Song evolution in Maluridae: influences of natural and sexualselection on acoustic structure

Emma I. GreigA,D, J. Jordan PriceB and Stephen Pruett-JonesC

ACornell Laboratory of Ornithology, and Department of Neurobiology & Behavior, Cornell University,Ithaca, NY 14850, USA.

BDepartment of Biology, St Mary’s College of Maryland, St Mary’s City, MD 20686, USA.CDepartment of Ecology & Evolution, University of Chicago, Chicago, IL 60637, USA.DCorresponding author. Email: eig9@cornell.edu

Abstract. Many factors may influence the evolution of acoustic signals, including sexual selection, morphologicalconstraints and environmental variation. These factors can play simultaneous and interacting roles in determining signalphenotypes. Here, we assess the evolution of song features in the Maluridae, a passerine family with significant variationamong taxa in levels of sperm competition, morphological features and breeding habitats ranging from arid grasslands inAustralia to tropical rainforests in New Guinea. We used phylogenetic comparative methods and a robust molecularphylogeny to compare song characteristics with a variety of other measures, including testes mass, body-size and latitude.Several aspects of the temporal and frequency structure of song were associated with relative testes mass, suggesting thatsexual selection may influence some song characteristics in this family. The lowest frequencies of song were stronglypredicted by body-size, indicating that morphological constraints have also likely influenced acoustic phenotypes. Songversatility, reflecting the diversity of note types in a song, was positively correlatedwith latitude, suggesting that complexitymay increase in association with more temperate or variable environments. Variation in song structure across the familyappears to reflect a complex interaction between natural and sexual selection.

Additional keywords: body-size, latitude, relative testes size.

Received 7 September 2012, accepted 10 April 2013, published online 15 August 2013

Introduction

Acoustic signals are subject to a wide variety of selective pres-sures. In birds, many factors have been shown to influence thestructure and rate of evolutionary change of vocalisations, in-cluding sexual selection (Irwin 2000; Badyaev et al. 2002; Priceand Lanyon 2004), body-size (Wallschläger 1980; Ryan andBrenowitz 1985; Price et al. 2006), habitat (Wiley 1991; Slab-bekoorn and Smith 2002; Patten et al. 2004), breeding latitude(Weir andWheatcroft 2011) and seasonal variability (Botero et al.2009; Medina and Francis 2012). These patterns may be evidentboth within and between species, and they illustrate the widevariety of selective pressures that may influence acousticphenotypes.

Given the range of selection pressures that can influencesongs, making specific predictions concerning the evolution ofacoustic signals is difficult. For example, sexual selection mayfavour the evolution of more complex or, alternatively, moresimple vocalisations, depending upon what features are mostattractive to females ormost successful inmale–male competitiveinteractions (Searcy andAndersson1986;Cardoso andHu2011).Body-sizemayconstrain the production of low frequencies (Ryanand Brenowitz 1985; Price et al. 2006), so evolutionary changesin body-size for social or ecological reasons may allow corre-

sponding changes in the frequency composition of song. Alter-natively, selection for low frequenciesmaycause a correspondingchange in morphology (Fitch 1999). Acoustic adaptation toclosedhabitatsmay favour songswith slow ratesofnote repetitionand lower sound frequencies, which transmit more effectivelythrough dense vegetation than rapidly repeated or higher-fre-quency notes (discussed in Handford and Lougheed 1991;Wiley1991; Patten et al. 2004; Seddon 2005). Finally, evidence sug-gests that birds living at higher latitudes may have songs that arelonger, more complex or more variable, presumably due to therelatively variable environments at these latitudes (Botero et al.2009; Weir and Wheatcroft 2011; Medina and Francis 2012).Ultimately, all of these selective pressuresmay interact and lead todifferent signal phenotypes in different taxa.

In examining potential factors influencing song evolution, it iscritical to consider such interactionswithin a phylogenetic frame-work, because behavioural traits such as song may show signif-icant phylogenetic dependence (Price and Lanyon 2002; Packertet al. 2003) and therefore trait values from related taxamay not bestatistically independent (Felsenstein 1985; Freckleton et al.2002). Accounting for phylogenetic relationships permits a sta-tistically appropriate evaluation of potential evolutionary corre-lates, aswell as an assessment of historical changes in traits. Thus,

SPECIAL ISSUE

CSIRO PUBLISHING

Emu, 2013, 113, 270–281http://dx.doi.org/10.1071/MU12078

Journal compilation � BirdLife Australia 2013 www.publish.csiro.au/journals/emu

a study group in which there is a well resolved phylogeny isessential for such an analysis, and such a phylogeny is availablefor the oscine familyMaluridae.Additionally, songs inMaluridaerange from fairly simple note repetitions to extremely complexassortments of rapidly repeated notes (Rowley andRussell 1997).Songs are usedduring territorial disputes andduringdawn-chorusdisplays (Rowley andRussell 1997;Dalziell andCockburn 2008;Greig and Pruett-Jones 2008, 2009, 2010; Hall and Peters 2008;Dowling andWebster 2013), and song playbacks are reported toelicit territorial responses in most studied species (Rowley andRussell 1997). Thus, song is likely important in mate attractionand territory defence in Maluridae, and other social functions ofsong such asmate guarding or within-group communicationmaybe important as well (Payne et al. 1988, 1991; Hall and Peters2008; Dowling and Webster 2013).

Here, we investigate associations of acoustic structure withlevels of sperm competition, body-size and latitude in a compar-ative framework, using the Maluridae. This family is ideal forsuch an investigation for at least four reasons. First, speciesexhibit extreme variation in potential sexual selection pressures;the occurrence of extra-pair paternity ranges from as low as 6%ofbroods in some species to as high as 95% (Mulder et al. 1994;Kingma et al. 2009; Rowe and Pruett-Jones 2013). Althoughextra-pair paternity rates have not been measured for all species,relative testes mass (testes mass controlling for variation in body-weight) is known for most species and can be used as a proxy forlevels of sperm competition (Rowe andPruett-Jones 2006, 2011).Second, the species vary in body-size, from small fairy-wrens andemu-wrens (e.g.Stipiturusmalachurus: 7.3 g) to largegrasswrens(e.g. Amytornis striatus: 19.2 g) (Rowley and Russell 1997).Third, for a reasonably small family, species inMaluridae occupya remarkable variety of habitats, from primary rainforest in NewGuinea to the driest desert grasslands in Australia. And, lastly, allof the species in Maluridae are non-migratory, permanent resi-dents of the habitats and territories on which they live, so vocalcommunication is presumably little affected by seasonal changesin habitat or by species interactions during migration. Overall,

therefore, this family allows us to examine the influences of, andcomplex interactions between, natural and sexual selection in theevolution of bird song.

Materials and methodsStudy taxaThe Maluridae consists of five genera, the grasswrens (Amytor-nis) and emu-wrens (Stipiturus) that inhabit arid habitats through-out much of Australia, and the fairy-wrens (Malurus,Clytomyiasand Sipodotus), which inhabit a variety of tropical and temperatehabitats throughout Australia and New Guinea (Rowley andRussell 1997). Malurids are small, primarily insectivorous pas-serines and are part of the large Australasian radiation of theMeliphagoidea (Gardner et al. 2010).All species of theMaluridaeare known, or believed, to exhibit cooperative breeding, althoughthe extent of cooperation varies across species and across popula-tions within a given species (Rowley and Russell 1997).

We obtained song recordings and collected published dataon body-weight (g), habitat and mean latitude for 16 of the 21malurid taxa that were included in a published molecular phy-logeny for the Maluridae (Table 1; phylogeny based on Driskellet al. 2011). We also obtained data on combined testes mass (g)(Rowe and Pruett-Jones 2011) for 14 of these 16 taxa (all exceptMalurus cyanocephalus and Sipodotus wallacii). We wereunable to obtain measurements for four of the taxa included byDriskell et al. (2011): Malurus splendens musgravi, M. grayi,M. campbelli and Clytomias insignis. The subspecies identity ofsong recordings for M. alboscapulatus was uncertain, so wecollapsed the two subspecies present in the analysis of Driskellet al. (2011) into one taxon representing the M. alboscapulatussubspecies complex. Note that there are now three publishedphylogenies for theMaluridae (Gardner et al. 2010;Driskell et al.2011; Lee et al. 2012). These three phylogenies generally agreein the overall arrangement of species and their topologyexcept that in the phylogeny of Lee et al. (2012) M. amabilis isplaced outside of the M. lamberti clade. Also, the position of

Table 1. Habitat, mean latitude of breeding range, body-weight and combined testes mass for all taxa included in phylogenetic analysesTaxa: M., Malurus; A., Amytornis; S., Stipiturus. Habitats are taken from Rowley and Russell (1997)

Taxon Habitat Meanlatitude (�S)

Body-weight (g)

Combined testesmass (g)

Body and testes mass source

M. alboscapulatus Grass and shrubland 6.0 11.2 0.25 Rowe and Pruett-Jones (2006)M. amabilis Closed forest 15.0 8.38 0.113 Rowe and Pruett-Jones (2013)M. coronatus coronatus Riparian and open forest 16.5 11.4 0.05 Rowley and Russell (1997) (body),

Rowe and Pruett-Jones (2006) (testes)M. cyaneus cyaneus Open forest 42.5 9.4 0.48 Rowe and Pruett-Jones (2006)M. cyaneus cyanochlamys Open forest 30.5 8.93 0.29 Rowe and Pruett-Jones (2013)M. cyanocephalus Open forest 6.5 13.4 Rowley and Russell (1997)M. elegans Closed forest 33.0 9.8 0.203 Rowe and Pruett-Jones (2013)M. lamberti assimilis Mallee 26.5 8.29 0.18 Rowe and Pruett-Jones (2013)M. lamberti lamberti Open forest 29.0 8.2 0.25 Rowe and Pruett-Jones (2006)M. leucopterus leuconotus Grass and shrubland 27.0 7.7 0.326 Rowe and Pruett-Jones (2013)M. melanocephalus melanocephalus Open forest 22.0 7.57 0.213 Rowe and Pruett-Jones (2013)M. pulcherrimus Open forest 31.0 9.27 0.13 Rowe and Pruett-Jones (2013)M. splendens melanotus Mallee 30.5 9.45 0.298 Rowe and Pruett-Jones (2013)A. striatus striatus Grass and shrubland 33.0 19.23 0.34 Rowe and Pruett-Jones (2011)S. malachurus Grass and shrubland 32.5 7.32 0.045 Rowe and Pruett-Jones (2011)Sipodotus wallacii Closed forest 5.0 7.9 Rowley and Russell (1997)

Song evolution in Maluridae Emu 271

M. coronatus varies across all three phylogenies (see also Josephet al. 2013). We chose the phylogeny of Driskell et al. (2011) inthis study because that phylogeny includes the largest number ofspecies of Malurus, the genus for which we had the mostrecordings.We do not test our results with the other phylogenetichypotheses, but the similarity of the phylogenies overall led us tosuspect that our results would be similar.

Measurements of testes mass were obtained from studiesby Rowe and Pruett-Jones (2006, 2011, 2013), and body-weightswere obtained from Rowley and Russell (1997). We calculatedthe mean breeding latitude of each taxon as the approximate mid-point between the highest and lowest latitudes of distributionmaps in Rowley and Russell (1997). We classified habitat intofive major types based on descriptions provided in Rowley andRussell (1997) (Table 1).

Song recordings

During the breeding seasons (October–December) of 2006–2010 we recorded songs during male dawn chorus displaysfrom Lovely Fairy-wrens (M. amabilis), Red-winged Fairy-wrens (M. elegans), Variegated Fairy-wrens (M. assimilis),White-winged Fairy-wrens (M. leucopterus), Red-backedFairy-wrens (M. melanocephalus), Blue-breasted Fairy-wrens(M. pulcherrimus) and Splendid Fairy-wrens (M. splendens)(details of recording locations in Supplementary material, seehttp://www.publish.csiro.au/?act=view_file&file_id=MU12078_AC.pdf). For these recordings we used a Marantz PMD 661 solid-state digital recorder at 96-kHz sampling rate, 24-bit depth, or aMarantz PMD 670 at 48-kHz sampling rate, 16-bit depth (D&MProfessional, Itasca, IL), combined with ME66 shotgun micro-phone capsules and K6 power modules (Sennheiser ElectronicCorporation, Old Lyme, CT; frequency response 0.04–20.0 kHz).Weobtainedadditional song recordings from theMacaulayLibraryofNatural Sounds (CornellUniversity, Ithaca,NY), the xeno-cantoonline database (www.xeno-canto.org, accessed 6 March 2013), acommercially available audio CD (Plowright 2007) and the per-sonal recording collection of Dr Michelle Hall (University ofMelbourne) (details of recordings in Supplementary material). Allrecordings were uncompressed WAV files except for the threerecordings from xeno-canto, which were compressed MP3 files.We included the latter three recordings in this analysis becausethose taxawere so poorly sampled; however, we acknowledge thatthe recording quality of those sampleswas lower than the quality ofthe other recordings in our database.

In many species of Maluridae both sexes are known to sing,especially members of the genus Malurus, in which at least onespecies (M. coronatus) performs coordinated male–female duets(Rowley and Russell 1997). Song in females is thought tofunction in territory defence (Cooney and Cockburn 1995; Halland Peters 2008). Singing is thought to be primarily by males inemu-wrens and grasswrens, although this has been little studied(Rowley and Russell 1997). We were confident that the sampledindividuals were males in all recordings collected by the seniorauthor (E. I. Greig), byMichelleHall and inmost recordings fromother sources; however, the sexwasuncertain for some recordings(detailed in Supplementary material). In some species (e.g.M. splendens and M. cyaneus), males produce simpler trilledsongs (Type II songs) in addition to their standard display

vocalisations (Type I songs) (Langmore andMulder 1992,Zelanoet al. 2001). We avoided including Type II songs in our compar-isons (see below).

We obtained recordings from five additional taxa that werenot included in the Driskell et al. (2011) phylogeny or for whichwe did not havemorphological data (M. coronatus macgillivrayi,M. splendens splendens, M. s. emmottorum, A. merrotsyi andStipiturus mallee: included in Table 2). We included these songsin analyses of note types (described below) and in assessingcorrelations between acoustic measurements, but we did not usethese taxa in any phylogenetic analyses.

Acoustic analysis

We chose at least one example song from each male that was ofhigh recording quality based on spectrograms digitised in RavenPro 1.4 (Bioacoustics Research Program 2004; 16-bit sampleformat; discrete Fourier transform; DFT= 512 samples; frequen-cy resolution = 124Hz; time resolution = 11.6ms; frame over-lap = 50%). We made an effort to use only one song from anyindividual; however, for eight poorly sampled taxa we usedmultiple songs from the same individualwhen calculating speciesmeans, for a total of 184 songs from 161 individuals (mean�s.e. = 8.6� 9.1 songs per taxon) (Table 2). ForM. splendens andM. cyaneus, we excluded Type II songs (trills) from our analysis.We did include display songs that contained trill components,because these were often the primary songs given by someindividuals during dawn chorus displays. For each song wemanually measured individual note characteristics in Luscinia(Lachlan 2007); we defined a note as a continuous tracing on aspectrogram, and we measured only the fundamental frequencyfor each note and avoidedmeasuring harmonics or sidebands.Wesummarised nine acoustic characteristics (listed in Table 2) foreach song. We chose these acoustic characteristics because theycapturedaspects ofvariation related to song frequency (peak, highand low frequency, song bandwidth and note bandwidth), tempo(song length, note length, number of notes and note rate) andcomplexity (note variety and song versatility). Our samplingmethod did not permit an estimate of song repertoire size, butsuch an analysis would be a valuable line of future inquiry.

We categorised all notes in the dataset (n= 9995 notes) intonote types using Ward’s agglomerative hierarchical clustering(Ward 1963) based on the peak, minimum and maximum notefrequency, note bandwidth, note duration and the change inbandwidth over time (bandwidth divided by duration). From thiscluster analysis we identified an optimal total number of clusters(i.e. note types) using a method in which the distances betweenadjoining clusters were plotted against the rank of those distances(described by Taft 2011). Clusters with a low rank have greaterdistances between them and their division explains a largeproportion of variance in the data, whereas clusters with a highrank have very short distances between them and represent itemsthat could be considered members of the same cluster. The resultof this plot is an L-shaped distribution of points, with the elbowofthis distribution representing a transition between numbers ofclusters that explain a large proportion of variance in the dataversus numbers of clusters that explain a small proportion ofvariance in the data (Taft 2011). To identify this elbow, wefollowed Taft (2011) and calculated linear fits for the distribution

272 Emu E. I. Greig et al.

Tab

le2.

Means

andstan

dard

deviations

ofallson

gva

riab

lesforalltax

aSam

plesizesof

songsandindividu

alsaregiven.

Asterisks

indicatetaxa

thatwereexclud

edfrom

phylogeneticanalyses,b

utwhich

wereused

forclassificatio

nof

notetypesandassessmento

fcorrelations

betweenacou

sticvariables.Taxa:M.,Malurus;A.,Amytornis;S.,S

tipitu

rus

Taxon

nsongs

(ind

ividuals)

Son

gleng

th(s)

Peak

frequency

(kHz)

High

frequency

(kHz)

Low

frequency

(kHz)

Son

gbandwidth

(kHz)

Note

bandwidth

(kHz)

Note

leng

th(m

s)

Num

ber

ofno

tes

Note

rate

(notes/s–1)

Note

variety

Song

versatility

M.a

lboscapulatus

2(2)

4.2±0.7

5.4±1.0

9.3±1.9

2.5±0.8

6.9±1.0

0.37

±0.09

40.6±5.9

59±21

.213

.9±2.8

10.0±1.4

0.18

±0.04

M.a

mab

ilis

9(9)

2.1±0.2

6.5±0.2

9.5±0.3

3.5±0.9

5.9±1.0

0.71

±0.12

46.4±11

.131

±6.0

14.6±2.8

7.7±1.7

0.25

±0.06

M.coron

atus

corona

tus

9(9)

4.6±1.9

4.7±0.5

8.3±0.7

1.4±0.2

6.9±0.7

1.27

±0.25

61.7±8.0

43±16

.89.4±1.2

6.1±2.0

0.15

±0.05

M.coronatus

macgillivrayi

3(2)*

3.9±1.0

4.8±0.2

8.0±0.5

1.3±0.2

6.7±0.3

1.12

±0.12

59.6±19

.737

±9.6

9.5±3.2

5.5±2.4

0.16

±0.06

M.cyaneus

cyan

eus

2(1)

3.5±0.2

5.6±0.7

9.7±1.1

2.3±0.2

7.4±1.3

0.96

±0.47

68.3±15

.435

±0.7

9.8±0.8

11.5±2.1

0.33

±0.05

M.cyaneus

cyan

ochlam

ys8(6)

3.1±0.9

5.8±0.5

9.6±0.7

2.9±0.5

6.7±0.8

0.97

±0.34

44.3±6.9

45±17

.214

.2±1.7

11.1±2.6

0.27

±0.08

M.cyano

cephalus

5(1)

4.7±1.4

2.4±0.0

3.8±0.3

1.3±0.3

2.5±0.4

0.38

±0.03

106.5±8.5

25±7.2

5.4±0.4

2.0±0.0

0.09

±0.03

M.elega

ns10

(10)

3.3±0.4

5.3±0.3

8.8±0.6

2.3±0.2

6.5±0.6

0.76

±0.15

62.1±15

.829

±7.6

8.8±2.3

9.3±1.6

0.33

±0.07

M.lam

bertiassimilis

23(23)

2.9±0.4

5.5±0.4

8.9±0.8

2.8±0.5

6.1±0.8

1.21

±0.36

39.4±6.8

53±8.9

18.0±2.4

11.6±1.6

0.22

±0.05

M.lam

bertilamberti

1(1)

36.3

9.3

3.5

5.8

1.03

37.7

5518

.513

0.24

M.leucopterus

leuconotus

22(22)

4.8±0.8

4.4±0.2

7.9±1.0

1.9±0.2

6.1±1.0

0.66

±0.17

21.4±1.9

119±21

.924

.8±2.0

10.0±1.6

0.09

±0.02

M.m

elanocephalusmelanocephalus

39(39)

3.8±0.6

6.3±0.3

10.5±0.4

3.0±0.5

7.5±0.6

0.86

±0.15

37.7±6.3

56±10

.714

.7±2.1

14.4±2.4

0.26

±0.05

M.p

ulcherrimus

6(6)

2.9±0.3

5.8±0.3

9.2±0.4

2.6±0.5

6.6±0.5

0.80

±0.17

36.4±6.4

39±10

.613

.2±2.6

10.0±2.5

0.27

±0.08

M.splendens

splend

ens

4(4)*

3.3±0.8

5.6±0.3

7.9±0.7

2.3±0.6

5.5±1.0

0.89

±0.18

52.8±11

.842

±10

.912

.6±1.9

9.8±2.2

0.24

±0.05

M.splendens

emmottorun

5(5)*

4.0±0.9

4.8±0.4

8.1±0.7

2.1±0.2

6.0±0.7

0.48

±0.13

47.8±5.5

53±11

.413

.3±1.2

9.2±1.3

0.18

±0.02

M.splendens

melanotus

12(12)

3.0±0.4

4.9±0.4

8.0±1.0

2.4±0.5

5.5±0.7

0.67

±0.39

45.2±4.1

45±8.9

14.9±1.5

9.3±1.9

0.21

±0.04

A.striatusstriatus

5(1)

3.0±1.7

6.7±1.2

11.6±0.3

1.3±0.2

10.3±0.5

1.00

±0.34

65.6±7.3

20±10

.36.7±0.8

6.2±0.8

0.39

±0.24

A.m

errotsyi

9(1)*

2.7±1.2

7.5±0.3

11.7±0.4

2.3±0.4

9.4±0.4

1.34

±0.36

79.3±12

.518

±6.9

6.8±1.0

5.9±1.6

0.36

±0.13

S.malachu

rus

3(1)

3.0±0.4

8.0±0.1

11.2±0.0

4.7±0.3

6.4±0.4

0.64

±0.07

58.7±2.0

35±2.5

11.8±0.8

8.0±1.0

0.23

±0.02

S.mallee

3(2)*

4.0±1.1

7.2±0.3

10.5±1.3

2.5±0.2

8.0±1.5

1.13

±0.37

89.4±9.8

26±5.9

6.5±0.5

7.3±0.6

0.30

±0.10

Sipodotuswallacii

2(2)

1.6±0.6

8.4±0.0

10.1±0.9

5.2±0.4

4.9±0.5

0.62

±0.03

71.8±11

.416

±8.5

9.8±1.9

3.0±0.0

0.22

±0.12

Song evolution in Maluridae Emu 273

of points to the left and to the right of all possibly pivot points; thepivot point that represents the best ‘hinge’ (and therefore anoptimal number of groups in the dataset) is the point with thelowest sum of the squared residuals from the regressions oneither side (Taft 2011). This classification method yielded 22note types as the optimal number for the dataset.

With notes classified into types, we then calculated the totalnumber of note types in each song (note variety), as well as thenumber of note types divided by the total number of notes in asong (song versatility). Both note variety and song versatilitywere qualitatively similar across species when we repeated thesecalculations using different cut-off points for the total number ofnote types in the dataset (20, 30, 40, 50, 100 and 150 note types).Thus, relative song complexity across Malurus using our twomeasurements (note variety and song versatility) was robust todifferent note type classification schemes.

Statistical analysisUsing the acoustic data collected above,we calculatedmean songvalues for each species (Table 2). We included testes mass andbody-weight as log-transformed independent variables andmeanlatitude as an untransformed variable. We assessed correlationsacross species using a generalised least-squares method in R(R Development Core Team 2012) with the package APE (Para-dis et al. 2004), which corrected for phylogenetic non-indepen-dence of traits (Pagel 1999; Freckleton et al. 2002). FollowingRowe and Pruett-Jones (2011), we estimated the magnitude ofthis phylogenetic dependence with the scaling parameter lambda(l); values of l close to 1 indicate greater phylogenetic depen-dence of a trait and values close to 0 indicate greater phylogeneticindependence (Pagel 1999, Freckleton et al. 2002). We ran twoiterations of this analysis, one with all 16 taxa for which we hadbody-weight and latitude data (using those two factors as pre-dictor variables), and one with the subset of 14 taxa for which wehad testes mass, body-weight and latitude data (with all threefactors as predictor variables).

To complement our estimations of l, we also calculated thedescriptive statistic K for all song traits (n = 16 taxa) to comparethe relative amount of convergence or divergence among traits(Blomberg et al. 2003). Values of K less than 1 indicate thatclosely related taxa are more different than would be expected bychance under a Brownian motion model of evolution, and valuesofK greater than 1 indicate that related taxa are more similar thanwould be expected under a Brownian motion model (Blomberget al. 2003).We used the software package Picante (Kembel et al.2010) implemented in R to calculate K for all traits, and we usedthe randomisation procedure implemented with the function‘phylosignal’ (using 10 000 repetitions) to determine whethertraits exhibited significantly more similarity between relatedindividuals than would be expected by chance.

We reconstructed evolutionary changes in song on the phy-logeny using MacClade 4.08 (Maddison and Maddison 2005).For this analysis, we converted our measurements of continuoussong features into discrete, ordered characters by plotting meansand standard errors for taxa and then dividing thesemeasures intostates where error bars did not overlap (see Price and Lanyon2002, 2004 for more detailed explanations of this technique).Only taxa for which a minimum of three song samples were

available (12 of the 16 taxa included in the phylogeneticanalysis: Table 2) were used in the defining of character statesto ensure a minimal degree of accuracy in the placement of errorbars, and divisions between states were positioned equidistantfrom nearest error bars. We then scored all taxa based on meanmeasurements of song features, including those with fewer thanthree representative songs, and mapped these discrete characterstates onto the tree using simple parsimony. This methodallowed us to identify statistically discontinuous evolutionarychanges in song patterns while controlling for within-taxonvariability. All mean measurements based on fewer than threesongs fell within the character states calculated using other taxa,so these measurements would not have generated additionalcharacter states.

Aswill be evident below, the results of our analyses examiningfactors related to song variation varied depending on whether weincluded two variables (body-weight and latitude) across 16 taxaor three variables (body-weight, testes mass and latitude) across14 taxa, as well as whether we consider the results of theregression analysis directly or adjusted the table-wide signifi-cance values with a Bonferroni correction. In the results sectionbelow,we present the results based on the regression analysis, butnote when these results are questioned due to the Bonferronicorrection.

Our habitat classifications for each taxon were very generaland do not include variation in microhabitats or intraspecificvariation in habitat use. Thus, we did not conduct a thoroughanalysis of habitat in relation to acoustic characteristics. How-ever, as a preliminary assessment of song variation in relation tohabitat type, we used one-way ANOVAs to compare mean songtrait values across habitat types.

Results

Song variables differed considerably among the 21 taxa forwhich we had recordings (ANOVA: n= 184 songs, P< 0.001for all 11 variables in Table 2). Naturally, several song variableswere correlated with one another (Table 3). For example, thenumber of notes, mean note rate and mean note length in a songwere highly correlated with each other. Thus, songs tended to becomposed of either many short, quickly repeated notes or fewerlong, widely spaced notes. Song bandwidth was strongly andpositively correlated with the highest frequencies but not thelowest frequencies of songs, suggesting that variation in songbandwidth among species was largely explained by the presenceor absence of high-frequency notes.

Most song traits were significantly more similar betweenclosely related taxa than would be expected by chance (i.e. moresimilar than if trait values were distributed randomly on the tree),but all had values of K that were less than 1 (Table 4), indicatingthat closely related taxa tended to be more divergent in thesecharacteristics than would be expected under a Brownian motionmodel of evolution. Our maximum-likelihood estimates of l forall song traits except note bandwidth were significantly differentfrom 1 but not significantly different from 0 (Table 5), indicatingsome degree of phylogenetic independence of these traits. Itshould be noted, however, that estimates of K and l were notalways consistent; in some cases traits exhibited significantphylogenetic signal based on analysis with ‘phylosignal’, but

274 Emu E. I. Greig et al.

not in our estimates of l in multiple regression models(Tables 4, 5, 6).

In both regression models, body-weight was strongly andnegatively correlated with lowest song frequency (P= 0.002 forthe model including 16 taxa, P = 0.026 for the model including14 taxa: Tables 5, 6, Fig. 1 top), indicating that in larger species,males tended to sing lower-frequency songs. This relationshipremained significant in the analysis with 16 taxa after sequentialBonferroni correction for 22 comparisons (Holm1979), but not inthe analysis with 14 taxa and 33 comparisons (Tables 5, 6). Body-weightwasnot a significant predictor of anyother songvariable inthe analysis with 16 taxa, and although body-weight showedsignificant relationshipswith some song traits in the analysis with14 taxa, these relationships were not significant after sequentialBonferroni correction. Overall, therefore, body-weight was as-sociated with the low frequency of songs but not stronglyassociated with other song variables.

Mean latitude was a significant positive predictor of songversatility in both models (P = 0.004 for the model including16 taxa, P = 0.003 for the model including 14 taxa: Tables 5,6, Fig. 1 bottom). Thus, more complex songs tended to occur athigher latitudes (i.e. farther south), although these relationshipsdid not remain significant after sequential Bonferroni correction

for multiple comparisons. Mean latitude was also significantlyassociated with note bandwidth; however, this was only in themodel with 16 taxa (P = 0.001 for the model including 16 taxa,which remained significant after sequential Bonferroni correc-tion,P= 0.141 for themodel including14 taxa:Tables 5, 6).Meanlatitude was associated with song length, note length and numberof notes (i.e. tempo-related song traits) in the model including14 taxa, but none of these relationships remained significant aftersequential Bonferroni correction.

Testes mass was a significant predictor of song structure forseveral song traits, including the length, number, rate and varietyof notes in a song (Table 6). Taxa with larger testes thereforetended to produce songswith shorter,more rapidly repeated notesthat included more note types. Additionally, these song traits hadvalues ofK < 1 and did not exhibit significant similarities betweenrelated individuals (i.e. traits appeared to be distributed randomlyon the tree relative to phylogenetic relationships). This departurefrom a Brownian model of evolution could potentially indicatethe presence of selection on these song traits. Larger testes masswas also significantly correlated with lower peak frequencies(P= 0.016), and a trend existed between testes mass and thehighest frequency (P= 0.058) and lowest frequency (P= 0.059)components of songs. All relationships with testes mass were,however, not significant after sequential Bonferroni correction.

Historical reconstructions of song evolution based on unam-biguous discrete changes complement these analyses (Fig. 2) andindicate the importance of latitude in song evolution. Althoughchanges in song structure have occurred throughout theevolution of the clade (Fig. 2), most changes appear to beassociated with historical movements between tropical and tem-perate environments. For example, the songs ofM. amabilis andM. coronatus (both tropical species) differ in a variety of waysfrom songs of closely related taxa with ranges farther south(M. lamberti, M. pulcherrimus and M. elegans). Similarly,M. alboscapulatus (tropical) and M. melanocephalus (subtrop-ical) differ from their sister taxon M. leucopterus (a temperatespecies), andM. cyanocephalus (distributed in New Guinea) hassongs that are strikingly different from any Australian species(see spectrograms in Fig. 2). Songs from S. wallacii, anotherspecies distributed in New Guinea, appeared visibly differentfrom other taxa in spectrograms (Fig. 2), but our quantitativeanalyses did not reveal any large historical changes in this lineageexplaining these differences. Song versatility (SV), a trait shown

Table 3. Correlation coefficients (r) between all song variables across all taxa (n= 184 songs)Strong correlations (r > 0.500) are highlighted in bold

Songlength

Peakfrequency

Highfrequency

Lowfrequency

Songbandwidth

Notebandwidth

Notelength

Numberof notes

Noterate

Notevariety

Peak frequency –0.351 –

High frequency –0.307 0.938 –

Low frequency –0.291 0.909 0.745 –

Song bandwidth –0.007 0.575 0.706 0.405 –

Note bandwidth –0.148 0.302 0.433 0.037 0.381 –

Note length –0.165 0.064 0.223 –0.120 0.041 0.173 –

Number of notes 0.623 –0.364 –0.446 –0.188 –0.119 –0.192 –0.708 –

Note rate 0.191 –0.278 –0.432 –0.074 –0.219 –0.153 –0.852 0.853 –

Note variety 0.174 0.188 0.075 0.271 0.246 0.069 –0.586 0.376 0.436 –

Song versatility –0.561 0.544 0.589 0.420 0.452 0.330 0.335 –0.615 –0.497 0.179

Table 4. Values of Blomberg’s K for all song variables and results ofrandomisation test assessing whether traits were more similar in relatedindividuals than would be expected by chance (i.e. if traits were

distributed randomly on the tree)SignificantP-values (P< 0.05) are highlighted inbold.ValuesofK<1 indicatethat relatives are less similar thanwould be expected under aBrownianmotion

model of evolution

Song variable K Z P

Song length 0.75 –1.57 0.012Peak frequency 0.73 –1.49 0.007High frequency 0.86 –1.24 0.024Low frequency 0.64 –1.58 0.008Song bandwidth 0.84 –1.41 0.005Note bandwidth 0.52 –0.54 0.339Note length 0.61 –1.02 0.104Number of notes 0.74 –0.82 0.219Note rate 0.58 –0.99 0.130Note variety 0.64 –1.33 0.045Song versatility 0.68 –1.42 0.026

Song evolution in Maluridae Emu 275

to correlate significantly with latitude, decreased in at least threeseparate lineages (M. alboscapulatus, M. coronatus andM. cyanocephalus), and note bandwidth (NB) also decreased intwo of these taxa. Each of these changes clearly correspondedwith movements from temperate to tropical breeding latitudes.Despite the relationships between song traits and latitude, wefound little evidence for an association of song traits and habitat.No song trait showed significant variation among habitats(P > 0.10 for all 11 traits).

Song changes have also been associated with changes inmorphology. For example, lowest frequencies (LF) have de-creased in the songs of M. coronatus and M. cyanocephalus(Fig. 2), and both of these species have large body-sizes incomparison to congeners (Table 1). Changes in song charactersalso complement correlations between song traits described inTable 3; for example, all four changes in note length (NL) wereassociated with changes in note rate (NR), and these variablesalways changed in opposite directions, with increases in notelengths occurring with decreases in note rates and vice versa.

Discussion

Signal evolution in any family of birds is obviously a complexinteraction of selection pressures and responses to environmentalvariability, and interspecific variation in the songs of the Mal-uridae illustrates this complex interaction. Species in the Mal-uridae experience variation along several gradients that shouldrelate to signal evolution. First, with respect to geographicaldistribution, malurid species occur from near the equator in NewGuinea south to a latitude of ~42�S. Second, with respect to

ecology, species occur in diverse habitats ranging from primaryrainforests to open and desolate deserts in Australia. Third, withrespect tomating behaviour, although all of the species are knownto be socially monogamous, the extent of reproductive promis-cuity in this one family varies as much as it does across all otherbirds of the world combined, from almost non-existent to morethan 90% of broods. Natural selection due to environmentalfactors (latitude and habitat) should be expected to influenceinterspecific variation in song structure (e.g. Wiley 1991; Boteroet al. 2009; Weir and Wheatcroft 2011); however, the effects ofextra-pair paternity on song evolution are less well understood(Garamszegi and Moller 2004; Price and Lanyon 2004; CardosoandHu 2011; Soma andGaramszegi 2011). Nevertheless, if songcharacteristics are important in mate choice and reflect rates ofextra-pair paternity within species, interspecific variation inpromiscuity might also be reflected in song variation acrossspecies.

We found that fairy-wrens, grasswrens and emu-wrens variedsignificantly in the acoustic structure of their songs and exhibiteda high degree of divergence in most song traits. Estimates ofK aswell as reconstructions of evolutionary changes on the phylogenysuggest thatmany traits have departed fromBrownianmotion andmay have been influenced by selection (Blomberg et al. 2003).This was complemented by our estimates of l, which for mostacoustic traits were not significantly different from 0 but weresignificantly different from 1, indicating phylogenetic indepen-dence for most traits.

Our result that low frequencies were influenced by body-size(Fig. 1 top, Tables 5, 6) complements other work comparingbody-size and sound frequencies in bird song, which suggest that

Table 5. Correlations of song variables with body-weight and mean breeding latitudes in multiple regression models (n= 16 taxa)Significant P-values (P< 0.05) are highlighted in bold. Asterisks indicate relationships that remained significant after sequential Bonferroni

correction for 22 comparisons. Maximum-likelihood estimates of l are shown and are compared to models in which l is set at 0 or 1

Song variable R2 F l P (l= 0) P (l= 1) Predictor Slope t P

Song length 0.38 3.92 0.00 1.000 0.001 Body 1.32 1.30 0.215Latitude –0.02 –1.61 0.132

Peak frequency 0.18 1.43 0.00 1.000 0.001 Body –2.50 –1.57 0.141Latitude 0.00 –0.15 0.882

High frequency 0.02 0.16 0.00 1.000 0.005 Body –1.09 –0.50 0.626Latitude 0.00 0.00 0.998

Low frequency 0.52 7.17 0.00 1.000 0.000 Body –3.40 –3.78 0.002*Latitude –0.02 –1.50 0.157

Song bandwidth 0.11 0.84 0.00 1.000 0.021 Body 2.31 1.28 0.223Latitude 0.02 0.75 0.466

Note bandwidth 0.64 7.15 0.98 0.014 0.859 Body –0.04 –0.20 0.845Latitude 0.01 4.55 0.001*

Note length 0.22 1.87 0.51 0.177 0.010 Body 40.73 1.74 0.106Latitude –0.11 –0.31 0.765

Number of notes 0.13 0.98 0.00 1.000 0.000 Body –25.80 –0.91 0.381Latitude –0.56 –1.36 0.197

Note rate 0.26 2.34 0.00 1.000 0.000 Body –10.55 –2.15 0.051Latitude –0.06 –0.79 0.444

Note variety 0.16 1.22 0.22 0.824 0.037 Body –3.63 –0.95 0.359Latitude 0.05 0.81 0.434

Song versatility 0.49 6.15 0.00 1.000 0.000 Body 0.08 1.02 0.324Latitude 0.00 3.45 0.004

276 Emu E. I. Greig et al.

body-size constrains the lower range of frequencies that a bird canproduce (Wallschläger 1980; Ryan and Brenowitz 1985; Priceet al. 2006; Martin et al. 2011). The significant negative rela-tionship between note length and note rate (Table 3) suggestsanother constraint on song production, that birds with longernotes are constrained in their abilities to produce these notesrapidly.Other frequencymeasureswere not correlatedwith body-size, but an apparent negative trend between the peak, highest andlowest frequencies of song and testes size suggest that these signalcharacteristics may have been influenced by sexual selection.Specieswith larger relative testesmass tend toproduce songswithgenerally lower frequencies; however, the reasons behind thisrelationship are unclear. The role that such frequency character-istics may play in male–male competition needs to beinvestigated.

Perhaps the most surprising relationship in our study was thecorrelation between song versatility and latitude (Fig. 1 bottom,Tables 5, 6). Birds at higher latitudes (i.e. with breeding rangesfarther south) tended to sing songs with a greater diversity of

note types, as well as songs with larger note bandwidths. Thiscomplements work in other avian systems that suggests that birdsat higher latitudes sing more complex or more variable songs(Irwin 2000; Botero et al. 2009; Weir and Wheatcroft 2011;Medina and Francis 2012). Possible reasons for this increasedacoustic complexity with increasing latitude include greateravailability of ‘acoustic space’ if species density is lower athigher latitudes (Podos et al. 2004), faster rates of song evolution(Weir and Wheatcroft 2011), higher sexual selection pressures(Irwin 2000) or greater environmental uncertainty (Medina andFrancis 2012). Although we did not find obvious influences ofhabitat on song structure, a more refined analysis of habitat andsongwould be valuable and could reveal additional relationships.Individually, or in combination, these factors may contribute togeographical variation in song complexity within the Maluridae.

Relationships between song and levels of sperm competition,measured as relative testes mass, suggest that taxa with largertestes tend to have songswith shorter,more rapidly repeated notes(Table 6). This is illustrated in Fig. 2 by species such as

Table 6. Correlations of song variables with body-weight, mean breeding latitudes and combined testes mass in multiple regressionmodels (n= 14 taxa)

Significant P-values (P< 0.05) are highlighted in bold. No relationships remained significant after sequential Bonferroni correction for33 comparisons. Maximum-likelihood estimates of l are shown and are compared to models in which l is set at 0 or 1

Song variable R2 F l P (l= 0) P (l= 1) Predictor Slope t P

Song length 0.45 2.74 1.00 0.242 1.000 Body –0.76 –1.25 0.240Testes 0.33 1.22 0.250Latitude –0.04 –2.24 0.049

Peak frequency 0.60 0.02 0.00 1.000 0.016 Body 0.33 0.28 0.788Testes –1.05 –2.90 0.016Latitude 0.02 1.12 0.290

High frequency 0.35 1.80 0.00 1.000 0.020 Body 1.82 1.25 0.240Testes –0.94 –2.14 0.058Latitude 0.02 0.70 0.497

Low frequency 0.71 8.18 0.71 0.302 0.145 Body –1.77 –2.61 0.026Testes –0.54 –2.13 0.059Latitude 0.01 0.87 0.403

Song bandwidth 0.61 5.29 0.00 1.000 0.006 Body 3.88 3.83 0.003Testes –0.27 –0.88 0.398Latitude 0.02 1.37 0.200

Note bandwidth 0.27 1.23 0.93 0.011 0.465 Body 0.17 0.85 0.415Testes –0.06 –0.65 0.528Latitude 0.01 1.60 0.141

Note length 0.69 7.47 0.00 1.000 0.026 Body 35.20 3.23 0.009Testes –10.65 –3.24 0.009Latitude 0.51 3.14 0.011

Number of notes 0.53 3.74 0.00 1.000 0.047 Body –52.08 –2.23 0.050Testes 15.82 2.24 0.049Latitude –0.79 –2.27 0.046

Note rate 0.64 5.81 0.00 1.000 0.041 Body –14.39 –3.90 0.003Testes 3.33 2.99 0.013Latitude –0.09 –1.65 0.130

Note variety 0.66 6.33 0.00 1.000 0.005 Body –6.71 –3.51 0.006Testes 2.21 3.82 0.003Latitude –0.03 –1.22 0.250

Song versatility 0.60 5.09 0.00 1.000 0.037 Body 0.15 2.05 0.067Testes –0.01 –0.30 0.774Latitude 0.00 3.85 0.003

Song evolution in Maluridae Emu 277

M. melanocephalus, M. leucopterus and M. splendens, whichhave large testes mass (Table 1) and reasonably note-rich songs.Note length, number of notes and note rate haveK values that are<1 and the traits donot exhibit significant phylogenetic signal (i.e.related taxa are not more similar than if traits were distributedrandomly on the tree), suggesting the possible presence ofselection on these traits. Additionally, maximum-likelihood esti-mates of l for these traits indicated phylogenetic independence(Tables 5, 6). Although note rate, number and length are notnecessarily aspects of song performance per se (e.g. Podos 1997;Ballentine et al. 2004;Cardoso et al. 2009), acoustic features suchas thesemaybeassociatedwithvarying levels of songdifficulty orstimulatory potential (e.g. Vallet et al. 1998) and are reasonablecandidates for sexually selected song traits in the Maluridae.

The lack of a clear relationship between song versatility andrelative testes mass suggests that sexual selection plays little rolein this aspect of song. In one sense, this is surprising becausevocalisations and vocal complexity appear to be so critical to theintegrity of social groups of fairy-wrens and, by extension, to theprocess of mate choice and male–male competition. Neverthe-less, the importance of song versatility in social behaviour maynot necessarily translate to variation among taxa that havedifferent rates of extra-pair paternity, especially if this songcharacteristic has multiple functions. Other factors, such as notenumber or rate, might play more direct roles in reproductivesuccess, leading to interspecific differences. Furthermore, previ-ous phylogenetic studies have shown that, although sexualselection can cause dramatic evolutionary changes in song, no

single aspect of song is clearly associatedwith variation inmatingsystems (Price and Lanyon 2004). Complexity may insteadchange stochastically, particularly in learned bird songs wherecopying errors and improvisation occur (e.g. Payne 1996;Kroodsma et al. 1999; Podos et al. 1999; Johnson 2006). Thus,our findings suggest that there may be many solutions to theproblem of effectively advertising to potential mates and com-peting with rivals.

One important aspect of interspecific variation in songs infairy-wrens that we did not analyse here is the use of predator-elicited Type II songs by males. In Superb Fairy-wrens(M. cyaneus) (Langmore and Mulder 1992), Splendid Fairy-wrens (M. splendens melanotus) (Zelano et al. 2001, Greig andPruett-Jones 2009, 2010) and Variegated Fairy-wrens(M. lamberti assimilis) (Greig et al. 2010), males give Type IIsongs in response to hearing predators call. Greig and Pruett-Jones (2009, 2010) showed that in Splendid Fairy-wrens thisvocalisation is not an alarm call but rather a display vocalisationthat is given when conspecifics are particularly alert (i.e. in thepresence of predators). In a broad, interspecific comparisonGreigand Webster (unpublished data) have shown that males of manyother species ofMalurus inAustralia respond tohearingpredatorsby giving a vocalisation similar to those described above. Thespecies that occur in New Guinea have not been surveyed withrespect to this vocalisation, and thus it is not clear whether it is atrait common to many members of the subfamily Malurinae orwhether it is unique to the Australian taxa. Overall, predator-elicited trills may be an important sexual or social signal and athorough study of the evolution of this song type would bevaluable.

Although our focus here was on interspecific variation, it isbecoming increasingly clear that the study of song dynamicswithin species is yielding remarkable insights into communica-tion in fairy-wrens (Hall and Peters 2008; Cockburn et al. 2009,Greig et al. 2012, Dowling and Webster 2013, Kleindorfer et al.2013). Song in males, besides having an obvious territorialfunction, is implicated in aspects of mate choice (Cockburnet al. 2009; Greig 2010), although a specific correlation betweensong and mating success, either within-pair or extra-pair, has yetto be determined. Similarly, although female song has beenstudied in two species (Cooney and Cockburn 1995; Hall andPeters 2008; Kleindorfer et al. 2013), the fact that female song issuch a prominent feature of communication in fairy-wrens sug-gests that this would be an extremely fruitful area for futureresearch. Finally, intraspecific geographic variation in songs andassociated conspecific responses of Red-backed and SuperbFairy-wrens suggests that song may be an important isolatingmechanism between divergent populations (Greig and Webster2013; Kleindorfer et al. 2013).

In summary, songs in Maluridae are influenced by morpho-logical constraints and show some clear patterns of evolution inresponse to environmental factors. The influences of sexualselection are less clear, but such influences are undoubtedlycomplex given the variety of social functions of song in thissystem. Our findings complement previous comparative andmeta-analyses that suggest that patterns of directional songevolution in response to sexual selection are not straightforward(Garamszegi and Moller 2004; Price and Lanyon 2004; Cardosoand Hu 2011; Soma and Garamszegi 2011). The correlation we

0.0

1.0

2.0

3.0

4.0

5.0

6.0

0.8 1.0 1.2

log(bodymass)

Low

freq

uenc

y

0.0

0.1

0.2

0.3

0.4

0.0 20.0 40.0

Mean latitude

Son

g ve

rsat

ility

Fig. 1. Relationships between low frequency and the logarithm of body-weight (top) and song versatility and mean latitude (bottom) in malurid taxa.Note that these figures do not correct for correlations with other predictorvariables or phylogenetic relationships among taxa.

278 Emu E. I. Greig et al.

observed between aspects of song complexity and latitude com-plements work in other avian systems that suggests that higherlatitudes support more variable, versatile or complex songs

(e.g. Weir and Wheatcroft 2011), for example by a general‘acoustic release’ if higher latitudes have lower species abun-dance (Podos et al. 2004), or if acoustic versatility is advanta-

M. alboscapulatus

M. melanocephalus

M. leucopterus

M. splendens melanotus

M. cyaneus cyanochlamys

M. cyaneus cyaneus

M. coronatus

M. elegans

M. pulcherrimus

M. lamberti lamberti

M. lamberti assimilis

M. amabilis

M. cyanocephalus

Sipodotus wallacii

Stipiturus malachurus

Amytornis striatus

2s

Fig. 2. Unambiguous evolutionary changes in song characters reconstructed on the molecular phylogeny of the Maluridae(from Driskell et al. 2011). Spectrograms on the right show typical song patterns for each taxon. Arrows and characterabbreviations above branches show increases (up arrows) and decreases (down arrows) in acoustic characters, abbreviated as:song length (SL), peak frequency (PF), high frequency (HF), low frequency (LF), song bandwidth (SB), note bandwidth(NB), note length (NL), number of notes (NN), note rate (NR), note variety (NV), song versatility (SV).

Song evolution in Maluridae Emu 279

geous in climatic conditions that are more variable (Botero et al.2009). Finally, song frequency characteristics in Maluridae ap-pear to be constrained by body-size in a manner similar to otheravian taxa. Overall, therefore, there appear to be a variety ofboth natural and social selective pressures that have influencedsongs in Maluridae, and these factors interact in complex waysto create the acoustic phenotypes characteristic of this avianfamily.

Acknowledgements

For assistance in the field, logistical or financial support, and advice ordiscussion we thank Mike Webster, Melissah Rowe, Michelle Hall, BenTaft, Dan Baldassarre, Becky Cramer, Jenélle Dowling, Sara Kaiser, ShaileeShah, DuncanYandell, JoanneHeathcote, John andLisa at Dryandra, LyanneBrouwer, Marina and Simon, Tim Daniels, Matt Medler, Jan Lewis, DougBarron, Karen and Angus Emmott, Graham Chapman, Leslie and MichaelBrooker, Mornington Wildlife Sanctuary, Macaulay Library of NaturalSounds, Alex White, Nick Brandley, Bryce Masuda and Karan Odom.Financial support was provided by the National Science Foundation (MSWand SP-J), the Cornell Laboratory of Ornithology, the University of ChicagoHinds Fund, American Ornithologists’ Union, Animal Behaviour Society,Frank M. ChapmanMemorial Fund and GAANNNational Training Fellow-ship (EG). All work was conducted with approval from appropriate animalethics and permitting agencies (University of Chicago Animal Care and UseCommittee approval #71708, Cornell University Animal Care and UseCommittee approval 2009-0105, James Cook University Ethics approval#A1340, Scientific Purposes Permit #WISP07773610, Regulation 17 licence#SF007698).

References

Badyaev,A.V.,Hill, G. E., andWeckworth, B.V. (2002). Species divergencein sexually selected traits: increase in song elaboration is related todecrease in plumage ornamentation in finches. Evolution 56, 412–419.

Ballentine, B., Hyman, J., and Nowicki, S. (2004). Vocal performanceinfluences female response to male bird song: an experimental test.Behavioral Ecology 15, 163–168. doi:10.1093/beheco/arg090

Bioacoustics Research Program 2004. Raven Pro: Interactive sound analysissoftware (Version 1.2). The Cornell Lab of Ornithology, Ithaca, NY.

Blomberg, S. P., Garland, T., and Ives, A. R. (2003). Testing for phylogeneticsignal in comparative data: behavioral traits are more labile. Evolution57, 717–745.

Botero, C. A., Boogert, N. J., Vehrencamp, S. L., and Lovette, I. J. (2009).Climatic patterns predict the elaboration of song displays in mocking-birds. Current Biology 19, 1151–1155. doi:10.1016/j.cub.2009.04.061

Cardoso,G.C., andHu,Y. (2011).Birdsongperformance and the evolution ofsimple (rather than elaborate) sexual signals. American Naturalist 178,679–686. doi:10.1086/662160

Cardoso, G. C., Atwell, J. W., Ketterson, E. D., and Price, T. D. (2009). Songtypes, song performance, and the use of repertoires in Dark-eyed Juncos(Junco hyemalis).Behavioral Ecology 20, 901–907. doi:10.1093/beheco/arp079

Cockburn, A., Dalziell, A. H., Blackmore, C. J., Double, M. C., Kokko, H.,Beck, N. R., Head,M. L., andWells, K. (2009). Superb Fairy-wrenmalesaggregate into hidden leks to solicit extragroup fertilizations beforedawn. Behavioral Ecology 20, 501–510. doi:10.1093/beheco/arp024

Cooney,R., andCockburn,A. (1995).Territorial defence is themajor functionof female song in the Superb Fairy-wren, Malurus cyaneus. AnimalBehaviour 49, 1635–1647. doi:10.1016/0003-3472(95)90086-1

Dalziell,A.H., andCockburn,A. (2008).Dawnsong inSuperbFairy-wrens: abird that seeks extra-pair copulations during the dawn chorus. AnimalBehaviour 75, 489–500. doi:10.1016/j.anbehav.2007.05.014

Dowling, J. L., and Webster, M. S. (2013). The form and function of duetsand choruses in Red-backed Fairy-wrens. Emu 113, 282–293.doi:10.1071/MU12082

Driskell, A. C., Norman, J. A., Pruett-Jones, S., Mangall, E., Sonsthagen, S.,andChristidis, L. (2011).Amultigene phylogeny examining evolutionaryand ecological relationships in the Australo-Papuan wrens of the sub-family Malurinae (Aves). Molecular Phylogenetics and Evolution 60,480–485. doi:10.1016/j.ympev.2011.03.030

Felsenstein, J. (1985). Phylogenies and the comparative method. AmericanNaturalist 125, 1–15. doi:10.1086/284325

Fitch, W. T. (1999). Acoustic exaggeration of size in birds via trachealelongation: comparative and theoretical analyses. Journal of Zoology248, 31–48. doi:10.1111/j.1469-7998.1999.tb01020.x

Freckleton, R. P., Harvey, P. H., and Pagel, M. (2002). Phylogenetic analysisand comparative data: a test and review of evidence. American Naturalist160, 712–726. doi:10.1086/343873

Garamszegi, L. Z., and Moller, A. P. (2004). Extrapair paternity and theevolution of bird song. Behavioral Ecology 15, 508–519. doi:10.1093/beheco/arh041

Gardner, J. L., Trueman, J. W. H., Ebert, D., Joseph, L., and Magrath, R. D.(2010). Phylogeny and evolution of the Meliphagoidea, the largestradiation of Australasian songbirds. Molecular Phylogenetics and Evo-lution 55, 1087–1102. doi:10.1016/j.ympev.2010.02.005

Greig, E. I. (2010). Vocal communication in Splendid Fairy-wrens(Malurus splendens). Ph.D. Dissertation, University of Chicago,Chicago, IL.

Greig, E. I., and Pruett-Jones, S. (2008). Splendid songs: the vocal behavior ofSplendid Fairy-wrens (Malurus splendens melanotus). Emu 108,103–114. doi:10.1071/MU07044

Greig, E. I., and Pruett-Jones, S. (2009). An alarm call that isn’t: the Type IIcall in the Splendid Fairy-wren. Animal Behaviour 78, 45–52.doi:10.1016/j.anbehav.2009.02.030

Greig,E. I., andPruett-Jones,S. (2010).Dangermayenhancecommunication:predator calls alert females to male displays. Behavioral Ecology 21,1360–1366. doi:10.1093/beheco/arq155

Greig, E. I., and Webster, M. S. (2013). Spatial decoupling of song andplumage generates novel phenotypes between 2 avian subspecies. Be-havioral Ecology 24, 1004–1013. doi:10.1093/beheco/art005

Greig, E. I., Spendel, K., and Brandley, N. C. (2010). A predator-elicitedvocalization in the Variegated Fairy-wren (Malurus lamberti). Emu 110,165–169. doi:10.1071/MU09107

Greig, E. I., Taft, B. N., and Pruett-Jones, S. (2012). Sons learn songs fromtheir social fathers in a cooperatively breeding bird. Proceedings of theRoyal Society of LondonB –Biological Sciences 279, 3154–3160. doi:10.1098/rspb.2011.2582

Hall, M. L., and Peters, A. (2008). Coordination between the sexes forterritorial defence in a duetting fairy-wren. Animal Behaviour 76,65–73. doi:10.1016/j.anbehav.2008.01.010

Handford,P., andLougheed, S.C. (1991).Variation indurationand frequencycharacteristics in the song of the Rufous-collared Sparrow, Zonotrichiacapensis, with respect to habitat, trill dialect and body size. Condor 93,644–658. doi:10.2307/1368196

Holm, S. (1979). A simple sequentially rejective multiple test procedure.Scandinavian Journal of Statistics 6, 65–70.

Irwin, D. E. (2000). Song variation in an avian ring species. Evolution 54,998–1010.

Johnson, S. L. (2006). Do American Robins acquire songs by both imitatingand inventing? Wilson Journal of Ornithology 118, 341–352. doi:10.1676/05-048.1

Joseph, L., Edwards, S., and McLean, A. (2013). The Maluridae: inferringavian biology and evolutionary history from DNA sequences. Emu 113,195–207. doi:10.1071/MU12081

Kembel, S. W., Cowan, P. D., Helmus, M. R., Cornwell, W. K., Morlon, H.,Ackerly, D. D., Blomberg, S. P., andWebb, C. O. (2010). Picante: R tools

280 Emu E. I. Greig et al.

for integrating phylogenies and ecology. Bioinformatics 26, 1463–1464.doi:10.1093/bioinformatics/btq166

Kingma, S. A., Hall, M. L., Segelbacher, G., and Peters, A. (2009). Radicalloss of an extreme extra-pair mating system. BMC Ecology 9, 15–25.doi:10.1186/1472-6785-9-15

Kleindorfer, S., Evans, C., Mihailova, M., Colombelli-Negrel, D., Hoi, H.,Griggio, M., Mahr, K., and Robertson, J. (2013). When subspeciesmatter: resident Superb Fairy-wrens (Malurus cyaneus) distinguish thesex and subspecies of intruder birds. Emu 113, 259–269. doi:10.1071/MU12066

Kroodsma, D. E., Liu, W. C., Goodwin, E., and Bedell, P. A. (1999). Theecology of song improvisation as illustrated by North American SedgeWrens. Auk 116, 373–386. doi:10.2307/4089372

Lachlan, R. F. (2007). Luscinia: a bioacoustics analysis computer program.Version 1.0. Available at http://luscinia.sourceforge.net/ [Verified 6March 2013].

Langmore, N. E., and Mulder, R. A. (1992). A novel context for bird song:predator calls prompt male singing in the kleptogamous Superb Fairy-wren, Malurus cyaneus. Ethology 90, 143–153. doi:10.1111/j.1439-0310.1992.tb00828.x

Lee, J. Y., Joseph, L., and Edwards, S. V. (2012). A species tree for theAustralo-Papuan fairy-wrens and allies (Aves: Maluridae). SystematicBiology 61, 253–271. doi:10.1093/sysbio/syr101

Maddison, D. R., and Maddison, W. P. (2005). ‘MacClade 4.08.’ (SinauerAssociates, Inc.: Sunderland, MA.)

Martin, J. P., Doucet, S.M., Knox, R. C., andMennill, D. J. (2011). Body sizecorrelates negatively with the frequency of distress calls and songs ofNeotropical birds. Journal of Field Ornithology 82, 259–268.

Medina, I., and Francis, C. D. (2012). Environmental variability and acousticsignals: a multi-level approach in songbirds. Biology Letters 8, 928–931.doi:10.1098/rsbl.2012.0522

Mulder, R. A., Dunn, P. O., Cockburn, A., Lazenbycohen, K. A., andHowell, M. J. (1994). Helpers liberate female fairy-wrens from con-straints on extra-pair mate choice. Proceedings of the Royal Society ofLondon. Series B, Biological Sciences 255, 223–229. doi:10.1098/rspb.1994.0032

Packert, M., Martens, J., Kosuch, J., Nazarenko, A. A., and Veith, M. (2003).Phylogenetic signal in the song of crests and kinglets (Aves: Regulus).Evolution 57, 616–629.

Pagel, M. (1999). Inferring the historical patterns of biological evolution.Nature 401, 877–884. doi:10.1038/44766

Paradis, E., Claude, J., and Strimmer, K. (2004). APE: analyses of phyloge-netics and evolution in R language. Bioinformatics 20, 289–290.doi:10.1093/bioinformatics/btg412

Patten, M. A., Rotenberry, J. T., and Zuk, M. (2004). Habitat selection,acoustic adaptation, and the evolution of reproductive isolation.Evolution58, 2144–2155.

Payne, R. B. (1996). Song traditions in Indigo Buntings: origins, improvi-sation, dispersal and extinction in cultural evolution. In ‘Ecology andEvolution of Acoustic Communication in Birds’. (Eds D. E. Kroodsma,and E. H. Miller.) pp. 198–220. (Comstock: New York.)

Payne, R. B., Payne, L. L., and Rowley, I. (1988). Kin and socialrelationships in Splendid Fairy-wrens: recognition by song in a cooper-ative bird. Animal Behaviour 36, 1341–1351. doi:10.1016/S0003-3472(88)80203-3

Payne, R. B., Payne, L. L., Rowley, I., and Russell, E. M. (1991). Socialrecognition and response to song in cooperativeRed-wingedFairy-wrens.Auk 108, 811–819.

Plowright, H. (2007). A Field Guide to Australian Birdsong. (Audio CD).(Bird Observers Club of Australia: Melbourne).

Podos, J. (1997). A performance constraint on the evolution of trilledvocalizations in a songbird family (Passeriformes: Emberizidae). Evolu-tion 51, 537–551. doi:10.2307/2411126

Podos, J., Nowicki, S., and Peters, S. (1999). Permissiveness in the learningand development of song syntax in swamp sparrows. Animal Behaviour58, 93–103. doi:10.1006/anbe.1999.1140

Podos, J., Huber, S. K., and Taft, B. (2004). Bird song: the interface ofevolution and mechanism. Annual Review of Ecology Evolution andSystematics 35, 55–87. doi:10.1146/annurev.ecolsys.35.021103.105719

Price, J. J., and Lanyon, S. M. (2002). Reconstructing the evolution ofcomplex bird song in the oropendolas. Evolution 56, 1514–1529.

Price, J. J., and Lanyon, S. M. (2004). Patterns of song evolution and sexualselection in the oropendolas and caciques. Behavioral Ecology 15,485–497. doi:10.1093/beheco/arh040

Price, J. J., Earnshaw, S. M., and Webster, M. S. (2006). MontezumaOropendolas modify a component of song constrained by body sizeduring vocal contests. Animal Behaviour 71, 799–807. doi:10.1016/j.anbehav.2005.05.025

R Development Core Team (2012). R: a language and environment forstatistical computing. (R Foundation for Statistical Computing: Vienna,Austria.) Available at http://www.r-project.org [Verified 6 March 2013].

Rowe, M., and Pruett-Jones, S. (2006). Reproductive biology and spermcompetition inAustralian fairy-wrens.AvianandPoultryBiologyReviews17, 31–47.

Rowe, M., and Pruett-Jones, S. (2011). Sperm competition selects for spermquantity and quality in theAustralianMaluridae.PLoSONE6(1), e15720.doi:10.1371/journal.pone.0015720

Rowe, M., and Pruett-Jones, S. (2013). Extra-pair paternity, sperm compe-tition and their evolutionary consequences in the Maluridae. Emu 113,218–231. doi:10.1071/MU12084

Rowley, I., and Russell, E. M. (1997). ‘Fairy-wrens and Grasswrens:Maluridae.’ (Oxford University Press: New York.)

Ryan, M. J., and Brenowitz, E. A. (1985). The role of body size, phylogeny,and ambient noise in the evolution of bird song.AmericanNaturalist 126,87–100. doi:10.1086/284398

Searcy,W. A., and Andersson,M. (1986). Sexual selection and the evolutionof song. Annual Review of Ecology and Systematics 17, 507–533.

Seddon,N. (2005).Ecological adaptationand species recognitiondrives vocalevolution in neotropical suboscine birds. Evolution 59, 200–215.

Slabbekoorn,H., andSmith, T.B. (2002).Habitat-dependent songdivergencein the Little Greenbul: an analysis of environmental selection pressures onacoustic signals. Evolution 56, 1849–1858.

Soma, M., and Garamszegi, L. Z. (2011). Rethinking birdsong evolution:meta-analysis of the relationship between song complexity and reproduc-tive success. Behavioral Ecology 22, 363–371. doi:10.1093/beheco/arq219

Taft, B. (2011). The role of dawn song in tree swallows and its place in thediversity of oscine song learning. Ph.D. Dissertation, University ofMassachusetts, Amherst, MA.

Vallet, E., Beme, I., and Kreutzer, M. (1998). Two-note syllables in canarysongs elicit high levels of sexual display.Animal Behaviour 55, 291–297.doi:10.1006/anbe.1997.0631

Wallschläger, D. (1980). Correlation of song frequency and body weight inpasserine birds. Experientia 36, 412–412. doi:10.1007/BF01975119

Ward, J.H. Jr (1963).Hierarchical grouping to optimize an objective function.Journal of the American Statistical Association 58, 236–244.doi:10.1080/01621459.1963.10500845

Weir, J. T., and Wheatcroft, D. (2011). A latitudinal gradient in rates ofevolution of avian syllable diversity and song length. Proceedings.Biological Sciences 278, 1713–1720. doi:10.1098/rspb.2010.2037

Wiley, R. H. (1991). Associations of song properties with habitats forterritorial oscine birds of eastern North-America. American Naturalist138, 973–993. doi:10.1086/285263

Zelano, B., Tarvin, K. A., and Pruett-Jones, S. (2001). Singing in the face ofdanger: the anomalous Type II song of the Splendid Fairy-wren.Ethology107, 201–216. doi:10.1046/j.1439-0310.2001.00645.x

Song evolution in Maluridae Emu 281

www.publish.csiro.au/journals/emu