs1 found that log(species number) ance and time, which I was unable to detect 131–142 (1994). but not clade age independently pre- in clades of passerine birds. The weaker time 6. Sibley, C. G. & Ahlquist, J. E. Phylogeny and Classification of the Birds of the World (Yale Univ. Press, New Haven, Connecticut, dicted morphological variance in multiple effect in Purvis’s second simulation with 1990). regressions, and concludes from this that constrained clade size reflects the distribu- 7. Sibley,C.G.& Monroe,B.L.Jr Distribution and Taxonomy of morphological evolution in birds seems to be tion of most nodes in each tree at similar Birds of the World (Yale Univ. Press, New Haven, Connecticut, associated with cladogenesis. However, my depth,regardless of the age of the root. 1990). 8. Barker, F. K., Cibois, A., Schikler, P., Feinstein, J. & Cracraft, J. simulations show that his significant results To determine whether time itself, and Proc. Natl Acad. Sci. USA 101, 11040–11045 (2004). for log(species number) are expected under hence gradual evolution, contributes to trait 9. Gould, S. J. & Eldredge, N. Paleobiology 3, 115–151 (1977).