The Auk 113(4):842-848, 1996 REPEATED EVOLUTION OF SEXUAL COLOR DIMORPHISM IN PASSERINE BIRDS TREVOR PRICE • AND GEOFFREY L. BIRCH Departmentof Biology0116, University of Californiaat SanDiego, La Jolla,California 92093, USA ABSTR•CT.--Wepresent a surveyof passefinebirds designed to investigatethe frequency with which sexualdimorphism in colorationor color pattern has evolved from monomor- phisin(or the converse).Based on the numberof generathat haveboth a monomorphicand a dimorphicspecies, and the minimumnumber of changesinferred to haveoccurred between genera,the transition between dimorphismand monomorphismhas occurredat least 150 times.Using the Sibley/Ahlquistphylogeny, we obtain maximumlikelihood estimatesof the probabilitythat one statewill be in the other after one million yearsof 0.01 to 0.02 (monomorphismto dimorphism)and 0.03to 0.04(dimorphism to monomorphism).The rate of transitionfrom dimorphismto monomorphismappears to be higher than the converse, and there are more monomorphicthan dimorphic species.We concludethat the transition betweenalternative states is notdifficult, and that the evolutionof sexualdimorphism, given appropriateselection pressures, is unlikelyto beconstrained. Received 15 August 1995, accepted I April 1996. MANYSPECIES OF BIRDS are sexuallydimorphic iceus),suggests that correlatedresponses do oc- in color or color pattern. This often is thought cur (Muma and Weatherhead 1989). However, to be a consequenceof sexual selection, and rudimentaryfemale traits,at leastin somespe- sexualdimorphism in color has been usedas an cies, may be maintained at their observedlevel index of the intensity of sexualselection in com- of expressionby selection.Several analyses have parative studies (Hamilton and Zuk 1982, Read implicated selectionon female plumagesas a and Harvey 1989,Fitzpatrick 1994,Moller and main cause of the presence or absenceof di- Birkhead 1994, Barracloughet al. 1995). Popu- morphism (Bj6rklund 1991, Irwin 1994). lation studieshave confirmed that male plum- Therefore, the degreeto which the evolution age patterns in sexually dimorphic speciesare of dimorphismis limited or preventedby a high subject to sexual selection (Price 1984, Moller genetic correlation between the sexesremains 1989, Hill 1990, Petrie and Halliday 1994),and unclear.In this paperwe estimatethe minimum Moller and Birkhead (1994) show that dimor- number of times dimorphism has evolved in phisin is correlatedwith the frequencyof ex- the Passeriformes. We find it has evolved at trapair copulationsacross species. least 150 times.We suggestthat this is a high The evolutionof sexualdimorphism requires number,and that the evolution of dimorphism not only appropriateselection pressures but also from monomorphism (or its converse,mono- sufficientgenetic variation. Lande (1980) sug- morphismfrom dimorphism)is not constrained gestedthat the evolution of sexualdimorphism greatly by an absenceof genetic variation for might be constrained,because many of the genes dimorphism. affectingvariation in malessimilarly affect vari- ation in females. Selection on males therefore METHODS resultsin a high correlatedresponse in females, and sex-limitation may require many genera- Throughout this paper, monomorphismand di- tions of selectionto separatethe expressionof morphism are used as shorthand for sexual monoch- romatism and sexual dichromatism. The trait "sexual thosefew genesthat affectthe male and female dimorphism"is particularlysuitable for study be- differently. The presenceof rudimentary male causeof the many specieswhose phenotypic state is traits in females, such as reduced epaulets in known. Therefore,even low frequenciesof transition female Red-winged Blackbirds(Agelaius phoen- between states should be detectable. We do not con- sidersexual dimorphism in size,or in ornamentssuch as wattles,plumes, and elongatetails. E-mail: [email protected] We studiedthe 5,398 passerinebird speciesin the 842 October 1996] SexualDimorphism in Birds 843 100 171 157 135 98 51 20 1,125 genera listed in Clements' (1981) check-list of E 439 birds of the world. A more recent list by Sibley and • 80 Monroe (1990) is similar to Clements in the classifi- cation of speciesto genera (but with 5,705 speciesin 1,164genera), and the resultsare identical regardless c• 6o of which compilation is used. Using field guides,we were able to determine the presenceor absenceof sexualdimorphism in color or color pattern for spe- • 4o cies in all but 54 of the genera (66 species).We clas- sified a speciesas sexually dimorphic if the sexeswere •O 20 describedseparately in species'descriptions, or if the sexesappeared different in illustrations. Therefore, •' 0 minor differences,such as more spotson a larger bird, I 2 3-4 5-8 9-16 16-32 >32 would not be counted. Nevertheless,the degree of dimorphismvaries from strongto mild. Examplesof NUMBER OF SPECIES IN GENUS specieswith weak sexualdimorphism are the Tropical FIG. 1. The proportion of passerinebird genera Parula (Parulapitayumi), in which the female lacksthe containing all mononaorphic(filled squares),all di- male'sorange breast band, and the White-browedTit- morphic(open squares ), or mixed mononaorphicand Warbler (Leptopoecilesophiae), in which the female is dimorphicspecies (filled circles)for generawith dif- paler and without much of the male's blue wash. ferent numbersof species.Numbers of generain each Speciesthat are seasonallydimorphic (e.g. Scarlet classare listed at the top of the figure. Tanager [Pirangaolivacea]) were classified as dimor- phic, whereasthose with polymorphismsthat are not sex limited (e.g. Scaly-breastedWren-Babbler [Pneo- mateof relationshipsamong genera currently avail- pyga albiventer],All and Ripley [1983]; and several able. Using parsimonywe then reconstructedances- Wheatear species[Oenanthe], Mayr and Stresemann tral stateson the tree using the ACCTRAN method [1949]) were classifiedas mononaorphic.We did not (Maddison1989). This gives a single parsimonious record the plumage state of every species,but rather estimate,and there may be other equally parsimo- whether a genusconsisted entirely of mononaorphic nious trees (Maddison 1989). We did not search for species,entirely of dimorphic species,or a mixture. these becausewe were interestedmainly in estimat- Therefore, we are unable to place an exact figure on ing minimum frequencyof change.The method as- the proportion of all speciesthat are mononaorphic. signs each node below the tips of the tree as mono- We estimatethat at least 60% of all speciesare mono- morphic (M), dimorphic (D), or mononaorphicor di- morphic (Fig. 1). Barracloughet al. (1995),in a smaller morphic(M/D) dependingon the stateof the genera sample of passerines,found that 69% were mono- above the node. One then works down the tree scor- morphic.Some species may be crypticallydimorphic ing eachnode according to the statesof the two nodes in ultraviolet light, in which casethe number of di- above(e.g. if the two nodesabove were M/D and D, morphic specieswould be underestimated.However, the node below is D). Finally, when all nodes had no specieshas been discoveredto be sexuallydimor- been assignedwe worked back up the tree assigning phic in ultraviolet light but mononaorphicto humans, M/D nodes as M or D depending on their inferred despite an extensive search (Staffan Andersson pers. ancestor. After all nodes had been reconstructed as comm.). mononaorphicor dimorphic, we tallied the number The main goal of the study was to estimate the of transitions between the two states. frequencyof transitionsbetween monomorphism and If changehas been frequent,parsimony is likely to dimorphism.To do this we used two approaches,i.e. produce a gross underestimateof the number of within generaand among genera. changes.For example,in the extremecase of a change Within genera.--We tallied the number of genera in state at every speciationevent, parsimony will re- containingboth a mononaorphicand a dimorphicspe- suit in a constructionwith no change below sister cies, excluding from considerationspecies that are pairs at the tips of the tree. Therefore, we used the both mononaorphicand dimorphic in different parts maximum likelihood method of Pagel (1994) to di- of their range. If genera are assumedto be mono- rectly estimatefrequency of change.The method ob- phyletic, then each genuscontaining both a mono- tains those probabilitiesof change from mononaor- morphic and a dimorphic speciesrepresents an in- phic to dimorphic and from dimorphic to mononaor- dependent evolutionary transition between the two phic that maximize the probability of the observed states. statesat the tips of the tree. The main assumptionis Amonggenera.--We located 399 of the genera on that the probability of changeper unit branch length Sibley and Ahlquist's(1990) phylogenetic tree based is constantthroughout the tree. We assumedthat the on DNA-DNA hybridization data. The tree clearly Sibley-Ahlquisttree and associatedbranch lengths will have inaccuracies,but it providesthe best esti- were a given parameter. Branch-lengthswere mea- 844 PRICEAND BIRCH [Auk,Vol. 113 suredwith a ruler. We usedall 594 tips in the phy~ timate of the probability that a dimorphic spe- logeny as input, so that when severalspecies for one cieswould be monomorphicafter 1 million years genuswere separatelyplaced on the tree they were varied from 3.4 to 4.4%. These estimates are con- treated separately.There were 26 tips for which we founded by many uncertainties,but they are were unable to identify plumagestate of the species consistentwith the