author’s proof author queries formdir.princeton.edu/pdf_dir/2019feeapgoatcallgroupco.pdf · q18...
TRANSCRIPT
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 [email protected].
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
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