Plumage Convergence in Picoides Woodpeckers Based on a Molecular Phylogeny, with Emphasis on Convergence in Downy and Hairy Woodpeckers Amy C

Home , Character evolution, Dendropicos, Picoides, Veniliornis

Wayne State University

Biological Sciences Faculty Research Publications Biological Sciences

11-1-2005 Plumage convergence in Picoides woodpeckers based on a molecular phylogeny, with emphasis on convergence in downy and hairy woodpeckers Amy C. Weibel Wayne State University, [email protected]

William S. Moore Wayne State University, [email protected]

Recommended Citation Weibel, A. C. and Moore, W. S. 2005. Plumage convergence in Picoides woodpeckers based on a molecular phylogeny, with emphasis on convergence in downy and hairy woodpeckers. Condor 107(4):797-809. http://dx.doi.org/10.1650/7858.1 Available at: http://digitalcommons.wayne.edu/biosci_frp/11

This Article is brought to you for free and open access by the Biological Sciences at DigitalCommons@WayneState. It has been accepted for inclusion in Biological Sciences Faculty Research Publications by an authorized administrator of DigitalCommons@WayneState. The Condor 107:797±809 ᭧ The Cooper Ornithological Society 2005

PLUMAGE CONVERGENCE IN PICOIDES WOODPECKERS BASED ON A MOLECULAR PHYLOGENY, WITH EMPHASIS ON CONVERGENCE IN DOWNY AND HAIRY WOODPECKERS

AMY C. WEIBEL1 AND WILLIAM S. MOORE Department of Biological Sciences, Wayne State University, Detroit, MI 48202

Abstract. Adult and juvenile plumage characters were traced onto a well-resolved molecular based phylogeny for Picoides woodpeckers, and a simple phylogenetic test of homology, parallelism, and convergence of plumage characters was performed. Reconstruction of ancestral character states revealed multiple events of independent evolution of derived character states in most characters studied, and a concentrated changes test revealed that some plumage characters evolved in association with habitat type. For example, there was a statistically signi®cant association between loss of dorsal barring and use of densely veg- etated habitats among Picoides species. Two analyses indicated that convergence, as opposed to parallel evolution or shared ancestry, underlies the similarity in plumage patterns between the Downy (Picoides pubescens) and Hairy (P. villosus) Woodpeckers. Possible causal explanations for convergence in plumage patterns may include mimicry and interspeci®c territoriality. Key words: adaptation, character evolution, convergence, Picoides, plumage, woodpeckers.

Convergencia en Plumaje en PaÂjaros Carpinteros del GeÂnero Picoides Basada en una Filogenia Molecular, con EÂ nfasis en la ConvergeÂncia entre Picoides pubescens y P. villosus Resumen. Se reconstruyoÂ la evolucioÂn de caracteres del plumaje de individuos adultos y juveniles con base en una ®logenia molecular bien resuelta de los paÂjaros carpinteros del geÂnero Picoides. El estudio provee una prueba ®logeneÂtica sencilla de homologõÂa, parale- lismo y convergencia para los caracteres de plumaje. Las reconstrucciones de estados de caracter ancestrales revelaron muÂltiples eventos de evolucioÂn independiente de estados de caracter derivados en casi todos los caracteres estudiados, y una prueba de cambios con- centrados reveloÂ que algunos caracteres del plumaje evolucionaron en asociacioÂn con el tipo de haÂbitat. Por ejemplo, existe una asociacioÂn estadõÂsticamente signi®cativa entre la peÂrdida de barras dorsales y el uso de ambientes con vegetacioÂn densa en las especies de Picoides. Dos anaÂlisis indicaron que la similitud en los patrones de plumaje entre Picoides pubescens y P. villosus puede explicarse por convergencia, no por evolucioÂn paralela o por ancestrõÂa comuÂn. Se discuten posibles explicaciones causales para la convergencia, como la imitacioÂn y la territorialidad interespecõÂ®ca.

INTRODUCTION forces in species that experience similar envi- The comparative method is a general and pow- ronments (Bell 1989, Harvey and Pagel 1991, erful approach for testing hypotheses of adap- Nee et al. 1996). Independent evolutionary tation (Ridley 1983) and identifying evolution- events can be identi®ed from a well-resolved ary trends (Ridley 1983, Harvey and Pagel phylogeny of the group of taxa under investi- 1991) across a range of taxonomic groups. Char- gation (Pagel and Harvey 1988, Losos 1990, acter states may be maintained in descendent Brooks and McLennan 1991, Omland 1999) and lineages and persist in extant sister species be- analyzed using the comparative method. cause of shared ancestry (Wilson 1975, Harvey Phenotypic resemblance among species may and Mace 1982, McKitrick 1993) or may evolve result from homology, parallel evolution, or con- independently in response to similar selective vergence. Homology of characters implies struc- tural resemblance due to shared ancestry (Boy- den 1973, Patterson 1982). In a review of his- Manuscript received 19 April 2005; accepted 10 Au- torical concepts and de®nitions of homology, gust 2005. 1 Present address: Academic Affairs, Grand Canyon Patterson (1982) noted that true homologies are University, Phoenix, AZ 85017. E-mail: acweibel@ synapomorphies, and thus the test for homology msn.com of similar character states is also a test of con-

[797] 798 AMY C. WEIBEL AND WILLIAM S. MOORE gruence within a monophyletic group. Similar METHODS character states resulting from parallel or con- The phylogeny that served as the comparative vergent evolution are homoplasies and, collec- framework for this study was estimated from tively, are distinguished from homologies by in- pooled DNA sequences of two mitochondrial congruence with other character states that de- protein-coding genes, cytochrome oxidase I ®ne a monophyletic group. Parallel and conver- (COI) and cytochrome b (cyt b), and a nuclear gent evolution are more dif®cult to distinguish gene intron, ␤-®brinogen intron 7 (␤-®bint7) from each other. Parallel evolution implies struc- (Weibel and Moore 2002b). The phylogeny is tural resemblance in derived character states the maximum likelihood topology for the data among closely related species in which devel- set based on a general time-reversible nucleotide opmental pathways were similarly yet indepen- rate variation model that included the proportion dently modi®ed after separation of evolutionary of invariable sites and the gamma distribution lineages from a common ancestor, whereas con- shape parameter (GTR ϩ I ϩ⌫). The portion of vergence in phenotypes results from modi®ca- the Weibel and Moore (2002b) tree relevant to tion of different antecedent characters or differ- this study is presented in Figure 2 and 3. ent developmental pathways in distantly related Adult plumage data were collected from study species (Futuyma 1998). Thus, the subtle dis- skins housed at the University of Michigan Mu- tinction between parallel and convergent evolu- seum of Natural History. Male and female adult tion is linked to development as well as phylog- plumages were evaluated using characters and eny, and inferences that distinguish the two can methods following Short (1971). Because of be made by studying developmental pathways substantial geographic variation in most Picoi- that transform juvenile to adult character states des species, we selected for study subspecies in a phylogenetic context. that were used by Winkler et al. (1995) for spe- In this study we used MacClade (Maddison cies descriptions. Juvenile plumage character and Maddison 1992) to examine plumage char- data were extracted primarily from Short (1971, acter data of Picoides woodpeckers in the con- 1982), Ehrlich et al. (1988), and Winkler et al. text of a well-resolved, DNA-sequence-based (1995). Data were arranged into two matrices phylogeny. Characters are traced or mapped (adult plumage and juvenile plumage). Charac- onto a phylogeny using the criterion of maxi- ters were individually traced onto the 3-gene mum parsimony; character state transformations species tree using the criterion of simple parsi- are polarized and ancestral character states are mony (unordered parsimony) in MacClade, version 3.0 (Maddison and Maddison 1992), in inferred. Thus, patterns of character evolution which gains and losses are given equal proba- are detected that allow hypotheses regarding the bilities and ancestral states are reconstructed direction of evolution, adaptation, and develop- based on character states of extant lineages. The mental programs to be tested. phylogeny is rooted (Weibel and Moore 2002a, A particularly intriguing example of apparent 2002b), thus character states are polarized al- convergence in plumage characteristics occurs lowing inference of ancestral versus derived in the Hairy (Picoides villosus) and Downy (P. states in the Picoides clade. pubescens) Woodpeckers. These two broadly Levels of similarity in plumage patterns sympatric species are so similar in plumage that among species pairs were evaluated, and the to- they can be dif®cult to distinguish in the ®eld tal number of character states (ancestral ϩ de- without side-by-side comparison, yet they are rived) held in common for all pairwise species not closely related (Weibel and Moore 2002a, comparisons in each of the two plumage matri- 2002b). Although each species belongs to a dis- ces were tabulated. The empirical distributions tinct clade, they more closely resemble each oth- of total character states held in common between er than other members of their respective clades pairs of species were represented in histograms (Fig. 1). Here we test the hypothesis of conver- for adult and juvenile plumage data using NCSS gence in plumage characters in Picoides wood- 2000 statistical software (Hintze 1999). Pairs of peckers and discuss potential adaptive causes of species for which high overall resemblance in convergence speci®cally in Hairy and Downy adult plumage is outwardly apparent were lo- Woodpeckers. cated on the histograms. Species pairs with the PLUMAGE CONVERGENCE IN PICOIDES WOODPECKERS 799

FIGURE 1. Sketches of North American ``small'' Picoides: P. nuttallii, P. scalaris, P. pubescens, P. minor (a Eurasian species); and ``large'' Picoides: P. albolarvatus, P. stricklandi, P. villosus, P. borealis. Note the ex- tensive similarity in plumage between the distantly related species, P. pubescens and P. villosus. Drawings by John Megahan from study skins housed in the University of Michigan Museum of Natural History. greatest level of plumage similarity are expected a binary dependent character (in this case, dense to share the greatest number of character states versus open habitat) are randomly distributed in (both ancestral and derived states). Hypotheses a de®ned clade regardless of the state of another of homology, parallelism, and convergence were independent character (in this case, barred ver- tested among species pairs having high levels of sus solid back, mantle, or wing patterning). As- similarity in adult plumage patterns by compar- signment of characters in this way stems from ing the total number of shared adult plumage the logic that phenotypic characters arise inde- character states with the total number of shared pendent of environment, and those characters juvenile plumage character states. that allow an individual to exploit some com- Field observations revealed a trend of de- ponent of the environment are favored by selec- creased dorsal barring or streaking in wood- tion (Westneat 1995). A probability distribution peckers that utilize dense vegetation rather than for gains and losses of the dependent character open habitat (WSM, unpubl. data). We tested is generated by simulation. An independent this hypothesis for Picoides by a concentrated character is traced onto the phylogenetic frame- changes test (Maddison 1990) in MacClade. The work, and signi®cant evolutionary association analysis tests the null hypothesis that changes in between dependent and independent character 800 AMY C. WEIBEL AND WILLIAM S. MOORE states is determined from the null probability Fig. 2) and mantle (character 7; Appendix, Fig. distribution using the observed number of gains 3b) patterning. There is an overall loss of head and losses of the dependent character state in patterning in Picoides species as shown by the lineages having a speci®ed independent charac- evolutionary reduction of a distinct malar patch ter state. (character 1), the narrowing of the auricular patch (character 5), the reduced expression of RESULTS male ornamentation (character 10), and the Plumage characters and character states (ances- transformation from distinct white nasal tufts tral vs. derived) are listed in the Appendix, with (character 13) to black or indistinct dusky nasal species grouped according to phylogenetic re- tufts (Appendix). Overall body patterning in Pi- lationships based on Weibel and Moore (2002b). coides has also decreased over evolutionary Data for species of a given clade are represented time: barring on the outer retrices (character 3; with a unique letter (A±G) such that character Appendix), back (character 6; Appendix, Fig. states in the context of phylogenetic relation- 3a), and wing (character 9; Appendix, Fig. 3c) ships can be readily identi®ed. Intraspeci®c var- are ancestral states, as is ventral speckling (char- iation occurred in all characters studied, thus acter 8; Appendix). However, ventral barring character state assignments are likely simplistic (character 8) has only recently evolved in Ve- and not descriptive of a species as a whole. niliornis (Appendix), and barring is derived in However, character state assignments do re¯ect Picoides in some less conspicuous characters: the general phenotype of a given species (West- inner retrices (character 4; Appendix, Fig. 4), neat 1995). The derived character state evolved mantle (character 7; Appendix, Fig. 3b) and multiple times independently in all characters throat (character 12; Appendix) patterning. studied; i.e., homoplasy is high in adult plumage Thus, adult plumage of the common ancestor to characters. Ancestral character states for some Picoides was likely dorsally barred or spotted characters are not resolved; these equivocal with speckled ventral patterning and well-devel- states are noted in the Appendix. oped head markings but with no markings on Reversals to the ancestral condition occurred the throat. Male ornamentation was expressed in two characters: crown (character 2; Appendix, on the crown with no streaking or spotting.

FIGURE 2. Relevant portion of the phylogenetic tree showing rejection of homology for four characters held in common between P. pubescens and P. villosus. Plumage similarity between these two species results from evolutionary reversals in two characters (crown streaking and mantle barring) and independent evolution of a full supercilium and loss of back barring. Characters are numbered according to their listing in the Appendix. Ancestral characters states are inferred from the species tree of all Picoides species studied thus far (Weibel and Moore 2002a, 2002b). PLUMAGE CONVERGENCE IN PICOIDES WOODPECKERS 801

FIGURE 3. Reconstruction of ancestral character state based on a simple parsimony criterion for plumage characters in Picoides woodpeckers: (a) back pattern, (b) mantle pattern, and (c) wing pattern of a folded wing. Species utilization of open versus dense habitat is mapped on the phylogeny. An asterisk is used in (a) to identify those nodes supported by bootstrap values that are between 50% and 70%.

Little is described or known about juvenile (character 3) in chicks evolved at least twice plumage patterns. Of the six characters studied, with two possible reversals to the ancestral state. four show evidence of multiple independent Diffuse ventral patterning (character 5) may evolution of the derived state (Appendix). have evolved independently several times, but Crown streaking (character 2) may have arisen the precise number of events is dif®cult to assess three times, and strong sexual dimorphism because of unresolved transformations in several 802 AMY C. WEIBEL AND WILLIAM S. MOORE

use also occurred in lineages of woodpeckers having a certain character state was calculated with a concentrated changes test by comparing the tracing of dense habitat use in the phylogeny with the tracing of the character state. For example, dense habitat use was gained along four lineages having solid back patterning and was lost in one lineage with solid patterning (Fig. 3a). Thus, the probability of observing four gains and one loss of dense habitat use in lineages of woodpeckers with solid back patterning by random chance is 0.003. Signi®cant evolutionary associations (Table 1) were detected between dense vegetation and solid back pattern- FIGURE 4. Relevant part of the phylogenetic tree ing (Fig. 3a), solid mantle patterning (Fig. 3b), showing rejection of homology for three characters and solid wing patterning (Fig. 3c). held in common between P. canicapillus and P. kizuki. Histograms from all (n ϭ 190) pairwise spe- Plumage similarity between these two species results cies comparisons for number of shared character from shared ancestral character states of reduced supercilium and lack of barring in inner retrices and states in adult and juvenile plumages are shown mantle; no perceptible evolution has occurred in these in Figure 5. In general, all Picoides species have characters. Characters are numbered according to their similar black and white plumage patterns with listing in the Appendix. Ancestral characters states are variable barring, streaking, or spotting (Fig. 1). inferred from the species tree of all Picoides species studied thus far (Weibel and Moore 2002a, 2002b). Each of the species pairs was mapped onto the distribution of shared adult plumage characters (Fig. 5a) and shared juvenile plumage characters ancestral lineages of the genus. However, it is (Fig. 5b). Three species pairs with visibly sim- clear that ventral speckling in chicks, regardless ilar adult plumage characteristics fall in the ex- of pattern intensity, is found in all species of treme tail of the empirical distribution. Species Picoides studied thus far. Juvenile plumage in pairs with the most similar adult plumage pat- the common ancestor to Picoides may have been terns are P. lignarius±P. mixtus (A) from South expressed as unstreaked crown coloration with America (13 of 13 characters), P. pubescens±P. subtle or no sexual dimorphism; both female and villosus (B) from North America (12 of 13 char- male chicks display adult male head ornamen- acters), and P. canicapillus±P. kizuki (C) from tation. The overall patterning of juveniles was Asia (11 of 13 characters). Two other species likely similar to adults, but with duller dorsal pairs sharing 11 of 13 character states (P. sca- coloration and heavier ventral patterning. laris±P. minor and Dendropicos fuscescens±P. Dense habitat use evolved at least four times maculatus) are not visibly similar in appearance in the Picoides phylogeny (Fig. 3), with one loss though their character states are the same for (or reversal to open habitat use) in the P. strick- most characters. For example, the absence of an landi lineage. The probability that an observed auricular patch on P. minor clearly distinguishes number of gains (and losses) of dense habitat the species from P. scalaris, and D. fuscescens

TABLE 1. Concentrated changes of observed character associations in Picoides woodpeckers. Decimal values are probabilities that evolutionary association of characters occur other than by random chance. Bold probabilities coincide with the observed loss of dense vegetation use and gains in this type of habitat use in lineages having solid back, mantle, or wing patterning based on Figure 3.

Gains Independent Observed characterÐtraced Dependent character losses 01234 Back pattern (solid) Habitat (dense vegetation) 1 0.040 0.154 0.152 0.045 0.003 Mantle pattern (solid) Habitat (dense vegetation) 1 0.014 0.127 0.255 0.135 0.016 Wing pattern (solid) Habitat (dense vegetation) 1 0.094 0.101 0.024 0.001 0.000 PLUMAGE CONVERGENCE IN PICOIDES WOODPECKERS 803

troduce a suite of problems for correctly recon- structing evolutionary relationships, as characters may not evolve independently and their genetic basis is poorly understood (Hillis 1987). The plumage patterns used by Short (1971) and examined here are clearly subject to evolutionary convergence and reversal (Christidis et al. 1988, Omland and Lanyon 2000). Furthermore, plumage patterns that are sexually selected (Burns 1998) may be driven by genetic expression but may also be in¯uenced by hormonal changes or environmentally induced developmental changes. Thus, plumage characters are considered highly labile and of dubious value for phylogenetic analysis (Christidis et al. 1988, Hackett and Rosenberg 1990, Kusmierski et al. 1997, Omland and Lanyon 2000, but see Live- zey 1991, Prum 1997, Chu 1998). Perhaps the greatest dif®culty with Short's (1971) evaluation of character evolution in Picoides woodpeckers is the lack of independence between the character data under study and the data used to re- FIGURE 5. Histograms of all pairwise species com- construct the phylogenetic framework. parisons (n ϭ 190) for the number of character states The veracity of ancestral character state re- held in common for (a) adult plumage characters and constructions based on comparative studies de- (b) juvenile plumage characters. Picoides species pairs with the most similar plumages are identi®ed on his- pends on the accuracy of the phylogeny and its tograms by upper case letters: A ϭ P. lignarius±P. correct rooting. The phylogeny used in this mixtus,Bϭ P. pubescens±P. villosus, and C ϭ P. study is the maximum likelihood tree recon- canicapillus±P. kizuki. structed under the GTR ϩ I ϩ⌫substitution model from a data matrix comprising concate- nated DNA sequences, totaling 3451 nucleotides has greenish rather than black and white plum- from two mitochondrial genes and a nuclear age (Winkler et al. 1995). Of the three species gene intron. Although the probabilistic meaning pairs with the most similar adult plumage pat- of bootstrap support values is debated, they un- tern, the P. lignarius±P. mixtus pair shares ®ve derestimate true probabilities that nodes are real of six juvenile plumage character states (Fig. when the bootstrap support value is greater than 5b); the P. canicapillus±P. kizuki pair has four 50% (Felsenstein and Kishino 1993). As a rough of six juvenile plumage character states in com- approximation, bootstrap values of 70% are in- mon and the P. pubescens±P. villosus pair terpreted as 95% certainty that nodes are real shares only two of six juvenile plumage char- (Hillis and Bull 1993). For the tree used in this acter states. study, 15 of 19 nodes have bootstrap values greater than 70%, and the remaining four nodes DISCUSSION have values between 54% and 70%. Because the Earlier studies of plumage character evolution in tree we used is based on maximum likelihood Picoides woodpeckers (Bock 1963, Short 1971, using a realistic substitution model and a large 1974) relied on plumage, behavioral, ecological, nucleotide sample, and because bootstrap sup- and external morphological characters to recon- port is generally strong, we believe the topology struct phylogeny. Bock (1963) and Short (1971) used for our comparative analysis is substantial- suggested that plumage characters are particu- ly correct. larly useful in systematic studies of groups such The tree depicted in Figure 3 is rooted at the as woodpeckers, in which specialized features common ancestor of Picoides as determined by have limited morphological variation. However, outgroup analysis (Weibel and Moore 2002b). A plumage and other phenotypic character data in- potential concern is some uncertainty as to 804 AMY C. WEIBEL AND WILLIAM S. MOORE which group of genera is sister to Picoides, the evolution of carotenoid plumage and epaulets clade comprising the genera Melanerpes and were associated with marsh nesting in a group Sphyrapicus or a clade including Colaptes, Pi- of blackbirds. They suggested these plumage culus, and Dryocopus (Moore et al. 2005). The characteristics are adaptive because of increased Colaptes clade is the sister clade to Picoides in sexual selection in variable marsh territories or the 3-gene tree we used as our comparative greater male±male interactions due to increased framework, and this relationship persists when density of territories in marshes. To reject these only cyt b and COI sequences are used in phy- or other evolutionary hypotheses, comparative logenetic analysis (Weibel and Moore 2002a). studies of the relationships of various characters However, phylogenetic analyses which included and habitats of many different species would be the mitochondrial 12S rRNA gene in addition to required (Johnson and Lanyon 2000). Thus, cyt b and COI (Webb and Moore 2005) and while the correlations between plumage and those using only ␤-®bint7 sequences (Weibel habitat components detected in Picoides suggest and Moore 2002b) infer the Melanerpes clade as an important evolutionary pattern in niche par- the sister clade to Picoides. However, this con- titioning, more comparative and experimental cern is obviated here by the fact that in all cases studies are necessary to understand how evolu- the basal split in Picoides is the same as depicted tionary forces gave rise to speci®c plumage in Figure 3, and thus this uncertainty should not traits, and how these traits are related to ecology. affect our analysis. Phylogenetic tests of homology were applied Polarization of character state changes in this to adult plumage character data to identify plau- study was based on rooting the Picoides part of sible evolutionary explanations for high plum- the tree with the DNA-based phylogeny and not age resemblance in three pairs of Picoides spe- with a phylogeny based on the plumage char- cies: P. lignarius±P. mixtus, P. canicapillus±P. acters, i.e., mapping the most parsimonious dis- kizuki, and P. pubescens±P. villosus. Much like tribution of plumage character state changes on the approach of cladistic compatibility (Mea- the tree was independent of the process of con- cham and Estabrook 1985), character states are structing the tree. Moreover, our reconstruction inferred to be homologous if they support a of ancestral character states depended only on monophyletic group speci®ed by other homolo- the distribution of character states among spe- gous characters; i.e., the test rejects evolutionary cies within Picoides and not on more distant out- homology if character states are not synapo- groups where homology of character states be- morphies for a speci®c clade. All character comes more dubious. states shared between P. lignarius and P. mixtus Tracing the most parsimonious character state pass this test, thus their plumage resemblance is transitions on the phylogeny revealed several in- attributable to shared ancestry (homology). This teresting patterns of convergence, parallelism, is consistent with the DNA-based phylogenetic and retention of ancestral plumage states, which inference that they are closely related sister spe- most plausibly result from ecological selection, cies. sexual selection, or social selection. In particu- However, homology is rejected for several lar, the overall loss of head patterning and dorsal character states held in common by P. canica- barring or streaking strongly suggests that these pillus and P. kizuki and by P. pubescens and P. suites of characters have been in¯uenced by se- villosus. Reconstruction of phylogeny based on lection, though identifying the selective agent these characters would be incongruent with the driving these evolutionary patterns is dif®cult. Picoides species tree (Weibel and Moore 2002a, Correlation in the evolution of a physical char- 2002b). Note that though they are not sister spe- acter and an ecological character narrows the cies, the similarity in adult plumage between P. ®eld of potential selective agents to examine, canicapillus and P. kizuki is attributable to ho- such as the correlation between lack of dorsal mology in the form of retention of shared an- barring and use of dense habitat in Picoides. cestral character states (sympleisiomorphies), However, a simple adaptive interpretation of the whereas P. maculatus, the sister species to P. relationship between plumage and habitat could canicapillus, has three autapomorphic character be complicated by sexual selection or the role states. Thus, no perceptible evolutionary chang- of plumage in territorial defense. For example, es have occurred in the plumage characters ex- Johnson and Lanyon (2000) determined that the amined here. P. pubescens and P. villosus,in PLUMAGE CONVERGENCE IN PICOIDES WOODPECKERS 805 contrast, share most adult plumage character probable because the plumage similarity is high- states but differ substantially in juvenile plum- ly detailed, involving a large suite of similar age. Thus, the patterns of adult and juvenile characteristics in these two distantly related spe- plumage in these two species in the context of cies. the phylogeny are most consistent with the hy- Several hypotheses have been proposed to ex- pothesis of convergent evolution of adult plum- plain why distantly related taxa may converge age. Perhaps the most compelling reason for re- in external appearance. For example ¯ocking jecting evolutionary homology (shared ancestry) species may converge to avoid predation as an explanation of plumage similarity between through mimicry (Barnard 1979) and sympatric P. pubescens and P. villosus is that these species species may mimic one another to elicit or avoid are phylogenetically arranged into two different behavioral interactions (Moynihan 1960, 1968, clades in which the most recent common ances- Diamond 1982). The Downy Woodpecker is a tor of the pair unites all New World Picoides much smaller bird than the Hairy Woodpecker, species. which is typically the more aggressive species, Given that the plumage similarity of P. villo- and convergence may reduce aggressive ex- sus and P. pubescens (hereafter Hairy and changes between the two species; Hairy Wood- Downy Woodpeckers, respectively) is not the re- peckers may attack larger rather than smaller sult of homology, it is of interest to infer wheth- competitors (Diamond 1982). However, it is un- er it results from parallel or convergent evolu- likely the Downy Woodpecker is mimicking the tion. Usually parallel evolution occurs in rela- Hairy Woodpecker under this hypothesis be- tively closely related species, whereas conver- cause neither is considered a ¯ocking species. gence occurs in more distantly related species Convergence may be explained by MuÈllerian (Futuyma 1998). Details of speci®c pathways in mimicry, with both species converging onto the development of juvenile and adult plumage similar features, but it is doubtful that either spe- in woodpeckers are lacking, and so these alter- cies harbors toxins or other factors that would native hypotheses cannot be directly tested. be dangerous to predators as their primary pred- However, a weaker, indirect test can be made by ator is likely hawks. comparing levels of similarity between adults Cody (1969) proposed the hypothesis of in- and juveniles of the two species in a phyloge- terspeci®c territoriality to explain convergence netic context. Strong similarity of adult plumage in avian species, speci®cally woodpeckers. The contrasted by strong dissimilarity of juvenile model premise is that it is economically advan- plumage between distantly related species is tageous for sympatric competitors to recognize consistent with the prediction of convergent evo- and exclude each other from territories in order lution and inconsistent with the prediction of to maximize available resources, and the process parallel evolution. is facilitated when competitors have similar phe- Convergence in characters among species is notypes thereby reducing the number of recog- thought to be a key indicator of adaptation, es- nition signals to learn. Convergent evolution by pecially as a response to similar environmental interspeci®c territoriality can only affect char- pressures (Patterson 1988). Nevertheless, Wake acteristics involved in visual or acoustic aggres- (1991) argued that there are alternative expla- sive displays, but cannot interfere with intersex- nations for phenotypic resemblance, for example ual recognition. The competitors must be con- similarities may arise by random chance where geners (but not sibling species) and territorial (or super®cially similar character states between at least demonstrate intrasexual aggressive be- taxa could result from the adaptive process in havior), and similarities are expected to be less one species but as a result of modi®cation of the pronounced where species are allopatric. developmental program in the other species Interspeci®c territoriality may be the best ex- (Wake 1991). However, these similar character planation for convergence between Downy and states, although having evolved by different pro- Hairy Woodpeckers; the hypothesis is supported cesses, may then be favored concurrently by se- by phylogenetic analysis (Weibel and Moore lection leading to even greater similarity in char- 2002a, 2002b) and evaluation of plumage traits acteristics; thus, close similarity ultimately aris- presented here. Moreover, it is likely that the es by adaptation. However, for Hairy and two species compete for a common limiting re- Downy Woodpeckers, Wake's argument is im- source, suitable nest-cavity substrate. Hairy and 806 AMY C. WEIBEL AND WILLIAM S. MOORE

Downy Woodpeckers are extensively sympatric comparative biology. University of Chicago Press, across the North American continent, but the Chicago. BURNS, K. J. 1998. Molecular phylogenetics of the ge- distribution of the Downy Woodpecker does not nus Piranga: implications for biogeography and extend into the Southwest or Central America. the evolution of morphology and behavior. Auk Further support for the interspeci®c territoriality 115:621±634. hypothesis comes from a plumage variant of the CHRISTIDIS, L., R. SCHODDE, AND P. R. BAVERSTOCK. Hairy Woodpecker with brownish rather than 1988. Genetic and morphological differentiation and phylogeny in the Australo-Papuan scrubwrens white underparts found in Central America (Sericornis, Acanthizidae). Auk 105:616±629. (Winkler et al. 1995), a region in which Downy CHU, P. C. 1998. A phylogeny of the gulls (Aves: La- Woodpeckers do not reside. Behavioral studies rinae) inferred from osteological and integumen- in the ®eld are required to further substantiate tary characters. Cladistics 14:1±43. interspeci®c territoriality as the causal explana- CODY, M. L. 1969. Convergent characteristics in sympatric species: a possible relation to interspeci®c tion for convergence in Downy and Hairy competition and aggression. Condor 71:222±239. Woodpeckers. DIAMOND, J. M. 1982. Mimicry of friarbirds by orioles. Many other Picoides species are sympatric Auk 99:186±196. with both Downy and Hairy Woodpeckers in EHRLICH, P. R., D. S. DOBKIN, AND D. WHEYE. 1988. parts of their ranges in western and southwestern The birder's handbook: a ®eld guide to the natural history of North American birds. Simon and North America, and most species have vivid Schuster, New York. black and white plumage patterns. However, FELSENSTEIN, J., AND H. KISHINO. 1993. Is there some- these species not only have different ecological thing wrong with the bootstrap on phylogenies? requirements, they also have distinguishing A reply to Hillis and Bull. Systematic Biology 42: plumage characteristics that make identi®cation 193±200. in the ®eld relatively easy. Thus, the evolution- FUTUYMA, D. J. 1998. Evolutionary biology. 3rd ed. Sinauer Associates, Sunderland, MA. ary processes that mold plumage patterns among HACKETT, S. J., AND K. V. ROSENBERG. 1990. Compar- Picoides species are likely as diverse as the spe- ison of phenotypic and genetic differentiation in cies themselves. This study provides strong sup- South American antwrens (Formicariidae). Auk port that the striking resemblance of Hairy and 107:473±489. Downy Woodpeckers is the result of convergent HARVEY, P. H., AND G. M. MACE. 1982. Comparisons between taxa and adaptive trends: problems of evolution, regardless of the underlying selective methodology, p. 343±361. In King's College So- factors. ciobiology Group [EDS.], Current problems in sociobiology. Cambridge University Press, Cam- ACKNOWLEDGMENTS bridge, UK. HARVEY, P. H., AND M. D. PAGEL. 1991. The compar- We thank J. Hinshaw for her assistance at the Univer- ative method in evolutionary biology. Oxford sity of Michigan Museum of Natural History and two University Press, Oxford. anonymous reviewers whose comments improved this HILLIS, D. M. 1987. Molecular versus morphological manuscript. Our special thanks to J. Megahan for his approaches to systematics. Annual Review of artistic contribution to this paper. This research was Ecology and Systematics 18:23±42. supported in part by the William A. Turner, Jr. Me- HILLIS, D. M., AND J. J. BULL. 1993. An empirical test morial Scholarship, Wayne State University Graduate of bootstrapping as a method for assessing con®- School Thesis/Dissertation Research Support, and two National Science Foundation grants (DEB-9316452 dence in phylogenetic analysis. Systematic Biol- and DEB-9726512) awarded to WSM. ogy 42:182±192. HINTZE, J. L. 1999. NCSS 2000 statistical system for Windows. Kaysville, UT. LITERATURE CITED JOHNSON,K.P.,AND S. M. LANYON. 2000. Evolutionary BARNARD, C. J. 1979. Predation and the evolution of changes in color patches of blackbirds are asso- social mimicry in birds. American Naturalist 13: ciated with marsh nesting. Behavioral Ecology 11: 613±618. 515±519. BELL, G. 1989. A comparative method. American Nat- KUSMIERSKI, R., G. BORGIA,A.UY, AND R. H. CROZIER. uralist 133:553±571. 1997. Labile evolution of display traits in bow- BOCK, W. J. 1963. Evolution and phylogeny in mor- erbirds indicates reduced effects of phylogenetic phologically similar groups. American Naturalist constraint. Proceedings of the Royal Society of 97:265±285. London Series B 264:307±313. BOYDEN, A. 1973. Perspectives in zoology. Pergamon LIVEZEY, B. C. 1991. A phylogenetic analysis and clas- Press, Oxford. si®cation of recent dabbling ducks (Tribe Anatini) BROOKS,D.R.,AND D. A. MCLENNAN. 1991. Phylog- based on comparative morphology. Auk 108:471± eny, ecology and behavior: a research program in 507. PLUMAGE CONVERGENCE IN PICOIDES WOODPECKERS 807

LOSOS, J. B. 1990. A phylogenetic analysis of character day [EDS.], Problems of phylogenetic reconstruc- displacement in Caribbean Anolis lizards. Evolu- tion. Academic Press, New York. tion 44:558±569. PATTERSON, C. 1988. Homology in classical and mo- MADDISON, W. P. 1990. A method for testing the cor- lecular biology. Molecular Biology and Evolution related evolution of two binary characters: are 5:603±625. gains or losses concentrated on certain branches PRUM, R. O. 1997. Phylogenetic tests of alternative of a phylogenetic tree? Evolution 44:539±557. intersexual selection mechanisms: trait macroevo- MADDISON,W.P.,AND D. R. MADDISON. 1992. Mac- lution in a polygynous clade (Aves: Pipridae). Clade: analysis of phylogeny and character evo- American Naturalist 149:668±692. lution, version 3.0. Sinauer and Associates, Sun- RIDLEY, M. 1983. The explanation of organic diversity: the comparative method and adaptations for mat- derland, MA. ing. Clarendon Press, Oxford. MCKITRICK, M. C. 1993. Phylogenetic constraint in SHORT, L. L. 1971. Systematics and behavior of some evolutionary theory: has it any explanatory pow- North American woodpeckers, genus Picoides er? Annual Review of Ecology and Systematics (Aves). Bulletin of the American Museum of Nat- 24:307±330. ural History 145:1±118. MEACHAM, C. A., AND G. F. ESTABROOK. 1985. Com- SHORT, L. L. 1974. The contribution of external mor- patibility methods in systematics. Annual Review phology to avian classi®cation. Proceedings of the of Ecology and Systematics 16:431±446. 16th International Ornithological Congress:185± MOORE, W. S., A. C. WEIBEL, AND A. AGIUS. 2005. 195. Mitochondrial DNA phylogeny of the woodpecker SHORT, L. L. 1982. Woodpeckers of the world. Dela- genus Veniliornis (Picidae, Picinae) and related ware Museum of Natural History, Greenville, DE. genera implies convergent evolution of plumage WAKE, D. B. 1991. Homoplasy: the result of natural patterns. Biological Journal of the Linnean Soci- selection or evidence of design limitations? Amer- ety, in press. ican Naturalist 138:543±567. WEBB,D.M.,AND W. S. MOORE. 2005. A phylogenetic MOYNIHAN, M. 1960. Some adaptations which help to analysis of woodpeckers and their allies using promote gregariousness. Proceedings of the 12th 12S, cyt b and COI nucleotide sequences (Class International Ornithological Congress: 523±541. Aves; Order Piciformes). Molecular Phylogenetics MOYNIHAN, M. 1968. Social mimicry; character con- and Evolution 36:233±248. vergence versus character displacement. Evolution WEIBEL,A.C.,AND W. S. MOORE. 2002a. Molecular 22:315±331. phylogeny of a group of cosmopolitan woodpeck- NEE, S., A. F. READ, AND P. H. HARVEY. 1996. Why ers (genus Picoides) based on COI and cyt b mi- phylogenies are necessary for comparative anal- tochondrial gene sequences. Molecular Phyloge- ysis, p. 399±411. In E. P. Martins [ED.], Phylog- netics and Evolution 22:65±75. enies and the comparative method in animal be- WEIBEL, A. C., AND W. S. MOORE. 2002b. A test of a havior. Oxford University Press, Oxford. mitochondrial gene-based phylogeny of wood- OMLAND, K. E. 1999. The assumptions and challenges peckers (genus Picoides) using an independent ␤ of ancestral state reconstructions. Systematic Bi- nuclear gene, -®brinogen intron 7. Molecular ology 48:604±611. Phylogenetics and Evolution 22:247±257. WESTNEAT, M. W. 1995. Feeding, function, and phy- OMLAND, K. E., AND S. M. LANYON. 2000. Recon- logeny: analysis of historical biomechanics in la- structing plumage evolution in orioles (Icterus): brid ®shes using comparative methods. Systematic repeated convergence and reversal in patterns. Biology 44:361±383. Evolution 54:2119±2133. WILSON, E. O. 1975. Sociobiology. Harvard University PAGEL, M. D., AND P. H. HARVEY. 1988. Recent de- Press, Cambridge, MA. velopments in the analysis of comparative data. WINKLER, H., D. A. CHRISTIE, AND D. NURNEY. 1995. Quarterly Review of Biology 63:413±440. Woodpeckers: an identi®cation guide to the wood- PATTERSON, C. 1982. Morphological characters and ho- peckers of the world. Houghton Mif¯in, Boston, mology, p. 21±74. In K. A. Joysey and A. E. Fri- MA. 808 AMY C. WEIBEL AND WILLIAM S. MOORE APPENDIX. Polarized character states for Picoides in the context of phylogenetic relationships. Picoides species and representative species of Veniliornis and Dendropicos that render Picoides paraphyletic are listed in order according to phylogenetic relationships inferred from the tree in Weibel and Moore (2002b) and shown in Figure 3. Data for species of a given clade are presented with a unique letter (A±G) such that character states

South American spp. A Small Picoides spp. B P. lignar- P. V. callo- V. nigri- P. nut- P. sca- P. pubes- P. ius mixtus notus ceps tallii laris cens minor ADULT PLUMAGE CHARACTERS (1) Malar patchc ancestral: speckled A A A A B B B derived: solid B absent (2) Crown ancestral: not streaked B derived: streaked A A A A B B B (3) Outer retrices (ventral view) ancestral: barred AAAA BBBB derived: not barred (4) Inner retrices (dorsal view) ancestral: not barred A A B B B B derived: barred A A (5) Auricular patchc aancestral: wide AA B B absent B derived: narrow A A B (6) Back pattern ancestral: barred A A B B B derived: not barred (solid black) not barred (solid white) B not barred (solid red, green)b AA (7) Mantle pattern ancestral: not barred A A B derived: barred A A B B B (8) Ventral patternc ancestral: speckled A A derived: solid B ¯ank speckling B B B barred A A (9) Wing pattern (folded view) ancestral: spotted/barred A A B B B B derived: solid A A (10) Male ornamentationc ancestral: crown A A B B derived: wide nuchal patch B narrow nuchal patch A A B supercilium (11) Supercilium aancestral: full A A B B B reduced A B derived: absent A (12) Throat pattern ancestral: not speckled/streaked A B B B B derived: speckled/streaked A A A (13) Nasal tufts aancestral: distinctly white B B B B distinctly black A derived: no color distinction A A A JUVENILE PLUMAGE CHARACTERS (1) Head coloration on crown ancestral: crown AAAA BBBB derived: nape (2) Crown pattern ancestral: not spotted/streaked A A A B B derived: spotted/streaked A B B (3) Sexual dimorphism ancestral: subtle A B derived: strong A B B none A A B (4) Dorsal color intensity ancestral: dull AAAA BBBB derived: vivid (adult-like) (5) Ventral pattern ancestral: heavy A A ? ? B B derived: diffuse B B (6) Overall pattern ancestral: adult-like A B derived: not adult-like A A A B B B a Estimation of ancestral character states are equivocal. b Character state for back pattern in V. callonotus and V. nigriceps is solid red and solid green, respectively. PLUMAGE CONVERGENCE IN PICOIDES WOODPECKERS 809 APPENDIX. Extended. in the context of phylogenetic relationships can be readily identi®ed. Character states that are not evolutionarily homologous in P. pubescens±P. villosus and P. canicapillus±P. kizuki species pairs (Fig. 2 and 4, respectively) are identi®ed by upper and lower borders, and ancestral character state estimates are fully resolved.

Large Picoides spp. C D Asian spp. E Asian spp. F 3-Toed spp. G P. albo- P. strick- P. villo- P. bo- D. fusc- P. leuco- P. P. cani- P. macu- P. P. arc- P. tridac- larvatus landi sus realis escens tos major capillus latus kizuki ticus tylus

C D FFFGG CCEE C

CCCCD EFFF EGG

C CDEEFFF G CC G

CCCC EEF FGG DF

CC FFGG CCDEE F CDE FFF CC E G C G

CCC EEF FG CD F G

C DE FFF CE CC GG

CCDEEFFFGG CC

EGG CE CC E C FFF

CDE F G CC EFF C G

CCCEE FGG CDFF

CCCE EGG C D FFF

CCCCDEE ? FGG F

CCCCDEEF ? F G G

DE G CC C F F CE?G

CCCCDEE GG FFF

CCC EF F CDE F GG

CCCDEEF FG C FG c Other potentially nonhomologous characters in which ancestral characters states could not be fully resolved throughout the species phylogeny.