1

RRH: Reudink et al. • Island color and song

2
3

Evolution of song and color in island birds

4

Matthew W. Reudink,1* Claudie Pageau,1 Marissa Fisher,1 Nathaniel Mount,1

5

Natasha Buckler,1 Ken A. Otter,2 and James V. Briskie3

6
7

1

8

Canada

9

2

Department of Biological Sciences, Thompson Rivers University, Kamloops, BC V2C 0C8,

Natural Resources and Environmental Studies, University of Northern British Columbia, Prince

10

George, BC, Canada

11

3

Department of Biological Sciences, University of Canterbury, Christchurch 8041, New Zealand

12
13

*Corresponding Author: mreudink@tru.ca

1

14

ABSTRACT—Island birds are often less colorful and sing different songs than their mainland

15

counterparts. Given that trade-offs can occur in the expression of costly sexually selected traits, it

16

is not clear whether island species with less colorful plumage than their mainland counterparts

17

have compensated with increased complexity of song structure. To examine potential trade-offs

18

in sexually selected traits in island birds, we examined concurrent patterns of coloration and song

19

structure in island and mainland (sub)species across 3 widespread and speciose songbird

20

families: Fringillidae, Meliphagidae, and Monarchidae. Phylogenetically controlled analyses

21

revealed a shift away from white plumage and towards melanin-based plumage in island

22

(sub)species but only in the Meliphagidae. Contrary to our predictions, we found no evidence

23

that song structure in island birds differed from their mainland counterparts. Although a number

24

of inverse correlations were identified between song structure and plumage coloration when all

25

birds were considered, this pattern became weaker when only island populations were examined.

26

Instead, we found island size affected song structure leading to a lower number of syllables and

27

lower minimum frequency on smaller islands. Birds on smaller islands also had reduced

28

carotenoid plumage but this pattern was only significant in the Fringillidae. These results

29

indicate that island environments can influence the evolution of both song and color in some

30

taxa, but that selection is tempered by island size, rather than through correlated evolution across

31

signaling modalities.

32
33

Key words: evolution, island, ornament, sexual selection, trade-off

2

34

Islands have long fascinated evolutionary biologists because they act as natural, replicated

35

experiments that can reveal parallel evolutionary mechanisms across taxonomic groups. Within

36

vertebrates, an ‘insularity syndrome’ has been widely documented, with island populations

37

exhibiting marked parallel changes compared to mainland counterparts. For example, the ‘island

38

rule’ (Lomolino 1985, 2005) describes a trend towards dwarfism in large vertebrates and

39

gigantism in small vertebrates on islands (but see Meiri et al. 2008, Durst and Roth 2012,

40

Lomolino et al. 2013). Island populations also appear to experience parallel shifts in life histories

41

towards K-selected reproductive traits, such as low fecundity, slow development, and high

42

survival (Alder and Levins 1994, Novosolov et al. 2013). Less well-studied and well-understood,

43

however, are apparent shifts in signaling systems such as coloration and vocal complexity.

44

Early studies on signaling in birds documented a preponderance of dull-colored species

45

on islands (e.g., Grant 1965) although most surveys were restricted to small geographic areas,

46

such as small island chains, or a small number of species (Figuerola and Green 2000, Fabre et al.

47

2012; but see Fitzpatrick 1998). Recently, Doutrelant et al. (2016) undertook an ambitious global

48

analysis of bird coloration on closely related island and continental species-pairs and clearly

49

demonstrated a reduction in brightness and intensity of coloration and reduction in the number of

50

color patches in island species. The authors also found evidence of a positive relationship

51

between the number of confamilial species and number of color patches, suggesting that a

52

reduction in the number of closely related sympatric species on islands may reduce selection on

53

species-recognition through plumage signals (Doutrelant et al. 2016).

54

If sexually selected traits, such as plumage coloration, are influenced by island effects,

55

signals in other sensory modalities are likely to be similarly influenced. Song is an important

56

inter- and intra-sexual signal in birds (Catchpole 1987) and may be similarly influenced by island

3

57

effects (Baker 1996, 2006). Bird song is an evolutionary labile trait and intra-specific variation in

58

bird song can be influenced by both inter- and intra-specific competition, habitat structure, and

59

changes to the acoustic environment (Searcy and Yasukawa 1996, Badyaev and Leaf 1997). In

60

one of the few broad-scale studies of song on islands, Morinay et al. (2013) demonstrated island-

61

specific changes in song structure, including reduced occurrence of broadband song and putative

62

aggressive song elements (rattles and buzzes) but increased bandwidth. Similar to the findings of

63

Doutrelant et al. (2016) on plumage coloration in birds, Morinay et al. (2013) found that song

64

structure also appears to be influenced by a decrease in the number of same-family sympatric

65

species. This supports the hypothesis that a reduction in the number of closely related species on

66

islands makes species recognition easier and lessens any constraints imposed on vocal signals in

67

continental birds that are sympatric with a larger number of related species (the ‘character release

68

hypothesis;’ Kroodsma 1985).

69

Although the colonization of islands appears to influence the evolution of both song and

70

plumage in birds, whether these changes happen concurrently remains unknown. If the

71

expression of sexually selected traits, such as increased song complexity (e.g., longer songs,

72

larger repertoires, greater diversity of notes) and more elaborate plumage (e.g., increased number

73

and size of carotenoid-based plumage patches, increased size of feather ornaments) are both

74

condition-dependent (Hill 1991), island environments might favor a trade-off such that changes

75

in one signal may be associated with changes in another signal. This trade-off has been observed

76

in continental species of fringillid finches, in which song complexity was negatively related to

77

elaboration of plumage ornaments (Badyaev and Weckworth 2002), although other researchers

78

have found either the reverse pattern or none at all in other groups of birds (e.g., Matysiokova et

79

al. 2017). Island birds might be especially subject to trade-offs between sexually selected

4

80

condition-dependent signals if a reduction in resources on islands places greater energetic

81

constraints on individuals than their counterparts in continental areas. For example, the

82

impoverished food webs found on small islands could place greater nutritional challenges on

83

colonizing birds, leading to trade-offs (Blanco et al. 2014). On the other hand, changes in both

84

coloration and songs could occur without any trade-off if the costs of sexually selected traits are

85

relaxed on islands; such alleviation of expression costs could be due to decreased predator

86

numbers or changes in habitat structure that favor traits that increase detectability to conspecifics

87

(Price et al. 1993, Shutler and Weatherhead 1990).

88

Here, we examine changes in both plumage coloration and song structure across 3

89

widespread, diverse, and ecologically distinct avian families: Fringillidae, Meliphagidae, and

90

Monarchidae. Using phylogenetically controlled analyses, we predicted a reduction in plumage

91

ornamentation (e.g., loss of color types) and changes in song structure (e.g., increased

92

bandwidth) on islands as found by previous authors (Doutrelant et al. 2016, Morinay et al. 2013),

93

but we also tested whether these changes occur concurrently or whether there is a trade-off

94

between plumage coloration and song structure in island birds. Finally, because islands vary

95

drastically in size, we asked whether island size was associated with song and color, with the

96

prediction that island effects would be more intense on smaller islands.

97

Methods

98
99

Data collection

100

We collected color data for 731 species and subspecies. Of those, we were also able to collect

101

song data from 145 species and subspecies. We collected data from species and subspecies

102

within Fringillidae (ncolor = 70 island, 265 mainland, nsong = 15 island, 41 mainland),

5

103

Meliphagidae (ncolor = 138 island, 87 mainland, nsong = 13 island, 23 mainland), and Monarchidae

104

(ncolor = 136 island, 35 mainland, nsong = 41 island, 12 mainland). For species list, see Table S6;

105

for reference, there are 227 recognized species in Fringillidae, 187 in Meliphagidae, and 100 in

106

Monarchidae (Rodewald 2020). These families were chosen because they are widely distributed

107

globally, speciose, and have colonized many island systems on multiple occasions, providing

108

replicates to examine patterns of change in the plumage and song of island birds.

109

Because we were unable to attain spectrophotomeric color measurements for island

110

subspecies (see below), we examined plates from the Handbook of the Birds of the World,

111

supplemented by photographs and field guides, to record the presence or absence (1/0) of 4 color

112

classes (carotenoid, melanin, structural, white) on 5 body regions: head, back, breast, wings, and

113

tail. Thus, each bird received a score from 0-5 for each color class; note that these scores were

114

non-exclusive and body regions could be scored in multiple color classes (e.g., the black and

115

yellow head of a male American goldfinch [Carduelis tristis] would receive a score of 1 for

116

melanin, 0 for white, 1 for carotenoid, and 0 for structural). We scored females and males of

117

each species separately.

118

We recognize that there are several limitations to our approach. First, scoring relies on

119

human observation and there is a risk of misclassification, especially in terms of reddish hues.

120

For example, though yellow, orange, and red plumage is typically produced by carotenoid

121

pigments, reddish hues can also be produced by the deposition of phaeomelanins (though these

122

are typically more ruddy; McGraw 2006a) or psittacofulvins, which produce a range of colors in

123

parrots (McGraw 2006b), but not passerines. As such, it is possible that some body regions

124

classified as containing carotenoid coloration were produced by other sources. Second, structural

125

colors depend on the arrangement of pigments, such as melanosomes, in the feather

6

126

microstructure; for our purposes, however, we classified iridescent, blue, purple, and green

127

feathers as ‘structural’, as opposed to structural and melanin-based. Greenish coloration,

128

however, which is produced by a combination of carotenoid pigments and structural properties

129

(Prum 2006), was scored as both carotenoid and structural. It is important to also note that

130

carotenoid coloration relies upon structural properties in order to produce colorful feathers

131

(Shawkey and Hill 2005), but, for our purposes, red/orange/yellow putatively carotenoid-based

132

coloration was only classified as ‘carotenoid.’ White coloration can also be considered a

133

structural color (Igic et al. 2018), but, because the mechanism of production differs from typical

134

structural colors, we classified it separately. Finally, we classified black/grey/brown/ruddy-

135

reddish colors as a single group (‘melanin’) rather than separated by eumelanin (black) and

136

phaeomelanin (rufous).

137

An ideal approach would be to measure plumage colors using spectrometry combined

138

with models of avian vision (e.g., Doutrelant et al. 2016). However, we were primarily interested

139

in subspecies on remote oceanic islands for which museum specimens are either limited,

140

unavailable, or difficult to access. Thus, while we recognize the value of a spectrophotometric

141

approach, it was not feasible for this study. An alternative approach for quantifying plumage

142

color employed by Dale et al. (2015) was to score digitally scanned images from Handbook of

143

the Birds of the World, which resulted in scores for nearly 6,000 passerines. Unfortunately, that

144

analysis did not take into account subspecific variation, which was of primary interest in our

145

study, and would have reduced our available sample by over 60% and not allowed us to address

146

our questions. The advantage to our approach over that of Dale et al. (2015) is that the loss or

147

gain of a small, but colorful, patch is more detectable than in an approach where color values are

148

averaged over an entire bird. However, our analysis is not as sensitive as a spectrophotometric

7

149

approach (sensu Doutrelant et al. 2016) that would reveal changes in hue (e.g., red to orange or

150

yellow carotenoid-based coloration) or other aspects of avian color. As such, we consider our

151

approach conservative; any patterns revealed should represent major changes in avian coloration

152

(gains/losses in color types).

153

We were able to obtain song recordings for 145 (sub)species from Xeno-canto

154

(www.xeno-canto.org). When possible, we obtained 5 recordings from each (sub)species and

155

averaged values across recordings. Only male song was analyzed as recordings of female song

156

were limited. We then used Avisoft-SASLab-Pro (Vs 5.2.13, Avisoft Bioacoustics, Germany) on

157

a Hamming window with 1024 FFT length to extract the following information from the

158

spectrogram of 1 song per recording: song length, minimum frequency, maximum frequency,

159

bandwidth, number of syllables per song, and syllable rate (Fig. 1). Syllables were defined as

160

temporally distinct elements, determined from visual inspection of a waveform/spectrogram

161

combination of each song. Syllable rate was calculated by dividing the total number of syllables

162

within a single song by the length of the song. Thus, a species with a low syllable rate produced

163

fewer syllables per song. To obtain a metric of song complexity, we plotted the species-average

164

Frequency Bandwidth across all syllables against average Syllable Rate (cf. Podos 2001) and

165

calculated the residual score for each species. These residual scores provide a relative measure of

166

modulation frequency of the song, which we use as an index of relative song complexity. We

167

considered species with positive residual scores to have a higher frequency modulation than

168

expected by their syllable rate as well as having a higher complexity score than species with

169

negative residual scores (cf. Crouch and Mason-Gamer 2019 “modulation index”).

170
171

Information on island size and latitude was obtained from a variety of web sources,
including the Island Directory (http://islands.unep.ch/isldir.htm) and Wikipedia pages for each

8

172

island (http://wikipedia.org). For species that occur on more than 1 island in an archipelago, we

173

used the total area of all islands to estimate the size of its range. Island size was transformed with

174

the natural log prior to analysis. At this stage there is not enough known about the patterns of

175

speciation in our sample to separate those (sub)species that are the result of a colonization event

176

from those that are the result of adaptive radiation within an archipelago. However, our statistical

177

analysis accounts for the non-independence of related species and does not assume that each

178

island species represents a separate colonization event.

179

We did not consider the degree of island isolation from other nearby islands or to the

180

nearest continent. Measuring degree of isolation is tricky as a simple distance is likely to be

181

biased since some species are more mobile than others, the direction of prevailing winds can

182

modify the ‘ease’ of colonizing islands, and as pointed out above, movement within archipelagos

183

could also be a factor. Thus, for the present analysis we assumed that the degree of isolation was

184

similar across taxa. As we restricted our analysis to taxonomically delineated subspecies and

185

species on islands, islands with high levels of gene flow would by definition not be included if

186

ongoing immigration was so high it would prevent the formation of distinct island morphologies.

187
188

Phylogenetic methods

189

We downloaded 10,000 possible phylogenies from birdtree.org (based on Jetz et al. 2012, 2014)

190

for 731 species belonging to Fringillidae, Meliphagidae, and Monarchidae. We then used

191

TreeAnnotator in BEAST v1.10 (Suchard et al. 2018) to construct a maximum clade credibility

192

tree (the most probable tree) using a 1% burn-in and median node heights. Because the

193

subspecific relationships for most taxa in our phylogeny remain unresolved, we subsequently

194

added subspecies to each species, resulting in polytomies at the tips for species with multiple

9

195

subspecies. To deal with issues arising from polytomies with branch lengths of zero, we used the

196

multi2di function in the phytools v6.99 R package (Revell 2012) to randomly assign the

197

arrangement of polytomies, with branch lengths set to 1e-6 of the total branch tree length.

198

Phylogenies were created separately for the 731 species for which we had color data and the 145

199

for which we had song data. We also created 3 separate phylogenies for the different families for

200

the color data. We used Pagel’s l (Harvey and Pagel 1991) as a measure of phylogenetic signal,

201

which we obtained from the phylosig function in the phytools v6.99 R package (Revell 2012).

202
203

Statistical analysis

204

All analyses were conducted in R 3.6.3 (R Core Development Team 2017). To ask whether

205

plumage coloration and song varied between islands and the mainland, we used phylogenetic

206

ANOVA with 10,000 simulations using the phytools v6.99 (Revell 2012), ape v5.3 (Paradis and

207

Schliep 2018), and geiger v2.0.6.2 (Harmon et al. 2008) R packages. For the color data, we

208

analysed all the families together and then separately for each color type.

209

Next, we used a phylogenetic least-squares (PGLS) approach using the nlme v.3.1.145

210

package in R (Pinheiro et al. 2018) to test (a) whether there was a relationship between plumage

211

coloration and song with an interaction of the family and (b) whether song or plumage coloration

212

on islands were associated with island size (km2). The latter analysis (b) was restricted to only

213

(sub)species present on islands. The analyses to test the relationship between plumage coloration

214

and song (a) included both mainland and island (sub)species and were repeated to only included

215

island (sub)species. Figures were created using the phytools v6.99 (Revell 2012) and ggplot2

216

v.3.2.1 (Wickham 2016) packages.

217

10

Results

218
219

Differences in color between island and mainland (sub)species

220

When we examined differences in plumage coloration between island and mainland (sub)species,

221

we found no difference when considering all 3 families together (Table S1), but the phylogenetic

222

signal λ was significant for every color (all p < 0.001) except for carotenoid in Monarchidae (λ =

223

0.55, p = 0.064). When analysing the families separately, only Meliphagidae showed differences

224

in coloration between island and mainland (Table S1). For the Meliphagidae family, male birds

225

on islands have significantly less white coloration than mainland species (F = 26.7, p = 0.001).

226

For the females, island species have more melanin (F = 16.82, p = 0.001) and marginally more

227

structural coloration (F = 7.18, p = 0.051) coloration, but less white coloration (F = 18.55, p =

228

0.005) than mainland species.

229
230

Differences in song between island and mainland (sub)species

231

We found no differences in song characteristics between island and mainland (sub)species for

232

any song variable (Table S2): syllable rate: F = 5.12, p = 0.58), minimum frequency (F = 1.07, p

233

= 0.80), maximum frequency (F = 1.08, p = 0.83), bandwidth (F = 0.31, p = 0.89), number of

234

syllables (F = 0.02, p = 0.98), song length (F = 0.96, p = 0.80), or our measure of complexity (F

235

= 0.22, p = 0.98). The phylogenetic signal λ was significant for every song characteristic (all p <

236

0.001).

237
238

Correlated evolution between color and song

239

For the analysis including both mainland and island (sub)species, we found an evolutionary

240

association between color and song variables for the carotenoid and melanin colors only (SOM

11

241

Table S3 A, B). Carotenoid coloration was positively associated with song length (t = 3.41, p =

242

0.0008) and number of syllables (t = 2.26, p = 0.03). Carotenoid coloration was also negatively

243

associated with maximum frequency for Meliphagidae (t = -2.58, p = 0.01), bandwidth for

244

Meliphagidae (t = -2.57, p = 0.01), and song complexity for Meliphagidae (t = -2.66, p = 0.009).

245

Melanin was positively associated with minimum frequency for Fringillidae (t = 2.47, p = 0.01).

246

Melanin was also negatively associated song length for Monarchidae (t = -2.47, p = 0.02). For

247

structural and white plumage coloration, we found no association with any song variable (SOM

248

Table 3 C, D; all p > 0.07).

249

For the analysis including only the island (sub)species, we found an evolutionary

250

association between color and song variables for the carotenoid and melanin colors only (SOM

251

Table 4A, B). Song length and number of syllables were both positively associated with

252

carotenoid coloration (t = 3.43, p = 0.001 and t = 2.58, p = 0.01, respectively). Carotenoid

253

coloration was positively associated with number of syllables for Fringillidae (t = 2.15, p =

254

0.035) and Meliphagidae (t = 2.20, p = 0.031), with bandwidth for Monarchidae (t = 2.34, p =

255

0.023), and negatively associated with syllable rate for Meliphagidae (t = -2.21, p = 0.03).

256

Melanin was positively associated with minimum frequency for Fringillidae (t = 2.54, p = 0.013)

257

and negatively associated with song complexity for Fringillidae (t = -2.01, p = 0.049). For

258

structural and white plumage coloration, we found no association with any song variable (Table

259

4C, D; all p > 0.19).

260
261

Color and song versus island size

262

No plumage colors were associated with the natural log of island size for the 3 families together

263

(all p > 0.13), Meliphagidae (all p > 0.094), or Monarchidae (all p > 0.094) families (Table S5).

12

264

However, for Fringillidae, males on smaller islands were evolutionary associated with less

265

carotenoid (T = 2.37, p = 0.02) and less white (T = 2.47, p = 0.02).

266

We detected significant relationships between island size and number of syllables (t67 =

267

2.45, p = 0.02) and minimum frequency (t67 = 2.13, p = 0.04), where birds on smaller islands

268

sang songs with fewer syllables and lower minimum frequencies (Fig. S1). We detected no

269

relationships between island size and syllable rate (t67 = 0.99, p = 0.33), song length (t67 = 1.53, p

270

= 0.13), maximum frequency (t67 = -0.02, p = 0.98), bandwidth (t67 = -0.09, p = 0.37), or

271

complexity (t67 = -1.20, p = 0.24).

272
273

Discussion

274

Our comparative analyses confirmed that the plumage of island Meliphagidae species differs

275

from that of their continental counterparts. Island Meliphagidae species were characterized by a

276

decrease in the extent of white colored plumage and, for females uniquely, there was an increase

277

in melanin-based plumage. For the other families, our comparative analyses were unsuccessful in

278

showing variation between island and mainland birds. Our results also did not indicate any

279

changes in song characteristics between island and mainland birds. However, smaller islands

280

were associated with a decrease in the number of syllables and minimum frequency of songs.

281

Plumage characteristics were not associated with island size for Meliphagidae and Monarchidae,

282

but carotenoid-based and white plumages were associated with increased island size for

283

Fringillidae. We found multiple evolutionary associations between carotenoid and melanin

284

plumages and song characteristics when all species (continental and island) were considered

285

together, supporting the hypothesis that birds might trade-off signaling modalities, but the pattern

13

286

was much weaker when restricted to just island species. These associations also varied among

287

families and song characteristics.

288

One explanation for a change in plumage on islands is an expected reduction in the

289

intensity of social and/or sexual selection in island populations of birds. The evolution of

290

carotenoid coloration and other plumage ornaments through sexual selection is well supported by

291

a variety of studies (reviewed in Hill 2006), and social selection may also be responsible for the

292

evolution of some conspicuous signals (reviewed in: Senar 2006). Given that species diversity

293

tends to be lower on islands, a reduction in number of sympatric species could select for less

294

conspicuous signals for species recognition through both plumage and vocal modalities

295

(Doutrelant et al. 2016). However, we found little evidence for a change in plumage colors, with

296

only the Meliphagidae displaying less white (both sexes) and more melanin (females only) on

297

islands. The loss of white and increase of melanin is likely to reduce contrast in plumage, which

298

is consistent with a reduction in the need for conspicuous plumage signals on islands, but the

299

lack of a similar pattern in the Fringillidae and Monarchidae indicates that the loss of plumage

300

signals is not present in all families.

301

Contrary to our expectation, we did not find any differences in song structure between

302

island and continental birds, even in the structures of song that have previously been found to

303

differ in previous studies (e.g., bandwidth; Morinay et al. 2013). The reason for this difference is

304

not clear. Due to a lack of recordings for many species in our sample (and particularly island

305

populations), our metrics of song structure were necessarily simplistic. Other measures of song

306

complexity, such as repertoire size, might provide a more realistic measure of song complexity

307

(and intensity of sexual selection) but have been lacking from comparative studies of song in

308

island birds. Recordings of female song were even less available and would also warrant further

14

309

study (Riebel 2019), especially since female song is well developed in some island birds (e.g.

310

New Zealand Bellbird [Anthornis melanura]; Brunton and Li 2006).

311

The downward shift in minimum frequency (but no change in maximum frequency) on

312

smaller islands may be due to increased body mass in island birds; unfortunately, body mass

313

measurements were unavailable for most island (sub)species and we were unable to test this

314

hypothesis. Alternatively, this downward shift may also reflect a reduction in interspecific

315

competition for acoustic space, as the risk of spectral overlap and interference from other species

316

is reduced on islands with lower species diversity (Kroodsma 1985; Morinay et al. 2013). In

317

contrast, population densities of island birds tend to be higher, suggesting insular species

318

experience greater intraspecific competition and leading to an increase in K-selected life history

319

traits (e.g., increased longevity, reduced rates of reproduction; Covas 2012).

320

The evolution of larger body size has also been proposed to be a direct adaptation to the

321

increased intraspecific competition on islands (Clegg and Owens 2002), but as larger birds have

322

lower pitched songs and syllable rates, traits posited as indices of quality (Ryan and Brenowitz

323

1985; Gil and Gahr 2002), changes in body size and song structure of island birds may be shaped

324

by either sexual selection or competition. Clearly, reconciling how the effects of increased

325

intraspecific competition, and reduced interspecific competition, affect the intensity of social and

326

sexual selection is needed before we can understand why birds on smaller islands show changes

327

in their song.

328

As both conspicuous plumage colors and elaborate song repertoires may be condition

329

dependent traits in birds, and thus costly to produce, it might be expected that trade-offs would

330

occur between these 2 modalities of communication under stressful conditions (Cooney et al.

331

2013). For example, if increased intraspecific competition on islands restricted resource

15

332

availability to males, it may not be possible for individuals to elaborate both their plumage and

333

song simultaneously, and selection may instead favor females to select mates based on only the

334

most reliable traits. Such a situation could arise if a species with extensive carotenoid-based

335

plumage colonized an island in which dietary sources of carotenoids were so rare or absent that

336

such plumage could not be expressed and this favored a switch to mate choice based on songs as

337

an indicator of male quality. In such a scenario, one would expect island birds to show an inverse

338

relationship between plumage color and song elaboration, with species losing plumage colors

339

showing increased complexity of their songs (and vice versa). Instead, we found associations

340

between plumage and song characteristics varied among the families and these did not indicate a

341

clear trade-off between song complexity and plumage coloration. For example, we found

342

increased song complexity with a decrease of carotenoid coloration in Meliphagidae when

343

comparing island and mainland birds, but we also observed an increase of number of syllables

344

with an increase of carotenoid in Monarchidae.

345

The multiple associations we found do not go in the same direction and, as a result, do

346

not support a consistent pattern of inverse trade-offs. This is similar to the lack of a trade-off

347

found by Mason et al. (2014) in a comparative analysis of sexual signals in the tanagers (Family

348

Thraupidae). It is possible both modalities are constrained, or that other selective pressures have

349

favored the independent changes in plumage and song we observed, a conclusion also reached by

350

Mason et al. (2014).

351

Our finding that songs became lower pitched (i.e., decrease in minimum frequency) with

352

fewer syllables on smaller islands could also be explained by differences in the signaling

353

environment of insular habitats compared to that inhabited by their continental relatives or

354

(sub)species living on larger islands (Morton 1975, Derryberry et al. 2018). A decrease in

16

355

minimum frequency is thought to be advantageous in dense forest habitats due to greater levels

356

of attenuation of high frequencies (Wiley and Richard 1978). Unfortunately, information on

357

forest structure is not available for most of the islands in our analysis, but as many of the

358

continental species in both the Meliphagidae and Monarchidae are found in the eucalypt forests

359

of Australia, a habitat that is generally more open than closed-canopy forests found on the nearby

360

islands of Indonesia and the South Pacific, the difference in song structure may be due to this

361

habitat difference rather than an island effect per se. Similarly, habitat may exert an influence on

362

the evolution of plumage coloration; however, the shift away from white coloration in

363

Meliphagidae on islands runs contrary to predictions based on the Light Environment

364

Hypothesis, which posits that long-wavelength colors should be favored in more closed

365

environments (Endler 1993). Measures of habitat structure need to be quantified before these

366

hypotheses can be tested.

367

Island size only seemed to influence coloration in Fringillidae for carotenoid and white

368

colors where species on bigger islands had more carotenoid and white colors. A decrease in

369

carotenoid-based plumage traits on smaller islands may reflect a greater difficulty faced by birds

370

in acquiring the necessary carotenoid precursors in the diet (Hill 1993). This could be due either

371

to the lack of the appropriate sources (e.g., due to a lower diversity of plants and animals on

372

small islands), or to the greater competition for the dietary precursors as a result of higher levels

373

of intraspecific competition. The decrease in the extent of red plumage in House Finches

374

(Haemorhous mexicanus) introduced to the Hawaiian Islands may be an example of this

375

phenomenon; shortly after their introduction from the west coast of North America, the

376

characteristic red wash over the head and breast of male house finches on the mainland was soon

377

replaced by a pale yellow or orange in Hawaii (Hill 1993). Supplemental feeding experiments in

17

378

the lab confirmed that a diet limited in carotenoids can reduce the extent of red in the plumage of

379

male house finches (Hill 1993). If a similar reversion to brighter carotenoid-based plumage could

380

be induced in birds on small island through food supplementation experiments, this would

381

support the hypothesis that at least some of the changes in ornamentation are due to

382

environmental constraints and not necessarily an adaptation to the insular environment.

383

Overall, we found a complex pattern of changes in plumage coloration and song structure

384

between island and continental birds. In contrast to previous studies, changes in plumage or song

385

signals were either not found, or if present restricted to just 1 or 2 families, suggesting that broad

386

generalisations found in previous comparative studies may mask interesting patterns at lower

387

taxon levels. As with previous studies, comparative analyses of island birds continue to suffer

388

from a lack of data at all levels, restricting comparisons to crude indices of plumage and song,

389

and again potentially masking more subtle patterns. Finally, our identification of island size as an

390

important covariate in comparative studies highlights the need to not treat all islands equally as

391

the selective pressures that drive changes in insular populations of birds may be the features of

392

the islands themselves and these need to be considered in understanding how island birds may

393

sometimes differ from their continental counterparts.

394
395

Acknowledgements

396

We thank the Erskine Visiting Scholar Program at the University of Canterbury, which provided

397

the opportunity for this project to come to fruition. A. Letten provided valuable help and advice

398

with statistical analysis. M.W.R. is funded by a Natural Sciences and Engineering Research

399

Council of Canada Discovery Grant.

18

Literature cited

400
401
402

Adler GH, Levins R. 1994. The island syndrome in rodent populations. Quarterly Review of
Biology 69:473–490.

403

Avisoft Bioacoustics. 2017. Glienicke, Germany.

404

Badyaev AV, Hill GE, Weckworth BV. 2002. Species divergence in sexually selected traits:

405

increase in song elaboration is related to decrease in plumage ornamentation in finches.

406

Evolution 56:412–419.

407
408
409
410

Badyaev AV, Leaf ES. 1997. Habitat associations of song characteristics in Phylloscopus and
Hippolais warblers. Auk 114:40–46.
Baker MC. 1996. Depauperate meme pool of vocal signals in an island population of singing
honeyeaters. Animal Behavior 51:853–858.

411

Baker MC. 2006. Differentiation of mating vocalizations in birds: acoustic features in mainland

412

and island populations and evidence of habitat-dependent selection on songs. Ethology

413

112:757–771.

414
415
416
417
418
419

Blanco G, Laiolo P, Fargallo JA. 2014. Linking environmental stress, feeding-shifts and the
‘island syndrome’: a nutritional challenge hypothesis. Population Ecology 56:203–216.
Brunton DH, Li X. 2006. The song structure and seasonal patterns of vocal behavior of male and
female bellbirds (Anthornis melanura). Journal of Ethology 24: 17–25.
Catchpole CK 1987. Bird song, sexual selection and female choice. Trends in Ecology and
Evolution 2:94–97.

420

Clegg SM, Owens IPF. 2002. The ‘island rule’ in birds: medium body size and its ecological

421

explanation. Proceedings of the Royal Society of London B 269: 1359–1365.

19

422

Cooney CR, MacGregor HEA, Seddon N, Tobias JA. 2018. Multi-modal signal evolution in

423

birds: re-examining a standard proxy for sexual selection. Proceedings of the Royal

424

Society of London B 285: 20181557.

425
426
427

Covas R. 2012. Evolution of reproductive life histories in island birds worldwide. Proceedings of
the Royal Society of London: B 279: 1531–1537.
Crouch NMA, Mason-Gamer RJ. 2019. Identifying ecological drivers of interspecific variation

428

in song complexity in songbirds (Passeriformes, Passeri). Journal of Avian Biology:

429

e02020.

430
431

Dale J, Dey CJ, Delhey K, Kempenaers B, Valcu M. 2015. The effects of life history and sexual
selection on male and female plumage colouration. Nature 527: 367–370.

432

Derryberry EP, Seddon N, Derryberry GE, Claramunt S, Seeholzer GF, Brumfield RT, Tobias,

433

JA. 2018. Ecological drivers of song evolution in birds: disentangling the effects of

434

habitat and morphology. Ecology and Evolution. Doi: 10.1002/ece3.3760.

435
436
437

Endler JA. 1993. The color of light in forests and its implications. Ecological Monographs 63:1–
27.
Doutrelant C, Paquet M, Renoult JP, Gregoire A, Crochet PA, Covas R. 2016. Worldwide

438

patterns of bird colouration on islands. Ecology Letters. Doi: 10.1111/ele/12588.

439

Durst PAP, Roth VL. 2012. Mainland size variation influences informs predictive models of

440

exceptional insular body size change in rodents. Proceedings of the Royal Society of

441

London B 282: 20150239.

442
443

Fabre P-H, Irestedt M, Fjeldså J, Bristol R, Groombridge JJ, Irham M. Jønsson KA. 2012.
Dynamic colonization exchanges between continents and islands drive diversification in

20

444

paradise-flycatchers (Terpsiphone, Monarchidae). Journal of Biogeography 39:1900–

445

1918.

446

Figuerola J, Green AJ. 2000. The evolution of sexual dimorphism in relation to mating patterns,

447

cavity nesting, insularity and sympatry in the Anseriformes. Functional Ecology 14:701–

448

710.

449

Fitzpatrick S. 1998. Intraspecific variation in wing length and male plumage coloration with

450

migratory behaviour in continental and island populations. Journal of Avian Biology

451

29:248–256.

452
453

Gil D, Gahr M. 2002. The honesty of bird song: multiple constraints for multiple traits. Trends
Ecology and Evolution 17:133–141.

454

Grant PR. 1965. Plumage and the evolution of birds on islands. Systematic Zoology 14:47–52.

455

Harmon LJ, Weir JT, Brock CD, Glor RE, Challenger W. 2008. GEIGER: investigating

456
457
458
459
460

evolutionary radiations. Bioinformatics 24:129–131.
Hill GE. 1991. Plumage colouration is a sexually selected indicator of male quality. Nature
350:337–339.
Hill GE. 1993. Geographic variation in the carotenoid plumage pigmentation of male house
finches (Carpodacus mexicanus). Biological Journal of the Linnean Society 49:63–86.

461

Hill GE. 2006. Female mate choice for ornamental coloration. In: Bird Coloration. Vol. 2.

462

Function and Evolution (Ed. by GE Hill & KJ McGraw), pp. 137–200. Cambridge,

463

Massachusetts: Harvard University Press.

464
465

Igic B, D’Alba L, Shawkey M. 2018. Fifty shades of white: how white feather brightness differs
among species. Naturwissenschaften 105:3–4.

21

466
467

Jetz W, Thomas GH, Joy JB, Hartmann K, Mooers O. 2012. The global diversity of birds in
space and time. Nature 491:444–448.

468

Jetz W, Thomas GH, Joy JB, Redding DW, Hartmann K, Mooers AO. 2014. Global distribution

469

and conservation of evolutionary distinctness in birds. Current Biology 24:919–930.

470

Kroodsma DE. 1985. Geographic variation in songs of the Bewick’s wren: a search for

471

correlations with avifaunal complexity. Behavioral Ecology and Sociobiology 16:143–

472

150.

473
474
475
476

Lomolino MV. 1985. Body size of mammals on islands: the island rule reexamined. The
American Naturalist 125:310–316.
Lomolino MV. 2005. Body size evolution in insular vertebrates: generality of the island rule.
Journal of Biogeography 32:1683–1699.

477

Lomolino MV, van der Geer AA, Lyras GA, Palombo MR, Sax DF, Rozzi R. 2013. Of mice and

478

mammoths: generality and antiquity of the island rule. Journal of Biogeography 40:1427–

479

1437.

480

Matysiokova B, Friedman N, Turcokova L, Remes V. 2017. The evolution of feather coloration

481

and song in Old World orioles (genus Oriolus). Journal of Avian Biology 48:1015–1024.

482

Mason NA, Shultz AJ, Burns KJ. 2014. Elaborate visual and acoustic signals evolve

483

independently in a large, phenotypically diverse radiation of songbirds. Proceedings of

484

the Royal Society of London B 281:20140967.

485

McGraw KJ. 2006a. Mechanics of Melanin-based Coloration. In: Bird Coloration. Vol. 1.

486

Mechanisms and Measurements (Ed. by GE Hill & KJ McGraw), pp. 243–294.

487

Cambridge, Massachusetts: Harvard University Press.

22

488

McGraw KJ 2006a. The mechanics of uncommon colors: pterins, porphyrins, and

489

psittacofulvins. In: Bird coloration. Vol. 1. Mechanisms and measurements (Ed. by GE

490

Hill & KJ McGraw), pp. 295–253. Cambridge, Massachusetts: Harvard University Press.

491

Meiri S, Cooper N, Purvis A. 2008. The island rule: made to be broken? Proceeding of the Royal

492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507

Society of London B 275:141–148.
Morinay J, Cardoso GC, Doutrelant C, Covas R. 2013. The evolution of birdsong on islands.
Ecology and Evolution 3:5127–5140.
Morton ES. 1975. Ecological sources of selection on avian sounds. The American Naturalist
109:17–34.
Novosolov M, Raia P, Meiri S. 2013. The island syndrome in lizards. Global Ecology and
Biogeography 22:184–191.
Paradis E, Schliep K. 2019. ape 5.0: an environment for modern phylogenetics and evolutionary
analyses in R. Bioinformatics 35:526–528.
Pinheiro J, Bates D, DebRoy S, Sarkar D, R Core Team. 2018. nlme: Linear and Nonlinear
Mixed Effects Models. R package version 3.1–137.
Podos J. 2001. Correlated evolution of morphology and vocal signal structure in Darwin’s
finches. Nature 409:185–188.
Price T, Schluter D, Heckman NE. 1993. Sexual selection when the female directly benefits.
Biological Journal of the Linnean Society 48:187–211.
Prum RO. 2006. Anatomy, physics, and evolution of structural colors. In: Bird coloration. Vol. 1.

508

Mechanisms and measurements (Ed. by GE Hill & KJ McGraw), pp. 354–397.

509

Cambridge, Massachusetts: Harvard University Press.

23

510

R Core Team. 2017. R: A language and environment for statistical computing. R Foundation for

511

Statistical Computing, Vienna, Austria. URL https://www.R-project.org/

512

Revell LJ. 2012. phytools: An R package for phylogenetic comparative biology (and other

513

things). Methods in Ecology and Evolution 3:217–223.

514

Riebel K, Odom KJ, Langmore NE, Hall ML. 2019. New insights from female bird song:

515

towards an integrated approach to studying male and female communication roles.

516

Biology Letters 15: 20190059.

517
518
519
520
521

Ryan MJ, Brenowitz EA. 1985. The role of body size, phylogeny, and ambient noise in the
evolution of bird song. The American Naturalist 126:87–100.
Shutler D, Weatherhead PJ. 1990. Targets of sexual selection: song and plumage of wood
warblers. Evolution 44:1967–1977.
Searcy WA, Yasukawa K. 1996. The reproductive success of secondary females relative to that

522

of monogamous and primary females in red-winged blackbirds. Journal of Avian Biology

523

27:225–230.

524

Senar JC. 2006. Colour displays as intrasexual signals in birds. In: Bird coloration. Vol. 2:

525

Function and fvolution (Ed. by GE Hill & KJ McGraw), pp. 87–136. Cambridge,

526

Massachusetts: Harvard University Press.

527
528
529

Shawkey MD, Hill GE. 2005. Carotenoids need structural colours to shine. Biology Letters
1:121–124.
Suchard MA, Lemey P, Baele G, Ayres DL, Drummond AJ, Rambaut, A. 2018. Bayesian

530

phylogenetic and phylodynamic data integration using BEAST 1.10. Virus Evolution 4:

531

vey016.

532

Wickham H. 2016. ggplot2: Elegant graphics for data analysis. Springer-Verlag, New York.

24

533

Wiley RH, Richards DG. 1978. Physical constraints on acoustic communication in the

534

atmosphere: implications for the evolution of animal vocalizations. Behavioral Ecology

535

and Sociobiology 3: 69–94.

536

25

537

Figure legends

538
539

Fig. 1. Examples of 2 species in the Family Monarchidae, Black Monarchs (Monarcha axillaris

540

axillaris) that occupy larger islands and Chestnut-bellied Monarchs (M. castaneiventris

541

obscurior) occupying smaller islands. Song length(s) is measured from start of first distinct

542

syllable in the song to the end of the last syllable—for single syllable songs song length equals

543

syllable length. Number of distinct syllables were identified as continuous sound pulses in

544

waveforms (S1, S2 etc). Maximum and minimum frequencies (Hz) were determined by

545

examining the Power Envelope at all points along the syllables in SAS-LAB Pro, and the

546

difference between maximum and minimum frequencies was calculated as the bandwidth (Hz).

547

26

548

Fig. 1

549

550

27