Revealing the factors that promote speciationTimothy G. Barraclough1, Alfried P. Vogler1,2and Paul H. Harvey31Department of Biology & NERC Centre for Population Biology, Imper ial College at Silwood Park, Ascot, Berkshire SL5 7PY, UK2Department of Entomology,The Natural History Museum, Cromwell Road, London SW7 5BD, UK3Department of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, UKWhat biological attributes of organi sms promote speciation, and ultimately, species diversity? Thisquestion has a long history of interest, with proposed diversity promoters including attr ibutes such assexual selection, ecological specialism and dispersability. However, such ideas are di¤cult to test becausethe time-scale of processes involved is too great for direct human observation and experimentation. Anincreasingly powerf ul solution is to investigate diversity patterns among extant groups to infer the natureof processes operating during the evolution of those groups. This approach relies on the use of robust,phylogenetically based null models to overcome some of the problems inherent in observational inference.We illustrate this area by (i) discussing recent advances in identifying correlates of diversit y among highertaxa, and (ii) proposing new methods for analysing patterns in species-level phylogenies, drawingexamples from a wide range of organisms.Keywords: species diver sity; species-richness ; evolution; diversi¢cation; phylogeny; sister group1. INTRODUCTIONLife is made up of species, and so understanding the evolu-tion of species-richness is fundamental to ourunderstanding of the natural world. Traditionally,however, this area is plagued by the di¤culties inherentin studying a phenomenon operating over time-scalesmany orders of magnitude greater than our own lifespans(Panchen 1992). Recent accumulat ion of phylogeneticinformation through the use of molecular techniques hasprovided novel possibilities for statistical tests of hypoth-eses concerning the evolution of species-richness in extanttaxa (Mooers & Heard 1997; Nee et al. 1996a; Purvis 1996;Sanderson & Donoghue 1996). We illustrate this approachby concentrating on two issues. First, can biological attri-butes of lineages in£uence t he tendency for those lineagesto accumulate species-richness? Second, can we usespecies-level phylogenies to identify modes of speciationand patterns of subsequent phenotypic change?2. BIOLOGICAL FACTORS PROMOTINGSPECIES-RICHNESSLineages vary in the number of species they contain.Possible causes for t his variation include di¡erences in theenvironment experienced by those lineages (C racraft1985; Ricklefs & Schluter 1993; Rosenzweig 1995; Kerr &Packer 1997), or mere chance variation in probabilities ofspeciation and extinction (Raup et al. 1973; Raup 1985).An additional possibility is that lineages may vary withrespect to biological attributes that in£uence the net rateof cladogenesis, either through an e¡ect on speciation rateor extinction rate, or the equilibrium number of species alineage can realize. For example, many authors havesuggested that strong sexual selection by female choicemay promote speciation and ultimately species-richness(Darwin 1871; West-Eberhard 1983). Similarly, speciesrange size has been postulated as an attribute with astrong e¡ect on extinction probability (Jackson 1974;Jablonski 1987), and small body s ize has b een prop osed asan attribute permitting high equilibr ium species numberswithin lineages (Hutchinson & MacArthur 1959; Morse etal. 1985).To test ideas of this kind we need to look for replicateevidence for an asso ciation between the trait of interestand species-richness. A recent approach is to comparesister groups that di¡er in their express ion of the trait inquestion. There are four reasons why sister-group compar-ison is currently the best approach for identifyingevolutionary correlates of species-richness (Mitter et al.1988; Zeh et al. 19 89; Barraclough et al. 1998).1. Replication. By including several s ister-group compari-sons in our tests we gain replicated statistical evidencefor any trend we observe. This replication increasesour ability to make general conclusions about ane¡ect, reducing the possibility of detecting accidental,hi storical associations between traits and species-rich-ness (Cracraft 1990).2. Rates of diversi¢cation. Our interest is in testingwhether lineages with a particular attribute accumulatemore species than those without it. However, di¡erenttaxa have di¡erent ages, and so we might expect oldertaxa to have more species than younger taxa if they aregrowing simply because they have had time to accumu-late more species. Sister taxa are, by de¢nition, thesame age, and so by comparing the number of speciesbetween si ster groups we obtain direct estimates of rela-tive net rates.3. Non-independence of taxa. Several taxa may share ahigh species accumulation rate and the same value of abiological attribute simply because they inherited bothPhil. Trans. R. Soc. Lond. B (1998) 353, 241^249 241 & 1998 The Royal Societyfrom a common ancestor. In this case these tax a do notrepresent independent data points supporting an evolu-tionary correlation b etween species diversity and thebiological variable as both traits have evolved onlyonce. Sister-group comparisons overcome this problembecause di¡erences between sister taxa have evolvedsubsequent to their divergence and so necessarily repre-sent independent evolutionary events (MÖller &Birkhead 1992).4. Confounding variables and noise. Taxa di¡er in manyattributes apart from the trait of pr imary interest, X,and diversi ¢cat ion rate. Some of these attributes mayin£uence diversi¢cation rate, causing error variationor noise in the data that may obscure any patternarising from an e¡ect of trait X on diversi¢cationrates. In addition, trait X may have no direct in£uenceon diversi¢cation rates, but we may observe a correla-tion through the in£uence of unaccounted-forconfounding variables. The shared common ancestryof sister taxa means they will tend to be similar toeach other in many respects. By comparing sister taxawe are comparing like with like, t hereby controlling formuch potential noise and c onfounding variation whichmight otherwise a¥ict our analysis (Read & Nee 1995;Harvey et al. 1995; Nee et al. 1996b).Two additional features of sister-group comparison shouldbe mentioned with respect to drawing conclusions. First,these tests detect a correlation and so
View Full Document
Unlocking...