Author’s Proof

Carefully read the entire proof and mark all corrections in the appropriate place, using the Adobe Reader commenting tools (Adobe Help). Do

not forget to reply to the queries.

We do not accept corrections in the form of edited manuscripts.

In order to ensure the timely publication of your article, please submit the corrections within 48 hours.

If you have any questions, please contact science.production.office@frontiersin.org.

Author Queries Form

Q1 The citation and surnames of all of the authors have been highlighted.

Please check all of the names carefully and indicate if any are

incorrect. Please note that this may affect the indexing of your article

in repositories such as PubMed.

Q2 Confirm that the email address in your correspondence section is

accurate.

Q3 Please ask the following authors to register with Frontiers (at https://

www.frontiersin.org/Registration/Register.aspx) if they would like their

names on the article abstract page and PDF to be linked to a Frontiers

profile. Please ensure to provide us with the profile link(s) when

submitting the proof corrections. Non-registered authors will have the

default profile image displayed.

“Tanujit Dey”

“Guy Cowlishaw.”

Q4 If you decide to use previously published, copyrighted figures in your

article, please keep in mind that it is your responsibility, as the author,

to obtain the appropriate permissions and licenses and to follow any

citation instructions requested by third-party rights holders. If

obtaining the reproduction rights involves the payment of a fee, these

charges are to be paid by the authors.

Q5 Ensure that all the figures, tables and captions are correct.

Q6 Verify that all the equations and special characters are displayed

correctly.

Q7 Ensure to add all grant numbers and funding information, as after

publication this is no longer possible.

Q8 Ensure, if it applies to your study, the ethics statement is included in

the article.

Q9 Please ensure that any supplementary material is correctly published

at this link: https://www.frontiersin.org/articles/10.3389/fevo.2019.

00073/full#supplementary-material (you may need to copy-paste the

link directly in your browser).

Please provide new files if you have any corrections. Note that ALL

supplementary files will be deposited to FigShare and receive a DOI.

Notify us of any previously deposited material.

Q10 Could you please confirm if all author affiliations are fine as listed?

Q11 Please add “Gros-Louis, 2008” to the reference list.

Q12 We have changed “Team, 2016” to “R Core Team, 2016” here and

elsewhere inside the text. Please confirm if this is fine.

Q13 The artwork of “Figure 1” does not have any part label “B,” however,

there is a citation for “Figure 1B” here. Could you please advice?

Q14 Kindly confirm if the details appearing in the “Author Contributions”

section are correct and make sure all authors listed in the Author’s list

are mentioned there.

Q15 Please cite “Datasheets 1, 2” inside the text.

Q16 Please provide doi for the following references.

“Rendall et al., 2000; Trillmich et al., 2014.”

Q17 Please cite “Supplemental Materials 1.1, 1.1.1, 1.1.2, 1.1.3” inside the

text.

Q18 Please confirm whether the heading levels in the article have been

identified correctly.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

ORIGINAL RESEARCHpublished: xx March 2019

doi: 10.3389/fevo.2019.00073

Frontiers in Ecology and Evolution | www.frontiersin.org 1 March 2019 | Volume 7 | Article 73

Edited by:

Claudia Fichtel,

Deutsches Primatenzentrum,

Germany

Reviewed by:

Odile PETIT,

Centre National de la Recherche

Scientifique (CNRS), France

Lucille Chapuis,

University of Exeter, United Kingdom

Q2

*Correspondence:

Lisa R. O’Bryan

obry0017@gmail.com

Specialty section:

This article was submitted to

Behavioral and Evolutionary Ecology,

a section of the journal

Frontiers in Ecology and Evolution

Received: 10 September 2018

Accepted: 26 February 2019

Published: xx March 2019

Citation:

O’Bryan LR, Abaid N, Nakayama S,

Dey T, King AJ, Cowlishaw G,

Rubenstein D and Garnier S (2019)

Contact Calls Facilitate Group

Contraction in Free-Ranging Goats

(Capra aegagrus hircus).

Front. Ecol. Evol. 7:73.

doi: 10.3389/fevo.2019.00073

Contact Calls Facilitate GroupContraction in Free-Ranging Goats(Capra aegagrus hircus)

Q1 Q3Lisa R. O’Bryan 1*, Nicole Abaid 2, Shinnosuke Nakayama 3, Tanujit Dey 4, Andrew J. King 5,

Guy Cowlishaw 6, Daniel Rubenstein 7 and Simon Garnier 1

1Department of Biology, New Jersey Institute of Technology, Newark, NJ, United States, 2Department of Biomedical

Engineering and Mechanics, Virginia Tech, Blacksburg, VA, United States, 3Department of Mechanical and Aerospace Q10

Engineering, New York University, New York City, NY, United States, 4Department of Quantitative Health Sciences, Lerner

Research Institute, Cleveland Clinic, Cleveland, OH, United States, 5Department of Biosciences, Swansea University,

Swansea, United Kingdom, 6 Institute of Zoology, Zoological Society of London, London, United Kingdom, 7Department of

Ecology and Evolutionary Biology, Princeton University, Princeton, NJ, United States

Many social animal species produce vocalizations believed to facilitate group contraction

when one or more group members have become distant. However, the mechanisms

underlying this function remain unclear for many species. We examined this question

with data on a semi-free ranging group of 16 adult domesticated goats (Capra aegagrus

hircus) inhabiting Tsaobis Nature Park, Namibia. All goats wore dataloggers consisting of

a GPS and audio recorder for 5–6 h per day for 10 days, providing continuous data on

their geolocations and vocal communication. We found that callers were farther from the

group centroid than expected by chance and that call production was associated with

the cessation of group expansion and subsequent group contraction. We did not find

strong evidence for antiphonal call exchange between distant and core group members.

Rather, we found that (i) call production by distant group members is associated with a

significant reduction of group movement away from the caller, and (ii) call production

by core group members is associated with greater, though not significantly greater,

group movement toward the caller. These findings suggest that calls may be used by

distant, and potentially core, groupmembers to facilitate the contraction of group spread.

Results from our study clarify the mechanisms through which social animals can regulate

collective movement behavior and the specific role that vocalizations play in this process.

Keywords: communication, group cohesion, wearable sensing devices, contact call, collective behavior

INTRODUCTION

Animals can gain many benefits from remaining cohesive with others, such as greater protection Q6

Q18

from predators, more opportunities for mating and access to social information on the locationsof resources (van Schaik, 1983; Dehn, 1990). In order to maintain such group cohesion, animalsmust coordinate their behavior with one another. Studies indicate that many groups can coordinatetheir behavior by responding to the positions and trajectories of near neighbors (Couzin et al.,2002; Herbert-Read et al., 2011; Farine et al., 2016). However, the efficacy of these mechanisms candiminish during contexts in which lines of sight are broken by vegetation or when visual attentionis focused on activities such as foraging (Strandburg-Peshkin et al., 2017). Accordingly, many social

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

O’Bryan et al. Contact Calls Facilitate Group Contraction

birds and mammals produce specific vocalizations (“contactcalls”) which are believed to play a role in aiding the maintenanceof group cohesion and/or coordination (Kondo and Watanabe,2009). One category of contact call includes loud calls whichindividuals produce primarily when they have become distantfrom some or all group members (Rendall et al., 2000; Gros-Louis et al., 2008). This type of contact call is often termed a“lost call” and is believed to play a role in helping the caller(s) toregain contact with group members (Cheney et al., 1996; Fischeret al., 2001; Digweed et al., 2007). Despite numerous studiesof loud contact calling behavior in a variety of species, howthese calls facilitate the contraction of group spread (hereafter“group contraction”) remains poorly understood, largely due tothe difficulty of monitoring the vocal and movement behaviors ofmultiple group members simultaneously and due to the typicalsubtlety of any behavioral changes (Cheney et al., 1996).

Here, we use wearable dataloggers to examine the role loudcontact calls play in group contraction among a herd of free-ranging adult goats (Capra aegagrus hircus). Domesticated goatsare a highly tractable study species that facilitated the readydeployment of our tracking devices on all groupmembers and themonitoring of these devices throughout the study. Furthermore,goats are gregarious animals that display herding behaviors(Escós et al., 1993), make collective decisions about the natureand timing of their activities (Shrader et al., 2007) and producecontact calls as their primary vocalization type (Briefer andMcElligott, 2011, 2012). Controlled experiments have shownthat partial social isolation induces contact call productionand increased levels of movement which could promote therestoration of contact with group members (Price and Thos,1980; Siebert et al., 2011). While differences in dominancemay exist between individuals, dominance does not correlatewith leadership (Stewart and Scott, 1947), suggesting that allindividuals have the capacity to influence group movement.Despite being domesticated, domestic goats have been found todisplay many behaviors similar to other bovids, as well as otherwild species (Shank, 1972; Dunbar et al., 1990; Saunders et al.,2005). Thus, we use our study system to test hypotheses posed forthe function of loud contact calls in a variety of social birds andmammals, and we expect our findings to be relevant to studies ofcontact calling behavior in these species as well.

A number of studies have sought to determine howloud contact calls facilitate group contraction. Antiphonal callproduction is common among separated parents and offspringin a variety of species, especially those in which parents leaveyoung behind while foraging, such as white-winged vampirebats (Diaemus youngi) (Carter et al., 2008) and king penguins(Aptenodytes patagonicus) (Aubin et al., 2000). Accordingly, acommon assumption about the role loud contact calls play ingroup contraction is that calls made by distant individuals elicita vocal response by group members located near the core ofthe group, enabling the distant individual to localize and movetoward the group. Indeed, in a study of white-faced capuchins(Cebus capucinus), Digweed et al. (2007) found that the durationof separation for distant individuals was shorter when their callswere answered than when they were not. However, studies havefound that such antiphonal responses to loud contact calls can be

rare and selective among adults. For example, the same study ofwhite-faced capuchins found that antiphonal responses to loudcontact calls occurred only 35% of the time, with the calls ofdominant individuals being more likely to elicit responses thanthe calls of lower ranking individuals (Digweed et al., 2007).Additionally, studies of chacma baboons (Papio ursinus) havefound that those individuals that do call, seemingly in responseto the loud contact calls of others, are also on the periphery of thegroup themselves (Cheney et al., 1996; Rendall et al., 2000). Inthis situation, if rejoining the group depends upon reciprocal callproduction between distant and core individuals, these patternsof call production would not be beneficial.

There is some evidence that loud contact calls can facilitategroup contraction in ways that do not depend upon antiphonalcall exchange. For example, descriptions of the vocal repertoireof white-faced capuchins by Gros-Louis (2008) indicate that Q11

these calls are sometimes produced by individuals in the coreof the group when other group members have become distant.Such spontaneously produced calls by core groupmembers couldprompt distant individuals to move toward the caller withoutthe need for a provoking vocalization by the distant individual.Similarly, a study of contact call production between femalebaboons (P. ursinus) and their infants found that, while femaleswere capable of recognizing the contact calls of their infants,they did not tend to respond vocally, but rather moved towardtheir infants in order to retrieve them (Rendall et al., 2000).Both studies suggest that contact calls can attract specific groupmembers toward the caller. While it remains unclear whetherdistant individuals use loud contact calls to attract the entiregroup toward themselves, there is evidence that vocalizationsproduced in other contexts can have a similar attractive effecton groups. For example, “trill” vocalizations produced by white-faced capuchins located toward the side or back of a travelingtroop are capable of changing the troop’s trajectory toward thatof the callers (Boinski, 1993). As another example, a study ofquiet contact calls (“close calls”) produced by foraging meerkats(Suricata suricatta) found that playbacks of these calls causednearby individuals to move toward the speaker and therebygravitate toward the “vocal hotspot” (Gall and Manser, 2017).

We first establish whether, in our study system, loud contactcalls are indeed more likely to occur following a period ofgroup expansion, when at least some individuals become distantfrom the group. We also test whether contact call productionis associated with subsequent group contraction while alsogaining insight into the timeframe over which this might occur.We then test two main hypotheses regarding the mechanismsthrough which call production could promote group contraction(Table 1). The “Vocal Beacon Hypothesis” poses that calls attractgroup members toward the caller. This could occur in twodifferent ways depending on the presence (1a) or absence (1b)of antiphonal calling: Prediction (1a) The production of aninitial contact call elicits the production of antiphonal call(s)by individuals closer to the center of the group, enabling theinitial caller to move toward the group. Prediction (1b) A groupmember produces a call which attracts other group memberstoward itself. While studies of domestic goats have found thatmothers and their offspring respond vocally to playbacks of

Frontiers in Ecology and Evolution | www.frontiersin.org 2 March 2019 | Volume 7 | Article 73

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

O’Bryan et al. Contact Calls Facilitate Group Contraction

TABLE 1 | Our predictions regarding call production patterns and the velocities of the caller relative to the group centroid, and vice versa, surrounding call production.Q5

Multiple callers

required?

Velocity of caller after

call

Change in

velocity of caller

after call

Velocity of centroid

after call

Change in velocity

of centroid after

call

Other

Vocal beacon

hypothesis (1a)

Yes Toward centroid More toward

centroid

Away from caller or

neutral

No prediction Subsequent caller is

closer to group

centroid than initial

caller

Vocal beacon

hypothesis (1b)

No Away from centroid or

neutral

No prediction Toward caller More toward caller

Straggler

hypothesis (2)

No Toward centroid No prediction Away from caller or

neutral

Less away from

caller

Change in velocity following call production is calculated relative to the caller’s or group centoid’s own velocity at the time of the call.

each other’s contact calls (Ruiz-Miranda et al., 1993; Briefer andMcElligott, 2011), counter calling behavior between adults isnot well studied. In addition to the Vocal Beacon Hypothesis,we propose a novel hypothesis—the “Straggler Hypothesis”—which proposes that calls slow or stop the movement of groupmembers away from the caller, enabling it to more readily catchup (Prediction 2). This hypothesis presumes that the caller isalready aware of others’ locations at the time of call production.A study of the contact barks of chacma baboons (P. ursinus)found that 80% of contact barks were produced by individualsin the last third of the group progression, even when they weresurrounded bymany groupmembers (Cheney et al., 1996).Whileit is unclear whether these calls cause baboons in the front of thegroup to slow theirmovement away from callers in the back of thegroup, this is a possible way in which calls could facilitate groupcontraction, and one hypothesis which we explicitly test here.

We tested our hypotheses using fine-scale GPS and audiodata collected from wearable dataloggers which we deployed ona group of 16 free-ranging goats in Namibia. Using these data,we examined the individual and group-level properties displayedduring periods of call production and tested the mechanismsthrough which contact calls may facilitate group contraction.Specifically, we examined measures of individual and groupspread at the time of call production, evidence for antiphonal callexchange, and changes in both group spread and the velocitiesof the callers and group members toward one another duringperiods surrounding call production.

MATERIALS AND METHODS

Study Subjects and Data CollectionFieldwork was carried out during September 2015 at TsaobisNature Park which is located on the edge of the Namib Desert incentral Namibia (22◦23′S 15◦45′W). The park is bordered to thenorth by the ephemeral Swakop River which supports patches ofriparian woodland along its banks. Beyond the banks, vegetationis sparse and consists primarily of small trees and shrubs (seeCowlishaw and Davies, 1997) for a full description of the ecologyat the site). At the beginning of the study, a group of 16 adultfemale goats were separated from a larger mixed-species herd ofgoats and sheep. The goats were housed in a corral in the evenings

and at night and were allowed to roam freely for 5–6 h perday from early morning to early afternoon. All group memberswere lactating during the study and had young offspring whichremained in the pen while the herd grazed. This practice, whichpredated the study, was for the protection of the young and issimilar to the infant caching behavior displayed by feral goats(McDougall, 2009). The goats had free access to their offspringwhile in the pen.

The goats were allowed to adjust to their modified group overa 5 day period at the start of the study. On the morning of thesixth day, all 16 goats were fitted with modified SHOALgroupF2HKv2 collars (see Fehlmann et al., 2017 for details). Eachcollar contained a TechnoSmArt GiPSy 4 GPS unit, recordinggeolocation at 1Hz, and an audio recorder which recorded datacontinuously at 16 kHz. This audio recorder was comprised ofanAdafruit ElectretMicrophoneAmplifier (MAX4466), AdafruitTrinketMicrocontroller, and Adafruit VS1053 Codec+MicroSDBreakout. The microphone was positioned on the outside of thecase surrounding the other collar components and was angled uptoward the wearer’s mouth. Audio data were saved as 1 h-longOGG files which were later converted to WAV files. The goatswore the collars for a 10 day period (09/09/2015–09/18/2015).Since our study species was a domesticated animal, we had readyaccess to the animals while in the pen, which enabled us tomonitor the status of the dataloggers and save battery by turningthe GPS and audio recorder on, and off, each day prior to, andfollowing, their free-ranging period. In order to synchronize theaudio and GPS data, we synchronized our watches to internet-derived Universal Coordinated time (UTC) time at the beginningof the study and spoke the current time out loud to each recordereach time we turned it on (see Supplemental Materials 1.1.3

for more information on our verification of synchronization).Any observed power failures were repaired following dailyinspection of devices. In addition, the collar of each groupmember was removed once in the evening between 09/12/2015and 09/14/2015 in order to back up the data and correct anyadditional malfunctions of the devices. Each removed collar wasreplaced the following morning before the goats were releasedfrom the pen.

In addition to the data collected by the collars, observationaldata were collected throughout the study. The goats were

Frontiers in Ecology and Evolution | www.frontiersin.org 3 March 2019 | Volume 7 | Article 73

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

430

431

432

433

434

435

436

437

438

439

440

441

442

443

444

445

446

447

448

449

450

451

452

453

454

455

456

O’Bryan et al. Contact Calls Facilitate Group Contraction

followed by two observers from a distance so as to not influencetheir behavior. In order to verify the audio data recorded by thecollars, all observations of vocalizations were documented as wellas the identity of the individual producing the vocalization. Onoccasions, the group would split into two subgroups (see sectionGPS Processing), and on these occasions, one observer wouldfollow each group. Occasionally, the movement of the goatshad to be influenced by the observers. These instances includedherding the goats out of the corral in the morning, herding themback toward the corral in the afternoon, and occasionally herdingthem back within the boundaries of the park, away from thelarger herd or if the herd began to split into more than twogroups. Data collected during these periods were not included inthe analysis (see section GPS Processing).

Data ProcessingGPS ProcessingData processing and analysis took place using R (version 3.1)and Matlab (version 2017b). Raw GPS data were manuallyerror corrected, first using the functions findLocErr, findLocNA,findTimeDup, and findMissing (swaRm package Garnier, 2016)to identify potential errors. The nature of the identified errorwas then assessed and manually corrected if it resulted froman obvious writing error, or otherwise omitted. Sequences ofNA’s were inserted when data were not available for a givenindividual for a given second. Sometimes missing values weredue to the collar having malfunctioned for a significant periodof time. However, sometimes missing values were simply dueto the GPS device temporarily losing satellite signal. When asequence of missing values occurred that was ≤10 s long, the“fixLocNA” function (swaRm package Garnier, 2016) was usedto linearly interpolate the missing values. Sequences of missingvalues longer than 10 s were not interpolated. In addition, thedata were cropped from the time the goats left the pen to the timethey arrived back at the pen. For a display of all GPS coveragetimes see Figure S1.

On five occasions, the herd was observed to split into twogroups which began to function separately from one anotherfor a period of time (mean: 171.7min, range: 59.6–300.7min).These periods were identified in the GPS data by calculating themean group member distance to centroid for each second andfinding the time periods in which this measure exceeded the thirdquartile of mean distances to centroid (87.5m) for study-widedata. All data falling within these intervals were omitted fromanalysis (858.6min). As mentioned above, we also omitted anytime periods when the group had to be herded (641.4min). Awindow of an additional±5min was added around each omittedperiod as a buffer. Thus, our analyses focus only on those timeperiods in which the herd was naturally functioning as a cohesivegroup (total observation time: 34.9 h).

Identification of Calls and Call BoutsIn order to determine the timing of call production and identityof the caller, we had to process the continuous audio streamscollected by our dataloggers (see Figure S1 for a display of audiocoverage times). We trained a support vector machine algorithm(Kernlab package v0.9-27, Karatzoglou et al., 2004) on the set

of loud contact calls that were documented in the observationaldata collected during the study (n = 248 vocalizations). Thesevocalizations were located in the audio data by visually andauditorally searching for calls around the times documentedin the observational data using Raven Pro (version 1.5). Oncethe calls were located, we annotated the start and end of eachcall. We also created an equivalent dataset of verified non-callsby randomly selecting sound clips from the same recordingsfrom which each verified call had been recorded. Each randomlyselected segment was manually verified to not contain any goatvocalizations by a group of trained listeners. We then extracteda 3 s clip for each manually verified call and non-call, whichconsisted of 1.5 s on either side of the center of the vocalizationor the center of the non-call sound clip. A pre-emphasis filterwas then applied to these segments which aids signal detectionby putting greater emphasis on high frequencies and less onlow frequencies, thus enabling greater disambiguation frombackground noise. We then calculated the amplitude envelopeand the fundamental frequency track for each clip and found theaverage values of each for both calls and non-calls. To examinethe ability of these twomeasures to enable differentiation betweencalls and non-calls, we fitted a Gaussian curve to the averageamplitude envelope and fundamental frequency track for callsand found the correlation with these curves for both callsand non-calls (Figure S2). We then trained a support vectormachine algorithm on half of the manually verified data set andvalidated it with the second half (Figure S3) which resulted inthe successful detection of 88.7% of observed calls. The machinelearning algorithmwas then applied to all audio recordings whichmoved a 3 s window along each recording in 1 s increments andassigned each interval a probability from 0 to 1 that that segmentcontained a call. We considered all segments with a probabilityof 0.5 or greater to be a call in order to reduce the number offalse negatives. However, since this threshold resulted in manyfalse positives, we proceeded to manually verify all positiveidentifications to determine whether they were goat vocalizationsor not and whether they were produced by the individual wearingthe collar or another individual. Since the microphone waspositioned very close to the throat of the wearer, the amplitude ofvocalizations produced by the wearer was much greater than callsproduced by nearby individuals. In our analysis we only considervocalizations produced by the individual wearing the collar. Thisresulted in a final dataset of 651 vocalizations, 244 of which fellduring the selected periods of movement data.

Rather than being produced intermittently throughout theday in a regular or random manner, call production in our studysystem was temporally clumped. Thus, we divided the calls intocalling bouts based on the amount of time between successivevocalizations by any group member. We only consideredintervals between calls that did not overlap with periods ofherding or group splits and where both calls occurred on thesame day. We fitted a Gaussian mixture model to the logtransformed time delays between successive calls to find thecutoff between bouts (Figure S4). Our inferred criterion was150 s which resulted in 57 call bouts. The number of calls in a boutranged from 1 to 57 with a median of 1 and mean (SD) of 4 (8.4).59.6% of bouts consisted of only one call while 40.4% consisted

Frontiers in Ecology and Evolution | www.frontiersin.org 4 March 2019 | Volume 7 | Article 73

457

458

459

460

461

462

463

464

465

466

467

468

469

470

471

472

473

474

475

476

477

478

479

480

481

482

483

484

485

486

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

505

506

507

508

509

510

511

512

513

514

515

516

517

518

519

520

521

522

523

524

525

526

527

528

529

530

531

532

533

534

535

536

537

538

539

540

541

542

543

544

545

546

547

548

549

550

551

552

553

554

555

556

557

558

559

560

561

562

563

564

565

566

567

568

569

570

O’Bryan et al. Contact Calls Facilitate Group Contraction

of more than one call. The duration of call bouts ranged from 0 s(for solitary calls) to 685 s, with a mean (SE) of 61.25 (17.5) s.

Assessing Synchrony of GPS and Audio DataTo assess the alignment of the audio data with the GPS data, weexamined events in which the goats visited a watering trough,since these events were both visually apparent when overlayingthe GPS trajectories over a map of the park and also acousticallyapparent due to the distinctive sound of the animal drinking(Lynch et al., 2013), which is amplified by the positioning of themicrophone at the animal’s throat. Individuals arriving and thendeparting from a watering trough displayed a sharp trajectorychange that aligned with the position of the trough.We identifiedthe time of arrival, to the second, at the change point in the GPSdata as well as the time of initiation of drinking sounds in theaudio data, with the aid of spectrograms generated in Raven Pro.Based on the GPS-derived start time of the drinking bouts and theamount of time known to have elapsed in the audio data up to theinitiation of drinking sounds, we calculated the GPS-derived starttime of the audio recording for that day. When we compared theset of audio start times calculated from the time spoken at thebeginning of the day with the audio start times inferred from theGPS data (n= 78) we found a range of time differences from−25to 34 s with a median difference of 0 s and a mean (SD) differenceof−0.1 (6.7) s. Based on this assessment, we concluded that therewas no need to shift the times relative to one another. We usedthe audio recording start times derived from the time spoken atthe beginning of each day for calculating the times of any callsthat occurred that day, since these times were available for themajority of recordings. If a given start time was not availableon a given day, for instance, due to a delay in the initiation ofrecording, we did not include any acoustic data from that devicefor that day in our analysis (see Figure S1 for an overview of allaudio data available for analysis).

Assessing Group Spread at the Time ofCall ProductionTo establish whether the goats’ contact calls do occur duringperiods of greater than normal group spread, we first assessedthe group’s spread at times of call production and comparedthese measures to randomly selected time points. We usedthe maximum individual distance to the group’s centroid (theaverage geolocation of all group members) as our measure ofgroup spread. We focused on the first call of each bout (n =57) and found the maximum distance to centroid at the timeof each call. We then randomly selected 57 time points fromthe dataset (without replacement) and repeated this selection10,000 times, finding the maximum distance to centroid at eachrandomly selected time. We used the “density” function from thebase package (R Core Team, 2016) to calculate the kernel densityQ12

estimates for the real data and each randomized selection of 57data points. The density estimates were calculated for a range ofmaximum distances to centroid from 20 to 100m by intervals of1m. For the randomized datasets, we then calculated the meandensity estimate for each distance across all 10,000 iterations andfound confidence intervals by calculating the 0.025 and 0.975quantiles for these data using the “quantile” function from the

stats package (R Core Team, 2016). The density functions for the Q12

real and randomized datasets were compared by examining areasof overlap between the real dataset and the confidence intervalsof the randomized dataset.

Investigating Caller Distances to Centroidat the Time of Call ProductionIn addition to examining the maximum distance to centroid atthe time of call production, we also examined the relationshipbetween call production and the caller’s own distance to thecentroid of the group at the time of its call. We compared thedistribution of distances to the centroid at the time of the first callin the bout for the individual who gave that call (n= 50 calls withGPS data for the caller) to the distribution of distances expectedif call production was randomly distributed between all groupmembers. This randomized dataset was generated by reassigningcaller identities for the selected calls (for instance, randomlyassigning all calls produced by individual #1 to individual #4 andall calls by individual #4 to individual #7) across 10,000 iterations,thereby breaking any correlation between call production andthe caller’s distance to centroid. The distances to centroid ofreassigned “callers” were then calculated for each iteration at eachcall time. From these data, we then calculated kernel densityestimates for distances to centroid ranging from 0 to 150m byintervals of 1m for both the real and randomized datasets. Forthe randomized dataset, we then found themean density estimateand confidence interval for each distance across all permutations.As in the previous analysis, the density functions for the realand randomized datasets were compared by examining areas ofoverlap between the real dataset and the confidence intervals ofthe randomized dataset.

In addition to distance to centroid, we also examined therelationship between calling and the caller’s ordinal distance fromthe centroid (n = 50 calls with GPS data for the caller). Tocalculate this measure, we ranked each group member basedon its distance to the centroid at the time of the first call in about, starting with the closest individual to the centroid. Sincewe did not always have GPS data for the same number of groupmembers throughout the dataset (Figure S1), we calculated eachindividual’s relative rank by dividing its ordinal rank by thenumber of individuals for whom we had GPS data for that timepoint. Thus, a relative distance of 1 means that the individual isthe furthest from the centroid (of those for whom we have data);with values close to zero indicating that the individual is one ofthe closest to the centroid. We then compared the distributionof observed ranked distances of callers to expectations based onthe ranks of randomized caller identities generated in the samemanner as the previous two analyses. Density estimates werecalculated for relative ranked distances to centroid within a rangefrom 0 to 1 by intervals of 0.05.

Determining the Relationship Between CallProduction and Group ContractionWe assessed changes in the group’s spread during time periodsbefore and after call production to establish whether callproduction is associated with the initiation of group contraction.

Frontiers in Ecology and Evolution | www.frontiersin.org 5 March 2019 | Volume 7 | Article 73

571

572

573

574

575

576

577

578

579

580

581

582

583

584

585

586

587

588

589

590

591

592

593

594

595

596

597

598

599

600

601

602

603

604

605

606

607

608

609

610

611

612

613

614

615

616

617

618

619

620

621

622

623

624

625

626

627

628

629

630

631

632

633

634

635

636

637

638

639

640

641

642

643

644

645

646

647

648

649

650

651

652

653

654

655

656

657

658

659

660

661

662

663

664

665

666

667

668

669

670

671

672

673

674

675

676

677

678

679

680

681

682

683

684

O’Bryan et al. Contact Calls Facilitate Group Contraction

We focused on calling bouts where an interval of 20min beforeand after the first call in the bout did not overlap with anyperiods corresponding to herding or group splits (n = 38 callingbouts). We found the maximum group member distance tocentroid every 15 s within the +/- 20min interval (40min total).We then subtracted the maximum distance to centroid at thetime of the call from the maximum distance to centroid at eachtime point and found the mean and 95% confidence interval forthese differences for all time points. Time points with confidenceintervals that did not include zero were considered to have groupspreads that were significantly different from that seen at the timeof the call. For comparison, we repeated the analysis by randomlyselecting 38 time points from the GPS data to use as focal pointsinstead of calls, ensuring that the +/- 20min window of eachpoint did not overlap with herding or group splits.

Investigating Calling Patterns Within BoutsIn order to assess whether calls by distant individuals elicit vocalresponses by those closer to the center of the group, as predictedby Hypothesis 1, we first calculated the total number of uniquecallers within each calling bout and determined the frequency ofmulti-caller bouts. For bouts with more than one caller, we firstcalculated the distance between subsequent callers at the time ofthe second caller’s first call. We calculated this distance usingthe “nnd” function from the swaRm package (Garnier, 2016)which calculates the distance between nearest neighbors giventheir geolocations at given timepoints. Since the first and secondcallers were not necessarily nearest neighbors, we reduced ourdataset to include only the geolocations of the first and secondcallers for each included timepoint before running the function.We then calculated the distance to centroid of both callers andused a chi-squared test to determine whether the second callerwas significantly more likely to be closer to, or further from,the centroid than the first caller and a Wilcoxon Rank-sum Testto test whether the distance to centroid of the second callerwas significantly different from the distance to centroid of thefirst caller.

Examining Velocity of Callers and CentroidWe measured both the caller’s velocity relative to the centroidof the remaining group members at times surrounding callproduction as well as the centroid’s velocity relative to the caller.We selected the first call of all bouts where an interval of20min before and after the call did not overlap with any periodscorresponding to herding or group splits and in which we hadGPS data for the first caller of the bout (n = 33). We found thegeolocation of the caller for each second during the time windowfor each call. We also calculated the geolocation of the groupcentroid of the remaining group members during this sameperiod. The position time series for the caller and the centroidwere smoothed over a moving 20 s window using the Matlabbuilt-in function “smooth” and a lowess local regression method.These time series were numerically differentiated to find velocityvectors for the caller and the group centroid. We then computedthe scalar projection of each caller’s velocity onto the unit vectororiginating at its position in the direction of the group centroid’sposition sc→g (where the “g” subscript denotes the group centroid

and the “c” denotes the caller), and the analogous quantity for thecentroid sg→c. These quantities are defined as:

Sc→g =Exg − Exc

||Exg − Exc||· Evc and Sg→c =

Exc − Exg

||Exc − Exg ||· Evg

and a schematic illustrating their physical meaning is shownin Figure 1. xg and vg represent the position and velocity ofthe group centroid, respectively, while xc and vc represent theposition and velocity of the caller. sg→c represents the velocityof the group centroid relative to the caller and sc→g representsthe velocity of the caller relative to the group centroid. Interms of the goats’ relative motion, sc→g is positive when thecaller moves toward the group centroid, and it is negativewhen it moves away from the group centroid. Similarly, sg→c

is positive when the group centroid moves toward the callerand negative otherwise. Therefore, these quantities capture therelative movements that could result in the caller and remaininggroup members rejoining, namely, the caller approaching thegroup centroid or the group centroid approaching the caller.

Across bouts, we found the mean (and 95% confidenceinterval) for the velocity of the caller to the group centroidand velocity of the group centroid to the caller for timepoints occurring every 15 s within each call’s ±20min interval.Furthermore, we found the mean difference (and 95% confidenceinterval) in the caller’s velocity to group centroid and groupcentroid’s velocity to caller relative to that displayed by eachat the time of the call. Lastly, we used the “ma” functionfrom the forecast package (Hyndman, 2017) to display a simplemoving average of these velocities with a window spanning from2.5min before to 2.5min after each timepoint. The Vocal BeaconHypothesis 1a predicts increased caller movement toward thegroup centroid following call production (presumably due to asubsequent call by a core group member). The Vocal BeaconHypothesis 1b predicts increased group centroid movement

FIGURE 1 | Schematic showing how the velocities of the caller toward the Q4

Q5group centroid sc→g and group centroid toward the caller sg→c are

calculated according to the positions of the group centroid (xg) and caller (xc)

and the velocities of the group centroid (vg) and caller (vg) at each timepoint.

Frontiers in Ecology and Evolution | www.frontiersin.org 6 March 2019 | Volume 7 | Article 73

685

686

687

688

689

690

691

692

693

694

695

696

697

698

699

700

701

702

703

704

705

706

707

708

709

710

711

712

713

714

715

716

717

718

719

720

721

722

723

724

725

726

727

728

729

730

731

732

733

734

735

736

737

738

739

740

741

742

743

744

745

746

747

748

749

750

751

752

753

754

755

756

757

758

759

760

761

762

763

764

765

766

767

768

769

770

771

772

773

774

775

776

777

778

779

780

781

782

783

784

785

786

787

788

789

790

791

792

793

794

795

796

797

798

O’Bryan et al. Contact Calls Facilitate Group Contraction

FIGURE 2 | Distributions of (A) maximum distance to centroid, (C) distances to centroid of callers, and (E) callers’ ranked distances from the centroid (1 = furthest) at

the time of the first call in a bout for real and randomized datasets. (A) nreal = 57 calls, nrand = 10,000 iterations of 57 randomly selected time points, (C,E) nreal = 50

calls, nrand = 10,000 permutations of caller identity for 50 calls. (B,D,F) Represent differences between density functions for real minus randomized datasets.

toward the caller after call production. The Straggler Hypothesispredicts caller movement toward the group centroid and areduction in centroid movement away from the caller followingcall production (Table 1).

RESULTS

Compared to randomly selected time points, the likelihoodof call production by any group member increased withincreasing maximum distance to centroid, reaching significancefor maximum distances to the group centroid >57m(Figures 2A, B). The likelihood that a given individual

would produce the first call in a bout also increased withthat individual’s own distance to the group centroid, withindividuals being significantly less likely to call when they werecloser to the group centroid (within 28m) and significantly morelikely to call when they were far from the group centroid (>95m;Figures 2C, D). Similarly, a group member was significantlymore likely to call when it was the furthest individual from thegroup centroid (normalized ranked distance to group centroid>0.95; Figures 2E, F).

Calls were produced by 13 out of the 16 group members witha mean (SD) of 3.6 (3.4) bouts initiated by each individual anda mean (SD) of 14.3 (18.5) calls per individual. The number ofunique callers in a bout (n = 57 bouts) ranged from 1 to 5 with

Frontiers in Ecology and Evolution | www.frontiersin.org 7 March 2019 | Volume 7 | Article 73

799

800

801

802

803

804

805

806

807

808

809

810

811

812

813

814

815

816

817

818

819

820

821

822

823

824

825

826

827

828

829

830

831

832

833

834

835

836

837

838

839

840

841

842

843

844

845

846

847

848

849

850

851

852

853

854

855

856

857

858

859

860

861

862

863

864

865

866

867

868

869

870

871

872

873

874

875

876

877

878

879

880

881

882

883

884

885

886

887

888

889

890

891

892

893

894

895

896

897

898

899

900

901

902

903

904

905

906

907

908

909

910

911

912

O’Bryan et al. Contact Calls Facilitate Group Contraction

a median of 1 and mean (SD) of 1.4 (0.9). 78.9% of call boutswere by a single caller while 21.1% were by multiple callers. Forbouts in which there was more than one caller (n = 9 out of 50with GPS data for both callers), the distance between the first andsecond caller at the time of the second caller’s first call rangedfrom 3.9 to 100.8m with a median of 42.2 and mean (SD) of42.0 (30.0) m. Our prediction for the Vocal Beacon Hypothesis1a, that the second caller would be closer to the centroid than thefirst (Table 1), was not supported. The second caller in the boutwas equally likely to be closer to, or farther from, the centroidcompared to the first caller (X2 = 0.11, df = 1, p-value = 0.74).The second caller was closer in 55.6% of the bouts and farther in44.4%. Furthermore, the distance to centroid of the second callerrelative to that of the first caller was not significantly differentfrom zero (Wilcoxon signed-rank test: V = 39, p-value = 0.93)and ranged from −38.0 to 36.9m with a median of −2.1 andmean (SD) of−3.8 (27.5) m.

The mean (SD) maximum groupmember distance to centroidat the time of the first call in a bout was 65.1 (37.6) m.When considering a ±20min time window before and after callproduction, calls fell in between periods of group expansionand contraction (Figure 3). The mean maximum distance tocentroid was significantly greater surrounding the time of callproduction (from 60 s before the call to 660 s after the call)compared all other times within the ±20min time window. Themean maximum distance to centroid was never significantlygreater than that observed at the time of the call. The mean (SD)maximum distance to centroid observed at the times of randomlyselected focal points, rather than real calls, was 36.7 (29.8) meters.This level of group spread was not significantly different frommean distances observed during any times within the ±20mintime interval (Figure S5). These patterns of group spread canbe examined visually on an event-by-event basis in Figure 4. Ascan be seen, call bouts tended to fall near times when maximumdistance to centroid had peaked.

The mean caller velocity relative to the centroid wasintermittently significantly positive (i.e., the caller moved towardthe centroid) surrounding the time of the call (from 300 sprior to call production to 45 s after call production). Afterthis period, the mean caller velocity had a negative trend andeventually became intermittently significantly negative (i.e., thecaller moved away from the centroid) between 615 and 975 s aftercall production (Figure 5A). The mean caller velocity relativeto the group centroid was never significantly greater than thatdisplayed at the time of the call, meaning that the caller wasnever moving significantly more toward the group centroidthan it was at the time of the call. The caller’s velocity relativeto the group centroid was intermittently significantly lowerthan that displayed at the time of the call both prior to andfollowing call production (prior to 405 s before call productionand following 450 s after call production; Figure 5C). The meangroup centroid velocity relative to the caller was significantlynegative (i.e., the centroid moved away from the caller) priorto call production (from 360 to 45 s prior to call production).Following this period, the mean group centroid velocity hada positive trend and eventually became significantly positive(i.e., the group centroid moved toward the caller) between 780

FIGURE 3 | The mean difference (with 95% confidence interval) in the

maximum distance to centroid at each time point in the ±20min interval

relative to the maximum distance to centroid at the time of the call (n = 38

calls). Relative time = 0 indicates the time of the calls.

and 960 s after call production (Figure 5B). The mean groupcentroid velocity relative to the caller was never significantlylower than that displayed at the time of the call, meaning thatthe group centroid never moved significantly more away fromthe caller than it did at the time of the call. The group centroid’svelocity relative to the caller was intermittently significantlygreater than that displayed at the time of the call both priorto and following call production (prior to 810 s before callproduction and following 240 s after call production; Figure 5D).These findings, of reduced group centroid movement away fromthe caller and intermittent caller movement toward the groupcentroid following call production, best support the predictionsof the Straggler Hypothesis (Table 1).

While we found that individuals that were distant from thecentroid were most likely to call during periods of wide groupspread (above), many calls were produced by individuals thatwere close to the centroid due to the greater number of groupmembers in these positions (Figure 1B). Thus, we examined Q13

whether calls might function differently when produced byindividuals distant from the group centroid compared to thoseproduced by individuals close to the centroid at the time of callproduction. We conducted a k-means cluster analysis on callbouts using the “kmeans” function from the stats package (R CoreTeam, 2016). The clustering variables we chose were the caller’s Q12

distance to centroid and the caller’s relative distance to centroidat the time of call production. We chose these variables since,together, they provide information on whether an individual maybe distant from the group (far from the group centroid in bothrank and distance) or more central (closer to the centroid in bothrank and distance). Before including these variables in the model,we scaled them to have a mean of 0 and standard deviation of1 using the “scale” function (R Core Team, 2016). We set the Q12

number of clusters to 2 in order to differentiate those individualsthat were distant from the group centroid from those that werecloser to the centroid at the time of call production. We ranthe analysis on all bouts in which we had GPS data for the firstcaller (n= 50).

Frontiers in Ecology and Evolution | www.frontiersin.org 8 March 2019 | Volume 7 | Article 73

913

914

915

916

917

918

919

920

921

922

923

924

925

926

927

928

929

930

931

932

933

934

935

936

937

938

939

940

941

942

943

944

945

946

947

948

949

950

951

952

953

954

955

956

957

958

959

960

961

962

963

964

965

966

967

968

969

970

971

972

973

974

975

976

977

978

979

980

981

982

983

984

985

986

987

988

989

990

991

992

993

994

995

996

997

998

999

1000

1001

1002

1003

1004

1005

1006

1007

1008

1009

1010

1011

1012

1013

1014

1015

1016

1017

1018

1019

1020

1021

1022

1023

1024

1025

1026

O’Bryan et al. Contact Calls Facilitate Group Contraction

FIGURE 4 | Maximum distance to centroid over time (hh:mm) for 2 days during the study period: (A) 09/10/2018 and (B) 09/15/2018. Vertical dotted lines indicate

the times of the first call in each call bout during these periods.

The results of the k-means cluster analysis are displayed inFigure 6. Cluster 1 consisted of 23 bouts and had a non-scaledmean (SD) caller distance to group centroid of 74.9 (37.9) mand a relative distance to group centroid of 0.90 (0.13) (1 =furthest from centroid). Cluster 2 consisted of 27 bouts and hada mean (SD) caller distance to group centroid of 13.2 (9.2) mand a relative distance to group centroid of 0.36 (0.21). Themaximum distance to group centroid of any group member atthe time of the call was significantly greater in cluster 1 thancluster 2 (Wilcox.test; V = 490, p < 0.001). The mean (SD)maximum group member distance to the group centroid incluster 1 was 98.2 (71.7) m while it was 47.1 (28.0) in cluster2. Due to this difference, we re-plotted the maximum distanceto group centroid over the time surrounding call productionin order to examine whether bouts in both clusters correspondwith peaks in group spread (Figure S6). We found that, forboth clusters, mean group spread was never significantly greaterthan it was at the time of the call. However, there was a morepronounced group contraction surrounding events in cluster 1,with significantly lower group spread for all times before 105 sprior to the call, intermittently following 30 s after the call, andfor all times following 525 s after call production (Figure S6A).Events in cluster 2 had significantly lower group spread between450 and 195 s prior to call production and then again following1,140 s after call production (Figure S6B).

We also re-examined counter-calling behavior within the twoclusters predicted by the k-means cluster analysis in order to

determine whether there was evidence for differences in behaviorbetween them. Cluster 1 had 5 multi-caller bouts out of 23 totalbouts (21.7%) while cluster 2 had 7 multi-caller bouts out of27 (25.9%). Due to the low sample sizes within each cluster (4events with GPS data for both callers in cluster 1 and 5 in cluster2), we simply calculated summary statistics in order to examinepotential trends. In cluster 1, the distances between the first andsecond callers ranged from 34.6 to 70.2m with a median of 42.4andmean (SD) of 47.4 (15.6)m. In 1 out of the 4 bouts, the secondcaller was further from the centroid than the first, while in 3 outof 4 it was closer. The distance to centroid of the second callerrelative to the first ranged from −38.1 to 25m with a median of−32.4 and mean (SD) of −19.4 (29.9). In cluster 2, the distancesbetween the first and second callers ranged from 3.9 to 100.8mwith a median of 30.3 and a mean (SD) of 37.7 (39.6) m. In 3 outof the 5 bouts, the second caller was further from the centroidthan the first, while in 2 out of 5 it was closer. The distancesto centroid of the second caller relative to the first ranged from−13.4 to 36.9mwith a median of 1.5 andmean (SD) of 8.7 (20.0).These results suggest that there was little difference betweenclusters in the tendency for counter calling to occur. There wasalso little evidence for differences in the distances between thefirst and second callers. However, there was some tendency forthe second caller to be closer to the centroid than the first callerin the first cluster and vice versa for the second cluster.

Finally, we re-ran our velocity analysis to examinelongitudinal changes in the callers’ and centroids’ velocities

Frontiers in Ecology and Evolution | www.frontiersin.org 9 March 2019 | Volume 7 | Article 73

1027

1028

1029

1030

1031

1032

1033

1034

1035

1036

1037

1038

1039

1040

1041

1042

1043

1044

1045

1046

1047

1048

1049

1050

1051

1052

1053

1054

1055

1056

1057

1058

1059

1060

1061

1062

1063

1064

1065

1066

1067

1068

1069

1070

1071

1072

1073

1074

1075

1076

1077

1078

1079

1080

1081

1082

1083

1084

1085

1086

1087

1088

1089

1090

1091

1092

1093

1094

1095

1096

1097

1098

1099

1100

1101

1102

1103

1104

1105

1106

1107

1108

1109

1110

1111

1112

1113

1114

1115

1116

1117

1118

1119

1120

1121

1122

1123

1124

1125

1126

1127

1128

1129

1130

1131

1132

1133

1134

1135

1136

1137

1138

1139

1140

O’Bryan et al. Contact Calls Facilitate Group Contraction

FIGURE 5 | The upper graphs show the mean velocity with 95% confidence interval of the (A) caller toward the group centroid and (B) group centroid toward the

caller, from 20min before to 20min after the production of the first call (relative time = 0) in a call bout (n = 33 calling bouts). The lower graphs show the mean

difference in velocity (with 95% confidence interval) of the (C) caller toward the group centroid and (D) group centroid toward the caller compared to their own

velocities at the time of the call (relative time = 0). The purple line represents the simple moving average of each plotted velocity calculated with a window of 301 s

centered on each datapoint.

FIGURE 6 | Calling bouts plotted according to the scaled relative distance to

centroid of the caller (1 = furthest individual) and the scaled distance to

centroid of the caller (meters) at the time of the first call in the bout (n = 50

calling bouts). Points are colored according to the cluster in which they were

assigned by the cluster analysis. The asterisks represent the center of each

cluster.

relative to one another for the two clusters predicted by thek-means cluster analysis. As before, we restricted analysis tothose bouts in which the ±20min window did not overlap withany periods of herding or group splits (n = 33, cluster 1 = 15,cluster 2 = 18). We found that events from cluster 1, where

the caller was far from the group centroid, were associatedwith significantly positive mean caller velocity relative to thecentroid surrounding the time of call production (from 15 sbefore the call to 105 s after the call; Figure 7A). The meancaller velocity was never significantly greater than at the timeof the call, but it was intermittently significantly lower both

prior to and following call production (prior to 60 s before

call production and following 270 s after call production;Figures 7C). Furthermore, these events were associated withgroup centroids that displayed significantly negative meanvelocity relative to the caller surrounding call production (forthe majority of timepoints from 330 s prior to the call to 135 safter the call; Figure 7B). Following call production, therewas an increase in mean group centroid velocity relative tothe caller which eventually became intermittently significantlypositive (between 825 and 945 s after call production). The groupcentroid’s mean velocity relative to the caller was at its mostnegative 45 s after the first call and became significantly greaterthan it was at the time of call production beginning 225 s afterthe call (Figure 7D). These results, of reduced group centroidmovement away from the caller and strong caller movementtoward the group centroid following call production, support thepredictions of the Straggler Hypothesis (Table 1).

Events in the second cluster, where the caller was close tothe group centroid, were associated with a neutral caller velocityrelative to the group centroid prior to call production and

Frontiers in Ecology and Evolution | www.frontiersin.org 10 March 2019 | Volume 7 | Article 73

1141

1142

1143

1144

1145

1146

1147

1148

1149

1150

1151

1152

1153

1154

1155

1156

1157

1158

1159

1160

1161

1162

1163

1164

1165

1166

1167

1168

1169

1170

1171

1172

1173

1174

1175

1176

1177

1178

1179

1180

1181

1182

1183

1184

1185

1186

1187

1188

1189

1190

1191

1192

1193

1194

1195

1196

1197

1198

1199

1200

1201

1202

1203

1204

1205

1206

1207

1208

1209

1210

1211

1212

1213

1214

1215

1216

1217

1218

1219

1220

1221

1222

1223

1224

1225

1226

1227

1228

1229

1230

1231

1232

1233

1234

1235

1236

1237

1238

1239

1240

1241

1242

1243

1244

1245

1246

1247

1248

1249

1250

1251

1252

1253

1254

O’Bryan et al. Contact Calls Facilitate Group Contraction

FIGURE 7 | The upper graphs show the mean velocity with 95% confidence interval of the (A) caller toward the group centroid and (B) group centroid toward the

caller, from 20min before to 20min after the production of the first call (relative time = 0) in a call bout. Data are presented for bouts assigned to cluster 1 (n = 15

calling bouts), which correspond with bouts in which callers are far from the centroid at the time of the call. The lower graphs show the mean difference in velocity

(with 95% confidence interval) of the (C) caller toward the group centroid and (D) group centroid toward the caller compared to their own velocities at the time of the

call (relative time = 0). The purple line represents the simple moving average of each plotted velocity calculated with a window of 301 s centered on each datapoint.

an intermittently significantly negative velocity beginning 45 safter call production (Figure 8A). However, for the majority oftimepoints, the mean caller velocity relative to the centroid wasnot significantly different from that displayed at the time of thecall (Figure 8C). For these events, the group centroid had anintermittently significantly negative mean velocity relative to thecaller prior to call production (between 375 and 165 s prior to callproduction) followed by an intermittently significantly positivemean velocity after call production (from 45 to 75 s following callproduction, and following 450 s after call production; Figure 8B).However, the group centroid’s mean velocity relative to the callerwas never significantly different from that displayed at the timeof the call (Figure 8D). These findings, of the caller moving awayfrom the group centroid and the group centroid moving towardthe caller following call production, support some predictions ofthe Vocal BeaconHypothesis (Table 1). However, our findings donot support the prediction that the group centroid’s movementtoward the caller following call production was significantlygreater than its movement toward the caller at the time of the call.

DISCUSSION

The aim of our study was to determine the function of goats’loud contact calls and explore how they might facilitate groupcontraction during periods of wide group spread. We first testedwhether these calls were indeed associated with periods of wide

group spread while also gaining insight into the positions ofgroup members producing these calls. We found evidence thatthese vocalizations were associated with contexts in which groupspread was especially wide, with group members that werefurther from the centroid being more likely to produce thecalls. These findings are consistent with the characterization ofsome loud contact calls, such as those produced by white-facedcapuchins (Gros-Louis et al., 2008), as “lost calls” and suggeststhat they do play a role in facilitating group contraction. Indeed,we found that the first call in a bout fell at an intermediarypoint between group expansion and contraction. From the timeof the call until a little more than 10min after the call, groupspread was significantly greater than it was at all other timessurrounding call production, but mean group spread was neversignificantly greater than at the time of the call. Randomlyselected time points did not fall at such peaks in group spreadsuggesting that it is not simply coincidental that calls fall duringthese periods. It also suggests that it is unlikely that the goatsproduced calls for reasons unrelated to the spread of their currentgroup. Overall, these results support our central hypothesis thatcall production plays a role in halting group expansion andprompting group members to adjust their behavior in order toinitiate group contraction.

Hypothesis 1a predicted that antiphonal call productionbetween distant and more central group members is necessary toenable group contraction. We found that the majority of calling

Frontiers in Ecology and Evolution | www.frontiersin.org 11 March 2019 | Volume 7 | Article 73

1255

1256

1257

1258

1259

1260

1261

1262

1263

1264

1265

1266

1267

1268

1269

1270

1271

1272

1273

1274

1275

1276

1277

1278

1279

1280

1281

1282

1283

1284

1285

1286

1287

1288

1289

1290

1291

1292

1293

1294

1295

1296

1297

1298

1299

1300

1301

1302

1303

1304

1305

1306

1307

1308

1309

1310

1311

1312

1313

1314

1315

1316

1317

1318

1319

1320

1321

1322

1323

1324

1325

1326

1327

1328

1329

1330

1331

1332

1333

1334

1335

1336

1337

1338

1339

1340

1341

1342

1343

1344

1345

1346

1347

1348

1349

1350

1351

1352

1353

1354

1355

1356

1357

1358

1359

1360

1361

1362

1363

1364

1365

1366

1367

1368

O’Bryan et al. Contact Calls Facilitate Group Contraction

FIGURE 8 | The upper graphs show the mean velocity with 95% confidence interval of the (A) caller toward the group centroid and (B) group centroid toward the

caller, from 20min before to 20min after the production of the first call (relative time = 0) in a call bout. Data are presented for bouts assigned to cluster 2 (n = 18

calling bouts), which correspond with bouts in which callers are close to the centroid at the time of the call. The lower graphs show the mean difference in velocity

(with 95% confidence interval) of the (C) caller toward the group centroid and (D) group centroid toward the caller compared to their own velocities at the time of the

call (relative time = 0). The purple line represents the simple moving average of each plotted velocity calculated with a window of 301 s centered on each datapoint.

bouts consisted of calls by only one caller. Furthermore, whenmulticaller bouts did occur, the second caller was equally likelyto be closer to, or further from, the group centroid comparedto the first caller, with only a small mean difference in distanceto group centroid between successive callers. Therefore, thesefindings do not support the prediction that call production elicitsantiphonal responses by more central group members, whichis consistent with others studies that have reported infrequentvocal responses to the loud contact calls of others (Cheney et al.,1996; Rendall et al., 2000; Digweed et al., 2007). While it ispossible that the low number of antiphonal responses in ourstudy system represents a failure to facilitate group contraction,this interpretation is not consistent with our finding that callproduction was correlated with the cessation of group expansionand subsequent group contraction.

When investigating the mechanisms that may facilitate groupcontraction, we found that the mean pattern across all bouts bestcorresponded with the predictions of the Straggler Hypothesis.We found evidence that the group centroid was moving awayfrom the caller prior to call production but reduced its movementaway from the caller following call production. We also foundsome evidence that the caller was moving toward the groupcentroid around the time of the call, but this movementwas less consistent. This inconsistency is likely due to ourfinding that callers that were far from the group centroid andcallers that were close to the group centroid displayed opposite

movement patterns relative to the group centroid around thetime of the call. We gained a more clear understanding ofthe mechanisms facilitating group contraction after dividingcalling bouts according to the initial caller’s distance from thecentroid. For bouts in which the caller was distant from thecentroid, we found that call production was associated with acessation of group centroid movement away from the caller,similar to the pattern observed in the data aggregated acrossall bouts. These bouts were also associated with a strong peakin caller movement toward the centroid at the time of callproduction which was not observed in the data aggregated acrossall bouts. These findings more clearly support the predictionsof the Straggler Hypothesis and suggest that callers far fromthe centroid can use loud contact calls to slow the groupcentroid’s movement away from themselves to enable them tocatch up more readily. This hypothesis is further supported byour finding that these bouts were associated with very widegroup spread and also with strong group contraction followingcall production. While we did not find support for Hypothesis1a, that counter calling plays a key role in group contraction,we did observe some instances of counter calling behavior inthese bouts. Furthermore, we found some evidence that thesecond caller tended to be more central than the first. Thiscounter calling behavior has the potential to play a role whenlagging group members are not aware of the group centroid’slocation and therefore cannot move toward it. In order to test this

Frontiers in Ecology and Evolution | www.frontiersin.org 12 March 2019 | Volume 7 | Article 73

1369

1370

1371

1372

1373

1374

1375

1376

1377

1378

1379

1380

1381

1382

1383

1384

1385

1386

1387

1388

1389

1390

1391

1392

1393

1394

1395

1396

1397

1398

1399

1400

1401

1402

1403

1404

1405

1406

1407

1408

1409

1410

1411

1412

1413

1414

1415

1416

1417

1418

1419

1420

1421

1422

1423

1424

1425

1426

1427

1428

1429

1430

1431

1432

1433

1434

1435

1436

1437

1438

1439

1440

1441

1442

1443

1444

1445

1446

1447

1448

1449

1450

1451

1452

1453

1454

1455

1456

1457

1458

1459

1460

1461

1462

1463

1464

1465

1466

1467

1468

1469

1470

1471

1472

1473

1474

1475

1476

1477

1478

1479

1480

1481

1482

O’Bryan et al. Contact Calls Facilitate Group Contraction

hypothesis, more data must be collected on instances of countercalling behavior.

For bouts in which the caller was close to the groupcentroid, we found that call production was associated with atransition period during which the caller switched from movingneutrally relative to the centroid to moving significantly awayfrom it. Call production was also associated with a transitionperiod during which the group centroid switched from movingsignificantly away from the caller to moving significantly towardit. However, group movement toward the caller followingcall production was not significantly greater than the group’smovement relative to the caller at the time of the call. Thesefindings support some predictions of Hypothesis 1b and suggestthat call production by central group members can boost thegroup centroid’s velocity toward itself, perhaps by attractingsome, but not all, group members. This mechanism couldfunction by drawing more distant group members, or thosemoving away from the centroid, toward the caller and the centerof the group, similar to the finding that central white-facedcapuchins sometimes produce contact calls when other groupmembers have become distant (Gros-Louis et al., 2008). Sinceevents in this cluster were associated with both lesser groupspread at the time of call production and lesser group contractionfollowing call production, as compared to the other cluster ofevents, it is possible that these calls serve a preventative function.Furthermore, as these calls were associated with the initiationof caller movement away from the group centroid, they couldpotentially function by alerting any inattentive group membersto the initiation of group movement. For instance, Trillmichet al. (2014) found that Sifakas (Propithecus verreauxi) frequentlyproduce contact calls prior to leaping, which the authors proposecould signal that “displacement is imminent.”We found a similarfrequency of counter calling behavior in these bouts comparedto those associated with distant initial callers. However, in thesebouts, the second caller tended to be more distant from thecentroid than the first. For these events, it is possible that twoor more callers, and any surrounding group members, couldsimultaneously move toward one another. Alternatively, sincethe initial callers in these bouts tended to move away from thecentroid following call production, these events could be moreprone to resulting in group splits. For instance, Gall and Manser(2017) found that simultaneous playbacks of quiet contact callsfrom both the front and back of meerkat groups expanded groupspread along that axis. Again, in order to test these hypotheses,more data needs to be collected on events in which countercalling occurs.

Overall, regardless of the caller’s location, our study providesevidence that loud contact call production can facilitate groupcontraction, by either reducing centroid velocity away fromlagging callers or boosting centroid velocity toward centralcallers. Our study is one of the first to continuously andsimultaneously monitor both the positions and vocalizationsof the majority of group members in a free-ranging animalgroup (For a similar methodological approach see Leighty et al.,2008). This methodology enabled us to detect changes in groupmovement patterns surrounding the time of call productionwhich may have been missed otherwise. Many studies of loud

contact calls have reported subtle responses or a lack of responsealtogether. For instance, a study of white-faced capuchinsreported that, when observing a given group member who waswithin hearing range of a calling distant individual, 77% of thetime it “ignored” the calls (Gros-Louis et al., 2008). Furthermore,a study of chacma baboons found that in 56% of trials involvingthe acoustic playback of a group member’s contact barks, thesubject simply briefly oriented toward the speaker (Cheney et al.,1996). These responses have led to the assumption that calls thatare not answered vocally may not aid the caller in returning tothe group (Digweed et al., 2007). Our study indicates that, indomesticated goats, unanswered calls can have the subtle effect ofadjusting the group centroid’s movement relative to the caller inways that could prevent separation or aid reunion. Asmany socialbirds and mammals produce loud contact calls when separatedfrom their group and/or specific individuals (Reviewed in Kondoand Watanabe, 2009), applying our methodology to studies ofcontact calling behavior in these species has significant potentialto improve understanding of the mechanisms underlying theirrole in the facilitation of group contraction.

Given the direct influence of calling on group dynamics inour study, further work would benefit from developing modelsof collective movement which can estimate the “forces” imposedby calls on individual group members based on their relativepositions, velocities, behaviors, and even relationships (Farineet al., 2016). For instance, a study of contact calls producedby captive elephants (Loxodonta africana) coming together afterperiods of fission found that vocal production by one groupmember was correlated with the movement of other groupmembers toward the caller, particularly if a given group membertended to spend more time in close proximity to the initial callerand also if that group member produced an antiphonal call inresponse to the initial call (Leighty et al., 2008). Future modelscould even account for the sensory abilities of group members(Strandburg-Peshkin et al., 2013) and the role the landscapemay play in both influencing movement patterns and interferingwith interindividual contact (Strandburg-Peshkin et al., 2017).Applying these analytical techniques to positional, behavioral andcommunicative data collected from wearable tracking deviceslike those used in our study will be especially useful for studyingthe softer class of contact calls which tend to be produced moreoften, by more individuals and in a greater variety of contexts(Harcourt et al., 1993; Meise et al., 2011; Reber et al., 2013). Inaddition to these techniques, playback experiments need to beconducted to test inferred causal relationships between contactcall production and behavioral changes in group members,ideally while group members are wearing tracking devices.

Finally, by taking advantage of new technological approaches,our findings provide some of the first insight into collectivebehaviors of animals under ecologically complex conditions(Hughey et al., 2018). Studies indicate that many groups, suchas bird flocks and fish schools, can coordinate their behavior byresponding to the positions and trajectories of near neighbors(Reynolds, 1987; Couzin and Krause, 2003; Herbert-Read et al.,2011). However, as these studies typically focus on short timescales and/or on individuals that are moving through relativelysimple environments (King et al., 2018), it is unclear whether

Frontiers in Ecology and Evolution | www.frontiersin.org 13 March 2019 | Volume 7 | Article 73

1483

1484

1485

1486

1487

1488

1489

1490

1491

1492

1493

1494

1495

1496

1497

1498

1499

1500

1501

1502

1503

1504

1505

1506

1507

1508

1509

1510

1511

1512

1513

1514

1515

1516

1517

1518

1519

1520

1521

1522

1523

1524

1525

1526

1527

1528

1529

1530

1531

1532

1533

1534

1535

1536

1537

1538

1539

1540

1541

1542

1543

1544

1545

1546

1547

1548

1549

1550

1551

1552

1553

1554

1555

1556

1557

1558

1559

1560

1561

1562

1563

1564

1565

1566

1567

1568

1569

1570

1571

1572

1573

1574

1575

1576

1577

1578

1579

1580

1581

1582

1583

1584

1585

1586

1587

1588

1589

1590

1591

1592

1593

1594

1595

1596

O’Bryan et al. Contact Calls Facilitate Group Contraction

these rules would be effective, or even beneficial, under moreecologically complex conditions, where dense vegetation, widegroup spread and/or occupied visual attention may limit thetracking of groupmates. Rather, individuals in some groups mayrequire flexibility in their movements relative to one another.For example, Strandburg-Peshkin et al. (2017) found thatwhen baboons were moving through dense habitats, individualmovements were more influenced by vegetation than socialfactors and levels of group polarization were lower, likely becausethe habitat constrained individuals’ abilities to track others’trajectories. As another example, Ginelli et al. (2015) documentedintermittent periods of group expansion and contraction amonga flock of sheep in an enclosed pasture and proposed that,by switching between these modes, individuals can enjoy thebenefits of both exploration and protection. In this study, sheepon the periphery appeared to trigger a mass group contractionevent by running toward the center of the group. While thismay be possible when all group members can see one another,it may not be effective otherwise. During contexts in which linesof sight between group members become broken, vocalizations,such as those described in our study, could be used to initiate andfacilitate group contraction.

ETHICS STATEMENT

This study was carried out in accordance with theQ8

recommendations of Rutgers Newark Animal Care and UseCommittee’. The protocol was approved by the Rutgers NewarkAnimal Care and Use Committee. Institutional Animal Care andUse Committee (IACUC) at Rutgers University-Newark.

AUTHOR CONTRIBUTIONS

LO and AK formulated the hypotheses. LO, AK, and SGdeveloped and built the dataloggers. GC and LOwere responsiblefor the fieldwork. LO performed the data collection. LO and SGprocessed the data. LO, NA, SN, TD, DR, and SG developed

techniques for analyzing the data. LO and NA conducted the Q14

analysis. DR and SG provided guidance on the progression ofthe work and LO wrote the manuscript with feedback fromall coauthors.

FUNDING

This material was based upon work partially supported by Q7

the National Science Foundation under Grant DMS-1638521to the Statistical and Applied Mathematical Sciences Institute.Any opinions, findings, and conclusions or recommendationsexpressed in this material are those of the author(s) and do notnecessarily reflect the views of the National Science Foundation.

ACKNOWLEDGMENTS

We thank Isak Ouseb for permission to work with his livestockand the Tsaobis beneficiaries, the Snyman andWittreich families,Johan Venter, the Gobabeb Research and Training Centre, theMinistry of Environment and Tourism, and the Ministry ofLands and Resettlement, for permission and support to conductour research at Tsaobis. We are grateful to Levi Arnoldus forassistance with fieldwork, Claudia Martina, Alecia Carter, the2015 Tsaobis Baboon Project team and Herman Strydom forsupport in the field, Mark Holton and Abid Haque for thedevelopment of our recording device, Julian Kivell, Phil Hopkins,and Gaelle Fehlmann for aid in the design and production ofour collars and Nicole Dykstra and undergraduates at NJITfor aiding the processing of our GPS and audio data. Thispaper is a publication of the ZSL Institute of Zoology’s TsaobisBaboon Project.

SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be foundonline at: https://www.frontiersin.org/articles/10.3389/fevo.2019.00073/full#supplementary-material

Q17

Q9

Q15

REFERENCES

Aubin, T., Jouventin, P., and Hildebrand, C. (2000). Penguins use the two-voice

system to recognize each other. Proc. R. Soc. B Biol. Sci. 267, 1081–1087.

doi: 10.1098/rspb.2000.1112

Boinski, S. (1993). Vocal coordination of troop movement among white-

faced capuchin monkeys,Cebus capucinus. Am. J. Primatol. 30, 85–100.

doi: 10.1002/ajp.1350300202

Briefer, E., and McElligott, A. G. (2011). Mutual mother-offspring vocal

recognition in an ungulate hider species (Capra hircus). Anim. Cogn. 14,

585–598. doi: 10.1007/s10071-011-0396-3

Briefer, E. F., and McElligott, A. (2012). Social effects on vocal ontogeny

in an ungulate, the goat, Capra hircus. Anim. Behav. 83, 991–1000.

doi: 10.1016/j.anbehav.2012.01.020

Carter, G. G., Skowronski, M. D., Faure, P. A., and Fenton, , B. (2008). Antiphonal

calling allows individual discrimination in white-winged vampire bats. Anim.

Behav. 76, 1343–1355. doi: 10.1016/j.anbehav.2008.04.023

Cheney, D. L., Seyfarth, R. M., and Palombit, , R. (1996). The function and

mechanisms underlying baboon ‘contact’ barks. Anim. Behav. 52, 507–518.

Couzin, I. D., and Krause, , J. (2003). Self organization and collective behavior in

vertebrates. Adv. Stud. Behav. 32, 1–75. doi: 10.1016/S0065-3454(03)01001-5

Couzin, I. D., Krause, J., James, R., Ruxton, G. D., and Franks, N. R. (2002).

Collective memory and spatial sorting in animal groups. J. Theor. Biol. 218,

1–11. doi: 10.1006/jtbi.2002.3065

Cowlishaw, G., and Davies, J. G. (1997). Flora of the pro-namib desert swakop

river catchment, Namibia: community classification and implications for desert

vegetation sampling. J. Arid Environ. 36, 271–290. doi: 10.1006/jare.1996.0203

Dehn, M. M. (1990). Vigilance for predators: detection and dilution effects. Behav.

Ecol. Sociobiol. 26, 337–342.

Digweed, S. M., Fedigan, L. M., and Rendall, D. (2007). Who cares who calls?

selective responses to the lost calls of socially dominant group members in

the white-faced capuchin (Cebus capucinus). Am. J. Primatol. 69, 829–835.

doi: 10.1002/ajp.20398

Dunbar, R. I. M., Buckland, D., and Millert, D. (1990). Mating strategies of

male feral goats: a problem in optimal foraging. Anim. Behav. 40, 653–667.

doi: 10.1016/S0003-3472(05)80695-5

Escós, J., Alados, C.L., and Boza, J. (1993). Leadership in a domestic goat herd.

Appl. Anim. Behav. Sci. 38, 41–47. doi: 10.1016/0168-1591(93)90040-V

Frontiers in Ecology and Evolution | www.frontiersin.org 14 March 2019 | Volume 7 | Article 73

1597

1598

1599

1600

1601

1602

1603

1604

1605

1606

1607

1608

1609

1610

1611

1612

1613

1614

1615

1616

1617

1618

1619

1620

1621

1622

1623

1624

1625

1626

1627

1628

1629

1630

1631

1632

1633

1634

1635

1636

1637

1638

1639

1640

1641

1642

1643

1644

1645

1646

1647

1648

1649

1650

1651

1652

1653

1654

1655

1656

1657

1658

1659

1660

1661

1662

1663

1664

1665

1666

1667

1668

1669

1670

1671

1672

1673

1674

1675

1676

1677

1678

1679

1680

1681

1682

1683

1684

1685

1686

1687

1688

1689

1690

1691

1692

1693

1694

1695

1696

1697

1698

1699

1700

1701

1702

1703

1704

1705

1706

1707

1708

1709

1710

O’Bryan et al. Contact Calls Facilitate Group Contraction

Farine, D. R., Strandburg-Peshkin, A., Berger-Wolf, T., Ziebart, B., Brugere, I.,

Li, J., et al. (2016). Both nearest neighbours and long-term affiliates predict

individual locations during collective movement in wild baboons. Sci. Rep.

6:27704. doi: 10.1038/srep27704

Fehlmann, G., O’Riain, M. J., Hopkins, P. W., O’Sullivan, J., Holton, M.

D., Shepard, E. L. C., et al. (2017). Identification of behaviours from

accelerometer data in a wild social primate. Anim. Biotelemetry 5:6.

doi: 10.1186/s40317-017-0121-3

Fischer, J., Hammerschmidt, K., Cheney, D. L., and Seyfarth, R. M. (2001).

Acoustic features of female chacma baboon barks. Ethology 107, 33–54.

doi: 10.1046/j.1439-0310.2001.00630.x

Gall, G. E. C., and Manser, M. B. (2017). Group cohesion in foraging

meerkats: follow the moving ‘Vocal Hot Spot.’ R. Soc. Open Sci. 4:170004.

doi: 10.1098/rsos.170004

Garnier, S. (2016). SwaRm. Available online at: https://github.com/swarm-lab/

swaRm

Ginelli, F., Peruani, F., Pillot, M. H., Chaté, H., Theraulaz, G., and Bon, R.

(2015). Intermittent collective dynamics emerge from conflicting imperatives

in sheep herds. Proc. Natl. Acad. Sci. U.S.A. 112, 1–6. doi: 10.1073/pnas.15037

49112

Gros-Louis, J. J., Perry, S. E., Fichtel, C., Wikberg, E., Gilkenson, H., Wofsy, S.,

et al. (2008). Vocal repertoire of Cebus capucinus: acoustic structure, context,

and usage. Int. J. Primatol. 29, 641–670. doi: 10.1007/s10764-008-9263-8

Harcourt, A. H., Stewart, K. J., and Hauser, M. D.(1993). Functions of wild gorilla

‘close’ calls. I. Repertoire, context, and interspecific comparison. Behaviour

124, 89–122.

Herbert-Read, J. E., Perna, A., Mann, R. P., Schaerf, T. M., Sumpter, D. J., and

Ward, A. J. (2011). Inferring the rules of interaction of shoaling fish. Proc. Natl.

Acad. Sci. U.S.A. 108, 18726–18731. doi: 10.1073/pnas.1109355108

Hughey, L. F., Hein, A. M., Strandburg-Peshkin, A., and Jensen, F. H. (2018).

Challenges and solutions for studying collective animal behaviour in the wild.

Philos. Trans. R. Soc. B Biol. Sci. 373:20170005. doi: 10.1098/rstb.2017.0005

Hyndman, R. J. (2017). Forecast: Forecasting Functions for Time Series and Linear

Models. R package version 8.2. Available online at: http://pkg.robjhyndman.

com/forecast

Karatzoglou, A., Smola, A., Hornik, K., and Zeileis, A. (2004). Kernlab

- an S4 package for kernel methods in R. J. Stat. Softw. 11, 1–20.

doi: 10.18637/jss.v011.i09

King, A. J., Fehlmann, G., Biro, D.,Ward, A. J., and Fürtbauer, I. (2018). Re-wilding

collective behaviour: an ecological perspective. Trends Ecol. Evol. 33, 347–357.

doi: 10.1016/j.tree.2018.03.004

Kondo, N., and Watanabe, S. (2009). Contact calls: information and social

function. Jpn. Psychol. Res. 51, 197–208. doi: 10.1111/j.1468-5884.2009.0

0399.x

Leighty, K. A., Soltis, J., Wesolek, C. M., and Savage, A. (2008). Rumble

vocalizations mediate interpartner distance in African elephants, Loxodonta

africana. Anim. Behav. 76, 1601–1608. doi: 10.1016/j.anbehav.2008.06.022

Lynch, E., Angeloni, L., Fristrup, K., Joyce, D., and Wittemyer, G. (2013). The use

of on-animal acoustical recording devices for studying animal behavior. Ecol.

Evol. 3, 2030–2037. doi: 10.1002/ece3.608

McDougall, P. (2009). The feral goats of kielderhead moor. J. Zool. 176, 215–246.

doi: 10.1111/j.1469-7998.1975.tb03194.x

Meise, K., Keller, C., Cowlishaw, G., and Fischer, J. (2011). Sources of acoustic

variation: implications for production specificity and call categorization in

chacma baboon (Papio ursinus) grunts. J. Acoust. Soc. Am. 129, 1631–1641.

doi: 10.1121/1.3531944

Price, E. O, and Thos, J. (1980). Behavioral responses to short-term

social isolation in sheep and goats. Appl. Anim. Ethol. 6, 331–339.

doi: 10.1016/0304-3762(80)90133-9

R Core Team (2016). R: A Language and Environment for Statistical Computing.

Vienna: R Foundation for Statistical Computing.

Reber, S. A., Townsend, S. W., and Manser, M. B. (2013). Social

monitoring via close calls in meerkats. Proc. Biol. Sci. 282:20131013.

doi: 10.1098/rspb.2013.1013

Rendall, D., Cheney, D. L., and Seyfarth, R. M. (2000). Proximate factors mediating

‘contact’ calls in adult female baboons (Papio Cynocephalus Ursinus) and their Q16

infants. J. Comp. Psychol. 114, 36–46.

Reynolds, C. W. (1987). Flocks, herds, and schools: a distributed behavioral model.

Comput. Graph. 21, 25–34. doi: 10.1145/37402.37406

Ruiz-Miranda, C. R., Szymanski, M. D., and Ingals, J. W. (1993). Physical

characteristics of the vocalizations of domestic goat does capra hircus in

response to their offspring’s cries. Bioacoustics 5, 99–116.

Saunders, F. C, McElligott, A. G., Safi, , K., and Hayden, T. J. (2005). Mating tactics

of male feral goats (Capra hircus): risks and benefits. Acta Ethol. 8, 103–110.

doi: 10.1007/s10211-005-0006-y

Shank, C. C. (1972). Some aspects of social behaviour in a population

of feral goats (Capra hircus L.). Z. Tierpsychol. 30, 488–528.

doi: 10.1111/j.1439-0310.1972.tb00876.x

Shrader, A. M., Kerley, G. I. H., Kotler, B. P., and Brown, J. S. (2007). Social

information, social feeding, and competition in group-living goats (Capra

hircus). Behav. Ecol. 18, 103–107. doi: 10.1093/beheco/arl057

Siebert, K., Langbein, J., Schön, P-C., Tuchscherer, A., and Puppe, B. (2011).

Degree of social isolation affects behavioural and vocal response patterns

in dwarf goats (Capra hircus). Appl. Anim. Behav. Sci. 131, 53–62.

doi: 10.1016/j.applanim.2011.01.003

Stewart, J. C., and Scott, J. P. (1947). Lack of correlation between leadership

and dominance relationships in a herd of goats. J. Comp. Physiol. Psychol. 40,

255–264. doi: 10.1037/h0060710

Strandburg-Peshkin, A., Farine, D. R., Crofoot, M. C., and Couzin, I.

D. (2017). Habitat and social factors shape individual decisions and

emergent group structure during baboon collective movement. ELife 6:e19505.

doi: 10.7554/eLife.19505

Strandburg-Peshkin, A., Twomey, C. R., Bode, N.W., Kao, A. B., Katz, Y., Ioannou,

C. C., et al. (2013). Visual sensory networks and effective information transfer

in animal groups. Curr. Biol. 23, R709–R711. doi: 10.1016/j.cub.2013.07.059

Trillmich, J., Fichtel, , C., and Kappeler, P. (2014). Coordination of group

movements in wild Verreaux ’ s Sifakas (Propithecus verreauxi). Behaviour 141, Q16

1103–1120.

van Schaik, C. P. (1983). Why are diurnal primates living in groups?” Behaviour

87, 120–144.

Conflict of Interest Statement: The authors declare that the research was

conducted in the absence of any commercial or financial relationships that could

be construed as a potential conflict of interest.

Copyright © 2019 O’Bryan, Abaid, Nakayama, Dey, King, Cowlishaw, Rubenstein

and Garnier. This is an open-access article distributed under the terms of the Creative

Commons Attribution License (CC BY). The use, distribution or reproduction in

other forums is permitted, provided the original author(s) and the copyright owner(s)

are credited and that the original publication in this journal is cited, in accordance

with accepted academic practice. No use, distribution or reproduction is permitted

which does not comply with these terms.

Frontiers in Ecology and Evolution | www.frontiersin.org 15 March 2019 | Volume 7 | Article 73