Barker 04

MOLECULAR
PHYLOGENETICS
AND
EVOLUTION
Molecular Phylogenetics and Evolution 31 (2004) 486504
www.elsevier.com/locate/ympev
Monophyly and relationships of wrens (Aves: Troglodytidae):

a congruence analysis of heterogeneous mitochondrial
and nuclear DNA sequence data
F. Keith Barker*
Bell Museum of Natural History, University of Minnesota, 100 Ecology, 1987 Upper Buford Circle, St. Paul, MN 55108, USA
Received 6 March 2003; revised 4 August 2003
Abstract
The wrens (Aves: Troglodytidae) are a group of primarily New World insectivorous birds, the monophyly of which has long been
recognized, but whose intergeneric relationships are essentially unknown. In order to test the monophyly of the group, and to
attempt to resolve relationships among genera within it, sequences from the mitochondrial cytochrome b gene and the fourth intron
of the nuclear b-brinogen gene were obtained from nearly all genera of wrens, from their relatives as suggested by traditional
taxonomy and DNADNA hybridization analyses, and from additional passerines. Maximum likelihood analysis of the two data
sets yielded maximal congruence between independently derived estimates of relationship, outperforming a variety of weighted
parsimony methods. Hierarchical likelihood ratio tests indicated that the two gene regions diered signicantly in every estimated
parameter of sequence evolution, and combined analysis of the two data sets was accomplished using a heterogeneous-model
Bayesian approach. Independent and simultaneous analyses of both data sets supported monophyly of the wrens (excluding one
recently added member, the monotypic genus Donacobius) and a sister-group relationship between wrens and the gnatcatchers
(Polioptila). Additionally, strong support was found for paraphyly of the genus Thryothorus, and for a sister-group relationship
between the genera Cistothorus and Troglodytes. Analyses of these data failed to resolve basal relationships within wrens, possibly
due to ambiguity in rooting with a distant, species-poor outgroup. Analysis of the combined data for wrens alone yielded results
which were largely congruent with relationships inferred using the complete data set, with the benet of stronger support for
relationships within the group. However, alternative rootings of this ingroup tree were weakly supported by nucleotide substitution
data. Insertiondeletion events suggest that the genus Salpinctes may be sister to all other wrens.
2003 Elsevier Inc. All rights reserved.
1. Introduction (Evans, 1997; Evans and Burn, 1996; Leonard and Pic-
man, 1987), vocal dialects and repertoires (Catchpole
The family Troglodytidae (sensu Mayr and Green- and Rowell, 1993; Kroodsma, 1977; Kroodsma and
way, 1960) comprises some 75 species of generally small, Canady, 1985; Kroodsma et al., 1999; Morton et al.,
active, highly vocal insectivorous birds. The family is 1986; Verner, 1975), vocal duetting (Brenowitz and
limited in distribution to the New World (except Trog- Arnold, 1986; Farabaugh, 1982; Levin, 1996a,b), and
lodytes troglodytes), with a center of diversity in Central cooperative breeding (Rabenold, 1990; Skutch, 1935).
America (Brewer, 2001). Members of this family have Since it has been the target of such intensive studies of a
been the subject of a wide variety of ecological and be- wide variety of signicant characteristics, the family
havioral studies, including studies of mating systems presents excellent opportunities for integrative, com-
(Johnson et al., 1993, 1994; Rabenold et al., 1990; parative analyses of ecological and behavioral evolution.
Leonard, 1990), male multiple-nest-building behavior However, complete understanding of variation within
the family must incorporate information on historical
associations among species and genera (e.g., Harvey and
*
Fax: 1-612-624-6777. Pagel, 1991). To date, no comprehensive hypothesis of
E-mail address: [email protected]. phylogenetic relationships among wrens exists, and the
1055-7903/$ - see front matter 2003 Elsevier Inc. All rights reserved.
doi:10.1016/j.ympev.2003.08.005
F.K. Barker / Molecular Phylogenetics and Evolution 31 (2004) 486504 487
relationship of the family to other passerines remains a hybridization study by Sheldon and Gill (1996) sup-
matter of contention. ported Sibley and Ahlquists arrangement of the gnat-
Hypotheses of the relationships of the Troglodytidae catchers and gnatwrens, treecreepers, and nuthatches
were until recently limited to circumscriptions of the (represented by Polioptila, Certhia, and Sitta respec-
familys contents, and associations of the family with tively) as the closest relatives to the wrens. However,
others in linear taxonomic series. The contents of the their data agreed with traditional taxonomy in placing
family have been stable over the last century, after sev- Auriparus with the penduline-tits. Also, their data
eral genera were eliminated from the family and recog- supported placement of the mimic-thrushes as the sis-
nized as members of the babbler subfamily Timaliinae ter-group to the starlings, well separated from the
and the family Sylviidae (sensu Mayr and Cottrell, 1986; wrens. Many of these hypothesized relationships among
Mayr and Paynter, 1964), though the enigmatic genus wrens, creepers, nuthatches, and mimic-thrushes were
Donacobius was recently assigned to the family (A.O.U., recently corroborated using nuclear DNA sequences
1983; Clench et al., 1982). The taxonomic placement of (Barker et al., 2002).
the family relative to other passerine families has varied, Hypotheses of relationship among genera of wrens
but several groups have traditionally been associated are extremely few, although some connections among
with the wrens, including the dippers (Cinclidae; e.g., genera have been suggested by past taxonomic practice.
Sharpe, 1881), mimic-thrushes (subfamily Miminae, Sumichrasts wren (Hylorchilus sumichrasti) was origi-
family Muscicapidae; sensu Mayr and Greenway, 1960; nally assigned to the genus Catherpes, and later removed
e.g., A.O.U., 1886), creepers (Certhiidae; e.g., Beecher, to its current genus based upon more extensive material
1953; Ridgway, 1904), and nuthatches (e.g., Beecher, (reviewed in Atkinson et al., 1993). Sibley and Monroe
1953). Most recent traditional classications place the (1990) suggested returning the genus to Catherpes (e.g.,
Troglodytidae near the Cinclidae and Mimidae in linear Sibley and Monroe, 1990), based upon vocal similarities,
series (Mayr and Amadon, 1951; Mayr and Greenway, but subsequent discovery of the vocally distinctive Na-
1960; Wetmore, 1960). vas wren (Hylorchilus navai) suggested retention of the
Recently, genetic data have been applied to the genus Hylorchilus (Atkinson et al., 1993). The ca~ non
problem of wren relationships. Sibley and Ahlquist wren Catherpes mexicanus has also been linked with the
(1990) proposed a hypothesis of relationships for the rock wren (Salpinctes obsoletus), and at least one widely
Class Aves, based upon analysis of DNADNA hy- used taxonomy (Mayr and Greenway, 1960) subsumes
bridization distance data, including explicitly phyloge- both species within Salpinctes. Recently Rice et al.
netic hypotheses of relationship among wrens and their (1999) found, based on mitochondrial DNA sequence
putative allies. Within wrens, Sibley and Ahlquist ob- data, that the timberline wren (Thryorchilus browni) was
tained very little resolution, with most branches shorter nested within the genus Troglodytes, suggesting para-
than estimated experimental error in hybridization phyly of the latter genus. Other than these, there are no
distance measurements. Regarding wren relationships hypotheses of relationship among wren genera except
to other groups, Sibley and Ahlquist proposed a novel the various linear series of taxa.
hypothesis: that wrens (including Donacobius, Sibley The work reported here was performed in order to
and Ahlquist, 1984) are the sister-group to a clade test previous notions of phylogenetic relationships of
containing the gnatcatchers and gnatwrens (genera wrens (Aves: Troglodytidae). Specically, this work at-
Polioptila, Microbates, and Ramphocaenus) and the tempts to test the monophyly of the family, to determine
verdin (genus Auriparus). The former genera have been its sister-group relationships, and to determine rela-
placed together as subfamily Polioptilinae of the Mu- tionships among genera within the family. As suggested
scicapidae (sensu lato) by Mayr (1946), and Mayr and by some authors (Hillis, 1987; Barrett et al., 1991; Flook
Paynter (1964), while the latter genus is generally con- et al., 1999; but see Bull et al., 1993; Reed and Sperling,
sidered the a member of the penduline tit family Re- 1999), data from two character systems which dier in
mizidae (e.g., A.O.U., 1998). Sibley and Ahlquist found their rates of divergence (the mitochondrial cytochrome
the Certhiidae (the creepers, Certhia and Salpornis) to b gene and the fourth intron of the nuclear b-brinogen
be the sister-group to the wren/gnatcatcher/verdin gene) were collected in the hope that they would provide
clade, and the Sittidae (the nuthatches and wallcreeper, complementary information at dierent hierarchical
Sitta and Tichodroma) as the sister to all of these. These levels. Though this strategy has been advocated in the
genetic results supported previous notions of a link literature, considerable disagreements have arisen over
between wrens and treecreepers and nuthatches, while methodological approaches to analyzing the data ob-
rejecting any anity with the babblers (which they tained (Brower et al., 1996; Bull et al., 1993; Chippindale
placed within a newly dened Sylviidae), mimic-thru- and Wiens, 1994; De Quieroz et al., 1995; DeSalle
shes (placed as sister-group to the traditional Sturnidae and Brower, 1997; Huelsenbeck et al., 1996; Krajewski
within another superfamily), and dippers (also placed et al., 1999). In particular, combining data sets with
in another superfamily). A subsequent DNADNA fundamentally dierent evolutionary dynamics can be
488 F.K. Barker / Molecular Phylogenetics and Evolution 31 (2004) 486504
problematic (Bull et al., 1993; Caterino et al., 2001; the traditional Troglodytidae (Mayr and Greenway,
Wilgenbusch and de Quieroz, 2000). The sequence data 1960), and Mimidae.
reported here were analyzed both independently and
in combination, using parsimony, maximum likeli- 2.1.2. Amplication and sequencing of cytochrome b and
hood, and Bayesian approaches, specically addressing b-brinogen intron 4
the issue of among-partition variation in evolutionary Total genomic DNA was extracted from 100 mg of
dynamics. each tissue sample by standard proteolytic digestion,
phenol/chloroform extraction, and concentration by
centrifugal dialysis (Centricon 100, Amicon); alterna-
2. Materials and methods tively, some samples were extracted using commercially
available protocols and reagents (PureGene, Gentra
2.1. Data collection and alignment Systems; QiaAmp, Qiagen). Preparation of cytochrome b
sequencing template was accomplished by primary PCR
2.1.1. Taxon sampling strategy amplication of a majority of the gene using the primer
Taxon sampling in this study was designed in order pair B1a/B8, and subsequent reamplication of three
to: (1) test the sister group relationships and monophyly segments (B1a/B2a, B3/B6, and L15507/B8; see Table 2)
of wrens, (2) test the placement of Donacobius relative to from a gel plug (5 lL of initial PCR run on a 1% low
the traditional families Mimidae and Troglodytidae and melting point agarose [FMC] gel, excised and melted in
(3) determine intergeneric relationships among wrens. 200 lL ddH2 O) of this initial amplication. Primary
Choice of taxa was guided primarily by the phylogenetic amplications were performed in 25 lL total reaction
hypotheses of Sibley and Ahlquist (1990) (see also volumes [2550 ng template DNA, 1 amplication
Harshman, 1994), Sheldon and Gill (1996), and Barker buer (1.5 mM nal MgCl2 , BoehringerMannheim),
et al. (2002). The taxa sampled, following the sequence 0.5 U Taq polymerase (BoehringerMannheim), 0.2 lM
of Sibley and Monroe (1990), are listed in Table 1. A in each primer, 80 lM in each dNTP], with thermal cy-
single suboscine (Pipra pipra) was chosen as an out- cling parameters for initial amplications as follows: an
group for estimation of interfamilial relationships. All of initial 3 min at 94 C, followed by 35 cycles of 15 s at
the superfamilial groupings of passeridan passerines 94 C, 30 s at 51 C, and 75 s at 72 C, followed by a nal
were sampled (Muscicapoidea, Sylvioidea, and Passe- 3-min extension at 72 C. Primary amplications of the
roidea), as well as two members of the putative sister- fourth intron of the b-brinogen gene were accomplished
group to the Passerida, the parvorder Corvida (Sibley as for cytochrome b, except using the FIB3/FIB4 primer
and Ahlquist, 1990). Within the Muscicapoidea, sam- pair (Table 2), and an extension time of 30 s. Reampli-
pling included a thrush (Muscicapidae: Turdinae), a cations were performed in 50 lL volumes (using 2 lL of
mimic-thrush (Sturnidae: Miminae), and a starling melted gel plug as template), with conditions as for pri-
(Sturnidae: Sturninae). Within the Sylvioidea, sampling mary amplications, excepting an increase of 2 in an-
included all of the putative wren outgroups, including nealing temperature, and an extension time of 30 s.
the nuthatches (Sittidae), treecreepers (Certhiidae: Cer- Reamplication products were puried by one of two
thiinae), and gnatcatchers and gnatwrens (Certhiidae: methods: (1) centrifugal dialysis using a Microcon 100
Polioptilinae). Within the wrens, all genera were sam- (Amicon), twice adding 500 lL ddH2 0 to the 50 lL re-
pled except the monotypic genera Ferminia, Thryorchi- action volume and spinning at 500g for 15 min; or (2)
lus, and Uropsila (13 genera, not including Donacobius). dsDNA extraction using the GeneClean protocol
It should be noted, however, that Thryorchilus has re- (Bio101). Dialyzed PCR products were quantied and
cently been allied to the genus Troglodytes in a molec- diluted to 25 ng/lL nal concentration, while gene-
ular study by Rice et al. (1999). Two representatives cleaned products were resuspended in 25 lL ddH2 O.
were included for the genera Campylorhynchus and Several intron 4 PCR products (Table 1) were cloned
Cistothorus, and ve species were sampled from the ge- in order to detect heterozygotes and length variants.
nus Thryothorus, which contains a large percentage of Samples selected for cloning were those for which clean
wren diversity (27/75 species; Sibley and Monroe, 1990). double-stranded sequence could not be obtained directly
All other genera were represented by an individual from from PCR product. Cloning was accomplished by liga-
one species each (10 genera, two of which are in any case tion of PCR product (using T4 DNA ligase [Promega])
monotypic). Two members of the Passeroidea were in- into EcoRV-digested (Pharmacia Biotech) pBluescript II
cluded in the analysis in order to complete sampling of (Stratagene), modied to contain thymine overhangs by
the superfamilies within the parvorder Passerida. Under treatment with dTTP and Taq polymerase. Ligated
the assumption that relationships among the sylvioid plasmids were used to transform Escherichia coli strain
taxa sampled here are similar to those indicated by DH5a cells (Gibco), and selected for inserts by growth
DNA hybridization, this sampling strategy allows rig- on LB plates with IPTG and X-Gal (50 30 ). Positive
orous testing of the membership of Donacobius in both colonies were grown overnight in LB, and the presence
Table 1
List of taxa included in this study, following the taxonomy of Sibley and Monroe (1990), including lengths of sequences obtained from b-brinogen
intron 4, and sample data
Species and taxonomy I4 Sourcea
Suborder Tyranni
Pipra pipra 609 LSUMNS B7079
Suborder Passeri: Parvorder Corvida

Meliphaga gracilis 603 de A.C. Driskell
Dasyornis broadbenti 609 de A.C. Driskell
Suborder Passeri: Parvorder Passerida

Superfamily Muscicapoidea
Sialia sialis 589b FMNH 350787
Sturnus vulgaris 585b FMNH 389606
Dumetella carolinensis 585 FMNH 350635
Superfamily Sylvioidea
Sittidae
Sitta pygmaea 593 FMNH 343324
Certhiidae: Certhiinae
Certhia familiaris 575 FMNH 351158
Certhiidae: Troglodytinae
Donacobius atricapillus 570 FMNH 1772 [SML88-246, MPEG]
Campylorhynchus brunneicapillus 590 FMNH 342076
C. megalopterus 590 MZAH 8711
Odontorchilus cinereus 593 FMNH 1789 [DW3682, MPEG]
Salpinctes obsoletus 585 MVZ 170087
Catherpes mexicanus 580 FMNH 5470 [BEHB033, MZAH]
Hylorchilus sumichrasti 581 MZAH [OMVP1132, MZAH]
Cinnycerthia peruana 588 ZMUC O2450 [NK13, MDS]
Cistothorus platensis 592 FMNH 350634
C. palustris 592 FMNH 333341
Thryomanes bewickii 591 MZAH 9734
Thryothorus coraya 585b FMNH 339666
T. maculipectus 583 MZAH 7828
T. ludovicianus 588 AMNH 20929
T. leucotis 592 MUSP 73431
T. guarayanus 589 FMNH 334541
Troglodytes aedon 592 FMNH 343273
Henicorhina leucosticta 594 FMNH 343285
Microcerculus marginatus 597b LSUMNS B11839
Cyphorhinus arada 595 FMNH 1775 [ATP86-142, MPEG]
Certhiidae: Poliptilinae
Polioptila caerulea 585b FMNH 343322
Cisticolidae
Prinia bairdii 444 FMNH 355824
Zosteropidae
Zosterops senegalensis 601 FMNH 346671
Superfamily Passeroidea
Acanthidops bairdii 580 LSUMNS B16267
Zeledonia coronata 580 LSUMNS B16178
a
Source indicates the specimen voucher for each sample. If voucher data were unavailable, then the tissue number is listed, the institution where the
voucher is located is indicated, and collectors number provided in brackets. ANSP, Philadelphia Academy of Natural Sciences; FMNH, Field Museum
of Natural History; LSUMNS, Louisiana State University Museum of Natural Science; MDS, Museo de Salango, Ecuador; MPEG, Museu Paraense
Emilio Goeldi; MUSP, Museu de la Universidad de Sa~ o Paolo; MVZ, Museum of Vertebrate Zoology, University of California, Berkeley; MZAH,
Museo de Zoologa Alfonso L. Herrera, Universidad Nacional Aut onoma de Mexico; and ZMUC, Zoological Museum, University of Copenhagen.
b
Cloned samples.
of appropriately sized inserts conrmed by electropho- of three positive colonies per PCR product were puried
resing 40 lL phenol/chloroform-extracted culture on an for sequencing using the PERFECTprep plasmid DNA
agarose gel, visualized with ethidium bromide. Cultures purication protocol and reagents (50 30 ).
Table 2 intron motifs (Perlman et al., 1990) identied. Contig

List of primers used in amplication and sequencing, numbered ac- alignments were created using Sequencher (Genecodes,
cording to location in the Gallus genome (Desjardins and Morais,
1990) for mitochondrial primers, and the human b-brinogen gene
Ann Arbor, MI). Multiple alignment of cytochrome b
(Chung et al., 1991) for intron primers sequences was accomplished by eye in Sequencher. Ap-
propriate coding of cytochrome b sequences was veried
Primer Sequence (50 30 ) Locationa
by translation into amino acids using the avian mito-
B1a ccatccaacatctcagcatgatgaaa L14990b chondrial code (Desjardins and Morais, 1990; imple-
B3 atctgcatctacctacacatcgg L15191c
L15506 ctcaccttcctacacgaaacagg L15506b
mented in MacClade v4.0, Maddison and Maddison,
B2a cccctcagaatgatatttgtcctca H15298b 2000). Multiple alignment of intron sequences was ac-
B6 gcgtaggcgaataggaagtatca H15709 complished by hand modication of an initial alignment
B8 ggagtcttcagtctctggtttacaagac H16065b obtained using Clustal W (Thompson et al., 1994).
FIB3 ctgtaatatcccggtggtttcagg S4706
FIB4 atttcagatgtttcacctccctttc AS5291
a
2.2. Phylogenetic analysis
For mitochondrial primers, L indicates binding to heavy strand,
H to light strand; for intron primers, S indicates binding to the sense
strand, and AS to the anti-sense strand. 2.2.1. Preliminary examination of data
b
From Helm-Bychowski and Cracraft (1993). All phylogenetic analyses were conducted using
c
From Lanyon and Hall (1994). PAUP* (v4.0b10; Swoord, 1998). The sequence data
were examined in a variety of ways prior to phylogenetic
analysis. Since heterogeneity in base composition is
Puried PCR products were cycle-sequenced using known to aect phylogenetic inference (Galtier and
50100 ng template (or 24 lL puried product for Gouy, 1998; Lockhart et al., 1994; Yang and Roberts,
GeneCleaned samples), amplication primers, and ei- 1995), variation among sequences in base pair compo-
ther the ABI PRISM Dye Terminator Cycle Sequencing sition was examined. Base composition homogeneity
Ready Reaction Kit (with AmpliTaq DNA Polymerase, was tested using a v2 analysis of base frequencies across
FS; PerkinElmer) or the ABI PRISM dRhodamine taxa, examining each coding position separately for the
Terminator Cycle Sequencing Ready Reaction Kit cytochrome b data. This test does not correct for phy-
(with AmpliTaq DNA Polymerase, FS; PerkinElmer), logenetic relationships among taxa, but should be con-
according the manufacturers directions, except using servative, since phylogenetic autocorrelation will reduce
10 lL reaction volumes. Puried plasmid preparations the probability of detecting the eect of shifts in base
were sequenced using dRhodamine sequencing chemis- composition. For base frequencies in partitions which
try and the primer pair SK/KS (SK 50 -CGCTCTAG showed signicant heterogeneity, v2 goodness-of-t
AACTAGTGGATC-30 , KS 50 -TCGAGGTCGACGGT analyses were also conducted for base frequencies
ATC-30 ). Sequencing reactions were either puried using of each sequence relative to the overall mean values,
CentriSep columns (samples sequenced with FS; correcting for multiple comparisons.
Princeton Separations, Princeton, NJ), or ethanol pre-
cipitated according to the manufacturers directions 2.2.2. Separate phylogenetic analyses: parsimony
(samples sequenced with dRhodamine). Reactions were A variety of parsimony analyses were performed in
then electrophoresed on an ABI 377 automated se- order to assess the impact of weighting on apparent
quencer (Applied Biosciences, PerkinElmer). All nu- conict between the data sets. Analyses of cytochrome b
cleotide positions were sequenced from both strands. data performed included: (1) equally weighted parsi-
Sequences obtained using the cytochrome b-targeted mony, (2) elimination of third position data, (3) elimi-
primers were between 1069 and 1093 bases in length nation of third position transitions only, (4) Rodrigos
(depending on the size of the intergenic spacer and (1992) modication of Wheelers (1990a) combinatorial
tRNA-Thr fragment in individual taxa), corresponding weights approach, and (5) analysis of amino acid
to positions 14,99116,064 of the Gallus mitochondrial translations. Intron 4 sequences were analyzed under
genome (Desjardins and Morais, 1990). Sequences ob- equally weighted parsimony alone, treating alignment
tained from direct sequencing of PCR product or clon- gaps in two dierent ways. Regardless of gap treatment,
ing using b-brinogen intron 4 primers varied from 468 indels were treated as missing data within the matrix,
to 633 bases in length, corresponding to positions 4707 though gapped regions were always included (excepting
5290 of the human gene (counting from the rst base of one, see Section 3). The matrix was analyzed either in
the glutamine codon which codes for the rst amino acid this form alone, or with potentially shared indel events
of the mature protein, equivalent to bases 52665849 of coded separately as potentially informative characters.
GenBank Accession No. M64983; Chung et al., 1991). Indel events were coded where the gapped region ex-
Intronexon boundaries were examined, exon sequences hibited: (1) uniformity of indel length, and (2) conser-
conrmed against available avian sequence (Weissbach vation of anking sequence. Taxa which shared a stretch
et al., 1991; unpublished data), and conserved 50 and 30 of indels in the nal alignment were coded as potentially
sharing an insertion/deletion event. During parsimony methods was assumed, with uniform interval priors,
analyses, all searches were heuristic, using the tree-bi- except for base frequencies, which were assigned a Di-
section and reconnection (TBR) branch-swapping al- richlet prior (Huelsenbeck and Ronquist, 2001). Anal-
gorithm, with 50 random addition sequences of taxa (or yses were conducted using Metropolis coupling with
more, depending on the discovery of multiple islands of four incrementally heated Markov chains (MC3 ; default
equally parsimonious trees). Support for inferred rela- heating parameter), each chain assigned to an indepen-
tionships was evaluated using the bootstrap (Felsen- dent processor using the distributed-memory parallel-
stein, 1985). ization implemented in MrBayes v3.0 (Altekar et al., in
press; compiled on a 16 processor Sun Fire 6800, with
2.2.3. Separate phylogenetic analyses: maximum likeli- 64-bit, 900 MHz UltraSparc III architecture, Sun Mi-
hood crosystems, Santa Clara, CA). Chains were run for
Analyses under the maximum likelihood criterion 2 106 generations, and sampled every 100. The log-
were preceded by an evaluation of alternative models of likelihood of each chain was evaluated as a function of
sequence evolution using a xed tree generated by the generation number, in order to determine the minimum
neighbor-joining method (Saitou and Nei, 1987) with number of generations to discard from the beginning of
JukesCantor distances (Jukes and Cantor, 1969). The the chain as burn-in. Posterior estimates of model
initial topology chosen had little eect on the parameter parameters and taxon bipartitions were derived from the
values estimated (results not presented). Models evalu- complete sample of chains minus those discarded as
ated included the F81 (Felsenstein, 1981), HKY85 burn-in. At least two runs from random starting points
(Hasegawa et al., 1985), and general time-reversible were performed for each data set, and the equilibrium
(Yang, 1994), allowing for invariant sites, C-distributed log-likelihood, parameter values, and bipartition fre-
rate heterogeneity, or a combination of both for each quencies of each run were compared to evaluate the
model (Gu et al., 1995). Additionally, each model was stability of posterior estimates (Huelsenbeck et al., 2002;
evaluated both with and without the assumption of a Huelsenbeck and Ronquist, 2001).
molecular clock, in order to determine the signicance
of rate heterogeneity among lineages (Felsenstein, 1981). 2.2.5. Evaluation of congruence
The signicance of increases in likelihood caused by Congruence of phylogenetic estimates from the indi-
addition of parameters (e.g., allowing for among-site vidual data sets was evaluated both qualitatively and
variation in rates) was tested using the likelihood ratio quantitatively. First, the results of separate analysis of
statistic 2 ln K 2ln k2 ln k1 , where k1 is the like- each data set were qualitatively compared in terms of
lihood of the restricted model), under the asymptotic topological congruence using the symmetric-dierence
assumption of a v2 distribution (Huelsenbeck and metric (Robinson and Foulds, 1981). In this way, the
Crandall, 1997; Huelsenbeck and Rannala, 1997). The analytical conditions and assumptions most favorable to
model selected was the least parameter rich which could the null hypothesis of homogeneity of phylogenetic
not be signicantly improved by additional parameters. signal were identied. The signicance of any in-
Subsequent to model evaluation and selection, a heu- congruence which could not be eliminated by varying
ristic search with ten random addition-sequence repli- analytical assumptions was evaluated by bootstrap
cates and TBR branch-swapping was performed, xing proportions estimated for the most congruent analyses
the model parameters inferred for the starting tree. In (Felsenstein, 1985), and by Bayesian posterior proba-
order to identify all equally likely resolved topologies, bilities. Signicant incongruence was inferred where
branches of eectively zero length were not collapsed conicting nodes were recovered in separate analyses,
(option lcollapse no). In order to test the robustness of with support exceeding criterion as estimated by either
the search to parameter value choice, the model pa- of these measures. Criterion for bootstrap was set at
rameters were reestimated on the initial optimal trees, 75%, in excess of the often-cited 70% level (Hillis and
and tree searches were repeated with the adjusted pa- Bull, 1993), and for the Bayesian posterior probabilities
rameter estimates. Nodal support was evaluated using at 0.95, as these latter are argued to have a straight-
the bootstrap, with model parameters xed across rep- forward statistical interpretation (Huelsenbeck and
licates, and initial trees for swapping obtained by Ronquist, 2001; Larget and Simon, 1999; but see Suzuki
neighbor-joining. et al., 2002).
2.2.4. Separate phylogenetic analyses: Bayesian 2.2.6. Combined analysis

Analysis of these data sets using Bayesian methods, Combined analyses were pursued under the analytical
as implemented in MrBayes (version 3.0; Altekar et al., conditions which most favored congruence in separate
in press; Huelsenbeck and Ronquist, 2001) was also analyses of the two data sets. Prior to combined analysis
performed. For each data set, the optimal model pa- using model-based approaches, process-parameter ho-
rameterization as determined by maximum likelihood mogeneity between the two data sets was evaluated by
maximum-likelihood model-tting with PAML v3.12, Meliphaga and Dasyornis were provided by A.C.
(Yang, 1997). A hierarchy of models enforcing param- Driskell [AY353241, AY353242]), corresponding to
eter homogeneity and relaxing this constraint were t, positions 14,99116,035 (1045 bases of 1143) of the
using the Mgene option of the baseml program. Gallus mitochondrial genome (Desjardins and Morais,
The following parameters were either constrained to 1990). Base composition of these sequences was fairly
homogeneity or allowed to vary between partitions: typical for mitochondrial DNA of birds in general
gene-specic substitution rates (c), base composition (Edwards et al., 1991; Hackett, 1996; Kocher et al.,
(pi ), relative rates of nucleotide substitution classes (rij ), 1989; Kornegay et al., 1993; Nunn and Cracraft, 1996),
and pattern of among-site rate heterogeneity (discrete showing highest skew at third positions, with succes-
approximation to the C-distribution, four rate catego- sively less skew at second and rst positions respectively
ries, parameter a). PAML does not currently allow es- (Table 3). Homogeneity across taxa was not rejected for
timation of a proportion of invariant sites (piv ), so this cytochrome b rst and second positions (Table 3).
parameter was not evaluated. The signicance of pa- However, the v2 test for third positions indicated sig-
rameter heterogeneity was evaluated by the v2 approx- nicant heterogeneity across all taxa, though heteroge-
imation to the likelihood ratio, as for the model neity was not signicant within the Troglodytidae alone
evaluations previously described. The likelihood of each (v2 57:11, df 54, p 0:36). Sequence-specic good-
heterogeneous model was compared to that of the more ness-of-t tests for third position base frequencies indi-
restrictive homogeneous model, for each parameter cated that only sequences of Pipra, Dasyornis, and
tested. Given that two data sets dier in evolutionary Dumetella deviated signicantly from the overall mean
dynamics, the assumption employed in maximum like- values (v2 101:5, 21.0, 18.6, respectively, df 3,
lihood analysis that site patterns are identically distrib- p < a 0:002 with correction for multiple compari-
uted is inappropriate. Unfortunately, simultaneous sons). Patterns of variation within cytochrome b also
maximum likelihood analysis of data sets using multiple corresponded to those previously noted for the gene in
models of substitution (heterogeneous-model likeli- other birds. Of codon third positions, 97.4% were vari-
hood), though theoretically feasible (e.g., Yang, 1996), able, with 91.8% of variable positions potentially in-
has not been implemented with an ecient search al- formative under the parsimony criterion. For rst and
gorithm to date. The closest approximation to such a second positions, 31.3 and 12.6% were variable, re-
search which has been implemented is a Bayesian anal- spectively (with 75.2 and 56.8% of variable sites poten-
ysis via Markov chain Monte Carlo, allowing signi- tially informative). Uncorrected pairwise distances for
cantly heterogeneous parameters to vary independently cytochrome b ranged from 6 to 20% divergence within
among partitions (MrBayes v. 3.0; Altekar et al., in the ingroup, and up to 25% in comparisons to Pipra.
press). Bayesian analysis of the combined data was Sequence of the fourth intron of the b-brinogen gene
performed as for the individual data sets, freeing sig- was obtained for all taxa (GenBank Accession Nos.
nicantly dierent parameters (as evaluated by likeli- AY352550AY352580; sequences of Meliphaga and
hood) to vary between the two partitions, using the Dasyornis were provided by A.C. Driskell [AY353243
unlink and prset options. and AY353244]). Clonal diversity for those taxa for
which multiple clones were sequenced was minimal (four
positions varied between clones, out of 2446 positions
3. Results sequenced for multiple clones: 0.16%), though one case
of apparent heterozygosity for a single base pair inser-
3.1. Sequence characteristics tiondeletion event was detected (position 160 of the
Sturnus sequence). The region of this indel event was
Sequences of the cytochrome b gene were obtained eliminated from phylogenetic analysis (see below). Base
for all taxa (GenBank Accession Nos. AY352520 composition of the intron was relatively constant across
AY352549; the sequence of Acanthidops was obtained taxa, and skewed toward adenine and thymine residues
from GenBank [AF489878], and the sequences of (Table 3), suggesting placement in an AT-rich isochore
Table 3
Average nucleotide composition of cytochrome b and b-brinogen intron 4
Partition A C G T v2
Cytochrome b-1st position 0.235 0.299 0.243 0.223 14.66
Cytochrome b-2nd position 0.195 0.262 0.130 0.413 2.60
Cytochrome b-3rd position 0.377 0.474 0.044 0.105 273.62
b-Fibrinogen Intron 4 0.312 0.169 0.177 0.342 14.07
Base proportions are shown for each coding position of cytochrome b. The v2 values refer to the two-way test of independence implemented in
PAUP* (Swoord, 1998), with df 96 (p < 0:01 indicated by an asterisk).
(Aota and Ikemura, 1986; Ikemura and Aota, 1988). Analytical Treatment of Intron 4
Sequences varied in length between 444 bp (Prinia) and Parsimony Parsimony+Gaps ML
609 bp (Pipra and Dasyornis), with a median of 590 bp.
40
Symmetric Difference
Alignment of these sequences was fairly straightforward,
yielding few regions of signicant ambiguity (only one
35
region, including bases 153165, was excluded from
analysis). An exception was the sequence obtained for
30
Prinia, which was signicantly shorter than those of
25
other taxa sampled, and which additionally included a
stretch of bases with questionable homology to other
EW
N3P
N3TI
AA
CW
ML
EW
N3P
N3TI
AA
CW
ML
EW
N3P
N3TI
AA
CW
ML
sequences in the alignment (bases 250460 of the align-
Analytical Treatment of Cytochrome b
ment). This region of the sequence was treated as
missing data for Prinia. The nal alignment indicated Fig. 1. Comparison of topologies from separate analyses of cyto-
the presence of 17 regions with potentially informative chrome b and intron 4 data. Labels on the abscissa indicate treatment
insertiondeletion events. Of these 17 regions, 13 ex- of the cytochrome b data (EW, equally weighted parsimony, N3P,
parsimony with elimination of third positions, N3TI, parsimony with
hibited characteristics (see Section 2) which justied elimination of third position transitions, AA, parsimony analysis of
coding and analysis as separate characters (see Appen- amino acid translations, CW, parsimony with combinatorial weights,
dix A). Levels of variation in this alignment of the intron and ML, maximum likelihood), and the three sections of the plot area
were relatively high in percentage of positions variable labelled above indicate treatments of the intron 4 data (parsimony
(59.6%), but most variability was due to singleton sub- both without and with informative gaps coded as separate characters,
ML, maximum likelihood). The boxplots summarize pairwise sym-
stitutions, as only 39.6% of all variable sites were po- metric dierences between all unique equally parsimonious or likely
tentially informative under the parsimony criterion. trees obtained for the two data sets with each combination of ana-
Uncorrected sequence divergence among taxa for intron lytical approaches.
4 ranged from 0.4 to 13.5% in the ingroup, and up to
17.9% in comparisons to Pipra. as the analysis criterion for both. For this reason, only
the results of likelihood-based analyses (maximum
3.2. Evaluation of congruence between cytochrome b and likelihood and Bayesian) are reported here in detail,
intron 4 and the quantitative evaluation of congruence is based
on these.
Topological comparisons between the trees derived Likelihood model evaluations of the two data sets in-
from separate phylogenetic analyses of the complete dicated substantially dierent evolutionary dynamics for
cytochrome b and b-brinogen intron 4 data sets are the two gene regions. Analysis of the cytochrome b data
summarized in Fig. 1. The minimum value possible for yielded the GTR + I + C model as the best t (versus
the RobinsonFoulds symmetric dierence metric is GTR + I, 2 ln K 659, df 1; GTR + C, 2 ln K 89,
zero (where no bipartitions dier between trees), while df 1; HKY85 + I + C, 2 ln K 842, df 4; all p
the maximum is twice the number of bipartitions in a 0:01). Additionally, the molecular clock could be rejected
fully-resolved unrooted tree (2n 6, which is in this for these data, though by a much narrower margin
case 60; when all bipartitions in the two trees conict). (2 ln K 50, df 31, p 0:02). The best-t model for
In general, three types of analysis of the cytochrome b the intron 4 data was GTR + C (versus HKY + C,
data yielded topologies most similar to those found in 2 ln K 29, df 4, p 0:01; the optimization algo-
analyses of the intron data: parsimony analysis with rithm of PAUP* would not assign a non-zero piv when
exclusion of third position transitions, combinatorial tting the GTR + I + C model). The GTR + I model yiel-
weights, and maximum likelihood (Fig. 1). These ded a slightly lower likelihood than GTR + C
treatments of the cytochrome b data yielded topologies (D lnk 6:2): this is probably not a signicant dierence
which had mean dierences from intron 4 topologies though this cannot be readily evaluated, as the two models
on average 8.2 dierence units better than compari- are not nested. For these data, the molecular clock could
sons of trees from equally weighted parsimony, par- not be rejected (2 ln K 41, df 31, p 0:11).
simony excluding third position data, or parsimony A heuristic search for the cytochrome b data under
analysis of inferred amino acid sequences. Addition of the GTR + I + C model (with initial estimates xed)
gap information to parsimony analysis of the intron 4 identied three equally likely trees, the strict consensus
data did not appreciably aect congruence with cyto- of which lacked resolution at one node (Table 4,
chrome b trees. However, maximum likelihood analy- Fig. 2A). Bootstrap analysis indicated that most recov-
sis of the intron 4 data did slightly increase ered nodes had very little support, as only 16 of the 29
congruence, and the maximum congruence between retained taxon bipartitions were recovered in more than
independent phylogenetic estimates from the two data 50% of bootstrap replicates, and only 11 in more than
sets was obtained when maximum likelihood was used 75% (the criterion value chosen for this study, Fig. 2A).
Table 4
Summary of model parameters and tree scores for maximum likelihood and Bayesian analyses of cytochrome b and b-brinogen intron 4 for all taxa
(rIJ refer to parameters of the GTR model of substitution, a is the parameter of the C-distribution for rate heterogeneity, and piv is the estimated
proportion of invariant sites)
Maximum likelihood Bayesian-separatea Bayesian-combinedb
Cytochrome b Intron 4 Cytochrome b Intron 4 Cytochrome b Intron 4
c
# Trees (# Nodes ) 3 (29) 405 (25) NA NA NA NA
lnL 11877.5 3944.1 11879.3 (10.4) 3987.5 (7.1) 15904.9 (10.1) joint estimate
Length 4.078 1.021 25.8 (3.6) 1.16 (0.06) 14.6 1.59 (0.01) 14.6 0.07 (0.01)
rAC 2.697 1.634 0.106 (0.05) 2.053 (0.49) 0.081 (0.04) 2.122 (0.49)
rAG 8.983 4.819 10.731 (3.89) 5.950 (1.17) 8.263 (2.55) 6.004 (1.19)
rAT 1.413 0.689 0.725 (0.31) 0.868 (0.21) 0.613 (0.23) 0.857 (0.20)
rCG 0.372 1.918 0.267 (0.16) 2.219 (0.56) 0.212 (0.12) 2.189 (0.55)
rCT 9.117 4.021 9.424 (3.42) 4.891 (1.01) 7.601 (2.23) 4.880 (0.96)
a 1.004 3.174 0.300 (0.03) 3.885 (1.83) 0.308 (0.03) 9.456 (8.64)
piv 0.481 NA 0.381 (0.02) NA 0.384 (0.03) 0.071 (0.05)
a
Means across retained Markov chain samples (ncytb 19; 400 generations, nI4 19; 700), standard deviations in parentheses.
b
As for a, ncombined 19; 900 generations.
c
Number of nodes retained in strict consensus.
A B
Pipra Pipra
Dasyornis 100 Dasyornis
Meliphaga 1.00 Meliphaga
99 Acanthidops 100 Acanthidops
66 1.00 Zeledonia 1.00 Zeledonia
0.65
Sitta Donacobius
19 89
60
0.51
Zosterops 1.00 73 Prinia
0.96 35
Donacobius 30 0.75 Zosterops
0.35 68
98 0.37
0.82 Prinia 1.00
Sialia
24 92
0.49 Dumetella 1.00 96 Dumetella
52
1.00 80 Sialia 1.00 Sturnus
0.85 Sturnus Sitta
44
0.99 Certhia 71 Certhia
0.93
39
Polioptila 79 Polioptila
0.84 Odontorchilus 0.99 96 Catherpes
27
Campylorhynchus brunneicapillus 100 1.00 Hylorchilus
97 0.21 82
0.95 0.98 1.00
C. megalopterus 60 Microcerculus
79
Troglodytes 93 0.76 Salpinctes
64 1.00
1.00 83 Cistothorus palustris Odontorchilus
1.00
0.98 C. platensis Troglodytes
95
98 Thryomanes 66 1.00 100 Cistothorus palustris
12 1.00 Thryothorus ludovicianus 0.95 1.00 C. platensis
11
0.26 Microcerculus Campylorhynchus brunneicapillus
0.19 100
76 64
Salpinctes 100 1.00 C. megalopterus
0.83 44 1.00
0.46 88 Catherpes 1.00 95 Thryomanes

9
1.00 Hylorchilus 1.00 Thryothorus ludovicianus
0.16 51
Cinnycerthia 0.80 Cinnycerthia
71 Cyphorhinus Thryothorus maculipectus
100
18 0.66 Henicorhina Cyphorhinus
1.00
0.79
100 Thryothorus coraya Henicorhina
20
8 1.00 T. maculipectus Thryothorus coraya
0.42
0.10 93 T. guarayanus 86 T. guarayanus
1.00 T. leucotis 1.00 T. leucotis
Fig. 2. Strict consensus of maximum likelihood trees obtained from separate analyses of cytochrome b (A) and b-brinogen intron 4 (B). Numbers of
trees, parameter estimates, and likelihoods are reported in Table 4. Numbers above each branch are the percentage of bootstrap replicates (n 100)
in which that branch was recovered, and numbers below each branch are estimated Bayesian posterior probabilities (19, 400, and 19,700 sampled
trees for each data set, respectively). Branches with either bootstrap percentages P75 or estimated Bayesian posteriors P0.95 are highlighted with
ellipses, and the values exceeding these criteria shaded.
The majority rule consensus of trees sampled in the were closely spaced in this data set, and search times
Bayesian analysis of the cytochrome b data was similar under the clock constraint increased to unrealistic
to the maximum likelihood trees recovered, and the es- lengths. Heuristic searches using the preferred model
timated posterior nodal probabilities yielded similar (without a clock, and with initial parameters xed)
conclusions. Searches enforcing the molecular clock for yielded 405 equally likely trees, the consensus of which
the intron 4 data were not feasible as many divergences yielded irresolution at four nodes (Table 4, Fig. 2B). The
strict consensus was surprisingly well-resolved (Fig. 2B, 3.3. Testing homogeneity of substitution process
25/30 possible nodes retained), given the large number
of trees obtained. Levels of bootstrap support for nodes Given that maximum likelihood analysis of the two
on the strict consensus was fairly high, with 23/25 re- data sets maximized congruence between them, analysis
tained nodes found in P50% of bootstrap replicates, of the combined data under maximum likelihood was
and 17/25 found in P75% (Fig. 2B). As for the cyto- preferred. Prior to this analysis, the validity of the as-
chrome b data, the consensus of trees obtained from sumption of evolutionary process homogeneity was
Bayesian analysis of the intron data was similar to the evaluated. Under the maximum likelihood criterion, the
consensus of maximum likelihood trees, and the pos- signicance of among-gene heterogeneity in best-t
terior probabilities obtained yielded essentially identical models of nucleotide substitution was estimated (Tables
conclusions. 5 and 6). Likelihood ratio tests indicated that the two
These analyses yielded only one case where signi- data sets diered signicantly in their best-tting sub-
cant conict was indicated. Both data sets recovered a stitution models and parameter values (Table 6), re-
group containing the genera Sialia, Dumetella, and jecting homogeneity between the two data sets in every
Sturnus. However, within this group, the cytochrome b parameter examined. Because of the high values of these
data consistently grouped Sialia and Sturnus (80% of statistics, their signicance is not aected by correction
bootstrap replicates), while the intron data yielded for multiple comparisons (28 comparisons suggests an
strong evidence for grouping of Dumetella and Sturnus a 0:002, and all values reported here have p < 0:001).
(96%). Bayesian analyses of the same data failed to The extremely high signicance values for these tests
recover signicantly conicting hypotheses of relation- prompted evaluation of the validity of the use of the v2
ship, as the estimated posterior probability of the approximation to the null distribution for 2 ln K. For
branch uniting Sialia and Sturnus was only 0.85 for the the simplest case of each model parameter being tested
cytochrome b data (Fig. 2A). Conversely, both data for homogeneity (highlighted in bold in Table 6), the
sets supported critical relationships, including mono- null distribution of the test statistic was evaluated by
phyly of the traditional Troglodytidae (i.e., exclusive of simulating the data under the simplied model (100
Donacobius; 64 and 93% of bootstrap replicates for simulated data sets per model comparison generated
cytochrome b and intron 4 respectively, estimated with Seq-Gen v1.1 [Rambaut and Grassly, 1997], using
posterior probabilities of 1.00 for both data sets), and a parameter values estimated by the baseml program of
sister-group relationship between the Troglodytidae PAML v3.12), and calculating the likelihood of the
and Polioptila (97 and 100%, 0.95 and 1.00). The simulated data sets under the null (homogeneous; the
presence of only a single point of signicant conict, condition of the generating model) and alternative (het-
the ambiguity of its signicance, and its irrelevance to erogeneous) hypotheses. In all cases except that of the
the primary phylogenetic hypotheses being tested (see 2 1C versus 1 comparison, the Monte Carlo-generated
Section 2), suggested that combined analysis of the two null distribution matched the v2 approximation extremely
partitions was warranted. well, as evaluated qualitatively by quantilequantile plots
Table 5
Likelihoods of heterogeneous model evaluations using PAML (pi are the base frequency parameters, R is the substitution rate matrix, a is the
parameter of the C-distribution of rates)
Model # pi R a Branch lengths Molecular clock lnk Ga Mgenea Malphaa
1 Equal Equal Equal Equal No 16261.6 No 0 0
2 + 1C Equal Equal Equal Proportional No 16214.9 Yes 0 0
2 + 2C Equal Equal Unequal Proportional No 16128.8 Yes 0 1
3 + 1C Unequal Equal Equal Proportional No 16151.2 Yes 2 0
3 + 2C Unequal Equal Unequal Proportional No 16053.7 Yes 2 1
4 + 1C Equal Unequal Equal Proportional No 16174.8 Yes 3 0
4 + 2C Equal Unequal Unequal Proportional No 16087.5 Yes 3 1
5 + 1C Unequal Unequal Equal Proportional No 16098.1 Yes 4 0
5 + 2C Unequal Unequal Unequal Proportional No 16018.3 Yes 4 1
6-NC Unequal Unequal Unequal Non-proportional No 15934.4 Yes 1 1
6-C Unequal Unequal Unequal Non-proportional Yes 15982.7 Yes 1 1
Under the headings pi , R, and a, an entry of equal indicates that the corresponding parameter was assumed uniform across partitions, and an
entry of unequal indicates that the parameter was estimated separately for each partition. Under the heading of branch lengths, equal indicates
that a single rate was assumed for the two partitions, proportional that a constant describing relative branch lengths for the two partitions (c) was
estimated, and non-proportional that all branch lengths were optimized independently for the two partitions.
a
These headings indicate settings used in analyses with PAML. Inclusion of the G statement with the input data matrix allows recognition of
multiple process partitions. The Mgene parameter indicates the non-rate-heterogeneity parameters to be allowed to vary between partitions, and
Malpha indicates when patterns of among-site rate variation are allowed to vary.
Table 6
Likelihood ratio tests of model parameter heterogeneity between cytochrome b and b-brinogen intron 4 (parameters and model numbers as in Table
5, 2 ln K likelihood ratio statistic, df degrees of freedom for v2 approximation)
Hypothesis tested General model Restricted model 2 ln K df
Homogeneous rate distribution (a) 2 2C 2 1C 172.1 1
3 + 2C 3 + 1C 195.1 1
4 + 2C 4 + 1C 174.6 1
5 + 2C 5 + 1C 159.5 1
Homogeneous base composition (pi ) 3 1C 2 1C 127.3 3
3 + 2C 2 + 2C 150.2 3
5 + 1C 4 + 1C 153.3 3
5 + 2C 4 + 2C 138.3 3
Homogeneous rate matrix (R) 4 1C 2 1C 80.2 5
4 + 2C 2 + 2C 82.7 5
5 + 1C 3 + 1C 106.2 5
5 + 2C 3 + 2C 70.7 5
Equal branch lengths 2 1C 1 93.5 1

Proportional branch lengths 6-NC 5 2C 167.9 36
Molecular clock 6-NC 6-C 96.6 62
All values are signicant at a 6 0:05, correcting for multiple comparisons. The v2 approximation to the null distribution of 2 ln K for model
comparisons highlighted in bold was validated via Monte Carlo simulation (see Section 3).
(results not shown). In the exceptional case, the Monte former. Within wrens, all genera except Thryothorus
Carlo-generated null distribution did not appear to were supported as monophyletic, and four strongly
follow a v2 distribution with 1 degree of freedom. Re- supported intergeneric groups were recovered. The rst
gardless of this discrepancy, the high value of the sta- of these comprised four of ve Thryothorus, Henicorh-
tistic obtained exceeded 100% of the empirically derived ina, Cinnycerthia, and Cyphorhinus. The second com-
null values, indicating heterogeneity in average substi- prised the genus Campylorhynchus, along with its sister-
tution rate between the two partitions. group of Thryomanes and Thryothorus ludovicianus. The
third comprised the genera Cistothorus and Troglodytes.
3.4. Combined analysis Finally, the genera Catherpes and Hylorchilus were
joined with high posterior probability. Additional nodes
Subsequent to model evaluation, the cytochrome b within the wrens received only weak support.
and intron 4 data were analyzed simultaneously using
Bayesian methods, allowing the partitions to vary in- 3.5. Additional analyses of the Troglodytidae and
dependently in their substitution parameter estimates. Polioptila
As optimal models varied between the partitions, the
more general of the two parameterizations was used Most of the lack of resolution in relationships among
(GTR + I + C, without a molecular clock), and all pa- the taxa sampled here could be isolated to three portions
rameters were allowed to vary between the data sets. of the tree. First, basal relationships among distantly
Although the parameter analysis above indicated that related passeridan groups (e.g., wrens and allies, mus-
independent branch lengths t the data signicantly cicapoids, sylvioids, and passeroids) were ambiguously
better than proportional branch lengths, only the latter resolved. This is perhaps unsurprising given the lack of
is implemented in MrBayes v3.0, and this option was dense taxon sampling at this level of comparison.
used. The majority rule consensus of 19,800 trees sam- Among closely related taxa, the clade containing four
pled from the Markov chains at stationarity is presented Thryothorus, Cinnycerthia, Cyphorhinus, and Henicorh-
in Fig. 3, and the parameters obtained in these analyses ina was unresolved, probably due to insucient varia-
are summarized in Table 4. The consensus tree obtained tion to resolve closely-spaced divergences. The nal
was extremely similar to those obtained in maximum major region of irresolution was among basal lineages
likelihood and Bayesian analysis of the intron 4 data within the Troglodytidae (especially Salpinctes, Micro-
alone. The traditional Troglodytidae was recovered. The cerculus, Hylorchilus/Catherpes, and Odontorchilus).
genus Donacobius was excluded from the family, and This lack of resolution could be due to a lack of su-
strongly supported as part of a group including cient variation to resolve these basal relationships.
Zosterops and Prinia. Polioptila and Certhia were Alternatively, the distant, species-poor sister-group to
strongly supported as successive sister-groups to the the wrens (Polioptila) could be problematic (Wheeler,
wrens, with weak support for Sitta as sister to all the 1990b; Smith, 1994). For this reason, phylogenetic
Fig. 3. Results of heterogeneous-model Bayesian analysis of cytochrome b and b-brinogen intron 4 for the complete data set. Shown is the 50%
majority rule consensus of 19,900 trees sampled from a single run of MC3 , with additional compatible nodes present in fewer than 50% of samples
retained. Branch lengths are proportional to the expected number of substitutions per site, estimated as the mean value of the branch in all samples of
the Markov chain where the branch appeared. All branches had estimated posterior probabilities P0.95, excepting those highlighted by dashing:
selected values relevant to hypotheses being tested are shown.
analyses were repeated using pruned data sets contain- Odontorchilus (Fig. 4). Thus, it seems likely that the
ing wrens plus their sister-group (as indicated by the weak support for basal wren relationships in analysis of
complete data set and both genes) and wrens alone. the complete data set is attributable to ambiguous
The combined Bayesian analysis was repeated for the rooting of a relatively robust ingroup network. In fact,
two data sets, including the traditional Troglodytidae pruning of Polioptila from trees obtained in Bayesian
(exclusive of Donacobius), both with and without the analysis of wrens plus this outgroup yielded nodal pos-
addition of Polioptila as the only outgroup (Fig. 4). The terior probabilities essentially identical to those ob-
nodal posterior values clearly indicate that placement of tained from analysis of wrens alone. Analysis of base
the outgroup is responsible for a great deal of the basal substitution data alone seems insucient to root the
irresolution observed in analysis of the complete data network of wren relationships conclusively.
set. Specically, posterior probabilities for the analysis
of wrens alone exceed 1 a 0:95 at a number of nodes 3.6. Indels in b-brinogen intron 4 and provisional rooting
which were not reliably resolved in analysis of the of wren phylogeny
complete taxon sample. Revealingly, posterior proba-
bilities for placement of the outgroup Polioptila on this Alignment of the intron sequences obtained required a
unrooted ingroup tree indicate substantial ambiguity number of insertion/deletion events which appear phy-
(Fig. 4). The majority of Markov chain samples which logenetically informative (indels showed ensemble
had the ingroup topology shown in Fig. 4 placed Po- CI 0.74 and RI 0.89 in equally-weighted parsimony
lioptila on the branch between Cistothorus/Troglodytes analysis of intron 4 alone), although inclusion of coded
and the remaining wrens. However, substantial numbers indels in phylogenetic analysis did not improve resolu-
of sampled trees placed the root on internodes among tion or support (results not shown). Since taxon sam-
Catherpes/Hylorchilus, Microcerculus, Salpinctes, and pling was focused on the wrens, it is unsurprising that the
Clade B
Hylorchilus Thryothorus
Catherpes coraya T. maculipectus
Cinnycerthia
Odontorchilus
Thryothorus guarayanus
1.00 12
Microcerculus 2
0.78 1.00 1.00
1.00 0.62
16
3
<1 1.00 28 T. leucotis
Salpinctes 1.00 0.24
0.85
1.00
1.00
<1
38 Henicorhina
<1
Troglodytes 1.00
1.00
1.00 Cyphorhinus
1.00
Cistothorus
palustris Campylorhynchus
C. platensis brunneicapillus
Clade C C. megalopterus
Thryomanes
Thryothorus
ludovicianus
Clade A
Fig. 4. Results of heterogeneous-model Bayesian analysis of cytochrome b and b-brinogen intron 4 from the Troglodytidae alone. Decimal values
near each branch indicate estimated posterior probabilities (19,800 sampled generations). Integer values in bold indicate the percentage of gener-
ations where the root fell on the associated branch of this unrooted tree, in a Bayesian analysis including the single outgroup Polioptila (19,900
sampled generations). The single black hash mark indicates the optimal placement of the indel in region 6 of the intron 4 alignment, the white hash
mark indicates optimization of the indel in region 16, and the two gray hash marks indicate two alternative optimizations of the indel in region 12 (see
Appendix). Clades AC are dened for discussion in the text.
majority of these informative indel events occur within Parsimony-based reconstruction of indel evolution in
the group. In particular, three taxon partitions presented both of these latter two regions of the intron suggest that
in Fig. 4 were supported by indel events in intron 4 basal relationships of wrens involve divergences among
(Appendix A). Towards the tips of the tree, the place- Salpinctes, Microcerculus, Catherpes/Hylorchilus, and
ment of the (Thryomanes, Thryothorus ludovicianus) the remaining wren genera. This result is consistent with
clade as sister-group to Campylorhynchus, was supported the unrooted consensus ingroup tree obtained in Bayes-
by a single base pair deletion (Fig. 4; Appendix A, region ian analysis, and with the distribution of optimal out-
16). Two other indel events within wrens provide po- group placement on this tree (see above).
tentially critical information regarding the most appro-
priate rooting of wren ingroup relationships. First,
Salpinctes is separated from all other wrens by a single 4. Discussion
unambiguous indel event (Fig. 4; Appendix A, region 6).
Based on outgroup comparison, this is clearly an inser- 4.1. Monophyly of the traditional Troglodytidae
tion which occurred in the common ancestor of all wrens
excepting Salpinctes, suggesting that this species may be Based upon DNA hybridization results (Sibley and
the sister group to all other wrens. However, a second Ahlquist, 1984), and an unpublished study of pterylosis
partially overlapping deletion in this same region of in- by Clench et al. (1982; cited in Kiltie and Fitzpatrick,
tron sequence from Thryothorus coraya suggests that 1984, Olson in Wetmore et al., 1984, p. 55), Sibley and
Salpinctes having reversed this synapomorphic indel Ahlquist (1990) suggested placement of the monotypic
cannot be discounted entirely. A second, less clearly in- genus Donacobius within the Troglodytidae. This con-
terpretable indel event in region 12 of the alignment clusion was adopted by the American Ornithologists
(Fig. 4, Appendix A) separates either (Salpinctes, Union in its list of North American Birds (1983 and
Microcerculus) or (Salpinctes, Microcerculus, Catherpes, subsequent). Donacobius atricapillus is a widely distrib-
Hylorchilus) from all other wrens. Based on outgroup uted South American marsh-nesting bird which cooper-
comparison, this region appears to have experienced a atively breeds (as do some wrens) and performs vocal
six base pair deletion in the common ancestor of oscines duets (as do most wrens). Traditionally, the genus has
(all sampled taxa except Pipra), followed by an over- been associated with the Mimidae (e.g., Mayr and
lapping three base pair deletion in all wrens excepting Greenway, 1960), though often noted as an aberrant
Salpinctes, Microcerculus, and possibly Catherpes and member of the family (e.g., Ridgway, 1907). Though the
Hylorchilus. The ambiguity in reconstruction of this species does exhibit some behavioral characteristics
latter indel is caused by an overlapping 15 bp deletion found in wrens, in others it is quite distinctive (e.g., it
event in the lineage leading to Catherpes and Hylorchilus. constructs a cup nest, whereas wrens typically construct
globular or pouch-like nests). Since the data on pterylosis tion of this pattern is somewhat complicated by ambi-
are unpublished, they are impossible to evaluate relative guity of the placement of the root node on the wren
to other evidence. Additionally, the genetic evidence cited ingroup network, but a number of phylogenetic con-
(a list of hybridization distances; Sibley and Ahlquist, clusions can be drawn regardless of the precise rooting.
1984) does not actually indicate a relationship between Notably, the relationships proposed here conict shar-
Donacobius and the wrens, rather demonstrating that ply with previous conceptions regarding the sequence
Donacobius is just as genetically distant from the mimic- from primitive to derived taxa within the family.
thrushes as it is from the wrens. Thus, the decisions by Traditional taxonomic series generally begin with the
Sibley and Monroe (1990), and the A.O.U. Check-list genus Campylorhynchus, which has been placed in its
committee (A.O.U., 1983) to refer the genus to the family own subfamily (e.g., Baird, 1858). However, this genus
Troglodytidae, are based upon evidence that is either is clearly nested well within the wrens (Figs. 3 and 4). An
unavailable for evaluation or equivocal in interpretation. assemblage of genera which tend toward a terrestrial
The analyses presented here, of both data sets sepa- behavioral and morphological habit (Thryorchilus,
rately and in combination, strongly support monophyly Henicorhina, Microcerculus, and Cyphorhinus) is typi-
of the traditional Troglodytidae. Specically, the genus cally listed at the end of the family. This assemblage is
Donacobius is excluded from membership within the clearly polyphyletic, with Thryorchilus belonging to
family, and cannot even be parsimoniously (or proba- Clade C (. Rice et al., 1999), Henicorhina and Cypho-
bilistically) placed as its sister-group. All analyses un- rhinus as sister-taxa within Clade B, and Microcerculus
equivocally place Polioptila as the sister-group to the as part of a probable basal radiation of the family (see
wrens, to the exclusion of Donacobius, strongly con- Fig. 4 for group denitions).
tradicting the conclusions of Clench et al. (1982), as well The single species of Catherpes, the canyon wren
as current taxonomies based on their result (A.O.U., (C. mexicanus) has been considered a close relative of the
1998; Sibley and Monroe, 1990). Also of note is that similarly petrophilous S. obsoletus, and even placed in
Donacobius is consistently excluded from relationship the latter genus (e.g., Mayr and Greenway, 1960). This
with the traditional family Mimidae (Sibley and Ahl- notion is clearly falsied by the analyses presented here,
quist, 1990, subfamily Miminae), the other group to as Catherpes and Hyorchilus are sister taxa. A close
which the genus has been commonly assigned. The af- relationship between Catherpes and Hylorchilus was
nities of Donacobius appear to lie with the Old World previously suggested based on similarity of the songs of
sylvioid passerine radiation, as it consistently clusters C. mexicanus and H. sumichrasti (Hardy and Delaney,
with Zosterops and Prinia in analyses presented here. 1987). The recently discovered dierentiation of H. navai
However, because of sparse taxon sampling, no partic- songs from the former two species (Atkinson et al., 1993;
ular hypothesis of relationship can be postulated for the G omez de Silva, 1997), probably represents a case of
genus based upon this result. secondary modication. In itself, this relationship does
Beyond the Polioptila/wren relationship, both the not argue against placing all three genera, which together
cytochrome b and intron 4 data support monophyly of comprise four petrophilous species, within a single genus.
Sibley and Ahlquists (1990) Certhiidae (as modied by However, data on indel events from the intron indicate
exclusion of Donacobius), in that Certhia is found as the that Salpinctes may actually represent the sister-group to
sister-group to the Polioptila/wren clade, though sup- all other wrens (see above). Until additional data are
port for this relationship was only strong in analyses of obtained, retaining all three genera seems warranted.
intron 4 or the combined data (Figs. 1 and 3). No sig- A more controversial hypothesis of relationship pro-
nicant support was found for Sibley and Ahlquists posed by these data is the placement of Thryothorus
proposed grouping of Sitta with this broad conception ludovicianus and Thryomanes bewickii as sister taxa to the
of the Certhiidae, though this relationship appeared in genus Campylorhynchus. Finding paraphyly of Thryoth-
the maximum likelihood analysis of the intron data orus per se would not be particularly surprising. For
(Fig. 2B; 71% bootstrap), and in the majority rule con- instance, nding the genus Thryomanes (or Henicorhina,
sensus of trees obtained in the Bayesian analyses of both etc.) nested within Thryothorus, given the behavioral and
the intron and combined data sets (0.93 and 0.87 pos- morphological similarities of these genera, would not
terior probabilities; Figs. 2B and 3). However, this latter profoundly disturb traditional notions of relationship.
relationship is well-supported by other nuclear sequence However, the placement of Thryothorus ludovicianus as
data (Barker et al., 2002). separate from other members of the genus, yielding
polyphyly of Thryothorus, merits some careful consid-
4.2. Relationships among wrens eration. Unfortunately, evidence for this relationship
is from the intron data alone. Comparison of the
The pattern of relationships among genera within the cytochrome b maximum likelihood trees with the best
Troglodytidae presented here is the rst proposed for trees found under the constraint of Thryothorus (or
this family, beyond linear taxonomic series. Interpreta- Thryothorus + Thryomanes) monophyly with the Shi-
modairaHasegawa test (Shimodaira and Hasegawa, site rate heterogeneity (Table 4). When analyses which
1999) failed to reject this alternative (1000 bootstrap take some of these diering dynamics into account are
RELL replications, d 6:47 p 0:18), whereas the same employed (parsimony weighting and likelihood), ob-
test with the intron 4 data is highly signicant (d 64:00, served incongruence between the loci is reduced (Fig. 1).
p < 0:01). One immediate objection which might be As measured by bootstrap values and Bayesian poster-
raised is the quality or nature of these sequence data; ior probabilities, the most congruent of these analyses
however, two individuals of Thryothorus ludovicianus (likelihood-based) indicates signicant conict at only
albinucha from the Yucat an yielded intron sequences one node (Fig. 2). This conict, rather than due to
identical to that presented here (including the unique conicting history of the genes sampled, is very likely
deletion event which unites them with Thryomanes and attributable to among-taxon heterogeneity in the dy-
Campylorhynchus [Fig. 4; Appendix A, indel 16], un- namics of sequence evolution (as measured by base
published data). Extensive hybridization between the composition, see Section 3).
common ancestor of Campylorhynchus and that of T. Likelihood evaluation of the optimal model for use in
ludovicianus and Thryomanes could provide an alterna- analysis of the two data sets suggests that they dier
tive explanation for this pattern, though hybridization signicantly in every parameter tested (Table 6). De-
among wren genera has not been reported: this expla- monstrably, the assumption of identically distributed
nation seems unlikely at best. Only evidence from addi- site pattern probabilities can be rejected for these data,
tional genetic loci (preferably nuclear) will resolve the and homogeneous-model likelihood analysis is unwar-
issue denitively. ranted. Previously, some authors have attempted to
Future work on relationships of wrens will be neces- evaluate a limited number of hypotheses using hetero-
sary in order to resolve a number of issues. Most critically, geneous-model maximum likelihood, with parameters
resolution of basal relationships of wrens will require independently estimated for each partition (Caterino
collection of additional sequence data and possibly more et al., 2001; DeBry, 1999; Wilgenbusch and de Quieroz,
dense taxon sampling. Additionally, acquiring nuclear 2000). Additionally, limited searches using heteroge-
data from additional outgroups (specically the gnatw- neous models are possible in some available software
rens Ramphocaenus and Microbates) might help stabilize (NNI branch swapping and star decomposition in
placement of the root on the wren ingroup network. PAML, Yang, 1997). These approaches are unsatisfying
Resolution of these relationships will allow biogeo- in that they provide little assurance that global optima
graphic analysis of the family, potentially resolving the have been obtained, and more importantly in that esti-
question of the continental origin of wrens (Mayr, 1946). mating branch support using resampling (i.e., the
This analysis would also benet greatly from sampling of bootstrap) is intractable. Ecient search options with
the remaining genera of wrens, the Central American heterogeneous likelihood models will become available
Uropsila and Thryorchilus, and (especially) the Cuban in the future (subsequent versions of PAUP*; Wil-
Ferminia. Finally, additional sampling from the taxo- genbusch and de Quieroz, 2000). Heterogeneous-model
nomically diverse genus Thryothorus and its close rela- Bayesian analysis with uniform interval priors provides
tives may reveal additional evidence for paraphyly of the a reasonable and ecient alternative approach to such
genus, and help resolve the history of morphological and problems, and oers the additional advantages associ-
behavioral diversication within the group. ated with Bayesian methods (Huelsenbeck et al., 2002).
Methods for addressing among-taxon heterogeneity in
4.3. Conict between data partitions and combined evolutionary process have also been developed, but have
analysis yet to be implemented with ecient search algorithms
(Galtier and Gouy, 1998; Yang and Roberts, 1995). As
Substantial incongruence was observed between esti- among-taxon compositional heterogeneity in this data
mates of relationship derived from analyses of the cy- set was limited, these approaches were not pursued.
tochrome b and intron 4 data (Fig. 1). This observed Bayesian analysis of other types of among-taxon process
incongruence appears to have a number of sources. heterogeneity has proven useful (e.g., the rate of
Comparison of trees derived from various parsimony sequence evolution; Huelsenbeck et al., 2000; Kishino
weighting schemes and maximum likelihood analysis et al., 2001; Thorne et al., 1998), and may be a fruitful
strongly suggests that one signicant factor contributing approach to addressing the problem of base composi-
to apparent incongruence is the contrasting evolutionary tion as well (Huelsenbeck et al., 2002).
dynamics of the two loci. Cytochrome b and intron 4
dier substantially in their rates and patterns of se-
quence evolution (Tables 46). In particular, cyto- Acknowledgments
chrome b does not evolve in a clock-like fashion (see
Section 3), accrues substitutions at a much higher rate This work was supported by grants from the Hinds
than intron 4, and has a very dierent pattern of among- Fund, University of Chicago; the F.M. Chapman
Memorial Fund, American Museum of Natural History; I thank Fredrik Ronquist (Uppsala) and Birali Runesha
and the American Ornithologists Union (Blake Award). (UMN Supercomputing Institute) for troubleshooting
For their generosity in providing samples, I thank Leo compilation of MrBayes v3.0. Various versions of the
Joseph (ANSP), Fred Sheldon (LSUMNS), Carla manuscript beneted from the comments of Allan Ba-
Cicero (MVZ), Jon Fjeldsa (MZUC), and Adolfo ker, Barry Cherno, Shannon Hackett, Sharon Jansa,
Navarro Sig uenza (MZAH); and for sharing unpub- Francois Lutzoni, the American Museums Ornithology
lished data I thank Shannon Hackett and Amy Driskell. Discussion Group, and an anonymous referee.
Appendix A
Location of potentially informative gapped regions in the alignment of b-brinogen intron 4 sequences (EMBL
Accession ALIGN_000588), and coding of indel data. All regions were coded separately as phylogenetic characters,
except regions in italics which were included in the matrix but uncoded, and the region in bold which was excluded
from analysis.
Locations. 1: 68-97, 2: 100, 3: 134-137, 4: 153-165, 5: 169-170, 6: 209-221, 7: 255, 8: 328, 9:344-345, 10: 359-371,
11: 361-363, 12:487-507, 13: 509-523, 14: 524-526, 15: 563-573, 16: 601, 17: 627-629, 18: 639, 19: 655-656.
Coding
00000001111111
12367891226789
Pipra pipra 00000000000000
Meliphaga gracilis 00001000000010
Dasyornis broadbenti 00001000000010
Sialia sialis 10000000000101
Sturnus vulgaris 10000100000100
Dumetella carolinensis 10000100000100
Sitta pygmaea 10000010000100
Certhia familiaris 10000000000100
Donacobius atricapillus 10000011000101
Campylorhynchus brunneicapillus 11010000011100
C. megalopterus 11010000011100
Odontorchilus cinereus 10010000010100
Salpinctes obsoletus 10000010000100
Catherpes mexicanus 100100001?0100
Hylorchilus sumichrasti 100100001?0100
Cinnycerthia peruana 10010000010100
Cistothorus platensis 10010000010100
C. palustris 10010010010100
Thryomanes bewickii 10010000011100
Thryothorus coraya 10010000010100
T. maculipectus 10010000010100
T. ludovicianus 10010000011100
T. leucotis 10010000010100
T. guarayanus 10010000010100
Troglodytes aedon 10010000010100
Henicorhina leucosticta 10010000010100
Microcerculus marginatus 10010010000100
Cyphorhinus arada 10010000010100
Polioptila caerulea 10000000000100
Prinia bairdii 1000????000100
Zosterops senegalensis 10000001000100
Acanthidops bairdii 10100000000100
Zeledonia coronata 10100000000100
References Desjardins, P., Morais, R., 1990. Sequence and gene organization of
the chicken mitochondrial genome: a novel gene order in higher
Altekar, G., Dwarkadas, S., Huelsenbeck, J.P., Ronquist, F., in press. vertebrates. J. Mol. Biol. 212, 599634.
Parallel metropolis coupled Markov chain Monte Carlo for Edwards, S.V., Arctander, P., et al., 1991. Mitochondrial resolution of
Bayesian phylogenetic inference. Bioinformatics. a deep branch in the genealogical tree for perching birds. Proc. R.
American Ornithologists Union, 1886. Check-list of North American Soc. Lond. B 243, 99107.
Birds, rst ed. University Press, Cambridge, MA. Evans, M.R., 1997. The inuence of habitat and male morphology on
American Ornithologists Union, 1983. Check-list of North American a mate-choice cue: the display nests of wrens. Anim. Behav. 54,
Birds, sixth ed. Allen Press, Lawrence, KS. 485491.
American Ornithologists Union, 1998. Check-list of North American Evans, M.R., Burn, J.L., 1996. An experimental analysis of mate
Birds, seventh ed. Allen Press, Lawrence, KS. choice in the wren: a monomorphic, polygynous passerine. Behav.
Aota, S.-I., Ikemura, T., 1986. Diversity in G + C content at the third Ecol. 7, 101108.
position of codons in vertebrate genes and its cause. Nucleic Acids Farabaugh, S.M., 1982. The ecological and social signicance of
Res. 14, 63456355. duetting. In: Kroodsma, D.E., Miller, E.H., Ouellet, H. (Eds.),
Atkinson, P.W., Whittingham, M.J., G omez de Silva Garza, H., Acoustic Communication in Birds, vol. 2. Academic Press, New
Kent, A.M., Maier, R.T., 1993. Notes on the ecology, conserva- York, pp. 85124.
tion and taxonomic status of Hylorchilus wrens. Bird Conserv. Felsenstein, J., 1981. Evolutionary trees from DNA sequences: a
Int. 3, 7585. maximum likelihood approach. J. Mol. Evol. 17, 368376.
Baird, S.F., 1858. Birds. With the co-operation of John Cassin and Felsenstein, J., 1985. Condence limits on phylogenies: an approach
George N. Lawrence. Vol. 9, Part 2. In: Reports of explorations using the bootstrap. Evolution 39, 783791.
and surveys, to ascertain the most practicable and economic route Flook, P.K., Klee, S., Rowell, C.H.F., 1999. Combined molecular
for a railroad from the Mississippi River to the Pacic Ocean. B. phylogenetic analysis of the Orthoptera (Arthropoda, Insecta) and
Tucker, Printer, Washington, DC. implications for their higher systematics. Syst. Biol. 48, 233253.
Barker, F.K., Barrowclough, G.F., Groth, J.G., 2002. A phylogenetic Galtier, N., Gouy, M., 1998. Inferring pattern and process: maximum-
hypothesis for passerine birds: taxonomic and biogeographic likelihood implementation of a nonhomogeneous model of DNA
implications of an analysis of nuclear DNA sequence data. Proc. sequence evolution for phylogenetic analysis. Mol. Biol. Evol. 15,
R. Soc. Lond., Ser. B 269, 295308. 871879.
Barrett, M., Donoghue, M.J., Sober, E., 1991. Against consensus. Syst. Gomez de Silva, G.H., 1997. Comparative analysis of the vocalizations
Zool. 40, 486493. of Hylorchilus wrens. Condor 99, 981984.
Beecher, W.J., 1953. A phylogeny of the oscines. Auk 70, 270333. Gu, X., Fu, Y., Li, W.-H., 1995. Maximum likelihood estimation of
Brenowitz, E.A., Arnold, A.P., 1986. Interspecic comparisons of the the heterogeneity of substitution rate among nucleotide sites. Mol.
size of neural song control regions and song complexity in duetting Biol. Evol., 546557.
birds: evolutionary implications. J. Neurosci. 6, 28752879. Hackett, S.J., 1996. Molecular phylogenetics and biogeography of
Brewer, D., 2001. Wrens, dippers and thrashers. Christopher Helm, tanagers in the genus Ramphocelus (Aves). Mol. Phylogenet. Evol.
London. 5, 368382.
Brower, A.V.Z., DeSalle, R., Vogler, A., 1996. Gene trees, species trees Hardy, J.W., Delaney, D.J., 1987. The vocalizations of the slender-
and systematics: a cladistic perspective. Annu. Rev. Ecol. Syst. 27, billed wren (Hylorchilus sumichrasti): who are its close relatives?
423450. Auk 104, 528530.
Bull, J.J., Huelsenbeck, J.P., Cunningham, C.W., Swoord, D.L., Harshman, J., 1994. Reweaving the tapestry: what can we learn from
Waddell, P.J., 1993. Partitioning and combining data in phyloge- Sibley and Ahlquist? Auk 111, 377388.
netic analysis. Syst. Zool. 42, 384397. Harvey, P.H., Pagel, M.D., 1991. The comparative method in
Catchpole, C.K., Rowell, A., 1993. Song sharing and local dialects in a evolutionary biology. Oxford University Press, New York.
population of the European wren Troglodytes troglodytes. Behav- Hasegawa, M., Kishino, M., Yano, T., 1985. Dating the humanape
iour 125, 6778. split by a molecular clock of mitochondrial DNA. J. Mol. Evol. 22,
Caterino, M.S., Reed, R.D., Kuo, M.M., Sperling, F.A.H., 2001. A 160174.
partitioned likelihood analysis of swallowtail buttery phylogeny Helm-Bychowski, K., Cracraft, J., 1993. Recovering phylogenetic signal
(Lepidoptera: Papilionidae). Syst. Biol. 50, 106127. from DNA sequences: relationships within the corvine assemblage
Chippindale, P.T., Wiens, J.J., 1994. Weighting, partitioning, combin- (Class Aves) as inferred from complete sequences of the mitochon-
ing characters in phylogenetic analysis. Syst. Biol. 43, 278287. drial DNA cytochrome-b gene. Mol. Biol. Evol. 10, 11961214.
Chung, D.W., Harris, J.E., Davie, E.W., 1991. Nucleotide sequences of Hillis, D.M., 1987. Molecular versus morphological approaches to
the three genes coding for human brinogen. In: Liu, C.Y., Chen, systematics. Annu. Rev. Ecol. Syst. 18, 2342.
S. (Eds.), Fibrinogen, Thrombosis, Coagulation, and Fibrinolysis. Hillis, D.M., Bull, J.J., 1993. An empirical test of bootstrapping as a
Plenum Press, New York, pp. 3948. method for assessing condence in phylogenetic analysis. Syst.
Clench, M.H., Gulledge, J.L., Parkes, K.C., 1982. The black-capped Biol. 42, 182192.
Donacobius is a wren, not a mimid. Abstracts, 100th meeting Huelsenbeck, J.P., Crandall, K.A., 1997. Phylogeny estimation and
Amer. Ornithol. Union. hypothesis testing using maximum likelihood. Annu. Rev. Ecol.
De Quieroz, A., Donoghue, M.J., Kim, J., 1995. Separate versus Syst. 28, 437466.
combined analysis of phylogenetic evidence. Annu. Rev. Ecol. Syst. Huelsenbeck, J.P., Rannala, B., 1997. Phylogenetic methods come of
26, 657681. age: testing hypotheses in an evolutionary context. Science 276,
DeBry, R.W., 1999. Maximum likelihood analysis of gene-based and 227232.
structure-based process partitions, using mammalian mitochon- Huelsenbeck, J.P., Ronquist, F.R., 2001. MrBayes: Bayesian inference
drial genomes. Syst. Biol. 48, 286299. of phylogeny. Biometrics 17, 754755.
DeSalle, R., Brower, A.V.Z., 1997. Process partitions, congruence, and Huelsenbeck, J.P., Bollback, J.P., Levine, A.M., 2002. Inferring the
the independence of characters: inferring relationships among root of a phylogenetic tree. Syst. Biol. 51, 3243.
closely related Hawaiian Drosophila from multiple gene regions. Huelsenbeck, J.P., Bull, J.J., Cunningham, C.W., 1996. Combining
Syst. Biol. 46, 751764. data in phylogenetic analysis. Trends Ecol. Evol. 11, 152158.
Huelsenbeck, J.P., Larget, B., Swoord, D., 2000. A compound Mayr, E., Cottrell, G.W., 1986. Check-list of Birds of the World, vol.
process for relaxing the molecular clock. Genetics 154, 18791892. 11. Mus. Comp. Zool., Cambridge, MA.
Huelsenbeck, J.P., Larget, B., Miller, R.E., Ronquist, F., 2002. Mayr, E., Greenway Jr., J.C., 1960. Check-list of Birds of the World,
Potential applications and pitfalls of Bayesian inference of vol. 9. Mus. Comp. Zool., Cambridge, MA.
phylogeny. Syst. Biol. 51, 673688. Mayr, E., Paynter Jr., R.A., 1964. Check-list of Birds of the World,
Ikemura, T., Aota, S.-I., 1988. Global variation in G + C content along vol. 10. Mus. Comp. Zool., Cambridge, MA.
vertebrate genome DNA: possible correlation with chromosome Morton, E.S., Gish, S.L., van der Voort, V., 1986. On the learning of
band structures. J. Mol. Biol. 203, 113. degraded and undegraded songs in the Carolina wren. Anim.
Johnson, L.S., Kermott, L.H., Lein, M.R., 1993. The cost of polygyny Behav. 34, 815820.
in the house wren Troglodytes aedon. J. Anim. Ecol. 62, 669682. Nunn, G.B., Cracraft, J., 1996. Phylogenetic relationships among the
Johnson, L.S., Kermott, L.H., Lein, M.R., 1994. Territorial polygyny major lineages of the birds-of-paradise (Paradisaeidae) using mito-
in house wrens: are females suciently compensated for the cost of chondrial DNA gene sequences. Mol. Phylogenet. Evol. 5, 445459.
mate sharing. Behav. Ecol. 5, 98104. Perlman, P.S., Peebles, C.L., Daniels, C., 1990. Dierent types of
Jukes, T.H., Cantor, C.R., 1969. Evolution of protein molecules. In: introns and splicing mechanisms. In: Stone, E.M., Schwartz, R.J.
Munro, H.N. (Ed.), Mammalian Protein Metabolism. Academic (Eds.), Intervening Sequences in Evolution and Development.
Press, New York, pp. 21132. Oxford University Press, New York, pp. 112161.
Kiltie, R.A., Fitzpatrick, J.W., 1984. Reproduction and social orga- Rabenold, K.N., 1990. Campylorhynchus wrens: the ecology of delayed
nization of the black-capped Donacobius (Donacobius atricapillus) dispersal and cooperation in the Venezuelan savanna. In: Stacey,
in southeastern Peru. Auk 101, 804811. P.B., Koenig, W.D. (Eds.), Cooperative Breeding in Birds: Long-
Kishino, H., Thorne, J.L., Bruno, W.J., 2001. Performance of a term Studies of Ecology and Behavior. Cambridge University
divergence time estimation method under a probabilistic model of Press, Cambridge, MA, pp. 157196.
rate evolution. Mol. Biol. Evol. 18, 352361. Rabenold, P.P., Rabenold, K.N., Piper, W.H., Haydock, J., Zack,
Kocher, T.D., Thomas, W.K., et al., 1989. Dynamics of mitochondrial S.W., 1990. Shared paternity revealed by genetic analysis in
DNA evolution in animals: amplication and sequencing with cooperatively breeding tropical wrens. Nature 348, 538540.
conserved primers. Proc. Natl. Acad. Sci. USA 86, 61966200. Rambaut, A., Grassly, N.C., 1997. Seq-Gen: an application for the
Kornegay, J.R., Kocher, T.D., et al., 1993. Pathways of lysozyme Monte Carlo simulation of DNA sequence evolution along
evolution inferred from the sequences of cytochrome b in birds. J. phylogenetic trees. Comput. Appl. Biosci. 13, 235238.
Mol. Evol. 37, 367379. Reed, R.D., Sperling, F.A.H., 1999. Interaction of process partitions
Krajewski, C., Fain, M.G., Buckley, L., King, D.G., 1999. Dynam- in phylogenetic analysis: an example from the swallowtail buttery
ically heterogenous partitions and phylogenetic inference: an genus Papilio. Mol. Biol. Evol. 16, 286297.
evaluation of analytical strategies with cytochrome b and ND6 Rice, N.H., Peterson, A.T., Escalona-Segura, G., 1999. Phylogenetic
gene sequences in cranes. Mol. Phylogenet. Evol. 13, 302313. patterns in montane Troglodytes wrens. Condor 101, 446451.
Kroodsma, D.E., 1977. Correlates of song organization among North Ridgway, R., 1904. The birds of North and Middle America. Bull.
American wrens. Am. Nat. 111, 9951008. U.S. Natl. Mus. 50 (3), 1801.
Kroodsma, D.E., Canady, R.A., 1985. Dierences in repertoire size, Ridgway, R., 1907. The birds of North and Middle America. Bull.
singing behavior, and associated neuroanatomy among marsh wren U.S. Natl. Mus. 50 (4), 1973.
populations have a genetic basis. Auk 102, 439446. Robinson, D.F., Foulds, L.R., 1981. Comparison of phylogenetic
Kroodsma, D.E., Sanchez, J., Stemple, D.W., Goodwin, E., Da trees. Math. Biosci. 53, 131147.
Silva, M.L., Vielliard, J.M.E., 1999. Sedentary life style of Rodrigo, A.G., 1992. A modication to Wheelers combinatorial
Neotropical sedge wrens promotes song imitation. Anim. Behav. weights calculations. Cladistics 8, 165170.
57, 855863. Saitou, N., Nei, M., 1987. The neighbor-joining method: a new method
Lanyon, S.M., Hall, J., 1994. Reexamination of barbet monophyly for reconstructing phylogenetic trees. Mol. Biol. Evol. 4, 406425.
using mitochondrial-DNA sequence data. Auk 111, 389397. Sharpe, R.B., 1881. Catalogue of Birds in the British Museum, vol. 6.
Larget, B., Simon, D.L., 1999. Markov chain Monte Carlo algorithms Brit. Mus., London.
for the Bayesian analysis of phylogenetic trees. Mol. Biol. Evol. 16, Sheldon, F.H., Gill, F.B., 1996. A reconsideration of songbird
750759. phylogeny, with emphasis on the evolution of titmice and their
Leonard, M.L., 1990. Polygyny in marsh wrens: asynchronous sylvioid relatives. Syst. Biol. 45, 473495.
settlement as an alternative to the polygyny-threshold model. Shimodaira, H., Hasegawa, M., 1999. Multiple comparisons of log-
Am. Nat. 136, 446458. likelihoods with applications to phylogenetic inference. Mol. Biol.
Leonard, M.L., Picman, J., 1987. The adaptive signicance of multiple Evol. 16, 11141116.
nest building by male marsh wrens. Anim. Behav. 35, 271277. Sibley, C.G., Ahlquist, J.E., 1984. The relationships of the starlings
Levin, R.N., 1996a. Song behavior and reproductive strategies in a (Sturnidae: Sturnini) and the mockingbirds (Sturnidae: Mimini).
duetting wren, Thryothorus nigricapillus: I. Removal experiments. Auk 101, 230243.
Anim. Behav. 52, 10931106. Sibley, C.G., Ahlquist, J.E., 1990. Phylogeny and Classication of
Levin, R.N., 1996b. Song behavior and reproductive strategies in a Birds: A Study in Molecular Evolution. Yale University Press, New
duetting wren, Thryothorus nigricapillus: II. Playback experiments. Haven, CT.
Anim. Behav. 52, 11071117. Sibley, C.G., Monroe Jr., B.L., 1990. Distribution and Taxonomy of
Lockhart, P.J., Steel, M.A., Hendy, M.D., Penny, D., 1994. Recov- Birds of the World. Yale University Press, New Haven, CT.
ering evolutionary trees under a more realistic model of sequence Skutch, A.F., 1935. Helpers at the nest. Auk 52, 257273.
evolution. Mol. Biol. Evol. 12, 605612. Smith, A.B., 1994. Rooting molecular trees: problems and strategies.
Maddison, D.R., Maddison, W.P., 2000. MacClade: Analysis of Biol. J. Linn. Soc. 51, 279292.
phylogeny and character evolution, Version 4.0. Sinauer, Sunder- Suzuki, Y., Glazko, G.V., Nei, M., 2002. Overcredibility of molecular
land, MA. phylogenies obtained by Bayesian phylogenetics. Proc. Natl. Acad.
Mayr, E., 1946. History of the North American bird fauna. Wilson Sci. USA 99, 1613816143.
Bull. 58, 341. Swoord, D.L., 1998. PAUP*. Phylogenetic Analysis Using Parsi-
Mayr, E., Amadon, D., 1951. A classication of recent birds. Amer. mony (*and Other Methods). Version 4. Sinauer Associates,
Mus. Nov. 1496, 142. Sunderland, MA.
Thompson, J.D., Higgins, D.G., Gibson, T.J., 1994. CLUSTAL W: Wheeler, W.C., 1990a. Combinatorial weights in phylogenetic analysis:
improving the sensitivity of progressive multiple sequence align- a statistical parsimony procedure. Cladistics 6, 269275.
ment through sequence weighting, position specic gap penalties Wheeler, W.C., 1990b. Nucleic acid sequence phylogeny and random
and weight matrix choice. Nucleic Acids Res. 22, 46734680. outgroups. Cladistics 6, 363367.
Thorne, J.L., Kishino, H., Painter, I.S., 1998. Estimating the rate of Wilgenbusch, J., de Quieroz, K., 2000. Phylogenetic relationships
evolution of the rate of molecular evolution. Mol. Biol. Evol. 15, among the phrynosomatid sand lizards inferred from mitochon-
16471657. drial DNA sequences generated by heterogeneous evolutionary
Verner, J., 1975. Complex song repertoire of male long-billed marsh processes. Syst. Biol. 49, 592612.
wrens in eastern Washington. Living Bird 14, 263300. Yang, Z., 1994. Estimating the pattern of nucleotide substitution. J.
Weissbach, L., Oddoux, C., Procyk, R., Grieninger, G., 1991. The b Mol. Evol. 39, 105111.
chain of chicken brinogen contains an atypical thrombin cleavage Yang, Z., 1996. Maximum-likelihood models for combined analysis of
site. Biochemistry 30, 32903294. multiple sequence data. J. Mol. Evol. 42, 587596.
Wetmore, A., 1960. A classication for the birds of the world. Yang, Z., 1997. PAML: a program package for phylogenetic analysis
Smithson. Misc. Coll. 139 (11), 137. by maximum likelihood. CABIOS 13, 555556.
Wetmore, A.W., Pasquier, R.F., Olson, S.L., 1984. The Birds of the Yang, Z., Roberts, D., 1995. On the use of nucleic acid sequences
Republic of Panama, vol. 4. Smithsonian Inst. Press, Washington, to infer early branchings in the tree of life. Mol. Biol. Evol. 12,
DC. 451458.

Barker 04

Uploaded by

Copyright:

Available Formats

Barker 04

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Barker 04

Uploaded by

Copyright:

Available Formats

MOLECULAR

Monophyly and relationships of wrens (Aves: Troglodytidae):

Received 6 March 2003; revised 4 August 2003

Suborder Passeri: Parvorder Corvida

Suborder Passeri: Parvorder Passerida

Table 2 intron motifs (Perlman et al., 1990) identied. Contig

2.2.4. Separate phylogenetic analyses: Bayesian 2.2.6. Combined analysis

609 bp (Pipra and Dasyornis), with a median of 590 bp.

0.46 88 Catherpes 1.00 95 Thryomanes

Equal branch lengths 2 1C 1 93.5 1

You might also like